UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Genetic and developmental studies of abnormal neural tube closure in SELH/Bc mice Gunn, Teresa Monique 1995

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1996-090884.pdf [ 23.76MB ]
Metadata
JSON: 831-1.0087283.json
JSON-LD: 831-1.0087283-ld.json
RDF/XML (Pretty): 831-1.0087283-rdf.xml
RDF/JSON: 831-1.0087283-rdf.json
Turtle: 831-1.0087283-turtle.txt
N-Triples: 831-1.0087283-rdf-ntriples.txt
Original Record: 831-1.0087283-source.json
Full Text
831-1.0087283-fulltext.txt
Citation
831-1.0087283.ris

Full Text

GENETIC AND DEVELOPMENTAL STUDIES OF ABNORMAL NEURAL TUBE CLOSURE IN SELH/Bc MICE by TERESA MONIQUE GUNN B.Sc, McGill University, 1990 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Genetics Graduate Program) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA November 1995 © Teresa Monique Gunn, 1995 ) In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. DE-6 (2/88) 11 ABSTRACT The SELH/Bc mouse strain produces the lethal cranial neural tube defect, exencephaly, in 10-20% of embryos and newborns and is a model for multifactorial human neural tube defects. The mesencephalic folds of all SELH/Bc embryos fail to elevate and make contact at Closure 2, a normal site of initiation of closure at the prosencephalon/ mesencephalon boundary. The cause of this defect was investigated. Histology of the cranial neural folds of early (3-11 somite) SELH/Bc embryos showed that morphological abnormalities are present as early as the 3-somite stage. Divergence of the folds from the neural groove was more angular and the lateral tips were consistently less elevated than in normal embryos. However, the primary defect was not obvious. The hypothesis that there is a defect in the cytoskeleton of neuroepithelial cells was tested, but not supported, as actin microfilament distribution was normal in 7-somite SELH/Bc embryos. The frequencies of exencephaly observed in F l , BC1, F2, and F2-testcross progeny from a cross between the SELH/Bc and LM/Bc (normal) strains suggested segregation of 2-3 additive exencephaly-liability loci. The genetic liability to exencephaly of 102 F2 sires (based on frequency produced in 100 testcross progeny each) was determined, and linkage of genetic markers to the exencephaly trait in 10 high-risk sires (produced the highest frequencies of exencephaly) and 10 low-risk sires (produced no exencephaly) was tested. Analysis suggested liability loci on chromosomes 10 and 13, and a suppressor locus in SELH/Bc on chromosome 2. Mapping in these regions was tested in exencephalic F2 embryos, where evidence of a liability locus was strong for chromosome 13, suggestive for chromosome 2, and weak for chromosome 10. Ul The panel of testcrossed F2 sires was used to test the hypothesis that lack of Closure 2 causes exencephaly, by demonstrating co-segregation of these traits in testcross progeny of high- and low-risk F2 sires. In addition, the hypothesis that liability to cleft cerebellum-ataxia, a trait observed in 5-10% of weaned SELH/Bc mice, shares a common genetic basis with liability to exencephaly was supported as low-risk sires produced neither trait and high-risk sires produced high frequencies of both. iv TABLE OF CONTENTS Abstract ii Table of Contents iv List of Tables vii List of Figures viii Acknowledgement xii Chapter One: General Introduction 1 Neurulation 2 Mechanisms of neurulation 8 Genes and molecules expressed during cranial neural tube closure 20 Models of neural tube closure defects 35 SELH/Bc mice 54 Genetic analysis of genetically complex traits 57 The genetic basis of exencephaly in SELH/Bc mice 66 The multifactorial threshold model of inheritance 70 Quantitative traits 75 Mapping loci that cause variation in quantitative traits 77 Mapping loci for genetically complex threshold traits 79 Strategies for mapping QTLs 80 Polymorphic linkage markers 85 The genetics and development of abnormal neural tube closure in SELH/Bc mice 87 V Chapter Two: General Materials and Methods 90 Animal maintenance and breeding 90 Mouse stocks 90 Statistical Methods 93 Chapter Three: Histological study of the cranial neural folds of SELH/Bc mice 95 Introduction 95 Materials and Methods 97 Results 100 Discussion 111 Chapter Four: Genetic Analysis of the Cause of exencephaly in SELHA/Bc mice 115 Introduction 115 Materials and Methods 116 Results 127 Discussion 140 Chapter Five: Mapping the SELHA/Bc exencephaly liability loci 147 Introduction 147 Materials and Methods 150 Results 161 Discussion 197 Chapter Six: Lack of Closure 2 causes exencephaly in SELH/Bc mice 216 Introduction 216 Materials and Methods 217 vi Results 222 Discussion 229 Chapter Seven: Studies on the cause of cleft cerebellum-ataxia 234 Introduction 234 Materials and Methods 236 Results 240 Discussion 252 Chapter Eight: Conclusions and General Discussion 259 References 272 Appendix A 305 Appendix B 315 Appendix C 318 LIST OF TABLES Table 1: Genes and molecules expressed in mouse cranial neural folds on D8.5-9 of gestation. Table 2: Mouse models of cranial neural tube closure defects. Table 3: Frequencies of exencephaly and embryonic death in genetic crosses between SELHA and LM/Bc. Table 4: Comparison of observed frequencies of exencephaly with those predicted by recessive epistatic models. Table 5: Comparison of observed frequencies of exencephaly with those predicted by the multifactorial threshold model. Table 6: Ranking of 21 highest-risk F2 sires by percentage exencephaly, mean-litter Freeman-Tukey arcsine frequencies of exencephaly, and corrected (for potential temporal fluctuations in exencephaly penetrance) mean-litter Freeman-Tukey arcsine frequencies. Table 7: Genotype of all F2 animals typed at the best chromosome 2, 10, and 13 markers. Table 8: Numbers of D8-D9 embryos scored for stage of cranial neural tube closure in SELH, LM/Bc, and F l , and from test-crosses of high-, intermediate-, and low-risk F2 sires. Table 9: Frequencies of exencephaly and cleft cerebellum produced in the testcross progeny of high- and low-risk F2 sires, and in SELH and LM/Bc embryos. Table 10: Twelve highest exencephaly-producing F2 sires ranked for production of exencephaly, cleft cerebellum, and both traits combined in their testcross progeny. LIST OF FIGURES viii Figure 1: Stages of neural tube closure in the mouse. 4 Figure 2: Multifactorial threshold model for neural tube closure. 71 Figure 3: Liability to exencephaly in SELH/Bc mice as a quantitative trait. 73 Figure 4: Transverse sections through headfold region of 3-5 somite stage SELH/Bc, ICR/Be, and LM/Bc embryos. 101 Figure 5: Overlaid tracings of profiles of 3-5 somite stage SELH/Bc and ICR/Be embryos. 102 Figure 6: Transverse sections through cranial neural folds of 6-8 somite stage SELH/Bc, ICR/Be, and LM/Bc embryos. 104 Figure 7: Overlaid tracings of profiles of 6-8 somite stage SELH/Bc and ICR/Be embryos, and of SELH/Bc and LM/Bc embryos. 105 Figure 8: Transverse sections through cranial neural folds of 9-11 somite stage SELH/Bc, ICR/Be, and LM/Bc embryos. 106 Figure 9: Lateral aspect of rostral mesencephalic neural folds of 9-11 somite stage SELH/Bc and ICR/Be embryos. 108 Figure 10: Transverse sections through mesencephalon of 7 somite SELH/Bc and ICR/Be embryos stained with fluorescein-phalloidin to visualize actin microfilaments. 109 Figure 11: Transverse section through mesencephalon of 7 somite SELH/Bc embryo stained with fluorescein-phalloidin to visualize actin microfilaments. 110 Figure 12: Distributions of exencephaly-liability of SELH/Bc, LM/Bc, and F l , BC1, F2 and F2-testcross generations from crosses between them on liability scale (probits). 123 Figure 13: Distributions of breeding values of individual sires for various crosses between SELHA/Bc and LM/Bc mice. 128 Figure 14: Fit of exencephaly-liability to multifactorial threshold model. 132 Figure 15: Expected distribution of breeding values of F2 sires if exencephaly-liability is caused by 2 additive loci. 135 IX Figure 16: Expected distribution of breeding values of F2 sires if exencephaly -liability is caused by 1 additive locus. 135 Figure 17: Distribution of mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly produced by F2 sires. 137 Figure 18: Smoothed frequency distribution of F2 sire breeding values. 138 Figure 19: Expected smoothed frequency distribution of F2 sire breeding values if 1 additive liability locus is segregating. 139 Figure 20: Plot of frequencies of exencephaly produced by SELHA sires over time. 154 Figure 21: Plot of frequencies of exencephaly produced by F2 sires over time. 154 Figure 22: Linear regression of exencephaly-production by F2 sires over time. 155 Figure 23: Polynomial regression of exencephaly-production by F2 sires over time. 155 Figure 24: Map of D_Mit_ SSLP makers used for linkage analysis. 162 Figures 25-27: Summed number of SELHA alleles at markers on chromosomes 1-19 in high- and low-risk F2 sires. 166-168 Figures 28-30: X 2 values at markers on chromosomes 1-19 for combined high-and low-risk sire data. 169-171 Figure 31: Genotype at D10Mitl64 vs genotype at D13Mit39 in high- and low-risk F2 sires. 173 Figure 32: Genotype at D3Mit22 vs genotype at D10Mitl64 + D13Mit39 in high-and low-risk F2 sires. 175 Figure 33: Genotype at D5Mitl68 vs genotype at D10Mitl64 + D13Mit39 in highl-and low-risk F2 sires. 175 Figure 34: Genotype at D17Mitl0 vs genotype at D10Mitl64 + D13Mit39 in highl-and low-risk F2 sires. 176 Figure 35: Genotype at D2Mit7 or D2Mit92 vs genotype at D10Mitl64 + D13Mit39 in high- and low-risk F2 sires. 177 Figure 36: Photographs of PCR products of high- and low-risk F2 sires at D2Mit7, D10Mitl80, and D13Mit39. 179 Figure 37: Graphical representation of the relationship between exencephaly-liability and number of SELHA alleles carried at exencephaly-liability loci. 180 Figure 38: Summed number of SELHA alleles at chromosome 2, 10 and 13 markers in 31 exencephalic F2 embryos. 181 Figure 39: Photographs of PCR products of exencephalic F2 embryos at D2Mit7, D10Mitl80, and D13Mit39. 182 Figure 40: X 2 values at markers on chromosomes 2, 10, and 13 in exencephalic F2 embryos. 184 Figure 41: Genotype at D10Mitl64 vs D13Mit39 or best genotype of D10Mitl80 or D10Mitl64 vs best genotype of D13Mitl-0 or D13Mit39 for 31 exencephalic F2 embryos. 185 Figure 42: Genotype at D2Mit7 or D2Mit92 vs genotype at D10Mitl64 and D13Mit39 in 31 exencephalic F2 embryos. 187 Figure 43: Best genotype of D2Mit7 or D2Mit92 vs best genotype of D10Mitl80 or D10Mitl64 + best genotype of D13MUT0 or D13Mit39 in 31 exencephalic F2 embryos. 188 Figure 44: Summed number of SELHA alleles in high- and low-risk F2 sires, using re-ranked high-risk group. 194 Figure 45: X 2 values at markers on chromosomes 2, 10, and 13 in high- and low-risk F2 sire data, using re-ranked high-risk group. 195 Figure 46: Genotype at D10Mitl64 vs genotype at D13Mit39 in high- and low-risk F2 sires, using re-ranked high-risk group. 196 Figure 47: Genotype at D2Mit7 vs genotype at D10Mitl64 + D13Mit39 in high-and low-risk F2 sires, using re-ranked high-risk group. 198 Figure 48: Genotype at D2Mit92 vs genotype at D10Mitl64 + D13Mit39 in high-and low-risk F2 sires, using re-ranked high-risk group. 198 Figure 49: Map of chromosome 13, showing candidate region. 202 Figure 50: Map of chromosome 10, showing most likely candidate region. 206 Figure 51: Map of chromosome 2, showing candidate region. 208 XI Figure 52: Diagrammatic side-view representations of normal and SELH-like embryos, showing closure pattern and location of ANP. 223 Figure 53: Comparison of pattern of cranial neural tube closure in SELH/Bc, LM/Bc, and F l embryos. 224 Figure 54: Comparison of pattern of cranial neural tube closure in testcross progeny of high- and low-risk F2 sires. 226 Figure 55: Relationship between frequency of exencephaly and frequency of Closure 2 in testcross progeny of F2 sires. 230 Figure 56: Position of cut made to examine cerebella of D14 embryos. 237 Figure 57:' Photographs showing section through head of LM/Bc and SELH embryos, demonstrating appearance of normal and cleft cerebella. 242 Figure 58: Relationship between frequency of exencephaly and frequency of cleft cerebellum in testcross progeny of F2 sires. 243 Figure 59: Distribution of LM/Bc embryos among stages of cranial neural tube closure, indicating timing of initiation of Closure 4 over rhombencephalon. 245 Figure 60: Distribution of SELH/Bc embryos among stages of cranial neural tube closure, indicating timing of initiation of Closure 4 over rhombencephalon. 247 Figure 61: Distribution of F l embryos among stages of cranial neural tube closure, indicating timing of initiation of Closure 4 over rhombencephalon. 249 Figure 62: Distribution of testcross progeny of low-risk F2 sires among stages of cranial neural tube closure, indicating timing of initiation of Closure 4 over rhombencephalon. 250 Figure 63: Distribution of testcross progeny of high-risk F2 sires among stages of cranial neural tube closure, indicating timing of initiation of Closure 4 over rhombencephalon. 251 Figure 64: Revised threshold model for neural tube closure defects in SELH/Bc mice. 268 XII Acknowledgement I would like to thank my supervisor, Diana Juriloff, for sharing her time and expertise, and for making my experience as a graduate student enjoyable. I would also like to thank Muriel Harris who, although not an official co-supervisor, has acted as one in many ways, and my other committee members, Dixie Mager and Wayne Vogl for welcoming me into their labs and providing useful comments and suggestions throughout the last five years. I would also like to acknowledge the help of my lab-mates, Randall Dreger, for helpful conversations, and Diana Mah, for patiently answering my PCR questions, typing the three additional high-risk sires for chromosome 2, 10 and 13 markers, and (most of all) her friendship; also, Arturo Malig, Eunah Makoni, Norah Moorehead, and Eduardo Alfonso Perez for their conscientious care of my mice. Finally, I would like to thank my family for their love and support, especially my husband, Dayne, and my parents, John and Ursule Betts. Chapter 1: GENERAL INTRODUCTION 1 Neurulation, the formation of the neural tube, is an important early morphogenic event in vertebrate development. It is from the neural tube that the brain and spinal cord differentiate to form the central nervous system. The elevation and fusion of the neural folds to close the neural tube is a complex process thought to involve the concerted effort of many different forces, but the origin and action of these forces is as yet poorly understood. Neural tube closure defects (NTD) are among the most common congenital abnormalities in human populations. Failure of the posterior neural tube to complete closure results in spina bifida, while failure of the cranial neural tube to close results in anencephaly. The incidence of spina bifida and anencephaly together varies between populations and ranges from less than 0.1% to more than 1.0% of births (Laurence, 1983; Fraser and Nora, 1986). The differences in incidence of NTD between ethnic groups, the increased incidence of consanguinity in parents of affected individuals, and the elevated recurrence risks in relatives of affected individuals suggest a genetic etiology to NTD, although environmental factors, such as folate nutrition (MRC, 1991; Czeizel and Dudas, 1994), also appear to be involved. NTD can also be caused by exposure to chemical teratogens such as valproic acid (Lammer et al., 1987). Maternal illness, such as diabetes (Soler et al.,-1976; Zacharias et al., 1984), can also increase the risk of having an affected child. Most human NTD do not fit simple Mendelian patterns of inheritance. The recurrence risk for sibs of affected individuals is about 1-6%, and about 3% for the child of an affected (spina bifida) parent (Carter, 1974; Campbell et al., 1986). The recurrence risk to second and third degree relatives is also elevated over the risk for the general population. Based on these observations, most NTD are considered to be multifactorial threshold traits (Leek, 1972; Carter, 2 1974; Laurence, 1983; Campbell et al., 1986; Fraser and Nora, 1986; Thompson and Thompson, 1986). Neurulation Neurulation in mammalian embryos occurs in two phases. Primary neurulation involves the formation and shaping of the neural plate and bending, elevation and fusion of the neural folds. Secondary neurulation involves the condensation of a cord of cells within the tail bud mesenchyme to form an epithelial tube that is continuous with the tube produced by primary neurulation. Basically, primary neurulation is the process whereby the flat layer of ectodermal cells that overlie the notochord is transformed into a hollow tube (Gilbert, 1988). A brief description of neurulation in general is given here; details of cranial neurulation will be given in the following discussion. The first step in neurulation is a change in cell shape. The neural plate is formed when the cells over the notochord become elongated, causing them to rise above the adjacent cells, which become more flattened. The elevated mid-line cells constitute the neural plate. The edges of the neural plate thicken (a process called "shaping") and begin to move upward ("bending"), forming the neural folds. This causes a V-shaped neural groove to appear in the centre of the neural plate, between the neural folds. The neural folds continue to curve and move upward, elevating to approach each other across the neural groove until their tips appose, make contact, and adhere to one another, fusing to form a continuous layer of cells across the neural groove. As the neural folds fuse, the cells that were originally adjacent to the neural plate also fuse across the midline to become the overlying epidermis. During neural tube closure, neural crest cells (NCCs) arise from the conversion of epithelial cells into mesenchyme cells at the lateral tips of the neural plate (Erickson and Weston, 3 1983; Tan and Morriss-Kay, 1985, 1986; Nichols, 1981, 1986, 1987; Chan and Tam, 1988; Fukiishi and Morriss-Kay, 1992; Serbedzija et al., 1992). These cells then migrate extensively through the embryo to give rise to many cell types, including pigment cells and certain cells of the peripheral nervous system. In mouse embryos, timing of NCC migration varies along the length of the neural tube. More rostrally located cells migrate earliest, during elevation of the neural folds, while the NCCs located more caudally, in the spinal region, migrate after fusion of the neural folds. Historically, it was thought that neural tube closure progressed by a "zipper-like" mechanism, where fusion begins at one point and proceeds in either direction to the ends of the neural tube. It has now been shown that there are several discrete sites of initiation of fusion in the mouse embryo, as represented in Figure 1 (Waterman, 1976; Geelen and Langman, 1977; Kaufman, 1979; Jacobson and Tam, 1982; Golden and Chernoff 1983, 1993; Macdonald et al., 1989; Sakai, 1989; Juriloff et al.,1991). It has been proposed that human embryos also initiate fusion of the neural tube at multiple sites (Van Allen et al., 1993; Busam et al., 1993; Seller, 1995), although embryological evidence for this is difficult to obtain. Based on the observations described by several groups (Waterman, 1976; Geelen and Langman, 1977; Kaufman, 1979; Jacobson and Tam, 1982; Golden and Chernoff 1983, 1993; Macdonald et al., 1989; Sakai, 1989; Juriloff et al.,1991), the closure of the cranial neural folds in mice takes place as follows. The cranial neural folds are first apparent as a small symmetrical pair of convex bulges in the cephalic region of the neural plate of pre- and early somite stage mouse embryos. In embryos with 2-4 somite pairs ("somites"), there is growth and lateral expansion of the neural folds as the rostral and rostrolateral edges of the prospective prosencephalic region begin elevation. The growth of the cephalic neural plate appears to be a result of an accumulation of mesenchyme cells and extracellular matrix in the neural folds 5 (Jacobson and Tam, 1982; Morris and Solursh, 1978a,b). By the 5-6 somite stage, the cranial neural folds become flattened and the curvature of the rostral and rostrolateral edges of the prosencephalon reverses as the folds become concave in shape. It is at this stage that the cranial flexure (a sharp ventral bend in the rostral neural plate) becomes clearly evident. This flexure develops at about the 1 somite stage at the most rostral end of the prosencephalon and the position of its apex moves caudally until about the 13 somite stage, at which time it is located at about the prosencephalon/mesencephalon boundary (Jacobson and Tam, 1982). The first site of initiation of fusion of the neural tube, Closure 1, occurs in the cervical region of 7 somite stage embryos (Fig. la-b). In these early stages, mouse embryos are "U" shaped, or doing a "backbend" so that their most caudal end is brought behind their most rostral end (see Fig. la-b). At the bottom of the U, there would likely be little stress on the neural tube in the direction opposite to fusion; it is in this region of dorsal reflection that Closure 1 takes place. Fusion in this region is achieved when the apices of the neural folds make contact across the mid-line and cell surface projections of cells at the surface epithelium/neuroepithelial boundary interdigitate (Copp et al., 1990). The cell surface of these cells is coated by carbohydrate-rich material (probably glycoproteins) that may play an important role in cell recognition and/or in mediating initial contact between the projections from opposing cells (Luft, 1976; Sadler, 1978). There is bidirectional continuation of Closure 1, with fusion spreading "zipper"-like rostral toward the caudal rhombencephalon and caudal through the spine to the zone of transition between primary and secondary neurulation. During this time, the folds of the prosencephalon and mesencephalon continue to elevate toward each other across the neural groove, and the cranial neural plate narrows. The first site of initiation of fusion in the cranial region, Closure 2, is at the prosencephalon/mesencephalon boundary at about the site of the cranial flexure (Fig. lc-d). 6 Presumably, the forces acting against closure of the neural tube at a location where it is being bent opposite to the direction of closure would be considerable, and it seems reasonable to suppose that the mechanisms of neural tube closure here would differ from those of Closure 1, which might be more passive. Closure 2 occurs in embryos with 10-15 somites, and initial contact of the apposed neural folds is between neuroepithelial cells (Geelen and Langman, 1977, 1979; Juriloff et al., 1991). Fusion continues in both directions (rostral and caudal) from the site of Closure 2. Continuation in the caudal direction may pause while fusion extends rostral into the prosencephalon (Waterman, 1976; Juriloff et al., 1991). Extension of fusion into the prosencephalon from Closure 2 also appears to involve initial contact between neuroepithelial cells (Waterman, 1976; Geelen and Langman, 1977, 1979). Scanning electron-microscopy (Waterman, 1976; Juriloff et al., 1991) suggests that caudal extension from Closure 2 into the mesencephalon may involve initial contact between projections that extend out from the surface ectoderm/neuroepithelium cell boundary; some projections appear to extend from neuroepithelial cells. Transmission electron-microscopy (Geelen and Langman, 1979) and histological sections (Geelen and Langman, 1977), however, suggest that fusion in this region involves interdigitation of extensions from the surface ectoderm of the apposed neural folds, followed immediately by contact between neuroepithelial cells. Whichever cells these ruffles, or microvilli, extend out from, they appear just prior to the meeting of the neural folds (Waterman, 1976; Geelen and Langman, 1977, 1979). These cells are coated by carbohydrate-rich material that appears to be the same as that found on the cells at the site of Closure 1. Extensive junctional complexes develop between the membranes of interdigitated microvilli. Subsequent interdigitation of neuroepithelial cells is also followed by (less extensive) membrane specializations, such as tight junctions. Approximately concurrent with Closure 2 is a third site of fusion, Closure 3, which 7. begins at the most rostral and ventral end of the neural tube and extends caudally to meet the advancing edge of Closure 2 (Fig. lc, e). Fusion in the prosencephalon by Closure 3 involves initial contact between numerous processes protruding from the apical ends of neuroepithelial cells (Waterman, 1976; Geelen and Langman, 1979). Unlike fusion at Closure 2, no distinct membrane specializations were seen to form between neuroepithelial cells following contact at Closure 3 (Geelen and Langman, 1979). The timing of Closure 3 relative to Closure 2 varies among inbred strains of mice. Closure 2 has been observed preceding Closure 3 in randomly-bred CFLP (Kaufman, 1979), LM/Bc, SWV/Bc (Juriloff et al., 1991; Golden and Chernoff, 1993), and AFJ/RkBc embryos (Juriloff et al., 1991), and Closure 3 appears to be delayed in ICPv/Bc embryos (Juriloff et al., 1991). In contrast, Closure 3 has been observed prior to Closure 2 in Jcl:ICR (Sakai, 1989) and CD-I mouse embryos (Waterman, 1976). The last of the cephalic regions to fuse is the rhombencephalon. Closure 4 initiates fusion over this region in embryos with 16-22 somites (Fig. lf-g). Fusion over the rhombencephalon is initiated by contact between surface ectoderm cells, followed by contact between what are probably neural crest cells (Geelen and Langman, 1977). Only after initial contact is established do neuroepithelial cells fuse in the dorsal midline (Geelen and Langman, 1977, 1979). Thus, unlike fusion in other regions of the neural tube, Closure 4 appears to occur by the elongation of a membrane rather than by fusion of the neural folds (see Fig. Ig; Golden and Chernoff, 1993). The neuroepithelium in the fused rhombencephalon appears to be only one cell layer thick, resulting in the covering over the rhombencephalon being thinner and more membranous in appearance than the opaque covering over the rest of the neural tube. Closure 4 appears to initiate fusion first from the caudal end of the rhombencephalon (Fig. Ig), from the termination point of Closure 1 (Golden and Chernoff, 1993). As this fusion proceeds, closure also begins at the rostral end of the rhombencephalon, from the caudal edge of Closure 2. 8 Fusion proceeds from each end of the rhombencephalon toward the middle (Fig. Ih) until they meet (Fig. 5c in Macdonald et al., 1989 and Fig. 6e,f in Copp et al., 1991; origin of fusion from the rostral rhombic lip is not described by Golden and Chernoff, 1993). It is possible that Closure 4 does not really exist, but that the rhombencephalon is simply closed by rostral extension from Closure 1 and caudal extension from Closure 2 (as in Sakai, 1989; Fig. 1 in Copp et al., 1994). Fusion from the rostral and caudal rhombic lips appears to be most advanced at the lateral aspects of the rhombencephalon, the advancing edge of fusion appearing "U"-shaped (see Fig. Ig). Thus, fusion proceeds toward the midline from all directions and the last region to close is a small oval-shaped area over the middle of the rhombencephalon (see Fig. Ih). Mechanisms of neurulation It is thought that several developmental events act together in the formation of the neural plate and neural folds to provide the mechanical force that elevates the folds, and that the specific combination of forces may differ at different levels of the neural tube. For example, the direction of embryo bending and the volume of mesenchyme of the neural folds is different in cranial and spinal regions, and these differences in the morphology of the neural folds at each of the sites of initiation of contact (Closures 1-4) are likely to require different forces or different combinations of forces to cause the neural folds to elevate and make contact. The forces that drive neural tube closure are as yet poorly understood, although several mechanisms have been proposed. The forces that are likely to influence primary neurulation of the head region will be the focus here. 9 Formation and shaping of the neural plate: Formation of the neural plate involves elevation of the neural plate above the surrounding prospective surface ectoderm, while shaping involves lengthening of the neural plate in the cranio-caudal direction, and narrowing (all levels) and subsequent widening (cephalic region) of the neural plate (Schoenwolf and Smith, 1990). These events are thought to involve forces intrinsic to the neuroepithelium, such as cell shape, cell division, and cell rearrangement. Cell elongation contributes to formation of the neural plate by causing the cells within the prospective neural plate to elongate and rise above the adjacent cells (Schoenwolf and Smith, 1990). Cell elongation is also thought to contribute to shaping of the neural plate, by contributing to its transverse narrowing, since cells become narrower as they get taller. The main force underlying cell elongation is thought to be the orientation of microtubules parallel to the long axis, although cortical tractoring and localized changes in intercellular adhesion of neuroepithelial cells have also been proposed to play a role (discussed in the following section). If microtubules are depolymerized in cultured chick embryo neuroepithelial cells, the elongated cells decrease in height by an average of 25%, eventually rounding up as they enter metaphase (Karfunkel, 1972; Schoenwolf and Powers, 1987). In addition, depolymerization and re-polymerization of the microtubules shows a direct relationship with neuroepithelial cell height and width of the neural plate (Schoenwolf and Powers, 1987), where the neural plate narrows as the cells become elongated. However, the degree of narrowing that can be accounted for by cell elongation due to microtubules is only about 30% of what one sees during shaping of the neural plate, indicating that other mechanisms must also be involved. Abnormal shaping of the neural plate would be expected to interfere with the subsequent elevation and fusion of the neural folds. Exposure of mouse (O'Shea, 1981) and hamster (Ferm, 1963) embryos to colchicine, which disrupts microtubules, results in failure of closure of the cephalic neural tube. 10 Exposure of mouse embryos to the anaesthetic xylocaine, which is thought to depolymerize microtubules, also inhibits cephalic neural tube closure (O'Shea, 1981), and scanning electron microscopy of the neuroepithelial cells shows that they are only elongated slightly. This supports the hypothesis that microtubules play an important role in neuroepithelial cell elongation, but that they are not the only force driving this change in cell shape. The cortical tractor model (Jacobson et al., 1986; Jacobson, 1994) is based upon the hypothesis that the cells of an epithelial sheet have motile activity, but are firmly attached to one another at their apical boundaries. Under the model, epithelial cell motion is characterized by a reverse "fountainhead" flow of the cell cortex from the basal and/or lateral aspects of the cell to the apical region. This flow pattern is the cortical tractor, and it can be triggered by changes in the local environment, perhaps involving ionic stimuli. For the cortical tractor to drive cell movement, there must be adhesion molecules that anchor the cell to its substrate or to neighbouring cells, and they would be cycled along with the cell cortex. If the speed of the cortical tractor differs between adjacent cells, the faster cell will try to crawl backwards out of the sheet. The apical adhesion complexes would prevent a cell from leaving the epithelial sheet, resulting instead in elongation of the faster cell. In this manner, cortical tractoring could cause cuboidal epithelial cells to become columnar, contributing to elevation of the prospective neural plate cells above the surrounding epithelium and to narrowing of the neural plate. As will be discussed later, cortical tractoring is also thought to contribute to bending of the neural folds. It has also been proposed that localized changes in intercellular adhesion could play a role in neuroepithelial cell elongation and neural plate narrowing (Gustafson and Wolpert, 1967; Karfunkel, 1974; Karfunkel et al., 1978). Under this hypothesis, increased lateral adhesion between neuroepithelial cells would increase the area of the lateral membranes, making the cells taller, and decrease their apical and basal areas, making the cells narrower. There is no 11 experimental evidence to support this as a mechanism of neuroepithelial cell elongation, but it has also been suggested (Gordon, 1985) that appropriate studies have not been performed. Cell rearrangement is thought to play a role in neural plate narrowing and lengthening by moving neuroepithelial cells mediolaterally. Cell division is also thought to contribute to the lengthening and narrowing of the neural plate, by inserting daughter cells in the longitudinal axis. In the cephalic neural plate, insertion of daughter cells in the transverse plane would contribute to the widening seen in this region. Schoenwolf and Alvarez (1989) showed that the amount of cell rearrangement and division seen in chimeras, where cephalic quail cells are transplanted into the corresponding cephalic region of chick embryos, is sufficient to approximate the shaping of the neural plate that is observed during neurulation. Intrinsic forces in bending and elevation of the neural folds: Bending of the neural plate to form the neural folds and elevation and convergence of the neural folds across the neural groove are thought to involve forces intrinsic and/or extrinsic to the neural plate (Karfunkel, 1974; Gordon, 1985; Schoenwolf and Smith, 1990). Historically, much attention has been paid to the role of intrinsic forces, particularly those that cause neuroepithelial cell wedging; if a flat sheet of cells changes shape from columnar to wedged and if the cells are firmly attached to their neighbouring cells, the sheet would be expected to curve laterally. Recent work has shown that only about one quarter of neuroepithelial cells become wedge-shaped during bending of the neural plate (Schoenwolf and Franks, 1984; Schoenwolf, 1984; Schoenwolf and Smith, 1990). The main areas of cell wedging are the median hinge point (MHP) and the dorsolateral hinge points (DLHPs). The MHP is formed by furrowing of the midline neural plate as it is anchored to the underlying notochord. Thus, the cells of the MHP line the neural groove. More than 70% of cells in the MHP are 12 wedge-shaped. The DLHPs, which are seen only in the cranial region, form by furrowing of the neural plate as it is anchored to the adjacent surface ectoderm. The DLHPs form near the juncture where the surface ectoderm and neuroepithelium meet, near the tips of the neural folds. As the neural folds elevate and converge, their lateral aspects rotate around the DLHPs until they meet in the dorsal midline. One of the possible causes of neuroepithelial cell wedging is basal positioning of the nucleus, causing basal expansion. As a neuroepithelial cell progresses through the cell cycle, the nucleus undergoes interkinetic migration, changing its position in the cell (Sauer, 1935; Schoenwolf and Smith, 1990). The nuclei of cells undergo M-phase at the apex of the neural plate, but reside at the base of neuroepithelial cells during S- and non-S phases. The position of the nucleus within the cell always coincides with the widest portion of the cell (Schoenwolf and Franks, 1984; Schoenwolf and Smith, 1990). It has been demonstrated that the cell cycle time is altered in MHP cells, with less time spent in M-phase in these cells than in the lateral neuroepithelial cells (Smith and Schoenwolf, 1987, 1988). Interkinetic nuclear migration involves microtubules (Messier, 1978), and disruption of microtubules in neurulating chick embryos by treatment with colchicine decreases the percentage of wedge-shaped cells within the neuroepithelium (Fernandez et al., 1987). Thus it is possible that cell wedging comes about by alteration of the cell cycle, causing the nucleus to spend more time in the base of the cell. This would lead to basal expansion of the cells, making them wedge-shaped. It has also been proposed that cell wedging may come about by the contraction of a ring of apically arranged actin microfilaments (Baker and Shroeder, 1967; Burnside, 1973; Karfunkel, 1974). Morphometry and computerized three-dimensional reconstruction of the neural tube demonstrated a strong relationship between the pattern of neuroepithelial bending and the degree of apical constriction of neuroepithelial cells (Bush et al., 1990). Along the length of the neural 13 tube, cells within regions of concave bending of the neuroepithelium showed a greater degree of apical constriction and more prominent bundles of microfilaments than cells in flat regions or areas of convex bending. Early in neurulation, when the cranial neural folds of mouse embryos are biconvex in shape, actin is localized to the basal aspect of neuroepithelial cells. As the folds become concave, the pattern of actin distribution shifts to the apical regions (Sadler et al., 1982). Actin microfilaments in the apical regions of neuroepithelial cells of chick embryos are associated with developing junctions and appear more organized than elsewhere in the cell (Nagele and Lee, 1980). Actin microfilaments in the chick are actually organized as a loose meshwork of several discrete bundles (Nagele and Lee, 1980) rather than as a ring as previously thought, but their arrangement suggests that microfilament contraction could cause apical constriction of neuroepithelial cells. The distribution patterns of fodrin, a form of spectrin that is thought to anchor actin to the cell membrane, have also been investigated using fluorescent antibodies (Sadler et al., 1986). The fluorescence pattern at the biconvex stage is strongest in the apical regions of neuroepithelial cells of the neural groove. Staining is weak and patchy in the rest of the neuroepithelium. As the neural folds become concave, and during elevation and convergence of the neural folds, there is a bright line of fluorescence along the apices of neuroepithelial cells. The strongest staining was at the lateral furrows, where the most bending occurs. These distribution patterns suggest that actin microfilaments are important in generating the concave shape of the neural folds and bending at the lateral furrows (also called the dorsolateral hinge points). The anaesthetic xylocaine is thought to depolymerize' microfilaments as well as microtubules. The neuroepithelium of mouse embryos exposed to xylocaine appears disorganized, and a smaller proportion of cells than in normal embryos are wedge-shaped 14 (O'Shea, 1981). That some cells still become wedge-shaped suggests that cell wedging may not be due solely to the action of actin microfilaments or microtubules, since xylocaine is thought to act on both. It was not specified whether cell wedging still occurred in any particular region or whether some wedging occurred throughout the neuroepithelium. Inhibition of microfilament structure and function by treatment of chick (Karfunkel, 1972) and rat (Morriss-Kay and Tuckett, 1985) embryos with cytochalasins B and D, respectively, leads to collapse of the neural folds. At the cellular level, treatment of mouse embryos with cytochalasin B (Webster and Langman, 1978) disrupts interkinetic nuclear migration as well as apical microfilaments. Exposure of mouse embryos to cytochalasin B does not appear to affect neural fold elevation, but the folds do not make contact and fuse across the midline (O'Shea, 1981). Exposure to cytochalasin D on days 7-11 of gestation produces exencephaly in about 5-30% of surviving fetuses (depending on dose) in C57BL/6J and BALB/c mice (Shepard and Greenaway, 1977); this suggests that neuroepithelial cell wedging may be more important for neural fold elevation in the head than in the trunk. Treatment of chick embryos with cytochalasin D (Schoenwolf et al., 1988) does not prevent MHP formation (including cell wedging) or neural fold elevation, but wedging of the DLHP cells and convergence of the neural folds is affected and the neural folds do not fuse across the midline. Neural crest cell migration is also prevented. This suggests that contraction of apically arranged actin microfilaments may be necessary for the wedging DLHP cells but not of MHP cells. It is possible that microtubules and microfilaments both play important roles in neuroepithelial cell wedging, possibly in different regions of the neural plate; for example, microtubules may be more important in causing wedging of MHP cells and microfilaments more important for DLHP cell wedging. It is also possible that microtubules may play a greater role in cell wedging than microfilaments at some level(s) of the neural tube (i.e., perhaps in the trunk) and that microfilaments may be 15 more important at other regions (i.e, in the cranial neural tube). Microsurgical separation of prospective MHP cells from more lateral neuroepithelial cells in chick embryos results in wedge shaped cells within the isolated MHP region (Schoenwolf, 1988). This suggests that cell wedging in this region is an active event that occurs independently of forces generated by other shaping and bending forces. Bilateral separation of lateral tissues (surface ectoderm and neuroepithelium) by microsurgery in chick embryos prior to neural fold formation allows normal shaping of the neural plate but inhibits subsequent neurulation at all levels except the forebrain (Schoenwolf, 1988). This indicates that forces intrinsic to the neuroepithelium are sufficient for shaping, bending, and fusion of the forebrain neural folds in chick. At all other levels, intrinsic forces appear to be sufficient for normal shaping (formation and thickening of the neural plate), while additional, extrinsic forces are required for neural fold elevation and fusion, as will be discussed below. Other intrinsic forces, in addition to those generated by microtubules and microfilaments, have been proposed to play a role in neural fold bending. Under the cortical tractoring model (Jacobson et al., 1986; Jacobson, 1994), as neural plate cells tractor, the cell surfaces that abut cells of the surface epiderm will either adhere to or be repelled by the epidermal cells. If there is a difference in the tractor speed between neural plate and epidermal cells, the neuroepithelial cells will crawl beneath the epidermal cells, which will be drawn over the neural plate cells toward the midline. The torque produced would force the edges of the neural plate up to form the neural folds. The expected sequence of events agrees well with what is seen in cross sections of newt embryos at appropriate stages of neurulation (Fig. 10 in Jacobson et al., 1986). Changes in cell adhesion may also result in epithelial invagination (Gustafson and Wolpert, 1962, 1967; Ettensohn, 1985). An increase in intercellular adhesiveness at one apex of the cells of a sheet would lead to an increase in the extent of lateral contact between the cells at that apex. 16 This would result in wedging of the cells, which could in turn cause the sheet to curve. Cell growth and cell division might also lead to invagination of an epithelial sheet (Ettensohn, 1985). This model assumes that cell growth and/or cell division lead to "crowding" of cells within the sheet, resulting in lateral pressure. If there are constraints on the lateral expansion of the apical aspect of the epithelium, the cells would become elongated and wedge-shaped, resulting in curving of the sheet. Extrinsic forces in bending and elevation of the neural folds: Extrinsic forces thought to play important roles in neural fold bending and elevation include longitudinal and axial stretching, expansion of mesenchyme and/or the extracellular matrix, and neural crest cell migration. If the neural plate is considered as an elastic sheet, then elongation of the notochord could result in stretching that sheet along a line, causing it to buckle and roll into a tube around the line of stretching (Jacobson, 1978). This phenomenon of transverse buckling has been referred to as Eulerian buckling (Wainwright et al., 1976). A similar force could be generated by the stretching that would accompany the elongation of the embryo and neural plate that was discussed earlier. This axial stretch force could cause dorsal buckling of the neural plate to form the neural tube (Jacobson and Gordon, 1976). In chick embryos, forces intrinsic to the neuroepithelium appear to be involved in bending and elevation of the forebrain neural folds, while extrinsic forces appear to be essential in more caudal regions (Schoenwolf, 1988; Schoenwolf and Smith, 1990). Microsurgical removal of chick MHP cells and some of the lateral neuroepithelial cells, as well as the tissues directly underlying them, results in formation of neural folds with subsequent elevation, convergence and fusion (Smith and Schoenwolf, 1991), suggesting that the extrinsic forces are generated by 17 tissues lateral to the neuroepithelium. The tissue underlying the neuroepithelium of mouse cranial neural folds consists of mesenchyme cells embedded in an extracellular matrix (ECM) that is rich in heparin and chondroitin sulphate glycosaminoglycans (HSPG and CSPG), hyaluronate, fibronectin and laminin (Solursh and Morriss, 1977; Morriss and Solursh, 1978a; Heifetz et al., 1980; Tuckett and Morriss-Kay, 1986). Strands of ECM stretch from the mesenchyme cells to the basement membrane of the neuroepithelium (Morriss and Solursh, 1978a). Heparitinase treatment of rat embryos shows that the presence of HSPG in and around the neuroepithelial basement membrane is necessary for bending of the neural folds from convex to concave during neural fold elevation (Tuckett and Morriss-Kay, 1989). Exposure to retinoic acid appears to delay cranial neural fold elevation (Tom et al., 1991; Morriss-Kay and Mahmood, 1992), and is associated with a loss of HSPG and CSPG (Morriss-Kay and Mahmood, 1992). The actual role of HSPG in neural fold bending is as yet unclear. Another ECM component that may play a role in cranial neural fold elevation is hyaluronate. Hyaluronate has been shown to cause hydration and expansion of intercellular spaces in biological and experimental systems (Comper and Laurent, 1978). During early neurulation in mammals, the mesenchyme of the cranial neural folds increases in volume, and it has been suggested that this expansion is due to hydration of the hyaluronate-rich ECM (Solursh and Morriss, 1977; Morriss and Solursh, 1978a). A study of the distributions of hyaluronate and mesenchyme cells in mouse mesencephalic neural folds during neurulation showed that they are non-randomly distributed and show distinct patterns that can be correlated with specific stages of neural fold elevation (Morris-Wiman and Brinkley, 1990a). Based on these observations, a model of neural fold elevation and convergence was proposed. Under this model, expansion of the central mesenchyme observed early in neurulation is 18 the result of the reorganization of a hyaluronate-rich ECM. Elevation and convergence of the neural folds is accompanied by reorganization of the mesenchyme. Mesenchyme cell density and hyaluronate concentration increase, and the mesenchyme cells change their orientation so that they are parallel to the neuroepithelial basement membrane (possibly due to tension created by lateral expansion of the neuroepithelium). The changes in mesenchyme cell and hyaluronate concentrations would be likely to result in increased cell-cell and cell-ECM interactions, limiting the ability of the mesenchyme to expand. The lateral pull of the neuroepithelium would eventually be counteracted by the medial pull of the mesenchyme as it becomes more rigid, and the resulting force would cause dorsal elevation of the neural folds. As the neural plate lining the neural groove is anchored to the underlying notochord, expansion of the central mesenchyme would result in medial movement of the neural folds", towards one another across the neural groove. When the central mesenchyme expands, hyaluronate concentration in this region decreases and mesenchyme cell density decreases as the cells are dispersed (Morris-Wiman and Brinkley, 1990a). An additional study was carried out in order to investigate the possibility that these observations were the result of decreased synthesis or increased degradation of hyaluronate (Morris-Wiman and Brinkley, 1990b). The lateral regions of the cranial mesenchyme have the greatest cell densities, largely due to the emergence of neural crest cells into this region from the neuroepithelium. Mitotic activity is consistently highest in the central mesenchyme, yet as neural fold elevation progresses, mesenchymal cell density decreases in the central mesenchyme. This decrease was not due to increased cell death. Therefore it is most likely that the change in cell density is due to displacement of the central mesenchymal cells. Displacement could occur by active migration or passive dispersal. As there is evidence for passive cell displacement of cells in other systems (Solursh et al., 1979; Sugrue, 1979; Brinkley and 19 Bookstein, 1986), and as there is circumstantial evidence for ECM expansion of the cranial mesenchyme of the neural folds, this seems the more likely of the two mechanisms. A study of in vitro hyaluronate synthesis and distribution patterns showed that hyaluronate levels decrease in the central regions as the cranial mesenchyme expands (Morris-Wiman and Brinkley, 1990c). Culture of rat embryos in the presence of the hyaluronate degrading enzyme Streptomyces hyaluronidase by one group (Smits-van Prooijie et al., 1986) inhibited neurulation, whereas in a similar study with the same enzyme, also in rat (Morriss-Kay et al., 1986), neural tube closure was delayed but not prevented. This data makes it hard to interpret whether hyaluronate is likely to play an important role in neurulation. Morris-Wiman and Brinkley (1990a) suggest that, as the effects of the enzyme were not apparent in the central mesenchyme until after 16 hours of culture in the latter study, the important events involving expansion of the central mesenchyme would have occurred before this time. Culture of mouse embryos in diazo-oxo-norleucine, which blocks glycosaminoglycan and glycoprotein synthesis, results in failure of elevation of the cranial neural folds (Morris-Wiman and Brinkley, 1990b,c). Elevation appears blocked early in elevation, leaving the embryos with either planar or slightly elevated neural folds. The increase in mesenchyme volume observed in the cranial neural folds is not seen in the cervical region of the neural tube (Morriss and Solursh, 1978a). This suggests that if hyaluronate, the ECM and expansion of the mesenchyme are important factors in neural fold elevation, they might represent a mechanism that would affect closure of the cephalic neural tube but not the more caudal regions. Neural crest cell migration from the midbrain/upper hindbrain neural folds appears to play an important role in elevation of the folds in this region. If rat neural crest cell emigration is inhibited by embryonic exposure to increased oxygen concentrations (Morriss and New, 1979) or by degradation of CSPG in the ECM (Morriss-Kay and Tuckett, 1989), fusion of the neural 20 folds in this region fails or is delayed. Neural crest cell migration in the midbrain/upper hindbrain region appears to be important for the flexibility of the lateral neural folds, allowing them to bend toward and fuse in the dorsal midline (Morriss-Kay et al., 1994). Nichols (1986) suggested that emigration of NCCs from the lateral neuroepithelium may enlarge the neural fold by pushing the neuroepithelium laterally, causing the underlying mesenchyme to expand. Thus, NCC emigration from the lateral neural folds may contribute to expansion of the mesenchyme and elevation of the neural folds. NCC emigration in the mesencephalon begins at about the 4-6 somite stage (Nichols, 1981, 1986; Chan and Tam, 1988), continues during neural fold elevation, and appears to end at about the time of Closure 2 (10-14 somites; Nichols, 1981, 1986; Serbedzja et al., 1992). Thus, the timing of this process is appropriate for it to be involved in neural fold elevation in this region. It is most likely that a number of the forces discussed, and perhaps others not yet identified, act in concert to shape the neural plate and to drive elevation of the neural folds. It is also likely that the combination of mechanisms varies at different levels of the incipient neural tube. Genes and molecules expressed during cranial neural tube closure The genes and molecules involved in neural tube closure are as yet poorly characterized. Many genes (Table 1) are expressed in specific spatio-temporal patterns in the cranial region of mouse embryos during neurulation, and they may therefore be involved in generating or interpreting the forces necessary for cranial neural fold elevation. Some of these, such as CSPG, HSPG, and hyaluronate in the ECM and actin and fodrin in the neuroepithelium, have already been mentioned. 21 Table 1. List of molecules and genes expressed in mouse cranial neural folds on D8.5-9 of gestation (during cranial neural tube closure)* cytoskeletal proteins: actin fodrin desmin vimentin Extracellular matrix components: chondroitin sulphate glycosaminoglycan heparin sulphate glycosaminoglycan hyaluronate basement membrane components: fibronectin laminin entactin types I, III, IV collagen genes encoding transcription Hoxal, Hoxa3, Hoxbl, Hoxb2, Hoxb5 factors: Emx2 Otxi, Otx2 Msx1, Msx2, Msx3 Pax3, Pax6, Pax7 En1, En2 DM, Dlx5, Dlx6 Gli, GH2, GH3 Krox20 Hnf3b, Hnf3a, Mf1, Mf2 p53 other genes: Rata, Rarb Crabpl, Crabpll Tan1 myc c-kit si Wnt1, Wnt3a, Wnt7b Fgf3, Fgf8 Fgfr2, Fgfrl Shh Sek Csk Macs Apob * These genes and molecules are discussed in depth and references are given in the accompanying text. 22 The.basement membrane at the lateral margins of the neuroepithelium is degraded to allow neural crest cells to migrate from the neuroepithelium into the mesenchyme. The main components of the basement membrane are fibronectin, laminin, and entactin (Tuckett and Morriss-Kay, 1986). Fibronectin is a glycoprotein found in the mesenchynial ECM and epithelial basement membrane in mouse (Schaart et al., 1989) and rat (Tuckett and Morriss-Kay, 1986) cranial neural folds during neural tube closure. In the rat, apposition of the neural folds in the prosencephalon is associated with an increase of fibronectin in the basement membrane, as detected by fluorescent anti-fibronectin antibodies (Tuckett and Morriss-Kay, 1986). Cranial neural crest cells appear to secrete fibronectin in chick (Newgreen and Thiery, 1980) and rat (Tuckett and Morriss-Kay, 1986) embryos. Fibronectin promotes cell migration (Yamada et al., 1982), and may provide a substrate for neural crest cells to migrate over. Laminin is another glycoprotein found in the ECM and all basement membranes, including those of the cranial neural folds of neurulating mouse and rat embryos (Leivo et al., 1980; Wu et al., 1983; Tuckett and Morriss-Kay, 1986). It is an important structural component of epithelial cells (Terranova et al., 1980). The sulphated glycoprotein entactin is also found in many basement membranes, including those of the cephalic neural folds of mouse (Wu et al., 1983) and rat (Tuckett and Morriss-Kay, 1986) embryos. It can also contribute to the ECM of mesenchymal tissues (Hogan et al., 1982) and may be involved in cell-matrix interactions (Carlin et al., 1981). Type IV collagen is also found in all basement membranes (Adamson and Ayers, 1979). Interstitial collagens type I and III are found in tissues of mesodermal origin, including the head and heart mesenchyme, and in mesoderm-bounded basement membranes of day 8 mouse embryos (Leivo et al., 1980). In addition to actin and fodrin, the cytoskeleton of neuroepithelial cells in 8 dpc (days post coitum; day of gestation) mouse embryos contains a variety of other molecules. Among 23 those found, at least transiently, are the intermediate filament components desmin and vimentin (Schaart et al., 1989). Cadherins are glycoproteins involved in calcium-dependent cell adhesion. There are at least three cadherin subclasses, including E-, N-, and P-cadherin; N-cadherin is expressed in the neuroepithelium and mesenchyme during neural fold elevation in the mouse (Takeichi, 1988). It has been suggested that cadherins associate with cortical actin bundles at adherens junctions, although whether this association involves signal transduction between actin and cadherins or whether it acts to anchor cadherin is not evident (Takeichi, 1988). It is interesting to note that agents that decrease permeability to calcium or displace membrane-bound calcium appear to inhibit actin microfilament function (such as neuroepithelial cell wedging) in neural fold elevation (O'Shea, 1981). In addition, for apical constriction of neuroepithelial cells to cause bending of the neural plate, cell adhesion between neuroepithelial cells would be required. This suggests that cadherins and actin microfilaments together could play a role in neural fold elevation. Many families of homeobox-containing genes have been identified. These genes encode DNA binding proteins thought to act as transcription factors, and many, including those listed here, are expressed in the cranial region during neural tube closure (8-9 dpc). Emx2 is expressed in dorsal neuroectodermal cells in the prosencephalon (Simeone et al., 1992a, 1992b). Otxl and Otx2 are expressed in the neuroectoderm of the prosencephalon and mesencephalon (Simeone et al., 1992b, 1993). Msxl (Hox-7.1; Chromosome 5), Msx2 (Hox-8; Chr 13), and Msx3 (unmapped) are expressed in the dorsolateral aspects of the cranial neural folds, in the neuroepithelium and mesenchyme in the region of NCC emigration (Robert et al., 1989; Mackenzie et al., 1991; MacKenzie et al., 1992; Liu et al., 1994, 1995; Davidson, 1995; Thomas Lufkin, personal communication). Pax3 (Chr 1) is expressed in the lateral 24 neuroepithelium of the prosencephalon, mesencephalon and rhombencephalon (Goulding et al., 1991), Pax6 (Chr 2) is expressed in the neuroepithelium of the prosencephalon and rhombencephalon (Walther and Gruss, 1991), and Pax7 (Chr 4) is expressed in the lateral tips of the neuroepithelium of the prosencephalon (Jostes et al., 1991). Hoxal and Hoxa3 (formerly HoxL6 and 1.5; Chr 6) and Hoxb5, Hoxbl, and Hoxbl (formerly Hox2.1, 2.9, 2.8; Chr 11) are expressed in the neuroectoderm of the caudal rhombencephalon (Gaunt, 1987; Holland and Hogan, 1988; Murphy et al., 1989; Wilkinson et al, 1989a; Frohman et al., 1990; Murphy and Hill, 1991). Interestingly, Hoxal (1.6) expression retreats caudally from the rhombencephalon between 8 and 8.5 dpc, and disruption of this gene results in delayed closure of the rhombencephalic neural tube in homozygotes (Murphy and Hill, 1991; Lufkin et al., 1991). Enl (Chr 1) and En2 (Chr 5) are expressed in a band of cells over the boundary between the mesencephalon and rhombencephalon (Joyner et al., 1985; Joyner and Martin, 1987; Davis et al., 1988; Davis and Joyner, 1988). Dlx2 (also called Tesl; Chr 2) is expressed in head mesenchyme, within a restricted domain of ectoderm along the frontonasal prominence, and in a section of cells in the region of the neural tube dorsal to the first branchial arch (Bulfone et al., 1993). Dlx5 and Dlx6 are also expressed in facial and branchial arch mesenchyme (Simeone et al., 1994). Although it appears that Dlx5 and Dlx6 have not been mapped, the human homologues, DLX5 and DLX6 map to 7q22, a region with homology to mouse chromosome 5. The proteins produced from the Gli, GUI and GU3 genes contain a DNA-binding zinc finger domain and are thought to be transcription factors. They are all expressed in the cephalic region during neural tube closure (Hui et al., 1994). Gli (Chr 10) is expressed in the ventral neural plate, flanking the midline, and in the paraxial cephalic mesenchyme. GUI (not mapped) and GU3 (Chr 13) are expressed in dorsolateral neuroepithelial and mesenchyme cells, in a 25 pattern that appears complementary to that of GIL They all have highest levels of expression in the prosencephalon and mesencephalon. Krox20, which also encodes a zinc-finger transcription factor, is expressed in the rhombencephalon in the region of prospective rhombomeres r3 and r5 in day 8-9 embryos (Wilkinson et al., 1989b; Wilkinson et al., 1989b). Another family of developmentally expressed genes that code for transcription factors are those containing the fork head domain DNA-binding motif. This family includes Hnf3fi (Chr 2), Hnf3cc (Chr 12), Mfl and Mf2. During cranial neural tube closure (8.0-9.0 dpc), these genes are expressed in some rostral structures: Hnf3fi is expressed in the mesencephalon and rhombencephalon, in the notochord and in neuroepithelial cells of the ventral midline (the floor plate), although expression in the neuroepithelium of the mesencephalon extends laterally (Sasaki and Hogan, 1993; Echelard et al., 1993; Monaghan et al., 1993; Sasaki and Hogan, 1994); Hnf3x is expressed in the rostral notochord and in the mesencephalic floor plate (Sasaki and Hogan, 1993; Monaghan et al., 1993); Mfl and Mfl are expressed in the non-notochordal mesoderm of the head and in neural crest cell-derived head mesenchyme (Sasaki and Hogan, 1993) . Ectopic expression of Hnf3fi under control of the midbrain/hindbrain-specific En2 promoter/enhancer in transgenic embryos caused abnormal brain development and some exencephaly and/or encephalocele in the midbrain/rostral hindbrain region (Sasaki and Hogan, 1994) . Ectopic expression of Hnf3fi also changes the spatial expression of other floor plate genes such as Hnf3x, Bmpl, and Steel factor (Sasaki and Hogan, 1994), suggesting that it is a regulator of floor plate development. Further evidence for this comes from targeted deletion (Ang and Rossant, 1994) and targeted mutation (Weinstein et al., 1994) of HnfJfi to create null alleles. Embryos lacking Hnf3fi expression die by day 11 of gestation, do not form a node or 26 notochord, and have severely abnormal (misshapen or truncated) rostral structures. Development of the neural tube is not prevented, but it is abnormal rostrally, being more like that of the spinal-cord in appearance. Dorsoventral patterning is severely altered, with no floor plate formation and altered expression patterns of floor plate markers. Retinoic acid (RA), a biologically active form of vitamin A, is a known teratogen and suspected morphogen. RA is thought to regulate the expression of genes, such as homeobox-containing genes, through the actions of nuclear receptors (RARs), and is therefore thought to play an important role in development (Eichele, 1993; Langston and Gudas, 1994). The cellular retinoic-acid-binding proteins (CRABP I and II; chromosomes 9 and 2, respectively) and the cellular retinol binding proteins (CRBP I and II; both on chromosome 9) are thought to be involved in the storage and/or transport of RA and its precursor, retinol, respectively. They are thought either to regulate the supply of RA to the RARs (Takase et al., 1979,1986) or to control the concentration of free cellular RA (Robertson, 1987; Smith, 1989). The RARs belong to the steroid/thyroid hormone receptor superfamily (Petkovich et al., 1987; Giguere et al., 1987). Three RARs have been identified, coded for by Rara (Chr 11), Rarb (Chr 14), Rarg (Chr 15), and each has its own temporal and spatial pattern of expression. Rara is expressed in the neuroepithelium of the prosencephalon in 8- to 10-somite stage embryos and is expressed almost ubiquitously by D8.5 (Ruberte et al., 1991), while Rarb is expressed in the neuroepithelium of the caudal hindbrain of 12- to 14-somite stage embryos (Ruberte et al., 1991). Rarg is not highly expressed in rostral regions of the embryo during cranial neural tube closure (Ruberte et al., 1990, 1991). Null mutations of Raral (Lufkin et al., 1993; Li et al., 1993), Rarbl (Mendelsohn et al., 1994), and Rarg2 (Lohnes et al., 1993) produce no phenotype in homozygotes. It has been postulated that the general lack of phenotype in mice with mutations in RARs is due to 27 functional redundancy of the RARs and their isoforms. To test this, double mutants were created (Lohnes et al., 1994; Mendelsohn et al., 1994). Many abnormalities were observed in these double mutants, including cranial NTD in Rard'lg'' embryos. In the cranial region, the Crabp genes are expressed in migrating neural crest cells and in the neuroepithelium of the rhombencephalon and mesencephalon of 8-9 dpc embryos. There is little or no expression in neuroepithelial cells of the floor plate, nor in the lateral edges following NCC emigration (Ruberte et al., 1991, 1992; Denker et al., 1990, Maden et al., 1990). Rostral expression of the Crbp genes during cranial neural tube closure is confined to the notochord and the neuroepithelium. In the latter, expression decreases in the caudo-rostral direction, being highest in the rhombencephalon and lowest in the prosencephalon (Maden et al., 1990; Ruberte et al., 1991). Interestingly, targeted disruption of Crabpl produces no abnormal phenotype in homozygotes (Gorry et al., 1994), CrabpII null mutant mice are essentially normal, showing only a minor post-axial limb defect (Lampron et al., 1995), and the only phenotype of Crabpl and II double mutants are a very slight decrease in viability and the same minor limb defect as CrabpII mutants (Lampron et al., 1995). No changes in expression of any of the RARs were detected, and the CrabpII single and CrabpI/CrabpII double mutants did not show an increased sensitivity to high or low doses of RA. This work suggests that the Crabps are not important regulators of RA concentration or transport, although it cannot be ruled out that they are important in situations where the supply of RA is limited. Tanl (formerly Motch, Notchl; Chr 2) is the murine homologue of the neurogenic Drosophila Notch gene, and appears to part of a gene family that also includes the Int3 gene (Chr 17; Robbins et al., 1992). Tanl is expressed in the cranial mesenchyme (abundant) and neuroepithelium (low expression) of 8.5 dpc embryos. During neural tube closure, expression 28 in the neuroepithelium increases to become a major site of Tanl expression by 9.5 dpc (Franco del Amo et al., 1992). The function of Tanl is as yet unknown. The Notch gene encodes a transmembrane receptor that contains 36 tandemly repeated copies of an epidermal growth factor-like sequence on its extracellular domain (Kidd et al., 1986; Wharton et al., 1985) and may be involved in cell-cell interactions important in cell fate determination (Xu et al., 1990; Fehon et al., 1991; Rebay et al., 1991). The proto-oncogene myc (Chr 12) is expressed in the neuroectoderm and cranial mesenchyme of 8.5 dpc mouse embryos (Downs et al., 1989). Proto-oncogenes are thought to contribute to cellular differentiation, and the distribution of myc suggests that it may play a role in embryogenesis. The nature of such a role has not yet been identified. A member of the TGF-fi family, dorsalin-1 (dsll), is expressed in the dorsal neural tube of chick embryos at the time of neural tube closure, but it is not clear whether it is expressed rostral to the hindbrain (Basler et al., 1993). It appears to regulate cell differentiation along the dorsoventral axis of the neural tube, possibly including that of neural crest cells. A homologous gene is likely to exist in mammals. Fgf3 (fibroblast growth factor 3, formerly int2; Chr 5) is expressed in the neuroectoderm of the rhombencephalon adjacent to the otic vesicles during cranial neural tube closure (Wilkinson et al., 1988). Targeted disruption of this gene produces inner ear defects and abnormally short and curly or kinked tails in homozygotes, but does not affect neural tube closure (Mansour et al., 1993). Fgf8 (unmapped) is expressed in a narrow band of neuroepithelial cells from the dorsal midline around the lateral walls of the neural tube at the mesencephalon/rhombencephalon junction during cranial neural tube closure (Crossley and Martin, 1995). It is also expressed in cells at the rostral limit of the prosencephalon, in the region of the commissural plate, prior to, 29 during, and following fusion of the neural folds in this region (Crossley and Martin, 1995). The fibroblast growth factor receptors Fgfr2 (also known as bek, Chr 7) and Fgfrl (also known as fig, flt2; Chr 8) are expressed in the neuroepithelium of the head folds during cranial neural tube closure (Orr-Urtreger et al., 1991). Mutations in the human FGFR1 and FGFR2 genes have been shown to cause a variety of craniosynostosis syndromes (premature fusion of the skull bone sutures) such as Pfeiffer, Crouzon, Apert and Jackson-Weiss syndromes (Lajeunie et al., 1995; Muencke et al., 1994; Rutland et al., 1995; Reardon et al., 1994; Wilkie et al , 1995; Jabs et al., 1994). These syndromes are inherited in an autosomal dominant manner, suggesting that most or all affected individuals are heterozygous for the mutations. It cannot, therefore, be ruled out that homozygotes might have NTD. Wnt genes comprise a family of at least 10 genes in the mouse. They were originally identified are proto-oncogenes and were later found to be homologous to the Drosophila segmentation gene wingless (Nusse and Varmus, 1982; Rijsewijk et al., 1987; Nusse and Varmus, 1992). It is interesting to note that expression of wg is regulated by the Drosophila homologs of vertebrate Pax genes (Parr and McMahon, 1994). Wnt genes encode cell signalling molecules and are thought to play important roles in vertebrate development (Nusse and Varmus, 1992). At least one Wnt gene, Wntl, has been shown to regulate the accumulation of the adherens junction complex components plakoglobin, E-cadherin, and/or 6-catenin in vitro (Bradley et al., 1993; Hinck et al., 1994), influencing cellular adhesion. The adherens junction complex acts to anchor the actin cytoskeleton, mediate adhesion between cells, and communicate positional information. The adherens junction and may therefore be important in neural fold elevation. Wntl (Chr 15) is expressed throughout the neural plate in the mesencephalon and in the lateral margins in the rostral rhombencephalon of 8 dpc embryos. By the time embryos have 30 about 15 somite pairs and cranial neural tube closure is almost complete, dorsoventral expression is limited to the caudal mesencephalon. Expression extends rostrally through the mesencephalon into the caudal prosencephalon along the dorsal midline (Wilkinson et al., 1987; McMahon et al., 1992; Parr et al., 1993). Disruption of the Wntl gene results in abnormal brain development in homozygotes, particularly in structures derived from the mesencephalon and rostral metencephalon (Thomas and Capecchi, 1990; McMahon and Bradley, 1990; Thomas et al., 1991). Mild midbrain hydrocephaly was observed in some homozygous fetuses (Thomas and Capecchi, 1990), but no exencephaly was observed in any of the studies. Wnt3a (Chr 11) and Wnt7b (Chr 15) are expressed in the caudal prosencephalon, just rostral of the prosencephalon-mesencephalon boundary, of 8-12 somite embryos (8.5 dpc), just prior to neural fold fusion in this region (Parr et al., 1993). Wnt3a is also expressed in the mesencephalon and rostral rhombencephalon of 8.5 dpc embryos, its distribution pattern overlapping that of Wntl (Parr et al., 1993). Sonic hedgehog (Shh) is a putative secreted protein, likely to be important in cell-cell interactions. It is first expressed in 8 somite mouse embryos on 8.5 dpc, in the ventral midline of the mesencephalon (Echelard et al., 1993). Expression extends rapidly into the prosencephalon, rhombencephalon and rostral spinal cord. Expression remains in the ventral midline except in the mesencephalon, where it extends ventrolaterally. Ectopic expression of chick Shh in the domain of Wntl in mouse embryos resulted in failure of neural tube closure in the mesencephalon and rostral rhombencephalon and sometimes in the prosencephalon as well (Echelard et al., 1993). The expression patterns of Hnf3j3 and mouse Shh were altered in transgenic embryos, suggesting that Shh plays a role in inductive interactions important in patterning of the ventral central nervous system. The proto-oncogene c-kit (W; chromosome 5), encodes a receptor tyrosine kinase. It is i 31 expressed in mesencephalic neuroepithelium and the surface ectoderm of the first and second branchial arches in 8.5-9 dpc mouse embryos (Orr-Urtreger et al., 1990). The SI (Steel) locus encodes the ligand for c-kit (Martin et al., 1990; Zsebo et al., 1990a, b). It is expressed in the midline floor plate cells of the mesencephalon and rhombencephalon in day 9.0 embryos (Matsui et al., 1990). Mutations at these loci have not been associated with craniofacial anomalies. Sek, a receptor protein tyrosine kinase, is expressed in the neuroepithelium of the prosencephalon and rhombencephalon during cranial neural tube closure. In the rhombencephalon, expression is strongest in prospective rhombomeres r3 and r5, with lower levels of expression in prospective rhombomeres r2 and r6 (Nieto et al., 1992). Transgenic mice homozygous for mutant or null alleles of a number of genes have NTD. These studies help to identify genes that are important in neurulation. Transgenic disruption of Csk (Chr 9), a cytoplasmic protein-tyrosine kinase, is lethal in homozygous embryos (Imamoto and Soriano, 1993; Nada et al., 1993). Homozygous csk' mutant embryos examined on day 9.5 of gestation are growth retarded, and their cranial neural tube is open over the mesencephalon but fused over the prosencephalon and rhombencephalon (Imamoto and Soriano, 1993). Many embryonic tissues, especially the neural tube, are disorganized, and the notochord consists of fewer cells than normal or is missing. This work supported the hypothesis that the Src family tyrosine kinases are normally negatively regulated by phosphorylation of a carboxy-terminal tyrosine by Csk, and they are activated in cskv~ mutant embryos, with their kinase activity being greatly enhanced. This in turn results in increased tyrosine phosphorylation of several proteins. Src family kinases have been found associated with the cytoskeleton and are thought to be involved in adhesion junctions (Shattil and Brugge, 1991; Wu et al., 1991). They may, therefore, affect the adhesion properties of neuroepithelial cells and 32 may be involved in cell-cell communication. Targeted modification of the apolipoprotein B gene (Apob; Chr 12) in mice to produce a truncated protein causes hypobetalipoproteinemia, and many mice carrying the modified Apob allele are exencephalic or hydrocephalic (Homanics et al., 1993, 1995). Postnatal hydrocephalus is characterized by a large, dome-shaped head with fluid accumulation in the lateral and third ventricles of the brain. Apolipoprotein B is a major structural component of several lipoproteins, and high plasma levels of these apo5-containing lipoproteins are associated with an elevated risk of coronary heart disease (Sniderman et al., 1980). The observation of NTD in animals homozygous for the modified allele was unexpected. Between 28% (13/47; Homanics et al., 1993) and 32% (20/62; Homanics et al., 1995) of homozygotes are exencephalic, and between 20% (38/168; Homanics et al., 1995) and 32% (24/75; Homanics et al., 1993) of surviving homozygotes develop hydrocephalus, as do 1% (6/515) of heterozygotes (Homanics et al., 1993, 1995). There is some variation in the extent of the open region of the neural tube in exencephalic embryos, with involvement of the midbrain and sometimes the rostral hindbrain (Homanics et al., 1995). Unaffected homozygotes can produce exencephalic and hydrocephalic offspring, demonstrating that the apoB'A genotype confers liability to both defects but that other factors are involved in determining phenotype (affected or normal). Humans with truncated forms of apoB often manifest symptoms of vitamin E deficiency (Kane and Havel, 1989), and mice homozygous for the truncated apoB allele show this deficiency (Homanics et al., 1995). Maternal vitamin E deficiencies have been shown to cause exencephaly and hydrocephalus in rodents (Thomas and Cheng, 1952; Cheng et al., 1957; Verma and King, 1967), suggesting that the effect of the modified apoB gene on vitamin E metabolism is the cause of exencephaly and hydrocephalus. Feeding homozygous mice a high vitamin E diet had no significant effect on the incidence of exencephaly in their offspring 33 although there was a trend to a slight decrease (Homanics et al., 1995). Vital staining, histology, and scanning electron microscopy showed evidence of excessive cell death in the alar plate of the hindbrain, including the region of the prospective cerebellum (Homanics et al., 1995). It is possible that this defect underlies delay and/or failure of neural tube closure in some embryos, resulting in exencephaly. In other animals, it is likely to lead to abnormal development of tissues of the fourth ventricle, including the developing cerebellum, choroid plexus, and surrounding mesenchyme, resulting in hydrocephalus. It is not yet known whether the Apob gene is expressed in the developing neural tube, and the possibility that the targeting event in the Apob locus affects the expression of nearby gene(s) that are involved in neural tube closure and brain development has not been ruled out. The cholesterol levels in homozygous animals may be decreased, and this could affect cell proliferation as cholesterol is a key component of the cell membrane. The Macs gene, which encodes the myristoylated, alanine-rich C kinase substrate (MARCKS) protein, is found on proximal Chromosome 10 (Blackshear et al., 1992). The precise function of the MARCKS protein has yet to be determined, but it is a substrate for multi-site phosphorylation by protein kinase C (PKC), has high affinity calmodulin binding, is involved in apparent cross-linking and bundling of actin, and has myristoylation-dependent membrane association (Blackshear, 1993). A related gene, Mrp (distal Chr 4), codes for a protein which shares most of the same characteristics as MARCKS (Blackshear, 1993; Abbott et al., 1993). Expression of the Macs gene is first detected in the mesencephalon (cranial neuropore) on day 8.5 of gestation (Stumpo et al., 1995). Replacement of the Macs gene using a targeting vector that carries the 5' untranslated sequences, first exon, and most of the first intron, results in a disrupted gene that does not produce any mRNA (Stumpo et al., 1995). Exencephaly affecting the fore-, mid-, and hindbrain is seen in 25% of homozygous MARCKS-deficient mice (Stumpo et al., 1995), with some variability in the extent of the affected region (Perry Blackshear, personal communication). Omphalocele (19%) and runting (29%) are also seen in homozygous Macs -I- embryos (Stumpo et al., 1995). Some runting was seen in the heterozygous and homozygous +/+ embryos. No exencephaly or omphalocele was observed in Macs +/+ embryos, while 0.9% of each was seen in heterozygotes. All Macs -I- homozygotes have abnormal brain and retina development and none survive beyond a few hours after birth. All the phenotypic abnormalities observed involve cell-cell interactions, suggesting that MARCKS may be involved in cell communication. The tumour suppressor gene, p53 (Chr 11), encodes a DNA-binding protein that can affect the transcription of other genes (Prives et al., 1994) and is thought to be involved in DNA repair (Bakalkin et al., 1994; Marx, 1994). It can also block progression through the cell cycle and help trigger programmed cell death (Shaw et al., 1992; Lowe et al., 1993; Levine et al., 1994). Most mice lacking the p53 gene are viable, although they are prone to the development of tumours by 4-6 months of age (Donehower et al., 1992). Between 8 and 16% of homozygous p53-/- animals were found to be exencephalic, the affected region encompassing the fore- and mid-brain, and sometimes the hindbrain (Sah et al., 1995). No heterozygous or wild type animals on any of the genetic backgrounds examined were exencephalic. All exencephalic p53-/-mice were female, as shown by PCR analysis using primers directed against the Zfy gene (Y chromosome-specific). This results in a skewed sex ratio in the surviving p53-/- mice (20-41% were female, depending on genetic background). NTD resulting from mutations in the p53 gene appear to affect only the cranial neural tube, and the incidence of NTD appears to be influenced by modifier loci as the incidence of exencephaly varies on different genetic backgrounds. 35 Models of neural tube closure defects There are several spontaneously occurring mutations and chromosomal abnormalities in mice that cause exencephaly (with or without spina bifida; Table 2) including Splotch (Sp), Curly tail (Cf), Extra toes (Xt), Exencephaly (xn), open brain (opb), Trisomy 12, Trisomy 14, Snell's translocation (T(2;4)lSn), Loop-tail (Lp), Crooked (Cd), Rib fusions (Rf), cranioschisis (cm), and Bent tail (Bn). There are other mouse mutations that produce only spina bifida, such as Vacuolated lens (vl), Axial defects (Axd), and T-Curtailed (Tlf5); they will not be discussed further. Unlike the majority of human cases, many of these animal models show autosomal recessive inheritance of the NTD. In addition, many of them have other associated anomalies in addition to the NTD. In some of these cases, the mutant genes have been identified, while in other cases, the genetic basis of the defect is unknown. It is hoped that the analysis of animal models of NTD and identification of the causative genes will identify genes that are involved in normal and abnormal development of the neural tube in humans. As the inheritance of human NTD is multifactorial and the genetic basis is thought to be heterogeneous, it will be difficult to identify genes involved in neural tube closure defects directly in human embryos. The study of animal models is therefore important in order to identify candidate genes and developmental mechanisms that are involved in human NTD. Although many of the animal models of NTD involve single genes with autosomal recessive patterns of inheritance, these genes could also be important in the development of human NTD; for example, it is possible that different kinds of mutations in the same gene may act additively or that the role of these genes may differ in human neural tube closure. It is also possible that these genes do not contribute to multifactorial NTD, making it important to study animal models with this mode of inheritance as well. 36 8 E c CO o <D "O Q) CO O o 0) J3 t_ OJ c 'c (0 o "6 w d) TJ O csi 0) to 8 cz co ?D CO o-•£ <" .£5 t3 {/) CO 10 T3 CO 8 cz CO co CD T3 O E (0 § g CO O E o> o> x; 15 co cn ac CO cn •e cz CO to Au Epst 10 CO c CO c E 0 T3 1 E co CO CO X CO E Q. £ } CO '18 to co C L CO o o C L co -<J- r - «N m o> S2 cn cn X J " CD E -9 o O E O LU o c cz co so .E a> E £ o "a 31 to CO Q Z 1 a . A CO i s (/) CO cc 3 O T3 cz cc '5 X r~" co o> (0 CD > « cz cc c E 0 T3 1 E CD co co co >» cc x: C L 8 c CO X CD CO CD O *-* e CD co o> 00 o> co +—> CD CO 0 JC 1 o cz CL < Q Z >> cc x: C L 8 c CO X CD c .X, co xz Q L 8 c CD X CO . -a t- cz CD CO J S = co 5S 0 CD 1 .o T " cz cz co co co > CO o> to CD >< c co c E 0 T3 1 "E CO to to . 'to >>:E ca o x: to Q- x: CD U a> co o a . o o is»' T- tO CD ^ to CD C co cz E 0 T3 1 'E CD IO Q z C O >. CC x: Q. CD O C CD X CO " D o X3 CD O P to CD >s C CO c E 0 T3 1 'E CD 10 >. CD x: C L CD O C CO X CD QL •—-to c o 'to 3 ^— n •sz co Is-a> !~ 1— co o> co" CD * S co T3 * CZ co 5 cn 2 " GO 10 CO CU < >> CO x: C L CD O C CO X CO E o v to «j !E o to o '5 co •o c cz o co to l_ CZ S o m co z CD > 'to to CD O •a CD cz Q Z X ^ co co "a x: S-5 CQ CD n S i cc co t to cn o to t -CO ^1 n cn CO Z < z >» co x: C L CD O C CD X CD m co 3 to O to 1-5 is" co N 3 » a. 2 3 co < z co x: C L 8 c CD X CO CM >. E E 0 0 10 to •c •c 1- I-T3 CO CZ Q) CO co x: co = CO cn 10 «S co E CD C ^ O C 2? co CO T3 si vz g-S o co cz cz s °-cz o "X3 s _o to c ca CO cz CO •a aj "co to —^1 o cz CO T3 CO co "S • D O CZ CO > 8 CO E o to o 3 co < 37 There are several splotch alleles which have arisen as spontaneous or radiation induced mutations. They include the original splotch (Sp), splotch delayed (Spd), splotch retarded (spr), Sp1", and Sp2". Splotch is a semi-dominant mutation; heterozygotes have white belly spots and lack pigmentation to distal body parts such as the tail tip and feet (Russell, 1947). Homozygotes for Sp, Sp1" and Sp2" have lumbosacral spina bifida, and over half also have exencephaly affecting the caudal mesencephalon/rostral rhombencephalon (Auerbach, 1954; Beechey and Searle, 1986; Franz, 1989). They die by day 14 of gestation (Auerbach, 1954). Al l Sp homozygotes also have brain abnormalities, including distortion and, in the mesencephalon, reduction of the lumen of the brain (Auerbach, 1954). These embryos also have diminished neural crest cell migration (Auerbach, 1954; Dickie, 1964; Kapron-Bras and Trasler, 1988), altered neural cell adhesion molecules (N-CAMs; Moase and Trasler 1991), a disorganized neuroepithelium (Morris and O'Shea, 1983), and increased levels of CSPG, HSPG, and hyaluronate (Trasler and Morriss-Kay, 1991; McLone and Knepper, 1986). Sp2" homozygotes and Spr heterozygotes were found to have a deletion in or encompassing the Pax3 gene (Epstein et al., 1991). Mice with the original Sp allele have a mutation in intron 3 of the Pax3 gene, leading to aberrant splicing of the mRNA and a non-functional protein (Epstein et al., 1993). Thus, mutations in Pax3 cause the phenotype observed in Splotch mice. As already mentioned, Pax3 is expressed in the lateral cranial neuroepithelium on days 8-9 of gestation. Neural crest cells migrate from this region during cranial neurulation. As Pax3 encodes a transcription factor, it is likely that a mutant allele of this gene affects the expression of other genes in the region of the lateral neuroepithelium, such as those coding for N-CAMs, CSPG, HSPG, and hyaluronate. Altered expression of these other genes would be likely to affect emigration of the neural crest, and elevation of the neural folds in the mesencephalon and rhombencephalon. The failure of elevation of the cranial neural folds could be a result of 38 diminished NCC migration, although this would not explain the occurrence of spina bifida as NCC migration occurs after neural tube closure in the trunk. The exencephaly does not involve the prosencephalon, but as already mentioned, closure of the neural tube in the prosencephalon of chick embryos (Schoenwolf, 1988) may proceed by a different, intrinsic, mechanism than the more caudal cranial regions. This may also be the case in the mouse and could account for the closure of the prosencephalon in Sp homozygotes with exencephaly. It is interesting that not all Sp homozygotes are exencephalic. This is not surprising as neural tube closure is thought to involve the action of several mechanisms; redundancy of these other mechanisms would be expected to compensate to some degree for lack of Pax3 expression. Stochastic factors may also play a role, as all splotch homozygotes on an inbred background are genotypically identical yet differ in phenotype. Clearly, identification of the mutant gene has not led to a simple explanation of the etiology of the (rostral) NTD in Sp mice, and more work is necessary to understand the role of Pax3 in neural tube closure. Mutations in the human PAX-3 gene cause Waardenburg Syndrome type I (WS-I; Baldwin et al, 1992; Morell et al., 1992; Tassabehji et al., 1992; Hoth et al., 1993), an autosomal dominant disorder that causes deafness and pigmentary disturbances. The phenotype (in heterozygotes) is consistent with a defect in the neural crest. Klein-Waardenburg Syndrome (WS-III), which is similar to WS-I but also causes limb abnormalities, is also caused by mutations in PAX-3 (Hoth et al., 1993). As it is mainly splotch homozygotes that have spina bifida and/or exencephaly, then by analogy WS-I homozygotes would be more likely to have NTD than heterozygotes. However, WS-I homozygotes are rare, and the one homozygous child that has been reported was more severely affected than heterozygotes but did not have NTD (Zlotogora et al., 1995). In addition, the majority of human NTD are isolated, with the neural tube being the only affected system. This would suggest that the study of Splotch mice will not 39 necessarily help in the direct understanding of how the majority of human NTD arise, although it will help to identify genes and gene interactions that can lead to NTD. It is possible that a gene or genes that act downstream to Pax3 are directly responsible for the defects that cause the NTD, and that mutations in these genes can cause isolated NTD. Since the first report on curly tail mice in 1954, it has been thought that the defects observed in this strain are caused by a recessive gene, but that genetic background modifies its expression (Gruneberg, 1954, Embury et al., 1979). It has recently been shown that some mice heterozygous for the curly tail mutation (ci) have an abnormal-tailed phenotype, demonstrating that ct is in fact a semi-dominant mutation (Neumann et al., 1994; Beier et al., 1995). The major causative gene, ct, has been mapped to distal Chromosome 4 (Neumann et al., 1994). At least three modifier loci appear to affect the incidence of NTD in ctlct mice and two of these modifiers have been mapped to distal chromosome 3 and mid-chromosome 5 (Neumann et al., 1994). Although the mode of inheritance of NTD in curly tail mice is multifactorial, it does not appear to parallel the situation of human NTD as there is no evidence for a major gene involved in liability to NTD in human populations (Carter, 1974; Laurence, 1983). Curly tail (ct) homozygotes develop tail flexion defects about 40-45% of the time, while 15-20% exhibit lumbosacral spina bifida, A small proportion (1-7%) of ct homozygotes also have exencephaly of varying severity. The most extreme cases have an open neural tube from the caudal forebrain to the rostral hindbrain, while only the hindbrain is affected in less severe cases (Gruneberg, 1954; Embury et al., 1979; Copp et al., 1982). Of all ct offspring that show an abnormal phenotype, including only tail flexion defects, 16-17% are exencephalic (Embury et al., 1979). As the incidence of exencephaly is low, experimental studies have virtually ignored the head and concentrated on the defects in caudal neurulation. Curly tail mice are 40 generally considered a good model for the study of human NTD as they show multifactorial inheritance of liability to NTD, and produce both spina bifida and exencephaly in the absence of other major defects. However, one would expect that the cause of exencephaly should also become an important direction of study. The tail flexion defects in curly tail mice appear to be a result of a cell-type specific decrease in the rate of cell proliferation in the notochord and gut endoderm that leads to increased ventral curvature and delayed closure of the posterior neuropore (PNP) and the development of caudal NTD (Copp et al., 1982; Copp et al., 1988; Brook et al., 1991). Although the foregut underlies much of the cranial neural folds, the topography of the latter is very different from that of the spinal neural folds and it seems unlikely that a mechanism involving differential rates of cell proliferation would underlie the exencephaly seen in curly tail embryos. It is possible that the effect of the ct gene(s) is different in the head than in the trunk. For example, the same gene that causes the differential rates of cell proliferation in the trunk might act differently in the head, where the molecular environment is different. Alternatively, the primary action of the gene in the trunk could be secondarily affecting the rates of cell proliferation. In the head, the primary effect may be the same but affect subsequent events differently. As it is likely that there are different mechanisms of neural fold elevation in the head and trunk, or that the same mechanisms exist but are not of equal importance in each region, it is possible that the one mutation could affect neural tube closure in the spine differently than in the head. It has also been observed that the accumulation of newly synthesized hyaluronic acid in the region of the posterior neuropore is reduced in affected ctlct embryos (Copp and Bernfield, 1988). It is possible that a similar situation exists in the cranial neural folds, but it has not been 41 investigated. As hyaluronic acid is thought to be involved in expansion of the mesenchyme and subsequent elevation of the cranial neural folds (Solursh and Morriss, 1977; Morriss and Solursh, 1978a; Morris-Wiman and Brinkley, 1990a), decreased levels of hyaluronate in the cranial neural folds could affect their elevation and closure. It is also possible that the exencephaly is not causally related to the spina bifida of curly tail mice, but that both occur in this strain by chance. This appears unlikely as these two traits are correlated in humans and as mice assumed to be homozygous for the ct mutation have been observed with caudal defects or exencephaly or both. In addition, there is evidence that the curly tail mutation may affect the incidence of exencephaly. In a study of the effect of inositol deficiency on the susceptibility of curly tail mice to NTD (Cockroft et al., 1992), ctlct and CBA (control) embryos showed a high incidence (71% and 61% respectively) of exencephaly. The curly tail strain and the CBA strain are related, suggesting the possibility that their shared genetic background (perhaps the ct modifier loci) affects the incidence of exencephaly. Inositol deficiency caused exencephaly in 26% of embryos of the PO strain, a normal strain that is not related to the CBA or curly tail strains. This suggests, under the multifactorial threshold model (discussed in greater detail below; also, see Falconer, 1989), that inositol deficiency shifts the distribution of the population of embryos closer to the developmental threshold, causing a greater proportion of the distribution to fall over the threshold and more embryos to be affected. Under this model, ctlct and CBA embryos would originally lie closer to the threshold as they show a greater response in exencephaly production to inositol deficiency than PO embryos. This could be due to the (possibly shared) modifier loci alleles in ctlct and CBA embryos. However, as a low dose of inositol halved the incidence of exencephaly in CBA and PO embryos, but had no effect on ctlct embryos, it is likely that the curly tail mutation is involved in this difference and therefore affects the incidence of exencephaly. 42 Another possibility is that the differential rate of cell proliferation in the curly tail strain may be a result of genetic background coincidentally present in this strain, not causally related to the NTD. That the growth differential is responsible for the caudal defects is supported by two studies. In the first, experimental growth retardation of embryos in vivo and in vitro leads to a greater decrease in the rate of proliferation of the neuroepithelium than of other cell types. The rate of closure of the posterior neuropore in these embryos is normalized and spina bifida is almost completely prevented (Copp et al., 1988). Implantation of a fine eyelash tip longitudinally into the hindgut lumen to prevent ventral curvature of ctlct embryos (Brook et al., 1991) also allows normal closure of the posterior neuropore. Extra toes (Xt) is a semi-dominant mutation that causes pre-axial Polydactyly, occasional hydrocephaly, and frequent presence of an enlarged interfrontal bone in heterozygotes. Homozygous embryos develop multiple abnormalities, including paddle-shaped feet with 8 or 9 digits, ectopia of the viscera, parietal brain hernia, hemimelia, brain and spinal cord abnormalities, edema, and defects of the sense organs, and die before birth (Johnson, 1967). Neural tube closure is delayed in Xt homozygotes, and it appears that there may be a lack of fusion of the neuroepithelium across the midline of the mesencephalon (p.567, Johnson, 1967). It appears that all homozygotes surviving at day 11 of gestation have closed their neural tube, although it always appears abnormal, but at 16 days the cranium over the mesencephalon often ruptures, allowing the brain to protrude (Johnson, 1967). This parietal brain hernia is a type of exencephaly, but it results from secondary disturbance of a closed neural tube rather than failure of closure. Mice heterozygous for Xt1, a spontaneous Xt allele, have a phenotype very similar to that of Xt heterozygotes. Homozygotes die before or within 2 days of birth, and display severe 43 Polydactyly and syndactyly of all feet, edema, and exencephaly or other brain malformations, as well as other abnormalities (Hui and Joyner, 1993). Xtbph (brachyphalangy) is allelic to Xt and produces a similar but distinct phenotype. Xtbph heterozygotes have abnormally broad paws, some syndactyly, but no Polydactyly. They also have an abnormal sternum and often show an interfrontal bone. Homozygotes are similar in phenotype to Xt homozygotes, but 27-57% of homozygous embryos of all ages (days 10-18) are exencephalic in the region of the midbrain (Johnson, 1969; Johnson, 1970). Studies to investigate the developmental origin of the exencephaly have not been published, but it is possible, based on the site of the NTD, that Xtbph/Xtbph embryos lack initiation of fusion at Closure 2, at the prosencephalon/mesencephalon boundary, and that closure of the cranial neural tube is never completed. This is in contrast to the NTD seen in Xt/Xt embryos, where a closed neural tube appears to re-open (Johnson, 1967). The mouse mutant Polydactyly Nagoya ipdn), a semi-dominant mutation that arose as a spontaneous mutation, also appears to be allelic to Xt (Schimmang et al., 1994). Pdn heterozygotes have mild digit anomalies. All homozygotes have preaxial Polydactyly, 21% are exencephalic, 12% have cleft palate, and 12% have open eyelids (Hayasaka et al., 1980). Non-exencephalic homozygotes have brain malformations including hydrocephaly and absent corpus callosum and olfactory bulbs (Hayasaka et al., 1980; Naruse, 1990). Pdn/Xt animals have a very similar but not identical phenotype to Pdn/Pdn or Xt/Xt; for example, no exencephaly was seen (Schimmang et al., 1994). Xt and Pdn map to proximal Chromosome 13 (Lyon et al., 1967; Schimmang et al., 1994). The zinc finger gene, GU3, has recently been found to be partially deleted in Xt and Xt1 mice (Vortkamp et al., 1992; Hui and Joyner, 1993). The mutation in Xt results in no expression in homozygotes and reduced expression (about 50% of normal) in heterozygotes 44 (Schimmang et al., 1992). Xt* also appears to be a null allele (Hui and Joyner, 1993). The Gli3 gene of Pdn mice does not appear to be deleted, and the length and expression of its RNA appears unaffected (Schimmang et al., 1994). The existence of point mutations has not yet been determined. This indicates that not only deletions (null alleles) of this gene cause semi-dominant mutations, as the Pdn mutation does not appear to affect GU3 expression but presumably affects protein function. As mentioned earlier, GU3 is expressed in the dorsolateral neural plate and mesenchyme during cranial neural tube closure, with highest expression in the prosencephalon and mesencephalon. It is also expressed in more caudal regions and through subsequent stages of development, including during brain and limb bud development (Hui et al., 1994). As GU2 has a very similar pattern of expression, it is possible that these two genes share some degree of functional redundancy and therefore null alleles of GU3 do not result in a phenotype in all parts of the body where it is normally expressed. It will be interesting to observe the phenotype of GU2 null alleles and to see whether homozygotes are exencephalic. A recessive mutation, add (axial digit-pattern deformity), is the result of a transgene insertion in the GU3 gene (Pohl et al., 1990). The phenotype of add is less severe than that of Xt as only the anterior forelimb is affected in homozygotes. It is interesting that the one recessive mutation in GU3 is the only one that does not affect craniofacial as well as limb development. As the deletion in Xt1 is intragenic, it is unlikely that a separate, closely linked gene is responsible for the craniofacial phenotype. For several years, it has been thought that the Xt mutant is a possible homologue to Greig's Cephalopolysyndactyly Syndrome (GCPS) in humans (Winter and Huson, 1988). GCPS is an autosomal dominant disorder that causes Polydactyly and cranial abnormalities. GCPS has 45 been shown to be caused by interruptions in the human GU3 gene, supporting the interpretation of homology between GCPS and Xt (Vortkamp et al., 1991). Like splotch and Waardenburg Syndrome I, the mutations in GH3/GLI3 that cause craniofacial abnormalities (but not NTD) act in an autosomal dominant manner. NTD, however, are generally found only in mice homozygous for GU3 mutations, and humans homozygous for GLI3 mutations are likely to be rare. Mutations in this gene are therefore not likely to make a significant contribution to the incidence of human NTD, even assuming that homozygotes would be anencephalic. Exencephaly (xn) is an autosomal recessive gene with partial (33-84%) penetrance, depending on genetic background (Wallace et al., 1978; Anderson, 1981). All xn embryos show delayed closure of the cranial neural tube, and on day 9 of gestation, all xn embryos have widely open cranial folds (Anderson, 1981). There is variation in both how far rostral and how far caudal the exencephaly extends (Wallace et al., 1978; Anderson, 1981). In the light of current understanding and hypotheses, this suggests that xn embryos omit initiation of fusion of the neural tube at Closure 2 and that rostral fusion is accomplished by extension of Closure 3 (resulting in variability in how far the rostral end is closed). It seems possible that Closure 4 may extend rostral from the rhombencephalon, resulting in caudal variation in the extent of the opening. A cross of an xn heterozygote to a curly tail homozygote produced no exencephalics in 50 offspring, suggesting that these two genes are not allelic (Wallace et al., 1978). Unfortunately, the xn gene was not mapped (Wallace et al., 1978) and the mutation is extinct. A new spontaneous mouse mutant with NTD, open brain (opb), was reported recently (Gunther et al., 1994). Heterozygotes appear normal, but homozygotes have multiple 46 malformations including exencephaly, malformations of the spinal cord and axial skeleton, defects in eye development, and preaxial digit duplications. The exencephaly generally affects the caudal forebrain, mid-brain, and rostral hindbrain. It appears to be the result of lack of initiation of fusion of the neural folds at the site of Closure 2, and rostral extension of Closure 3 must also be affected as only the most rostral region of the prosencephalon fuses (Gunther et al., 1994). Completion of Closure 4 over the rostral rhombencephalon also appears to be affected (Gunther et al., 1994) as the exencephaly extends into the rostral hindbrain. The opb mutation exhibits a segregation pattern consistent with full penetrance of a single autosomal recessive gene (Gunther et al., 1994). The trait is being moved onto a C57BL/6 background, and while this does not appear to be affecting the penetrance, the expressivity of cranial NTD seems to have decreased (Gunther et al., 1994). This suggests that gene(s) in the C57BL/6 background might modify the expression of opb. Gunther et al. (1994) report that preliminary linkage studies suggest that opb segregates independently from the Sp and Xt loci. Mouse embryos with trisomy 12 have exencephaly involving the midbrain and hindbrain,, and about half of embryos with trisomy 14 have exencephaly, affecting only the midbrain. In both cases, the cranial neural folds form normally but the edges of the neural plate then fail to bend in and converge in the midline (Putz and Morriss-Kay, 1981; Morriss-Kay and Putz, 1986). In addition, there is failure of initiation at Closure 2 (Morriss-Kay and Putz, 1986), and in trisomy 12 embryos, initiation or completion of Closure 4 must also be affected as some portion of the neural tube over the hindbrain remains open. In the affected regions in each case, the basal lamina of the neuroepithelium is disrupted, there is an increase in the number of necrotic cells in the neuroepithelium, and there are fewer mesenchyme cells, which are abnormally distributed and show a low number of intercellular contacts (Morriss-Kay and Putz, 1986). 47 Apical localization of actin microfilaments and transition of the neural folds from a convex to concave shape are delayed (Morriss-Kay and Putz, 1986). It is possible that the defect in trisomy 14 embryos is less severe than the defect caused by trisomy 12, enabling half of the former to close their rostral neural tube successfully. Matings of mice heterozygous for Snell's translocation (T(2;4)lSn) to normal mice produce exencephalic offspring. The translocation arose in an X-ray irradiated male, and breeding of his semi-sterile male descendants to normal females produced exencephalic embryos, with variation in the rostral and caudal limits of the exencephaly (Snell et al., 1934; Snell and Picken, 1935; O'Shea and Kaufman, 1983). Exencephalic offspring have a duplication of distal chromosome 2 and a deletion of distal chromosome 4 (Eicher and Washburn, 1977). Caudal NTD, including lumbosacral spina bifida, are seen in the offspring of translocation carrier males mated to CFLP females (O'Shea and Kaufman, 1983). A homozygous translocation stock has been developed. The frequency of exencephaly appears to depend on the genetic background of the normal strain crossed to individuals heterozygous for the translocation: about 23% of term embryos are exencephalic from matings to C57BR/cdJ, 7% from matings to DBA/2J mice (Eicher and Washburn, 1977), and 26% of day 10 embryos are exencephalic and 7% have caudal NTD in crosses to CFLP (O'Shea and Kaufman, 1983). This suggests that there are other loci, at which different normal strains carry different alleles, that affect the penetrance of exencephaly caused by the duplication of distal chromosome 2 and deletion of distal chromosome 4, though other (i.e., environmental) factors might also explain the different frequencies of exencephaly observed in different studies. There is variability in the extent to which the cranial NTD of the offspring of Snell's translocation heterozygotes mated to normal mice remains open in exencephalics. In some cases, 48 only the rhombencephalon remains open, while in others the entire rostral neural tube fails to close (Snell et al., 1934; O'Shea and Kaufman, 1983; O'Shea, 1986). The neural folds of these embryos exhibit several abnormalities, including small neuroepithelial cells, increased NCC and neuroepithelial cell death, disruptions of the basal lamina, and abnormally elongated mesenchyme cells with reductions in their cell area and intercellular contacts (O'Shea, 1986). It is interesting that Snell's translocation involves Chromosome 4 (Miller et al., 1972), and that the breakpoint appears to lie near the location of the curly tail gene, which maps to about 4 cM proximal the DNA marker D4Mitl3 (Neumann et al., 1994; Beier et al., 1995). This puts it at about 65 cM, while the breakpoint in Snell's translocation is about 21 cM distal to the misty locus, putting it at about 67.1 cM (Snell, 1946; Miller et al., 1972). Given the inaccuracies of the recombination linkage map, it is possible that the Snell's translocation breakpoint is at the ct locus. It is also perhaps interesting that the MARCKS-related protein gene, mrp, maps near this region (59 cM). The incidence of NTD in ct also appears to be affected by modifier loci, and both spina bifida and exencephaly are seen in homozygotes. It has been reported that the exencephalic offspring of Snell's translocation carriers are 2,2,42,4, having a duplication of distal Chromosome 2 and a deletion of distal Chromosome 4 (Eicher and Washburn, 1977). The deleted region of Chromosome 4 is therefore likely to carry the ct locus, and ct mutations can act in a dominant manner (Beier et al., 1995). A null allele of the ct gene, as would be generated by deletion of the chromosomal region in which it lies, may be more severe than the mutation found in ct mice, causing a higher incidence of NTD in the affected offspring of Snell's translocation heterozygotes. In addition, it is possible that other loci involved in the translocation (on Chromosome 2 or 4) affect the development of NTD, whether in co-operation with ct or alone, particularly as the frequency of translocation-associated exencephaly is much higher than that observed in curly tail homozygotes. 49 Loop-tail {Lp; Chr 1) is a semi-dominant mutation with incomplete penetrance (Strong and Hollander, 1949). Heterozygotes have looped or curly tails, the extent of which varies widely. They also demonstrate a wobbling of the head and enlarged lateral ventricles of the brain (Strong and Hollander, 1949; van Abeelen and Raven, 1968). The neural folds of homozygotes fail to complete elevation and converge along much of the neural tube, and the NTD seen vary from exencephaly in the caudal midbrain and rostral hindbrain to craniorachischisis, where only the forebrain and rostral midbrain close (Strong and Hollander, 1949; Stein and Rudin, 1953; Wilson and Finta, 1980a). A recent study (Copp et al., 1994) used a microsatellite marker closely linked to the Lp locus to identify homozygotes and heterozygotes before the appearance of morphological abnormalities; without such a marker, genotype can only be inferred subsequent to the event that causes the abnormalities. In this study, it was found that Lp homozygotes do not initiate fusion of the neural tube at Closure 1 and that only they develop craniorachischisis. The rostral sites of initiation of fusion, Closures 2 and 3, do take place. Closure 4, over the rhombencephalon, does not take place, suggesting that the neural tube caudal to its initiation site must take place for Closure 4 to succeed. As the entire rhombencephalon always remains open, this suggests that caudal fusion from the rostral rhombencephalon does not occur, or does not go very far, in these embryos. Heterozygotes show a delay in initiation of Closure 1 and in closure of the posterior neuropore, resulting in tail defects (Copp et al., 1994). It would be interesting to know whether initiation of fusion over the rhombencephalon is also delayed in heterozygotes, given their mild brain abnormality and wobbly head phenotype. There are several abnormalities in the neuroepithelium of Lp/Lp embryos, although as * they have been observed after the time of normal neural tube closure, it is unclear whether they cause the NTD or are a secondary effect of it. The apical cell surfaces of lateral neuroepithelial 50 cells are flatter and have fewer microvilli than normal embryos (Wilson and Finta, 80a). There are increased intercellular spaces (Wilson, 1978) and gap junctional vesicles (Wilson and Finta, 1980b), and there are gaps in the basal lamina (Wilson, 1985a,b). The angle between the elevating neural folds is consistently wider than in normal embryos, and it has been suggested that Lp/Lp embryos may have a defect in the cellular mechanisms that cause neuroepithelial cell wedging as they have a higher ratio of wedge-shaped cells than do normal embryos (Wilson and Wyatt, 1992). The localization of actin microfilaments in the neuroepithelium is normal (Wilson and Finta, 1980b), but quantitative or qualitative changes cannot be ruled out. It has been proposed that the Lp mutation may reduce axial elongation of the embryo. By day 9.5 of gestation, it is obvious that the body axis of homozygotes is shorter than that of normal embryos, as are the notochord, neural plate, and gut, and this difference persists through the remaining gestational period (Smith and Stein, 1962). The primitive streak, however, is longer (Smith and Stein, 1962). As axial stretching is thought to play a role in neural fold elevation (Jacobson and Gordon, 1976), a negative effect of the Lp mutation on axial elongation could result in failure of initiation of fusion of the neural folds at the site of Closure 1. There is a premature shift from a predominance of hyaluronate to a predominance of sulphated glycosaminoglycans in the mesenchymal extracellular matrix and neuroepithelial basal lamina in the hindbrain of Lp/Lp embryos during neural tube closure in this region (D9-9.5; Copp and Wilson, 1981; Wilson 1985c). This change in the relative abundance of extracellular matrix proteoglycans is expected to be associated with a decrease in cell proliferation and motility, which could affect axial development and NCC migration. This change may also alter normal interaction between the cells of the mesenchyme and neuroepithelium and decrease mesenchymal support of the neural folds, affecting their ability to elevate. 51 Crooked tail (Cd) has been mapped to chromosome 6. Heterozygotes may have crooked or normal tails and show an increased sensitivity to insulin-induced exencephaly (Morgan, 1954; Cole and Trasler, 1980), while homozygotes show a variety of phenotypes (Morgan, 1954). When homozygous, Cd is a partial lethal with about 2% presumed pre-implantation- and about 12% post-implantation-mortality. About 4% of homozygotes have exencephaly without spina bifida, and 7% have crooked tails and abnormal tail fur, small, abnormally shaped heads with thin cranial bones and nervous head movement, abnormal eyes, and retarded eruption of their lower incisors, resulting in overgrowth of the upper incisors and impairing their ability to eat and hence grow and survive (Morgan, 1954). The crooking of the tail in heterozygotes is due to abnormal caudal vertebrae (Morgan, 1954). Rib fusions (Rf) is a semi-dominant gene which affects development of the ribs and the spinal column (Theiler and Stevens 1960). There appears to be a paucity of detectable heterozygotes and homozygotes, indicating either early embryonic mortality or incomplete penetrance. In heterozygotes, the proximal regions of at least two ribs are fused, and in some cases several vertebrae are irregular. Homozygous have extensively fused ribs and neural arches, malformed somites, and wavy neural tubes. They are smaller than normal littermates and rarely survive to birth. Exencephaly is seen in approximately 70% of presumed homozygotes and approximately 10% of presumed heterozygotes, and it generally involves the hindbrain, midbrain, and sometimes the forebrain. Rf heterozygotes show an increased susceptibility to insulin-induced exencephaly (Cole and Trasler, 1980). It should be noted that as genotype (homozygous RflRf or heterozygous) is inferred from the severity of the abnormal phenotype, it is possible that some animals were mis-classified as homozygous or heterozygous by Theiler and Stevens (1960), although this would not account for a deficiency of both these genotypic classes. It could, 52 however, affect the data on exencephaly frequencies. Although Rf homozygotes have asymmetrically wavy neural folds in the cranial and cervical regions, the posterior regions are not usually affected. Examination of the published data suggests that the exencephaly caused by the Rf gene is the result of failure of initiation of fusion of the neural folds at the forebrain-midbrain boundary (Closure 2), and while fusion at and extension from Closure 3 appears to occur in some embryos (Fig. 4 in Theiler and Stevens, 1964), the fact that exencephaly can include the forebrain suggests that Closure 3 does not always succeed in fusing the entire prosencephalon. Further work on the development of exencephaly in Rf embryos does not seem to have been done, and the gene is not mapped. The gene cranioschisis {cm), which arose in a strain carrying oel (open eyelids with cleft palate), causes exencephaly in homozygotes (Brown and Harne, 1973, 1974; Kalter, 1981). The severity of exencephaly varies among embryos, some having only a very small (1 mm in diameter) opening over the midbrain while others fail to close most or all of the rostral neural tube (Brown and Harne, 1973; Brown et al., 1984), suggesting that any or all of the rostral sites of fusion, Closures 2, 3, and 4, can fail. The cm gene is unmapped. It fits an autosomal recessive pattern of inheritance, although the segregation ration is 0.9 of the expected (Brown and Harne, 1973, 1974). This suggests reduced penetrance. The segregation ratios observed in litters where cleft palate and exencephaly are both segregating indicate either very close linkage or allelism (Brown and Harne, 1973). Both types of abnormal embryos also have fewer calcified caudal vertebrae than their normal sibs (Brown and Harne, 1973), indicating allelism. In addition, this suggests that the cm/oel gene(s) are expressed in rostral and caudal structures during development. The neural folds of cm homozygotes appear to be smaller than those of normal embryos 53 (Waterman, 1979). There are no obvious defects in the neuroepithelium, but the mesenchyme cells of the cranial neural folds of day 9 embryos are rounded up, with numerous cytoplasmic blebs, and do not extend the long filipodia that interconnect normal mesenchyme cells (Waterman, 1979; Brown et al., 1984). It is possible that the mesenchyme cells have lost their ability to attach to the ECM. The abnormal morphology of the mesenchyme suggests reduced cell communication between mesenchyme cells and a decreased contribution by the mesenchyme to the structural support of the neural folds. The blebbing of mesenchyme cells appears to be a transient event as it was not seen in a small number of presumptive exencephalics during cranial neural tube closure or in exencephalic embryos on days 10 or 11 of gestation (Waterman, 1979). Bent tail (Bn) maps to the X chromosome. It has not been well studied, but it was reported to often produce spina bifida and exencephaly in hemizygous and homozygous individuals (J. Butler and M.F. Lyon, personal communication in Johnson, 1976). It is apparent from the above discussion that mutations in many different genes can cause exencephaly in mice. Among those whose gene products have been identified are genes that code for transcription factors (Sp, Xt, p53), a protein kinase, a kinase substrate, and apolipoprotein B. Chromosomal duplications can also result in NTD (i.e., trisomies 12 and 14). While the vast majority of NTD are caused by mutations in single genes, the penetrance of many of these mutations is affected by background effects, consistent with a multifactorial nature of inheritance. There is heterogeneity in the NTD phenotype (exencephaly, spina bifida, or both) caused by mutations in different genes as well as in the type of NTD caused by mutation of a single gene. The developmental mechanisms by which NTD arise are generally vague, even 54 when the mutant genes have been identified, but they appear to be heterogeneous. In addition, the mutations known to cause NTD in mice show threshold effects (which will be discussed in greater detail below) such as variation in penetrance or expressivity on a fixed genetic background and increased susceptibility to teratogens that cause NTD. It is interesting that there are mutations that appear to result in lack of Closure 1, 2, or 4, or both Closures 2 and 3; this supports the hypothesis that different mechanisms of elevation and fusion are important at different levels of the body axis. Closure 3 would appear to be functionally redundant to Closure 2, being most important when Closure 2 fails. These observations on the sites of exencephaly also support the hypothesis that Closure 4 does exist and that fusion over the rhombencephalon is not merely due to rostral extension of Closure 1 and caudal extension of Closure 2. If the latter case were true, it is unlikely that embryos would be exencephalic only in the hindbrain. SELH/Bc Mice The SELH/Bc (SELH) mouse stock has been developed by Drs. Diana Juriloff and Muriel Harris in the Department of Medical Genetics at the University of British Columbia and the history is described in Juriloff et al. (1989). The SELH stock was derived from a cross of a partially inbred stock of mixed genetic background (BALB/cGa, 129/-, CBA/-), homozygous for the lidgap-Gates (lgGa) mutation, to "random-bred BLU:Ha(ICR)". The lgGa mutation was backcrossed onto this ICR background by intercross-backcross to N3, at which time a new recessive mutation, spherocytosis British Columbia, sph2Bc (Unger et al., 1983) appeared in the second intercross generation (N3 X N3). The parents of the affected pups were used to begin a new stock segregating for sph2Bc. This stock was maintained by brother-sister inbreeding with selection for sph2Bc carriers and against the lgGa mutation. Subsequently, exencephaly was 55 observed in F5 newborns of this new stock. Exencephaly-producing parents were selected for in subsequent generations, and when it was noted that some adults were ataxic, parents that produced both ataxia and exencephaly were selected when possible (D.M. Juriloff, personal communication). At F14, one breeding pair produced a new mutation at the nude locus, nuBc (Koehn et al., 1988), and since that time, SELH has also produced three independent mutations at the albino locus, c80, c2Bc, c3Bc (Juriloff et al., 1992; Juriloff et al., 1994), and independent mutations at unidentified loci that cause the lens of the eye to go opaque, curly hair, and white belly spot (D.M. Juriloff, personal communication). The apparently high spontaneous mutation rate in SELH indicates that this stock may be genetically unstable. SELH mice are studied as a mouse model of neural tube closure defects. The liability to exencephaly is genetic, involving few (2-3) loci and best fitting a multifactorial threshold model of inheritance (discussed below; Juriloff et al., 1989). All SELH mice close their cranial neural tube by an abnormal mechanism. They lack initiation of fusion at Closure 2 at the prosencephalon/mesencephalon boundary. Most SELH mice complete cranial neurulation by extension of Closure 3 (from the most rostral aspect of the neural tube) through the prosencephalon and mesencephalon to the rostral rhombencephalon (Macdonald et al., 1989). About 10-20% of SELH embryos fail to complete fusion by this mechanism, leaving the neural folds unelevated and unfused over the mesencephalon. These embryos are subsequently exencephalic and die shortly after birth. The exencephaly in SELH mice is generally over the midbrain region, with a little variation in how far rostral or caudal the neural tube may remain open. Several of the other mouse models of NTD with exencephaly (with or without spina bifida) have an open neural tube in a region that includes the region left open in SELH: Csk'', Apob', Macs'', p53'', Splotch (Sp), Extra toes (Xt), Exencephaly (xn), Curly tail (ct), Crooked 56 (Cd), Rib fusions (Rf), and embryos with Trisomy 12 or 14. The site of exencephaly in some Loop tail (Lp) embryos and in the offspring of Snell's translocation (T(2;4)lSn) carriers overlaps only slightly with that seen in SELH embryos. Histology of the cranial neural folds of SELH mice during their elevation showed no gross abnormalities in the mesenchyme or neuroepithelium (Macdonald et al., 1989), although some differences between the neural folds of SELH and ICR/Be (normal) strain embryos were apparent. The folds of the rostral mesencephalon of SELH embryos were widely flared and the lateral mesenchyme, ventral to the neuroepithelium/surface ectoderm boundary, was abnormally collapsed (Macdonald et al., 1989). The neural folds of SELH embryos contained slightly more pyknotic cells in the neuroepithelium and mesenchyme and slightly more small unstained inclusions in the neuroepithelium than those of ICR/Be embryos (Macdonald et al., 1989), but these differences have no obvious role in failure of mesencephalic neural fold elevation. About 5-10% of weaned SELH mice are ataxic and have a mid-line cleft between the two halves of the cerebellum (Juriloff et al., 1993; Harris et al., 1994). During cranial neural tube closure, the region of the prospective cerebellum is just rostral to the rhombencephalon. In some cases where rostral closure is particularly delayed, Closure 4 has been observed to extend from the rostral rhombencephalon into the mesencephalon (Harris et al., 1994). Closure over the rhombencephalon initiates between surface ectoderm cells, unlike the more rostral regions of the head where initial contact involves neuroepithelial cells (Geelen and Langman, 1977, 1979). In the rhombencephalon, neuroepithelial cells do not make contact until some time after fusion of the surface ectoderm. If Closure 4 extends rostral into the mesencephalon, initial contact may be made between surface ectoderm cells in this region, and it is possible that the neuroepithelial cells never make contact across the midline. This is consistent with what is seen in histological sections through the prospective cerebellar region of potentially ataxic SELH embryos on days 57 10 and 11 of gestation; the surface ectoderm is fused but the neuroepithelium tapers out and never reaches the midline (Harris et al., 1994). Failure of the neuroepithelium to fuse across the midline would result in a cleft cerebellum and ataxia later in development. The abnormal mechanism of cranial neural tube closure employed by SELH embryos may cause both exencephaly and cleft cerebellum. A pilot study has suggested that there is a genetic correlation between production of exencephaly and production of ataxia (Juriloff et al., 1993), but it did not reach statistical significance. Genetic analysis of genetically complex traits There is no one method of genetic analysis available to uncover the mode of inheritance and the number of loci involved in causing a trait. Each of the existing methods tests a limited number of possible genetic models, and even when all these methods are applied, only a fraction of the possibilities have been examined. These methods can, however, suggest whether there is dominance or additive gene action and whether a single gene, a few loci, or many genes are responsible for the trait. It is not always necessary to identify the absolute number of loci; in particular, differentiating between few or many loci indicates whether mapping of the underlying loci is feasible or not. Success in mapping is more likely if there are few loci to be found; their individual effects on the trait are likely to be greater and fewer animals need to be typed. The genetic basis of a trait can be simple (caused by a single gene) or complex (caused by more than one gene). The former is most likely to be caused by a mutation. On some genetic backgrounds, the expression of this gene may be influenced by modifier loci, which seem likely to be polymorphisms ("neutral mutations") rather than mutations that affect gene function. Genetically complex traits could be the result of mutations or polymorphisms. In the former case, the genetic analysis of the trait should be the same in any crosses to other inbred strains. 5 8 If the trait is caused by polymorphisms, it is likely that some other strains will either share the same allele at one or more loci or have different alleles at yet other loci that can be involved in the trait. The results of genetic analyses of the trait in crosses to various strains would therefore be expected to differ. As it is not possible to know which situation is present, the results of any genetic analysis should be considered as specific for the cross analysed. Many of the existing methods of analysis fall into two main classes; parametric and nonparametric. Parametric methods assume that the data follow a known distribution (e.g., a normal distribution). For threshold traits, which are binomially distributed, this assumption is not met unless the data are transformed by an appropriate method. In addition, the means and variances of binomial data are not independent as in a normal distribution, and this cannot be addressed by transformation. Nonparametric, or distribution-free, methods have been developed in order to avoid the problems of using parametric methods on traits that are not normally distributed. A relatively simple first approach is to compare (for example, by X 2 tests) the observed frequencies of the trait of interest in the parental, F l , reciprocal BCls, and F2 generations to the frequencies expected based on Mendelian segregation ratios under simple genetic models, such as a single autosomal recessive locus or a single codominant locus. There is no conclusive evidence for or against these models, however, unless breeding tests (testcrosses) have been made of a segregating generation, such as the BC1 or F2, as the segregation ratios of more complex genetic models can simulate those of a single locus in the BC1 and F2 generations themselves. This was shown by Sewall Wright in his analysis of the inheritance of Polydactyly in guinea pigs (1934), where the parental, F l , BC1, reciprocal BC1 (rBCl) and F2 generations showed phenotypic segregation ratios consistent with the Mendelian segregation of a single autosomal recessive locus. Testcrosses of individual BC1 animals were not consistent with this 59 model; for example, BC1 animals with the recessive phenotype produced a substantial proportion (22.6%) of offspring with the dominant phenotype when mated to the recessive parent. Despite this, the error of not including breeding tests of segregating generations is often made (i.e., analysis of adrenal structure in mice, Shire 1970; ethanol consumption and preference in mice, Fuller and Collins, 1972; hypertension in rats, Yen et al., 1974; nephrogenic diabetes insipidus in mice, Virgo and Miller, 1977; dactylaplasia in mice, Chai 1981; head spots in mice, Wildman and Doolittle, 1986; insulin-dependent diabetes in NOD mice, Prochazka et al., 1987; epilepsy in mice, Rise et al., 1991; dietary obesity in mice, West et al., 1994). In these studies, any proposed genetic models cannot be considered as conclusive and require further breeding tests to confirm them. A second approach is to compare the observed frequencies of the trait in the F l , BC1, rBCl , F2 and any other available generations (including testcrosses of segregating generations) to the frequencies expected for multiple autosomal recessive loci with duplicate epistasis (Juriloff and Roberts, 1975; Juriloff, 1980; Juriloff and Harris, 1982; Juriolff, 1995; D.M. Juriloff, personal communication). As the mammalian genome is thought to have evolved by duplications of ancestral genomes, it is possible that a number of duplicated genes (that make up gene families) have retained the same or overlapping (redundant) functions. In this case, each of the related genes would need to lose its function in order to produce the mutant phenotype. The expected frequencies of the trait of interest under models with multiple autosomal recessive loci with duplicate epistasis in the multiple homozygote are easy to calculate, and reduced penetrance can be factored in. Support for this type of genetic model also requires breeding tests of a segregating generation. Quantitative traits are often assumed to be caused by many loci. While this is likely to be true for many complex traits, such as body weight or height, it would be possible for even 60 a single gene to result in a continuous distribution of phenotype if there is a significant amount of environmental variation. Stewart (1969) proposed that, as the number of loci that cause a trait may not always be large, it is desirable to have a method to test the hypothesis that segregation at one or two loci can account for the variation seen in a quantitative character in a particular experimental cross. The method he presented involves the comparison of the frequency distribution of a BC1 generation with a composite distribution made up from the parental and F l generations. In the case of a single locus, the BC1 distribution is expected (by Mendelian segregation) to be the same as a distribution consisting of 50% of the parental and 50% of the F l distribution. Stewart (1969) gave a X 2 test to compare the observed and expected distributions. It is also possible to compare the observed data to that expected under a polygenic model (many additive loci) by testing the fit of the observed data to a normal distribution. The observed distribution of an F2 generation can be used in place of or in addition to that of the BC1. This approach is particularly useful when the distributions of the F l and parental generation overlap, making it difficult to classify BC1 or F2 individuals with phenotypic values in the region of overlap. It is also possible to compare the number and position of observed modes in the observed frequency distribution to that expected under some genetic models. For example, if one locus is segregating, a BC1 distribution should be bimodal, with one mode around the mean of the F l generation and the second around the mean of the parental generation. If 2 additive loci with equal effects are segregating, a BC1 is expected to have three modes, where one mode is around the F l generation mean, a second mode of equal size is around the parental generation mean, and a third mode of twice the size is intermediate to these two positions. Stewart's approach assumes the parental and F l distributions to be normal. If they do not meet this criterion, the data need to be transformed by an appropriate method. While this 61 method aids the evaluation of trends in the observed data, it is limited in the number of genetic models it can test. The ability to distinguish different modes is limited by the distance between the means of the parental populations; if the modes are close together, intermediate peaks will tend to overlap and combine to form single modes. The ability to distinguish modes is also limited by the error in estimating the metric (measureing the trait), the relative contribution of each locus and potential interactions between them, and by sample size. In addition, if this approach suggests that 1 or 2 loci are involved in determining a trait, the hypothesis needs to be confirmed by testcrossing a segregating generation. Stewart's approach has been used to look at inheritance of several traits in mice, including adrenal structure (Shire, 1970), a number of physiological characteristics (Stewart and Elston, 1973), nephrogenic diabetes insipidus (Virgo and Miller, 1977), 6-aminonicotinamide- and cortisone-induced cleft palate (Biddle, 1977; Biddle and Fraser, 1977), and spontaneous cleft lip (Juriloff, 1980). It has also been used in conjunction with a maximum likelihood method of analysing a number of genetic models for BC1 data (Elston and Stewart, 1973; Stewart and Elston, 1973; Shire, 1970; Virgo and Miller, 1977; and McCall and Frierson, 1981). Under this latter approach, the log likelihood of observing each of the expected genotypes under several genetic models is calculated from the observed parental, F l and backcross data. Each generation is assumed to be represented by a normal distribution with the observed mean and variance of the generation, and the environmental variance is expected to be the same for each genotype. As the testcross of segregating generation is not included in this model testing, further breeding tests would be necessary to verify the selected preferred model. The Castle-Wright formula (Castle, 1921a, b; Wright, 1934) estimates the number of loci responsible for the difference between two inbred strains in their values for a quantitative trait. Its basis is that the variance of a normally distributed F2 generation is inversely related to the 62 number of loci segregating; with fewer loci, more individuals lie at the extremes of the distribution and so the variance is greater than if there are many loci. The data are assumed to be normally distributed, with the variance determined by independent factors (unlinked loci) with equal and additive effects. In addition, the parental strains are assumed to show the extreme high and low values; i.e., all the. "plus" alleles are in the high parent and all the "minus" alleles are in the low parent. Dominance and epistasis alter the estimate, making it larger or smaller depending on their effect on the variance. For low frequency threshold traits, in which variance is not independent of the mean and is often truncated, the Castle-Wright formula gives an overestimate of the number of loci involved (Biddle, 1975). Application of the Castle-Wright formula is severely limited by its assumptions, which are biologically unrealistic and unlikely to be met by many traits. It is often used and taken as a minimum estimate of gene number (i.e., in Dagg et al., 1966; West et al., 1994), but when the formula is used despite violation of its underlying assumptions, the results are not necessarily convincing. Collins' non-parametric method is based on the tenet that "the individual organism is the solely relevant quantum of biological information" (Collins, 1970). Rather than looking at the data as means of generations, he uses the actual phenotypic distributions to test genetic models (Collins and Fuller, 1968; Collins, 1970; Fuller and Collins, 1972). Looking at the data as the proportion of animals in each generation with a qualitative trait, rather than the mean frequency of the trait in each generation, makes no assumptions about the underlying distribution. Collins often uses Bruell diagrams to represent his data (Collins, 1970; Fuller and Collins, 1972). In such diagrams, the proportion of animals with the trait (Y axis) is plotted against the proportion of genes from one parental strain (X axis). The apices of the genetic triangle in the diagram are formed by the overall proportion of the three genetically uniform populations (parentals and Fl). The observed and expected proportions for intermediate generations under chosen genetic 6 3 model(s) are also plotted. The expected proportions are calculated based on Mendelian segregation ratios and using information available from the parental and F l generations. X 2 tests of goodness-of-fit between the observed and expected values can then be calculated. Collins' approach also shows whether there is additivity or dominance by the position of the F l and segregating generations relative to the parental generations; for example, if the gene(s) causing the trait act additively, the F l and other generations should fall on the straight line between the two parental populations. If the gene(s) are additive, Collins' method cannot be used to determine the number of loci as the position of the observed proportions for the F l and intermediate generations will always fall on the same line. As with any other method, conclusive evidence for any genetic model requires breeding tests of a segregating generation. Inclusion of additional segregating generations, such as F l x BC1, F l x rBCl , or BC1 x rBCl , reduces the number of assumptions to be made about intermediate genotypes in multiple-gene models and also allows more comparisons to be made, increasing the number of degrees of freedom in the X 2 tests. If a trait is caused by gene(s) with additive effects, then the mean of the F l population is expected to lie mid-way between those of the parental generations and the BC1 mean is expected to lie mid-way between the F l and parental values. With Collins' method, the proportion of F l individuals with the trait are expected to lie mid-way between the proportions seen in the parental populations if there is additive gene action. This is not true for threshold traits. Threshold traits are generally assumed to be continuous traits, normally distributed with respect to liability for the trait (Falconer, 1989), but the threshold imposes a discontinuity on the visible expression of the trait. The result is binomial data, giving the proportion of affected and normal offspring. The incidence of the trait therefore represents the proportion of the normal distribution that falls over the threshold and the mean incidence of a trait for a population does 6 4 not represent the mean of the underlying normal distribution. In order to test a threshold trait for additive gene action, it is therefore necessary to transform the data in such a way that uncovers the mean of the underlying normal distribution (i.e., by probits, Finney, 1971). The means can then be plotted relative to genotype; if there is additive gene action, they should show a linear relationship. This approach is not foolproof, however. The proportion of the normal distribution that falls over the threshold depends upon the variance of that distribution (the larger the variance, the wider the distribution, and so more individuals are expected to fall over the threshold). The variance of a BC1 generation includes genetic and environmental components and is therefore expected to be larger than that of the parental or F l generations, which have only genetic variance. The actual mean of a BC1 generation should therefore be closer to the normal parent (the one with fewer individuals over the threshold) than will be determined by probit transformation. This approach also assumes that the F l and both parental strains have equal variances, which may not be true. Recombinant inbred strains have also been used to determine the number of genes that affect many traits in mice and rats, including penetrance of the curly-tail trait in diet mice (Neumann et al., 1994), susceptibility to isoniazid-induced seizures (Taylor, 1976), catalase activity (Schisler and Singh, 1991), hepatocarcinogenesis (Drinkwater and Ginsler, 1986), and the development of autoimmunity (Raveche et al., 1981), audiogenic seizures (Seyfried et al., 1980; Neumann and Seyfried, 1990; Neumann and Collins, 1991), epilepsy (Frankel et al., 1994), and hypertension (Pravenec et al., 1988). Recombinant inbred strains are developed by brother-sister inbreeding of an F2 generation from a cross between two inbred strains. Thus, a set of recombinant inbred strains approximates a segregating generation, where each strain represents a single F2 individual (Bailey, 1981). As these strains are inbred, not all F2 65 genotypes will be recovered as all the mice will be homozygous at all loci. If a single gene causes a trait, then only two types of recombinant inbred strains are possible, where each is like one parental strain or the other. A single autosomal recessive gene cannot be distinguished from a single codominant gene as the F l genotype (heterozygous) will not be represented. If a non-progenitor-strain phenotype is seen, more than one locus must be involved in causing the trait. The more genes that are responsible for a trait, the less likely it is that the progenitor-strain phenotypes will be seen as the progenitor-strain genotypes will be the rarest (Bailey, 1981). Most recombinant inbred strains do not survive to reach the level of an inbred strain (21 generations of consecutive brother-sister inbreeding), and so only a relatively small number of strains are available; rarely more than about 25. The probability of not finding one recombinant inbred strain with a progenitor-strain phenotype is therefore quite high as the number of underlying loci increases. For example, only 1/8 of recombinant inbred strains are expected to have a progenitor-strain genotype if three loci cause the trait, and the probability of not finding one in 25 strains is fairly high. In addition, recombinant inbred strains have only been made from crosses between a limited number of strains, so they can only be used when two progenitor strains differ for the trait of interest. A relatively common tautology in genetic analyses is to test a single gene model (often not including a testcross of a segregating generation) and, when it does not fit the data, to propose an alternative model to fit the data. The alternative model is then tested on the same data that was used to create it (i.e., analyses of: dactylaplasia in mice by Chai, 1981; audiogenic seizures in mice by Ginsburg and Miller, 1963; and head spots in mice by Wildman and Doolittle, 1986). Simply rejecting a single locus model and showing fit to an alternative model does not constitute a thorough genetic analysis, particularly when there are other methods of analysis available. As previously mentioned, the best approach is to test a number of genetic 66 models, using some of the methods discussed above. Even using these approaches, the best model is only the best of those examined; there are such a large number of possible models, incorporating various number of loci, linkage relationships, dominance or additivity, and epistatic effects, that it is not feasible to test them all. The best fitting model will, however, give an estimate of gene number, and this estimate can be used to determine whether or not mapping of the loci is feasible. The genetic basis of exencephaly in SELH/Bc mice The previous genetic analysis of the cause of exencephaly in SELH mice (Juriloff et al., 1989) involved a cross between SELH mice and the normal but related strain, ICR/Be, and used several of the methods of analysis described above. In this previous study, classical genetic crosses were made to determine the frequencies of exencephaly recovered in reciprocal F l , reciprocal F2, BC1, and BC2 embryos. The low, near zero, frequency of exencephaly in F l embryos indicated that exencephaly in SELH mice is not caused by an X-linked recessive or autosomal dominant gene, nor is it dependent on maternal genotype. The frequencies of exencephaly observed in F l , F2, BC1, and BC2 embryos were compared to the frequencies expected for a single autosomal recessive locus with reduced penetrance (16.9%, as this was the frequency of exencephaly observed in SELH embryos), but all differed significantly. The frequencies of exencephaly observed in these generations were also compared to those expected for two or three autosomal recessive loci with duplicate epistasis and 16.9% penetrance in the multiple homozygote. In each case, the data for three of the four generations differed significantly from expected, indicating poor fit of the data, overall, to these models. The observed frequencies of exencephaly in the parental, F l , BC1, and BC2 generations were also compared to the values expected under an additive polygenic threshold model. The fit of the 6 7 mean liability values of the various generations to linear regression when plotted against the proportion of SELH genes in each of the generations was almost perfect, indicating very good fit of the data to additive genetic inheritance and the multifactorial threshold model. This method of analysis does not, however, provide insight into the number of genes that cause exencephaly. This previous genetic analysis used three methods to determine the number of genes responsible for exencephaly-liability in SELH mice. First, Stewart's curve smoothing approach was applied to the distribution of frequencies of exencephaly produced in the BC2 progeny of BC1 sires. Whether examined as raw percentages, mean litter arc-sine-transformed frequencies, or smoothed frequencies based on the transformed frequencies, there was no evidence of bimodality, as would be expected if a single locus were segregating in BC1 sires. Instead, the frequency distribution appeared asymmetrical, with a mode near the lower end of the frequency scale. The second method used to estimate gene number was to look at the proportion of BC1 sires that produced SELH-like frequencies of exencephaly in their testcross (BC2) progeny. For example, if a single locus is segregating, one-half of BC1 sires are expected to be SELH-like with respect to exencephaly; for two loci, one-quarter of BC1 sires are expected to be SELH-like; for three loci, one-eighth of BC1 sires are expected to be SELH-like. Three of twenty-six BC1 sires were similar to SELH in exencephaly production (produced at least 15% exencephaly in BC2 progeny), and a total of seven sires did not differ significantly from 17% exencephaly in their progeny. Thus, between one quarter (7/26) and one eighth (3/26) of BC1 sires were SELH-like in exencephaly production, consistent with the segregation of two or three exencephaly-causing loci. Finally, for an additive genetic trait, the mean frequency of exencephaly observed in the 68 F2 generation is expected to be the same as in the F l generation on the liability scale, but the variances of these two generations will differ. The size of the variance in the F2 generation is inversely related to the number of loci segregating for the trait under study. Thus, the larger the number of loci, the smaller the variance, and the smaller the proportion of the F2 distribution that will fall over the threshold. Thus, if many loci are segregating, the frequency of exencephaly in the F2 generation is expected to be low, similar to that seen in the F l . The frequency of exencephaly observed in the F2 (2.4%) was more than one-half that expected for a single locus (4.25%), indicating that the number of loci involved is small. As exencephaly in SELH mice appears to fit a multifactorial threshold model of inheritance (Juriloff et al., 1989), a genetic maternal effect could be one of the factors that influences the distribution of the embryo's liability relative to a threshold value (Fraser, 1976). Although development is generally considered to be controlled by embryonic genes, many developmental processes are influenced by parental genotype. Most commonly, these influences are maternal effects, where genes acting in mothers exert an influence upon the expression of a trait in their offspring. A variety of mechanisms have been proposed to explain maternal effects in mice, including mitochondrial inheritance (Pollard and Fraser, 1973), stored maternal mRNAs (McLaren, 1979), uterine environment (McLaren and Michie, 1958), transplacental transport of maternal growth factors (Juriloff, 1987) or antibodies (Bonner and Slavkin, 1975; Billington and Wild, 1979; Popp, 1979), and maternal genetic imprinting (Cattanach, 1986; Sapienza et al., 1986; Monk, 1988). Risk of NTD in humans may be influenced by maternal age, maternal illness, and maternal diet (reviewed by Campbell et al., 1986), and maternal folate metabolism appears to influence risk of NTD in offspring (Czeizel and Dudas, 1994). The presence of large genetic effects on liability to SELH-type exencephaly was tested in a previous study (Gunn et al., 1992). SELH mice were crossed to the normal, unrelated 69 SWV/Bc strain. Litters of reciprocal F l (SELH.SWV and SWV.SELH) and every combination of reciprocal BC1 (SWV.SELH x SELH, SELH.SWV x SELH, SELH x SWV.SELH, and SELH x SELH.SWV) embryos were scored for frequency of exencephaly. SWV.SELH F l animals inherit their cytoplasm and mitochondria from SWV, while SELH.SWV F l mice inherit SELH cytoplasm and mitochondria. Each type of SWV.SELH F l female was backcrossed to SELH sires; this cross can detect cytoplasmic (including mitochondrial) inheritance of factors that influence exencephaly. Each type of F l male was backcrossed to SELH females; this cross can detect Y-linked and additive X-linked factors as the female progeny differ in the strain origin of one of their X chromosomes and the male progeny differ in the origin of their Y chromosome. Comparison of frequencies of exencephaly in litters from SELH females with those from F l females tests for maternal effects due to nuclear genetic differences expressed in mothers. There were no statistically significant differences between the frequencies of exencephaly observed in BC1 embryos from SELH or F l mothers, or in the BC1 offspring from reciprocal F l females or males. Thus, the difference in liability to exencephaly between SELH and SWV mice does not appear to be influenced by genetic maternal effects. This suggests that the defect in SELH embryos is more likely to depend on a cell-autonomous mechanism, since mechanisms influenced by intercellular agents such as growth factors or hormones might be predicted to be susceptible to influences from genes expressed in maternal tissues. The crosses made between SELH and SWV/Bc also provided the opportunity to extend and test the conclusions of the previous genetic analysis of exencephaly liability (Juriloff et al., 1989) in a cross between SELH and a different normal strain (SWV/Bc rather than ICR/Be). The frequency of exencephaly observed in the pooled BC1 generation differs significantly from that expected if a single autosomal recessive locus were responsible for exencephaly liability. The results are generally consistent with those of the previous genetic analysis (Juriloff et al., 70 1989), with good fit to a multigenic additive threshold model of inheritance. However, as extensive genetic crosses were not made and a segregating generation was not testcrossed, it was not possible to exclude the hypothesis of the joint effect of three or four recessive loci with duplicate epistasis and 23.9% penetrance. The similarity of the results between crosses of SELH to a normal, related strain (ICR/Be) and to a normal, unrelated strain (SWV/Bc) suggests that the exencephaly-causing genes in SELH are mutations, not polymorphisms. The multifactorial threshold model of inheritance The best fitting genetic model for liability to exencephaly in SELH mice is multifactorial inheritance with a development threshold (Juriloff et al., 1989). Under this model (Wright, 1934; Fraser, 1976; Fraser and Nora, 1986; Falconer, 1989), populations are normally distributed for a continuous variable (i.e., liability to NTD), with a developmental threshold separating the continuous distribution into two discontinuous segments. For prenatal processes such as neural tube closure, it is possible to tell whether individuals in a population fall over the threshold, but not where on the distribution they lie. Genetic, environmental, and stochastic (random) factors influence the position of the distribution relative to the threshold as well as influence the position of the threshold itself, and the genetic component may involve the action of one or more genes. The means of the distributions of different inbred strains of mice will lie at different points on the liability scale due to the different alleles they carry at loci that affect neural tube closure, but each of these distributions can be moved closer to the threshold by environmental and/or stochastic factors (Fig. 2). It is for this reason that mouse strains with an increased genetic liability to NTD have an increased susceptibility to teratogens that cause NTD; the teratogen is an environmental factor that moves all distributions closer to the threshold. Thus, if a distribution of one strain lies closer to the threshold than another due to genetic factors, the 71 soAjqwe jo jeqwnu 72 greater the proportion of the normal distribution of the former strain that will be moved over the threshold and the more individuals will be affected compared to the latter strain. For neural tube closure defects such as exencephaly in SELH mice, a continuous variable (liability scale) could be timing of (cranial) neural fold elevation; embryos with delayed fold elevation have a higher probability of not completing fold elevation and fusion within the time during development when they are biochemically competent to do so. This biological limit, the latest point in development when the neural folds are able to complete elevation and fuse, would represent the developmental threshold. Under the multifactorial threshold model, the abnormal closure mechanism in SELH embryos puts their mean liability near the threshold for developing exencephaly, and a portion (18%) of the normally distributed population falls over this threshold (i.e., the dashed-line distribution in Fig. 3). The abnormal pattern of neural tube closure is the result of mutations in a small number (two or three) of genes (Juriloff et al., 1989). As all SELH embryos are genetically uniform and share a common environment, it is likely that stochastic factors (Kurnit et al., 1987) affect liability to exencephaly. SELH mice show an increased liability to retinoic acid induced exencephaly (Tom et al., 1991), with the effects of teratogen and genotype being additive. Threshold traits include those caused by single genes, where the expression of these genes is influenced by genetic background, environment and/or stochastic factors. Thus, populations can show a normal distribution of liability to a trait caused by a single gene, but only a portion of the population will fall over the threshold and be affected. The proportion of individuals that are affected represents the penetrance of the trait (Fraser and Nora, 1986). The severity (expressivity) of the trait may also be determined by the position of an individual relative to the threshold, where individuals near the threshold are mildly affected and those far beyond the threshold are severely affected (Fraser and Nora, 1986). Many of the single gene CO O o ) c •E a) = > CO CO >. "co ~ i s cu -3 CD E o CO SZ TJ CO C CD 73 _CO _ _ CO > 3 ^ CD « Jc > O "° E CO co i l .E "55 CO CD . £ CO J ) CO CD — O co it c .a ~j J I \ ttt m E CD CD CO CO a> o > = > £ >- * r CD TJ x -C - '= CD o o_ CD o c 0 X 0 •— o "O CM II a." o 0 il= CO CN + * \- o o _c C/) 0 E o M — CO c o as > 0 "D •a co T - "D i c CO CT) >> co CNJ — 1 M — o 0 CO o CO CO I TJ — Q . .!= TJ ±3 74 mutations in mice that cause NTD discussed above show threshold effects such as reduced penetrance, variable expressivity, and increased susceptibility to teratogens that cause NTD. Since most human NTD appear to be multifactorial in etiology, SELH mice are a valuable multifactorial animal model for the study of NTD. The NTD seen in most of the other mouse models are caused by single genes, often with associated abnormalities (models of "syndromes") and often with some abnormal phenotype in heterozygotes. It seems likely that these genes act in a different manner from those that contribute to polygenic NTD; the latter cause a "predisposition" but require additional genetic, environmental, or stochastic factors to produce the NTD. For example, the single genes that cause multiple defects seem more likely to affect basic cellular processes that are involved in many morphogenetic events, whereas mutations that only affect neurulation seem more likely to affect a cellular process specific for neurulation. The primary defect caused by some single gene mutations may be in another system but affect neurulation indirectly, such as the effect of the curly tail gene on cell proliferation in different cell types in the region of the posterior neuropore. It therefore seems likely that many of the single gene mutations that cause NTD do not specifically cause defects in the driving force(s) of neurulation. A mutation that compromises one mechanism of neural fold elevation and/or fusion may not result in a significant incidence of NTD if there is redundancy between the different driving forces of neurulation at each level of the neural tube. Thus, mutations that cause defects in these driving forces may be more likely to show multigenic inheritance, where neural tube closure is only affected when two or more genes involved in neurulation are mutated. Alternatively, it has been suggested (Tanksley, 1993) that while single gene mutations are most likely to be null alleles, polygenic traits are more likely to involve allelic forms of a gene that modify its product rather than eliminate it; in this case, the same gene could be involved in both monogenic and 75 polygenic disorders, the mode of inheritance depending on the type of mutation. Animal models such as SELH are important to the understanding of the various mechanisms by which NTD arise as experimental studies of human embryos are not possible. In addition, heterogeneity in human populations and in the causes of NTD in human embryos makes the mapping of genes that underlie a complex trait very difficult, but the genes that cause these traits can be mapped using inbred strains of mice. The homologs of these murine loci can then be examined for mutations in susceptible human populations. The mouse-human comparative maps based on linkage homology (e.g., Nadeau et al., 1992) can also be used to look for candidate genes in homologous regions of the human genome as different information may be available for each genome. Even when a candidate gene cannot be identified, mouse-human comparative maps can often be used to identify the most likely map location of the human homolog. Quantitative traits Many traits do not exhibit classic Mendelian patterns of inheritance. These "complex" traits often show reduced penetrance and their phenotype may be caused by multiple loci, each with multiple alleles, both within individuals (multigenic or polygenic inheritance) and within a population (genetic heterogeneity). Environmental and stochastic factors may also be involved determining phenotype. This class of trait includes neural tube closure defects. Variation in complex traits often shows complex inheritance. Quantitative traits demonstrate complex inheritance, where phenotypic variation follows a continuous distribution caused by variation in more than one gene and by variation in non-genetic factors. Genes concerned with Mendelian traits ("major genes") are distinguished from those concerned with quantitative traits ("minor genes") by the magnitude of their effects relative to other sources of 76 variation (Falconer, 1989). Some genes, however, have intermediate effects and cannot be classified as major or minor. Variation caused by the simultaneous segregation of many genes with small, additive effects is said to be polygenic, and the genes are sometimes referred to as quantitative trait loci (QTLs), or polygenes (Falconer, 1989). The term "QTL" will be used in this discussion in a more general sense to represent any locus that causes variation in a quantitative trait. Although there is no genetic variation within an inbred strain of mice, non-genetic sources of variation can create a continuous distribution for a genetically determined trait. When different inbred strains are intercrossed, the resulting F l animals are also genetically homogeneous, but if they are intercrossed or backcrossed to either parental strain, the variation in the resulting F2 or BC1 animals will be greater than that of the parental or F l populations as these generations have both genetic and non-genetic sources of variation. Thus, although all SELH mice are genetically homogeneous, they can show continuous variation for liability to exencephaly (i.e., timing of cranial neural fold elevation; see Fig. 3); this variation is greater in segregating generations from crosses between SELH and other strains. Many characters vary in a discontinuous way, but are not inherited in a simple Mendelian manner. Sewall Wright (1934) first recognized that inheritance of a discontinuous character where there are two phenotypic classes (i.e., affected or normal) could be accounted for by multifactorial inheritance of a continuously distributed variable where a developmental threshold divides the continuous distribution into two discontinuous segments. Thus; the underlying liability to traits that display multifactorial inheritance with a developmental threshold may show continuous variation, but the observed phenotype is qualitative, or binomial, in nature (affected or normal). In the case of exencephaly in SELH mice, liability to exencephaly is caused by more than one gene (Juriloff et al., 1989) and appears to be influenced by non-genetic factors, 77 making it quantitative in nature. The developmental threshold that represents the biological and/or temporal constraints on neural tube closure divides the continuous distribution of liability into two portions; the animals that lie in the portion of the distribution that falls over the threshold fail to complete cranial neural tube closure and become exencephalic, while those that lie in the portion of the distribution that remains below the threshold do complete cranial neural tube closure and are normal (Fig. 3). Mapping loci that cause variation in quantitative traits Mapping the loci that cause variation in quantitative traits is more difficult than mapping single genes because the presence of a particular allele at any one QTL is not necessarily required for development of the trait. This means that one affected individual in a population segregating for the QTLs will not necessarily share all of its "liability" alleles in common with other affected individuals, making it virtually impossible to find perfect cosegregation of the trait with even a genetic marker that is very closely linked to a locus carrying a liability allele. The strategy of mapping loci that cause variation in quantitative traits is to detect linkage, or non-random association, between genetic markers and loci that have a major effect on a quantitative trait (Soller et al., 1976; Edwards et al., 1987; Soller and Genizi, 1978; Lander and Botstein, 1989; Haley and Knott, 1992; Luo and Kearsey, 1992; Stuber et al., 1992; Darvasi et al., 1993; Jansen, 1993; Moreno-Gonzalez, 1993; Rodolphe and Lefort, 1993; Zeng, 1993, 1994; Haley et al., 1994; Jansen and Stam, 1994; Kruglyak and Lander, 1995). The major requirements for mapping QTLs are an appropriate genetic cross, a dense map of polymorphic genetic markers, and a suitable analytical approach. It must also be possible to score the trait quantitatively, even if its resulting phenotype is qualitative, as for multifactorial threshold characters. 78 By crossing two inbred strains of mice that differ significantly with respect to the trait and intercrossing their progeny, an F2 generation segregating for the QTLs can be obtained. These animals can then be scored for the trait to estimate their underlying genetic liability and typed for genetic markers. The data can then be analysed for a statistical correlation or linkage between degree of liability to the trait and genotype. For example, an F2 individual with a high phenotypic "score" is expected to inherit most or all of its marker alleles in the region of a QTL from the parental strain that demonstrates the trait, whereas an F2 individual with a low "score" should inherit most or all of its alleles from the unaffected parental strain. The number of progeny to be typed can be reduced by selective genotyping (Lander and Botstein, 1989), where only the animals at the extreme ends of the distribution, for example those with phenotype values in the top and bottom 10%, are typed for genetic markers. The animals in the top 10% should have most or all of their alleles at the underlying QTLs from the "high" parental strain, while those in the bottom 10% should inherit most or all of their QTL alleles from the "low" parent. This involves examining the phenotype of a large population, but then only genotyping a fraction of it. The individuals whose genotype can most easily be inferred from their phenotype contribute the most linkage information (Lander and Botstein, 1989). In addition, the larger the phenotypic effect of the QTLs underlying a trait, the fewer the number of progeny that need to be typed to map them (Lander and Botstein, 1989; Friedman et al., 1991). The smaller the number of QTLs, the larger their effect should be for a strain difference of fixed size. It is therefore important to have an estimate of the number of causative genes differing between the parental strains chosen for the original genetic cross, as this helps determine the number of progeny to be scored. The methods described above in "Genetic analysis of genetically complex traits" can be used to estimate the number of loci segregating. It is best to choose parental strains such that the alleles in the "high" (liable) strain all 79 increase the phenotype while those in the "low" (normal) strain all decrease the phenotype. This is usually possible by choosing inbred strains where artificial selection for ("high" strain) and against ("low" strain) the trait has occurred. If there are high and low alleles in one strain, segregating progeny will exhibit phenotypes beyond the extremes of the parental strains. This also confuses the data analysis as a "high" individual could inherit any combination of liability alleles from both parents, making linkage analysis more difficult. Mapping loci for genetically complex threshold traits Mapping the loci that cause variation in genetically complex threshold traits requires an additional step to determine the phenotype of individuals at the quantitative level, as the observed phenotype is discontinuous (affected or normal). For example, mice segregating for the genes that cause a trait such as exencephaly in SELH/Bc mice are either normal or exencephalic, depending on where they lie relative to the threshold, but their position in the continuous distribution is influenced by genetic and non-genetic factors. This can cause animals with a lower genetic liability to fall over the threshold and be affected, whereas animals with a higher genetic liability may be normal due to non-genetic factors that allow them to complete cranial neural tube closure. Thus, unaffected (non-exencephalic) mice may carry alleles that confer high risk at all the loci involved in causing exencephaly while exencephalic mice do not always carry alleles that confer the highest genetic liability to exencephaly. The underlying genetic liability to exencephaly can be measured indirectly by progeny testing. Measuring the underlying quantitative phenotype of individuals in a population segregating for genes that cause a threshold trait or complex lethal trait by progeny testing involves scoring the testcross progeny of individuals segregating for the trait of interest. The mean value of an individual's offspring represents the individual's breeding value, and this 80 breeding value represents the sum of the average effects of the genes carried by that individual (Falconer, 1989). For lethal traits, such as exencephaly, only surviving (non-exencephalic) individuals can be testcrossed; however, as mentioned above, these unaffected animals may carry high-liability (i.e., SELH) alleles at all loci involved in causing the trait. Strategies for mapping QTLs There are two general methods of data analysis to detect linkage between QTLs and genetic markers. The first of these is single point analysis, where one marker is analysed at a time (Sax, 1923; Geldermann, 1975; Soller and Genizi, 1978; Tanksley et al., 1982; Edwards et al., 1987). If there are enough markers, closely spaced and covering the entire genome, it is likely that a QTL will show linkage to at least one marker. Association between a QTL and marker can be detected by calculations based on the normal distribution, looking for a significant difference in the mean phenotypic value of different genotypes (Tanksley et al., 1982), or using standard linear regression (Edwards et al., 1987). Other statistical tests can also be used, such as X 2 , t- and F-tests (i.e., in Todd et al., 1991; Soller et al., 1976; Soller and Genizi, 1978; Stuber et al., 1992) to look for an association between parental origin of marker alleles and phenotype. The second approach, termed interval mapping, analyzes sets of linked markers simultaneously using an adaptation of maximum likelihood methods and LOD scores (Paterson et al., 1988; Lander and Botstein, 1989; Luo and Kearsey, 1992). For each genetic marker, the most likely phenotypic effect of a putative QTL is calculated. This is the effect that maximizes the likelihood of obtaining the observed data. A Lod score (log10 of the odds ratio) is determined, where the numerator of the odds ratio is the probability that the observed data would arise from a QTL with this effect and the denominator is the probability it would arise given no linked QTL. Presence of a QTL is inferred when the Lod score exceeds a pre-81 determined threshold where there is a 5% or less chance of a single false positive (based on genome size and density of the genetic markers typed). A computer program, QTL-MAPMAKER (Lander and Botstein, 1989), has been written to perform the complicated calculations necessary for this method. Modifications of this approach include using multiple regression analysis of flanking markers (Haley and Knott, 1992; Jansen, 1993; Zeng, 1993, 1994). Interval mapping compensates for recombination between markers and a QTL; if a QTL is between two markers, both will show linkage to the trait, and the markers that flank these markers are also likely to show some linkage. Interval and single point analysis appear to give equivalent results if markers are less than 15 cM apart (Stuber et al., 1992; Haley and Knott, 1992; Darvasi et al., 1993). The power of interval mapping to detect QTLs is decreased if the following assumptions are not met: (1) that all the alleles that "increase" the quantitative phenotype are fixed in one parent and all the alleles that "decrease" the phenotype are fixed in the other; (2) that one (or few) QTLs of relatively large effect cause variation in the trait, as opposed to many loci with small, equal, and additive effects (polygenes); and (3) that the phenotype follows a normal distribution with equal variance in both parental strains. The latter two assumptions are particularly important, as failure to meet them could in fact result in failure to detect QTLs or in unacceptably high false positive rates. In order to address the difficulty of finding polygenes using interval mapping methods, Rodolphe and Lefort (1993) have proposed an alternative method which considers the genome as a whole and detects chromosomal regions involved in causing the difference between two homozygous lines and their BC1 or F2 progeny for a quantitative trait phenotype. They suggest that the simultaneous use of different methods, each of which is optimal for a specific situation, 82 improves the efficiency of detecting QTLs. Detecting loci that cause variation in nonparametric traits using interval mapping methods may be possible if a mathematical transformation is found that converts the trait into an approximately normal distribution with equal variance in both parental strains (Wright, 1968). It is not always possible to find an appropriate transformation, however, and even if one is found, the effect of outliers may be too great (Kruglyak and Lander, 1995). An alternative method is to apply distribution-free methods of analysis, such as that proposed by Kruglyak and Lander (1995). This approach extends the use of the nonparametric Wilcoxon rank-sum rank test (Wilcoxon's signed-ranks test in Sokal and Rholf, 1981) from single point analysis, where it is used to determine whether the distribution of phenotypes for a quantitative trait differs between groups of animals that have been classified to groups based on genotype at a marker locus. In order to extend this approach to a genome-wide search for QTLs, Kruglyak and Lander (1995) have generalized the Wilcoxin rank-sum statistic to the situation of interval mapping and provided a means of determining an appropriate significance threshold that maintains a low false positive rate. Unlike the Lander and Botstein (1989) approach, this nonparametric method requires genotyping of all individuals in the segregating generation, rather than just those at the extremes of the distributions. This nonparametric approach has been incorporated into the QTL mapping package MAPMAKER/QTL (version 2). As this nonparametric approach only tests for the presence of a QTL, while the parametric method described by Lander and Botstein (1989) also provides a direct estimate of the phenotypic effect of the QTL, the use of both approaches is recommended by Kruglyak and Lander (1995). Interval mapping, using Lander and Botstein's approach (1989) and the QTL-MAPMAKER program, has been successful in finding QTLs underlying several traits. Two 83 (Bpl and Bp2) of the estimated 3-5 genes that influence blood pressure in rats (Jacob et al., 1991) have been identified, as has one locus {dietary obesity-], Do-1) of the 1-4 genes that cause the difference in adiposity between male AKR/J and SWR/J mice (West et al., 1994) and one locus (Ter) that influences testicular size (weight) and causes increased susceptibility to testicular teratomas (Asada et al., 1994). The map position of one of the genes (Szfl) that causes high frequency of epilepic-like seizures in the SWXL-4 recombinant inbred strain has been identified (Frankel et al., 1994), but it is unclear how many loci are involved in causing this trait or what their mode of inheritance may be. One locus (Pas-1) that causes susceptibility to lung tumours in mice has been identified (Gariboldi et al., 1993); it is not clear whether more loci were not found because marker loci were not available for all chromosomal regions or because there is only one locus that causes variation in this trait in the cross made (A/J x C3H/He). One locus (Mom-1) of the estimated 1-2 loci that influence the number of tumours found in mice homozygous for a mutation of the Min-l (multiple intestinal neoplasia) gene (Dietrich et al., 1993) has also been identified using Lander and Botstein's interval mapping methods (1989) and the QTL-MAPMAKER program. This latter study was reanalysed using Kruglyak and Lander's nonparametric approach (Kruglyak and Lander, 1995), with the same results. Rise and colleagues (1991) mapped two genes (El-1, El-2) that cause epilepsy in mice using both traditional (X 2 and t-tests) and interval mapping (QTL-MAPMAKER) approaches. The original analysis involved looking for cosegregation between individual genetic markers and mean seizure frequency, using X 2 and t-tests to test for a significant difference between groups. This analysis indicated strong association between seizure susceptibility and chromosome 9 and chromosome 2 markers. Subsequent application of interval mapping methods, using the QTL-MAPMAKER computer program, also indicated seizure susceptibility loci on chromosomes 2 and 9. This demonstrates that interval mapping methods are not necessarily superior to 84 traditional, single-point, methods of analysis. Spontaneous development of insulin-dependent diabetes mellitus in the NOD (non-obese diabetic) mouse strain is determined by at least nine unlinked loci (ldd-1 through Idd-9). Traditional methods of analysis have been used to map two of these loci (Idd-3, and -4; Todd et al., 1991), using X 2 tests to evaluate associations between phenotype and genotype (at individual marker loci). Loci that cause variation in genetically complex traits have also been mapped using other approaches involving congenic and recombinant inbred strains. Cleft lip with or without palate (CL(P)) in mice appears be caused by either a single major recessive locus (Juriloff, 1980; Biddle and Fraser, 1986) or two recessive loci with duplicate epistasis (Juriloff, 1980; Juriloff, 1995). A major locus for CL(P) has been mapped to chromosome 11 in a congenic strain that carries its alleles at the major CL(P)-causing gene(s) from the genetically liable strain, A/WySn, on an AEJ/RkBc (normal strain) genetic background (Juriloff and Mah, 1995). This approach is based on the fact that a congenic strain pair (Flaherty, 1981) should differ only in small regions containing the trait-causing gene(s). Congenic animals are expected to be heterozygous at the trait causing locus or loci, as they are generated by backcrossing to the normal strain. Thus, when typed for genetic markers, congenic animals that produce CL(P) in their offspring (i.e., are CL(P) carriers) are expected to be heterozygous for A/WySn and AEJ/RkBc alleles at and near a CL(P)-causing gene. An appropriate pattern of inheritance was seen in these animals for markers on chromosome 11. In addition, the affected testcross progeny of these animals were homozygous for A/WySn alleles at the DllMit lO marker, as expected if it lies near a CL(P)-causing gene. Recombinant inbred strains have also been used to map the loci that cause variation in genetically complex traits. For example, three of the loci that influence the penetrance of NTD 85 in ctlct mice (Neumann et al., 1994) were mapped in backcrosses of ctlct mice to the BXD/Ty (C57BL/6J X DBA/2J) set of 23 recombinant inbred (RI) strains. The distribution of the relative frequencies of the curly-tail trait was close to continuous, with no frequencies similar to those of backcrosses to C57BL/6J and DBA/2J themselves. This suggests that more than one or two "modifier" loci influence the frequency of NTD in these backcrosses, as supported by an estimate of 3.2 loci obtained from an analysis of variance. Tests of association between the frequency of affected mice in backcrosses to the BXD RI strains and strain distribution patterns of over 550 marker loci were performed. In addition, the crosses were made of ctlct mice to conventional linkage stocks and X 2 tests used to detect significant deviation from Mendelian proportions (i.e., association between genotype and phenotype) in affected offspring. These studies indicated association between the frequency of affected animals and two loci, on chromosomes 3 and 5. The loci that influence susceptibility to audiogenic seizures in mice have also been mapped using RI strains (Neumann and Seyfried, 1990; Neumann and Collins, 1991). Polymorphic linkage markers In order to look for an association between genotype and phenotype, a dense map of highly polymorphic genetic markers is required. There are several types of genetic markers available, such as the mouse coat colour loci, electrophoretic mobility variants of proteins, restriction fragment length polymorphisms (RFLPs), and simple sequence length polymorphisms (SSLPs; also known as microsatellites). SSLPs (Love et al., 1990; Aitman et al., 1991; Cornall et al., 1991; Hearne et al., 1991; Dietrich et al., 1992; Hearne et al., 1992) are simple sequence nucleotide repeats, in which a 1-6 basepair motif is repeated 10-50 times. They are abundant, spaced throughout the mouse genome, and are highly polymorphic, even between inbred strains, due to variation in the number of repeat units. SSLPs are also fast and easy to 86 type. Unique-sequence PCR (polymerase chain reaction) primers that flank the nucleotide repeat region can be used to amplify the SSLP region and the PCR product can be visualized by electrophoresis through agarose gels. PCR primers for amplification of SSLP sequences have been developed by several groups, and it is now possible to purchase mouse "MapPairs"™, PCR primers that flank simple sequence repeats, from Research Genetics, Inc. There are over six-thousand mouse MapPairs™ now available, and the map positions of the SSLPs amplified using these primers have been determined. This information is being integrated with existing genetic maps, making SSLPs an invaluable tool as genetic markers for mapping and linkage studies. PCR is an in vitro method for the exponential amplification of a specific region of DNA (Saiki et al., 1985, 1988; Mullis et al., 1986; Mullis and Faloona, 1987). It involves the use of two oligonucleotide primers that hybridize in opposite orientation to unique DNA sequences on the two strands of target DNA, flanking the region to be amplified. Repeated cycles of heat denaturation of the target DNA, annealing of the primers, and extension of the primers by a thermostable (Taq) DNA polymerase result in amplification of the region between the primers (the target DNA). In each subsequent round, the DNA synthesized in preceding rounds as well as the original target DNA are used as templates for amplification of this region, resulting in exponential amplification after the first round. In 25-30 cycles, the target DNA can be amplified by several millionfold. The temperature used for annealing and the concentration of the chemical components required by the Taq polymerase can be varied to increase or decrease the specificity of the amplification reaction. Visualization of PCR product is based on the exponential amplification of a specific region of DNA. If genomic DNA is amplified, there will be several million times as many copies of the target DNA as of the original genomic DNA. The original concentration of target DNA is quite low. If 1 ng of target DNA is used, over 1 mg can be produced by PCR. 87 Electrophoretic separation of DNA fragments of different size in an agarose gel, followed by staining of the gel with ethidium bromide and observation under ultraviolet light, allows visualization of any bands of DNA containing at least 10 ng of DNA (Sambrook et al., 1989). Thus the amplified DNA can be seen as discrete band(s), the position of which depends on size (in nucleotides) of the DNA fragments. The Genetics and development of abnormal neural tube closure in SELH/Bc mice This thesis presents work on the genetics and development of exencephaly and the abnormal mechanism of neural tube closure in SELH/Bc mice. In order to determine whether there is a detectable morphological defect in the neural folds of SELH mice at earlier stages of development than previously examined, a histological study was performed, comparing the morphology of the cranial neural folds of 4-11 somite-stage SELH, ICR/Be, and LM/Bc mice. In order to examine the possibility that there is a defect in some component of the neuroepithelial cell cytoskeleton in SELH embryos, frozen sections of 7-somite pair SELH and ICR/Be embryos were stained with a fluorescent probe to filamentous actin. While actin itself is not necessarily expected to be abnormal in SELH embryos, its distribution pattern could be altered by mutations in the genes encoding any of the proteins with which it interacts within the cell. As the primary defect leading to lack of cranial neural fold elevation in SELH embryos was not obvious, it seemed logical to map the genes that cause exencephaly in SELH mice, in the hope that this would eventually lead to the identification of the genes themselves. The development of SSLP markers promised to increase the density of polymorphic genetic markers to a level that would provide enough densely spaced informative markers to test for linkage between them and QTLs across most or all of the mouse genome. This suggested that it would 88 be possible to map the loci that cause quantitative traits such as exencephaly in SELH mice. Restriction fragment length polymorphisms, electrophoretic (enzyme) variants, and the mouse coat colour loci would provide additional markers, if necessary. It therefore seemed logical to begin to map the genes that cause exenecephaly in SELH mice, in the hope that this would eventually lead to the identification of the genes themselves. This seemed a more appropriate method of determining the cause of exencephaly than testing a seemingly endless number of potential candidate genes and/or molecules using methods similar to those used to test for cytoskeletal defects (see preceding paragraph). In addition, the use of comparative maps (based on linkage homology) of the mouse and human genomes would suggest candidate regions in the human genome that could be investigated for linkage to multifactorial NTD. In order to map the exencephaly-liability loci, a cross was made between SELH/Bc mice and the unrelated LM/Bc strain, which has a normal pattern of cranial neural tube closure. A number of analytical techniques were used to estimate the number of exencephaly-causing genes that differ between these strains and their most likely mode of inheritance. Identification of the most likely map locations of these exencephaly-liability loci involved the development of relatively simple methods for the analysis of complex traits. Many of the already existing methods are mathematically complex and make assumptions that are not valid for nonparametric, low frequency, lethal threshold traits (like exencephaly in SELH/Bc mice). In addition, they generally require custom-written software and are not easily adapted to investigate complex genetic models (i.e., where the trait is not caused by many loci with small, equal and additive effects). Traits determined by a small number of QTLs, such as exencephaly in SELH/Bc mice, are relatively simple and should not require such complicated methods of analysis. The work presented here also determined whether the abnormal mechanism of cranial neural tube closure in SELH/Bc embryos is the cause of exencephaly by looking for an 89 association between lack of Closure 2 and incidence of exencephaly in the testcross progeny of mice segregating for the SELH/Bc exencephaly-liability loci. Demonstration of a causal relationship would show that this mechanism of closure is not a polymorphic variant coincidentally present in SELH/Bc mice, but is indeed an abnormal event that results in a high liability to exencephaly. The cleft cerebellum-ataxia observed in adult SELH/Bc mice appears to be developmentally related to exencephaly as both traits seem to be the result of the abnormal mechanism of cranial neural tube closure in SELH/Bc embryos. This question was addressed by looking for a genetic correlation between production of exencephaly and embryos with cleft cerebella in the testcross progeny of mice segregating for the SELH/Bc exencephaly-liability loci. Demonstration of a correlation would indicate that at least some of the same genes that cause exencephaly also cause cleft cerebellum-ataxia. The pattern of neural tube closure in the prospective cerebellar region of testcross embryos from mice segregating for the exencephaly-liability genes was also examined. Thus, there are three main objectives to the work presented in this thesis. The first is to identify the primary cause of exencephaly, using histological (Chapter 3) and genetic (Chapters 4 and 5) approaches. The second is to address the question of whether lack of Closure 2 is the cause of exencephaly (Chapter 6), and the third is to determine whether this abnormal mechanism of cranial neural tube closure in SELH embryos is also responsible for the cleft cerebellum-ataxia trait observed in 5-10% of adult SELH mice (Chapter 7). This work will advance our understanding of the development of exencephaly and cleft cerebellum-ataxia in SELH mice and should also contribute to the general understanding of the mechanisms by which the mammalian neural tube closes and NTD arise. Chapter 2: GENERAL MATERIALS AND METHODS 90 This chapter describes the Materials and Methods that are common to more than one of the studies described in the subsequent chapters. Animal Maintenance and Breeding All mice were maintained in the animal unit in the Department of Medical Genetics at the University of British Columbia in windowless rooms on a 12 hour light (6:00 am to 6:00 pm), 12 hour dark cycle. The temperature was controlled at about 22°C (20-24°C). The mice were housed in standard polycarbonate cages with dried corncob bedding and supplied Purina Laboratory Rodent Diet (#5001) and acidified water (pH 3.1, HC1) ad libitum. Timed matings were obtained by placing one to three nulliparous females in cages with single males at approximately 3:00 pm. Females were checked for vaginal plugs by 10:30 am each morning. Females were either left with sires continuously until a plug was found (to produce testcross litters), or separated from the males if unmated after 3-6 days and replaced after a break of several days (to produce litters for the study of stage of cranial neural tube closure on days 8-9 of gestation and for histology). Females were approximately 2 to 4 months old when bred. As ovulation generally occurs at the mid-point of the dark cycle (Bronson et al., 1966), 10 a.m. of the day a plug was observed was designated as day 0 hour 10 (DO/lOh) of gestation. Mouse stocks The SELH/Bc (SELH) mouse stock has been developed by Diana Juriloff and Muriel Harris, and its history is described in Juriloff et al. (1989) and has been discussed in the 91 Introduction of this thesis (pages 54-56). In 1989, it was shown that all SELH mice produce exencephaly in their progeny (Macdonald et al., 1989), and the exencephaly liability trait is fixed in the SELH stock. SELH/Bc females were used in the histological study and to begin the SELHR/Bc colony (see below). In addition, two sublines of SELH/Bc (SELH) mice were used in this study. SELHA/Bc (SELHA) is an inbred strain. From a point at which the SELH colony was calculated to have reached about 90% homozygosity, a subline was developed with 10 subsequent consecutive generations of brother-sister inbreeding. Offspring of one breeding pair from this tenth inbred generation were used to start the SELHA subline in 1990, and brother-sister inbreeding was continued. All SELHA mice used in this study have at least 18 consecutive generations of brother-sister inbreeding after the SELH colony had reached about 90% homozygosity, and are expected to have reached at least 99% homozygosity. SELHR/Bc (SELHR) mice were generated to produce a large random-bred SELH stock to provide the many SELH females needed to collect testcross litters from SELHA, LM/Bc, and LM.SELHA F l and F2 sires, to determine the incidence of exencephaly in these crosses. This substock was begun in 1992 with 6 nonsib pairs of SELH dams and sires taken from families of the SELH colony that had been observed to produce exencephaly in the previous generation. None of the founders came from sublines carrying any. of the other (single locus) spontaneous mutations (Unger et al., 1983; Juriloff et al., 1992; Juriloff et al., 1994; D.M. Juriloff, personal communication) present in some SELH mice. The SELHR breeding colony was subsequently maintained by non-sib SELHR breeding pairs, except for one infusion of SELH females approximately one year after beginning the colony. This was deemed necessary when it was realized that more mice were needed to complete the testcross studies in a reasonable time. Five nonsib SELH females from the SELH breeding colony from parents that had been observed to 92 produce exencephaly were bred to SELHR sires. The female progeny of these new breeding pairs were only used as dams for testcrosses and no progeny from these new pairs were used to set up additional breeding pairs. A total of 2010 female and 2186 male SELHR mice were weaned, for a sex ratio of 1:1.1 (X2=7.38, significantly different from expected 1:1 Mendelian ratio at 0.01>p>0.005). The most likely explanation for this deviation from a 1:1 Mendelian segregation ratio is that more females offspring than male were exencephalic; it has previously been shown that approximately 70% of exencephalic SELH embryos are female (Juriloff et al., 1989). A deficiency of females would therefore be expected. SELHR breeding pairs were set up in a random fashion, using a chart of random numbers to select sires and dams (by pedigree number). Sib pairs were, however, avoided. When not used for breeding, SELHR males were retired soon after weaning, first being observed for ataxia and morphological abnormalities. As some SELH mice with split cerebella do not appear ataxic until they are a month or more in age, the incidence of ataxia in SELHR mice, at 3.6% (152/4196; 99/2010 females (4.9%) and 53/2186 males (2.4%)), is likely to be underestimated. Of the 2010 adult SELHR female mice observed, 2 were seen with a small midline fur tuft on their foreheads and 2 had odd "tufts" at the dorsal base of their tails (both traits similar to the "dorsal excrescences" described by Center, 1960). One might speculate that these two traits may reflect some form of neural tube closure defect. One SELHR male had malocclusion, where the teeth of the upper jaw overgrew those of the lower jaw because they were not correctly lined up. SELHR females were bred at 6-10 weeks of age; ataxic mice were not bred. LM/Bc is a highly inbred strain (the animals used in this study were at F61 or higher) that has been used in other developmental studies (Harris et al., 1984; Harris and Juriloff, 1986; Finnell et al., 1986) and has normal neural tube closure, including initiation of cranial neural tube fusion at the site of Closure 2 (Juriloff et al., 1991; Golden and Chernoff, 1983, 1993) and 93 virtually no spontaneous production of exencephaly. These mice are homozygous for /g^, a mutation that causes open eyes at birth. LM/Bc mice were derived by brother-sister inbreeding of mice that carried 1/8 of their genes from the SWV strain and 7/8 of their genes from an inbred strain that had been derived from an unpedigreed stock of "C3H" mice. ICR/Be is a highly inbred strain developed from the (BLU:Ha)ICR mouse stock (Juriloff et al,, 1989). ICR/Be mice have been used in a number of studies of development (Harris and McLeod, 1982; Juriloff et al., 1987; Juriloff et al., 1989) and show no unusual incidence of spontaneous exencephaly or other malformations, despite their relation to the SELH/Bc mouse stock through their common "ancestor", the (BLU:Ha)ICR stock. Statistical Methods The Chi-square (X2) test of goodness of fit is used to test the fit of observed frequencies ("O") to those expected ("E") under the hypothesis being tested, where X 2= (0-E)2/E (Sokal and Rohlf, 1981). The probability of obtaining a given X 2 value can be obtained using tables of the X 2 distribution (Rohlf and Sokal, 1981) or computer programmes such as "CDC Saber". Fisher's exact test (Sokal and Rohlf, 1981) gives the probability of obtaining a result under the assumptions that the row and column classes of a 2x2 table are independent and their totals are fixed. The probabilities of obtaining the observed cell frequencies (a, b, c, and d) and all cell frequencies representing a greater deviation from expectation are calculated using the formula: P= (a+b)!(c+d)!(a+c)!(b+d)!/a!b!c!d!n!. These probabilities are summed to give the probability of obtaining the observed result as well as less likely outcomes. An alternative test of independence in a 2x2 table is the X 2 test of independence, where: X 2= (ad-bc)2n/(a+b)(c+d)(a+c)(b+d). The probability of the resulting X 2 value can be obtained using tables of the X 2 distribution (Rohlf and Sokal, 1981) or computer programmes such as 94 "CDC Saber", and gives the probability of two events occurring together if they are independent. Spearman's correlation of rank correlation (Sokal and Rohlf, 1981) is computed directly from the differences between the ranks, R, and R 2 of paired variables using the formula: r = l - n(nM). It is a non-parametric test for association, as is necessary when the data are not bivariate normally distributed. It gives greater weight to pairs of ranks that are further apart and so is most appropriate when there is less certainty about the reliability of close ranks. Exencephaly is a binomial trait and has a truncated distribution (at 0% exencephaly) when the mean is near zero. The variance of some generations can therefore be dependent on the mean. The Freeman-Tukey arcsine transformation is used to stabilize the variance of binomial distributions when it might otherwise depend strongly on a parameter (i.e., the mean). It moves the position of the maximum variance toward the extremes of the distribution, making the variance more uniform over a broad range of values and converting the distribution into an approximately normal one (Mosteller and Youtz, 1961). However, the effect of extreme outlying values is sometimes too great for the variance to be stabilized. Mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly produced by individual animals were calculated using the tables in Mosteller and Youtz (1961). 95 Chapter 3: HISTOLOGICAL STUDY OF THE CRANIAL NEURAL FOLDS OF SELH/Bc MICE Introduction The first histological study of the neural folds of SELH embryos (Macdonald et al., 1989) compared 14 SELH and 16 ICR/Be embryos with 8-13 somites, embedded in paraffin-and sectioned at 7um, and 4 SELH and 4 ICR/Be embryos with 6-8 somites, embedded in plastic and sectioned at 2um. While the thick (7 u.m) sections allow observation of the overall morphology of the neural folds, identification of the cell types present, and observation of the amount of cell death and cell division, cell shape cannot be accurately distinguished due to cell overlap. In all embryos, the folds of the rostral mesencephalon of SELH embryos were widely flared and the lateral mesenchyme, ventral to the neuroepithelium/surface ectoderm cell boundary, was abnormally collapsed. The cellular basis of these morphological abnormalities could not be detected: the total area of neuroepithelium was similar in SELH and ICR/Be embryos, as were cell shape and density. In the 2 urn sections, more pyknotic cells were apparent in the neuroepithelium and mesenchyme of SELH embryos than in ICR, but the amount of cell death was not exceptional. In these sections, SELH embryos had more small unstained inclusions in their neuroepithelial cells than did ICR/Be embryos. The significance, if any, of these observations relative to failure of neural tube closure is not evident, and it is possible that these differences are not involved in failure of neural fold elevation in the mesencephalon of SELH embryos. In an attempt to elucidate further the mechanism leading to lack of mesencephalic neural fold elevation in SELH embryos at the cellular level, a second histological study of the aberrant cranial neural tube closure of SELH mice compared to ICR/Be and, to a lesser extent, LM/Bc, was undertaken. LM/Bc embryos were included, because this unrelated strain with normal neural tube closure was being crossed to SELHA mice for the genetic and mapping studies of exencephaly-liability. Fewer LM/Bc embryos were sectioned, however, and their angle of section was not always comparable to that of SELH 96 and ICR/Be. A total of 35 SELH, 25 ICR/Be, and 17 LM/Bc embryos covering three stages of development (3-5 somite pairs, 6-8 somite pairs, and 9-11 somite pairs) were embedded in methacrylate (plastic) and sectioned at 3um. Thus, this study looked at more embryos than the previous one, and also examined embryos at earlier stages of development, including the 3-5 somite stage when neural fold elevation is just beginning. As the exencephaly observed in the SELH/Bc mouse stock is a result of failure of elevation of the neural folds, the possible involvement of the cytoskeleton was explored. As discussed in Chapter 1, one of the possible driving forces for neural fold elevation is cell wedging, caused by the contraction of apically arranged actin microfilaments within the neuroepithelium. This would require that the actin microfilaments be anchored to the cell membrane, and a number of cytoskeletal proteins such as ankyrin, fodrin (spectrin), and band 3 and band 4.1 proteins, are involved in process (Darnell et al., 1986). Thus, a defect in one component of the cytoskeleton might alter the organization and/or location of actin microfilaments in neuroepithelial cells or the effect of their contraction on neuroepithelial cell shape. The location of actin microfilaments within the neuroepithelium, as indicated by a fluorescently labelled phallotoxin probe for filamentous actin, was therefore used as a "marker" system to determine whether SELH mice carry a defect in the cytoskeleton of neuroepithelial cells at the prosencephalon/ mesencephalon boundary that is evident during the time of normal neural fold elevation. This study has been published in Teratology (Gunn et al., 1993); the presentation of this work is based on that manuscript which I wrote under the guidance and with the help of Diana Juriloff and Muriel Harris. I collected and prepared all embryos for histology, and I did all but a small fraction of the methacrylate sectioning myself. I also did all of the frozen histology and fluorescent staining. Diana Juriloff, Muriel Harris and I shared in the interpretation of the sections for differences between SELH, ICR/Be, and LM/Bc. 97 Materials and Methods Mouse maintenance and breeding All mice were maintained and timed pregnancies were obtained as described in Chapter 2 (General Materials and Methods). Females with plugs were killed at various times on D8 of gestation in order to obtain litters with the desired numbers of somite pairs. Mouse stocks Sixty-four percent of the SELH embryos used for histology in this study were from SELHA sires mated to SELHA dams; the rest, distributed among the somite classes, were from SELH/Bc dams mated to SELHA/Bc sires. These strains have been described in Chapter 2. Approximately 20% of SELH X SELHA F, embryos are exencephalic (Chapter 4, Table 3). All embryos used for studies of actin microfilaments were from SELHA/Bc dams and sires. ICR/Be and LM/Bc are inbred strains with normal neural tube closure (Juriloff et al., 1991); they are described in Chapter 2. ICR/Be mice share a common genetic origin with SELH mice, whereas LM/Bc mice are not related to SELH or ICR/Be mice. Histology A histological study of the cranial neural folds of SELH embryos during the time of normal neural tube closure was done. As the previous major analysis of the genetic cause of exencephaly in SELH (Juriloff et al., 1989) was based on a cross to ICR/Be mice, ICR/Be is used in this study for intensive histological comparison with SELH. Because a cross of SELH to the unrelated normal strain, LM/Bc, was to be the basis of a new study to map the SELH exencephaly-liability loci, some LM/Bc embryos were also included in these histological preparations. They are included for interest, but they are not well matched for angle of section 98 at all stages, and therefore are not the focus of comparison with SELH. Litters were collected by removing the uterus and placing it in 0.85% NaCI. The intact deciduae were dissected out and fixed in Bouin's fixative for at least 24 hours before being transferred to 70% ethanol. The embryos were then carefully removed from the deciduae under 70% ethanol and scored for somite pairs. Embryos with 3-5, 6-8, and 9-11 somite pairs were dehydrated in 80% then 90% ethanol and transferred to infiltrating solution (methacrylate catalyzed solution A, Polysciences Inc., Warrington PA) for at least 3 hours. Embryos were embedded in methacrylate resin (JB-4 embedding kit, Polysciences Inc.), oriented so that transverse sections would be taken perpendicular to the region of Closure 2. Sections of 3 microns in thickness were made on a Sorval JB-4A microtome and stained with 1.0% toluidine blue in 1.0% sodium borate. At the 3-5 somite stage, 11 SELH embryos (from 5 litters), 11 ICR/Be embryos (from 6 litters), and 3 LM/Bc embryos (from 2 litters) were examined. At the 6-8 somite stage, there were 13 SELH embryos (from 7 litters), 7 ICR/Be embryos (from 5 litters), and 12 LM/Bc embryos (from 4 litters). At the 9-11 somite stage, 11 SELH embryos (from 8 litters), 7 ICR/Be embryos (from 4 litters), and 2 LM/Bc embryos (from 2 litters) were examined. The level and angle of sectioning of embryos in the figures are represented by sketches, after Kaufman (1990). Fluorescent staining of actin microfilaments Frozen sections of 7 somite SELH and ICR/Be embryos were stained with fluorescently labelled phalloidin to visualize the location of actin microfilaments in the neuroepithelium of the cranial neural tube, particularly in the region of Closure 2. Embryos were obtained by removing the uterus to 2 mM phosphate buffered saline (PBS), where the embryos were carefully dissected 99 from the surrounding membranes and scored for somite count. Embryos with 7 somite pairs were transferred to 3.7% paraformaldehyde (in 2mM PBS) for 30 minutes, then stored at 4°C in 25% sucrose (in 2 mM PBS) for up to 3 months. As the small size of D8 embryos makes them difficult to handle and orient for frozen sectioning, the embryos were inserted into slits in small pieces of liver tissue (also removed from the pregnant females during dissection) just prior to freezing. The fixation and storage of the liver pieces was the same as for the embryos. The tissue was placed in a small rectangular "boat" fashioned out of aluminium foil, then covered with Tissue-Tek Optimum Cutting Temperature (O.C.T.) Compound and immersed into a metal cup containing isopentane. This cup was then immersed into liquid nitrogen until the OCT compound had frozen (i.e., it had solidified and become opaque). The "boat" was then removed to an American Optical Model 830-C Cryo-cut Cryostat Microtome at -18°C, where the OCT compound-embedded tissue was removed from it and allowed to come to the temperature of the cryostat. The sample was mounted on a chuck, using OCT, then cut in 8urn sections and mounted serially onto 5 different poly-l-lysine coated slides. The slides were kept at least overnight, and up to 2 weeks, in a -20°C freezer before staining as this "aging" seemed to prevent the sections from floating off the slides during staining. Slides were brought to room temperature, then rehydrated for ten minutes in 2 mM PBS before staining. Each series of slides was stained as follows: one with haematoxylin and eosin; one with either 1.65 x 10"6 M fluorescein-phalloidin (Molecular Probes Inc.) or 1.65 x 10"6 M FT.TC-phalloidin (Sigma) as a fluorescent probe for filamentous actin; one with 1.65 x 10"6 M fluorescein-phalloidin (or FITC-phalloidin) and 1.039 x 10"4 M phallacidin (competitive specificity control; Sigma); one with 1.04 x 10"4 M phallacidin (control for phalloidin treatment); and one with 2 mM PBS (control for autofluorescence of the tissue). Slides were incubated at 100 room temperature for 30 minutes with the treatment solutions, rinsed three times for ten minutes each with PBS, and mounted with 50% glycerol/50% PBS/0.02% sodium azide. Each step of the staining procedure was carried out in a covered pan, to keep the light out so that fluorescence of the stain was optimal when the slides were examined. Some sets of slides were also stained with 1.65 x 10"6 M NBD-phallacidin (Molecular Probes Inc.) in place of the fluorescein-phalloidin and with 1.04 x 10"4 M phalloidin (Sigma) in place of the phallacidin. The results were similar to those obtained with fluorescein-phalloidin and are not shown. The stained sections were viewed on a Leitz Diaplan microscope equipped with a Leitz 13 filter block and photographed using Kodak TMax-400 film taken at 1600 ASA. Results Histology In transverse sections of 3-5 somite embryos (Fig. 4), SELH differs from ICR/Be in the shape of the cranial neural folds. In 11/11 SELH embryos with 3-5 somites, the lateral tips of the folds are more elongated (Fig. 4a-d) and the lateral mesenchyme is contained in a longer but narrower space (Fig. 4a-d) than in 11/11 ICR/Be embryos examined at this stage (Fig. 4e-h; see also Fig. 5). In 10/11 of these SELH embryos, the lateral aspects of the neuroepithelium also appear to curve downward, giving the tips of the folds an abnormally "hooked" shape. This hooking is most apparent in the prospective mesencephalon (Fig. 4b) and the prospective posterior prosencephalon (Fig. 4c). Although there is some hooking apparent in ICR/Be embryos (Fig. 4g), it is much less obvious and while it is seen more rostrally (Fig. 4g), it does not extend into the prospective mesencephalon (compare Fig. 4b and 4f). The consistent difference in position of the lateral tips of the neural folds of SELH and ICR/Be is demonstrated in tracings 101 Fig. 5. a-d: Overlaid tracings of profiles of the 3-5 somite S E L H and ICR/Be sections shown in Fig. 4. The solid line with dark shading of the neuroepithelium and surface ectoderm represents S E L H embryos (from Fig. 4a-d). The dashed line with pale shading represents ICR/Be embryos (from Fig. 4e-h). Note the difference between the lateral tips of S E L H and ICR/Be neural folds (arrow), and the slightly more convex profile of ICR/Be neuroepithelium. 103 of the profiles of the sections (Fig. 5). The folds of all 3-5 somite SELH embryos also appear more flattened than those of ICR/Be (compare Fig. 4a with 4e; 4b with 4f; and see Fig. 5a and 5b). In transverse section, the shape of the neuroepithelium normally is a smooth convex curve, but in SELH, there is a "dogleg" bend as the folds diverge from the medial region (compare Fig. 4b and 4f; and see Fig. 5b). Fewer 3-5 somite LM/Bc embryos were examined and the angle of section is not well matched with SELH and ICR/Be. However, the folds in LM/Bc are neither obviously hooked at their lateral tips nor angular or flattened as in SELH (compare Fig. 4j with 4b). At the 6-8 somite stage (Fig. 6), in transverse sections the lateral neuroepithelium of SELH neural folds (Fig. 6a-d) appears to be elongated in comparison to ICR/Be (Fig. 6e-h; also Fig. 7a-d) and LM/Bc (Fig. 6i-l; also Fig. 7e-h) embryos. There is also very little elevation of the lateral edges in the 13 SELH embryos examined, and the lateral aspects of the folds often are curved downward rather than elevating as in the 7 ICR/Be embryos observed (compare Fig. 6b with 6f; and see Fig. 7b). These differences are less obvious between SELH and LM/Bc, although there is an appreciable difference in the degree of elevation of the lateral tips in the prosencephalon (compare Fig. 6c and 6k, Fig. 6d and 61; and see Fig. 7g-h). At the 9-11 somite stage (Fig. 8) the cranial folds of SELH continue to differ from those of ICR/Be. In 9/11 SELH embryos at this stage, there is a sharp bend in the neuroepithelium lateral to the neural groove (most obvious in Fig. 8b,d), beyond which the folds appear splayed horizontally (Fig. 8a-d). In the prosencephalon, the folds lateral to this bend are particularly flat (Fig. 8c,d). In contrast, the folds of 7/7 ICR/Be embryos sectioned are more upright and show no horizontal splaying (Fig. 8e-h). In 11/11 SELH embryos examined, the surface ectoderm just below the lateral tips appears abnormally indented (Fig. 8a,c,d and Fig. 9); there is some indentation of the lateral surface ectoderm in ICR/Be (Fig. 8d), but it is much less pronounced 104 105 Fig. 7. Overlaid tracings of profiles of the 6-8 somite SELH, ICR/Be, and LM/Bc sections shown in Figure 6. a-d: SELH (from Fig. 6a-d) and ICR/Be (from Fig. 6e-h). The solid line with dark shading of the neuroepithelium represents SELH; the dashed line with light shading of the neuroepithelium represents ICR/Be. Note the lack of elevation of SELH lateral tips (arrow), e-h: overlaid profiles of SELH (from Fig. 6a-d) and LM/Bc (from Fig. 6i-l). The solid line with dark shading of the neuroepithelium represents SELH; the dashed line with light shading represents LM/Bc. 106 co CO CD •H o °> c m „ (D CQ CO ^ CD > O o 1 * O) CX) In =5 c ' CO 03 .SZ TO 0 5 ™ J= * " CD 03 £ ® JZ •= c 5 co — c o 2 ° « 03 03 > • 2 CO O 03 t CO t 03 .> CO ~ o =5 ^ 3 g * £ | 03 "O S CO CJ c c aj CO 03 <2 _ » T3 0) CO i- it: CO 3 co M _ _ „ i- o ^ O CO 03 c cn CD co c -o c co c = ® ® -1 0 o ™ o to co CO — c o CJ3 Q . JO F i s < 0 0 P -g - -« CO 03 £ 107 than in SELH. The position at which the surface ectoderm meets the lateral tip of the neuroepithelium is different in 11/11 SELH embryos (Fig. 9, arrowhead), being more ventral or more lateral than in the 7 ICR/Be embryos. As the angle of section of the 2 LM/Bc embryos sectioned at this stage (Fig. 8i-l) is very different to that of SELH and ICR/Be, the shape of the folds generally cannot be compared. The 2 LM/Bc embryos observed appear to be less advanced in neural fold elevation than the ICR/Be embryos examined. In sections d, h and 1 (Fig. 8), where the angle of section of all three strains is similar, it can be seen that there is an indentation of the surface ectoderm below the lateral tips of the SELH folds that is much less pronounced in ICR/Be and LM/Bc. Fluorescent staining of actin microfilaments In all phalloidin-stained frozen sections of various angles of sectioning, there are no detectable differences in staining between the 24 SELH embryos and the 17 ICR/Be controls observed. Embryos of both strains show a distinct band of staining along the apices of the neuroepithelial cells at all levels of sectioning (Figs. 10 and 11). At a gross level, neuroepithelial cells of SELH embryos are normal in shape and size. In some of both SELH and ICR/Be embryos, condensations of fluorescently stained actin could be seen at the junctions between neighbouring neuroepithelial cells (Figs. 10 and 11). The competitive specificity controls showed faint or no fluorescence, while the controls for phalloidin treatment and autofluorescence of the tissue showed no fluorescent staining in the embryos. 108 LU t co o • o 0 CD CO 3 O CO CD 0 * E | 0 E CO 3 ^ 0) 1 x: C L CD 0 0 - 2 ° 3 © c > 0 M £ 0 o to c £ o C L -ti 0 co O co C L e © X l — S j= 0 2 i l l ° . x 2 CO LU £ g,W co LL .£ 0 O C 0 - ° 5 co . c TJ XI CO CD Q. - „. 0 P ?5 r- ,CD 0 £ T= co c 3 0 0 CO E O T3 L L - 05 CD i i - c £ . c u_ O CO O CO ^ *- • 0 1 O C O T3 Q . CD c x : P 2 CD CD co - i . _ l . CT O CO E o U_ 0 X l 109 110 x i a wit rofi o CD ' f i 2 C stai stai u o LH -a J S CU £ C LH CO w fS CO u u CJ CO ' 0 ) E V -O o =3 CO c : r - <4-H ed O 4 - a c o ion ributi ributi OJ -*—1 '-3 o , , hal ica OH ex ed 6 _>> <D C3 OJ S a aj E J S do J S QC u cx O OJ i — J S J S c to ctio No E ,2 OJ ithel uO <D OO lidin. ithel 1 - lidin. ex _u CU > 15 o to U q J S i — ex s f— _ts he I ' 5 he I esc ithin t oi uor ithin t 53 I l l Discussion A previous, limited histological study (Macdonald et al., 1989) of mostly thick (7um) sections of 6-13 somite SELH embryos showed widely flared cranial neural folds and collapsed lateral mesenchyme, but did not lead to an explanation of the failure of Closure 2 in SELH mice. The aim of the present study was to extend the preliminary observations by examining thin (3um) sections in more embryos, at three developmental stages, including earlier stages than had previously been examined. Embryos of the ICR/Be strain, the strain most closely related to SELH, provide appropriate normal controls. Some LM/Bc embryos were also examined because this strain is the normal strain used in a genetic study of exencephaly. In addition, as the site of Closure 2 varies among normal strains (Juriloff et al., 1991), the use of ICR/Be and LM/Bc provides the opportunity to look for morphological differences in SELH that are consistently unlike two normal strains. However, comparison to LM/Bc is limited as sections comparable to those of ICR/Be and SELH were not obtained at all somite stages. In normal 3-5 somite embryos, the cranial neural folds form symmetrical biconvex bulges (Sadler et al., 1986). During this time, there is growth and lateral expansion of the folds as the rostral and rostrolateral edges of the prosencephalon begin elevation, and they gradually lose their convexity by the 5-6 somite stage (Kaufman, 1979; Jacobson and Tam, 1982; Sadler et al., 1986). By the 3-5 somite stage, there are already several differences between SELH and normal strains. In SELH embryos, the normally smooth convex arc of the neuroepithelium is changed into three straighter segments with two clear inflexion points, one forming a dorsal edge of the neural groove and one near the lateral tip. This gives the folds an appearance of being flatter with a hook at the tip. In 6-8 somite normal embryos, the neural folds appear flattened (planar fold in Fig. 5 of Morris-Wiman and Brinkley, 1990a) as they lose their convexity and continue elevation (Morriss 112 and Solursh, 1978a; Jacobson and Tam, 1982). The cranial neural folds of 6-8 somite SELH embryos continue to differ in shape from those of ICR/Be and LM/Bc embryos. The lateral tips of SELH neural folds appear abnormally elongated and are often curved downward instead of beginning elevation, particularly in sections through the prosencephalon. At the 6-8 somite stage, it is interesting to note that the boundary between the neuroepithelium and surface ectoderm and the mesoderm at the lateral tips of SELH neural folds is unclear, in contrast to ICR/Be and LM/Bc (see Fig. 6a,b and Fig. 10a,b). The boundary is difficult to define in SELH, and the transition zone is disorderly; this raises the possibility that there may be a difference in the timing or characteristics of neural crest cell emigration from the lateral tips of SELH cranial neural folds, compared to normal embryos. In normal 9-11 somite embryos, the folds of the prosencephalon continue elevation, prior to initiation of Closure 2 (Morriss and Solursh, 1978a; Kaufman, 1979; Jacobson and Tam, 1982), but in SELH embryos, the cranial neural folds are splayed horizontally. The lateral tips of the SELH cranial neural folds also continue to appear abnormal. The surface ectoderm is indented adjacent to the neuroepithelial/surface ectoderm boundary. In addition, the neuroepithelial/surface ectoderm boundary in SELH embryos is displaced ventro-laterally compared with normal embryos. In summary, at all stages of embryos studied (3-11 somites), SELH cranial neural folds have an abnormal morphology which is apparent not only in the region of Closure 2, but also in the most rostral sections examined. At least two hypotheses can explain the abnormalities. The first hypothesis involves factors extrinsic to the neuroepithelium. Work in avian embryos suggests that forces generated by lateral, non-neuroepithelial tissues are required for elevation of the neural folds in some regions (Schoenwolf, 1988; Smith and Schoenwolf, 1991). Parasagittal cuts at the junction between the prospective neural plate and the prospective surface 113 ectoderm inhibit elevation and convergence of the neural folds in the midbrain and cranial hindbrain but not in the forebrain (Schoenwolf, 1988). This pattern is like that of SELH mice, where closure of the forebrain is normal but the folds of the midbrain fail to elevate. The morphology of SELH neural folds is abnormal at the lateral tips, which could reflect a defect in lateral forces needed for elevation of the midbrain folds. A second hypothesis for the abnormal morphology of SELH cranial neural folds involves factor(s) intrinsic to the neuroepithelium, and a possible defect in its cytoskeleton. The shape of SELH neural folds suggests that in cross-section, the neuroepithelium is longer than normal and extends beyond its normal location at the tips (Figs. 4-9). If the structure and location of the surface ectoderm is fixed, then a lengthened neuroepithelium would need to bend to meet it, resulting in the overhang or hook-like shape observed at the lateral tips of the cranial neural folds of 3-5 somite SELH embryos. A lengthened neuroepithelium could result in the more ventro-lateral position of the neuroepithelial/surface ectoderm boundary seen in the neural folds of 9-11 somite SELH embryos. A lengthened neuroepithelium could also cause the indentation of the surface ectoderm observed at this stage; extension of the neuroepithelium beyond its normal location at the lateral ends could cause the adjacent surface ectoderm to buckle. The normal elevation of the neural folds may require the contraction of apically-arranged actin microfilaments within the neuroepithelium (Sadler et al., 1982). A failure of contraction could lead to elongation of the neuroepithelium as observed in SELH cranial neural folds. Therefore the present study examined the possibility that a cytoskeletal abnormality, reflected in actin distribution, is involved in the failure of Closure 2 in SELH mice. The distribution patterns of actin in the cranial folds of normal mouse embryos have been described by Sadler et al. (1982). In 2 somite embryos, actin is localized along the basal aspect of the neuroepithelium (Sadler et al., 1982). By the 4-5 somite stage, actin is localized apically and 114 no longer basally (Sadler et al., 1982). Distribution patterns of fodrin (non-erythrocyte spectrin), an actin binding protein, are similar (Sadler et al., 1986). The actin and fodrin patterns suggest that the contraction of basal actin microfilaments in the neuroepithelium plays a role in the maintenance of the biconvex shape of early neural folds, while the contraction of apical actin microfilaments is important in subsequent elevation of the folds. The present study shows that filamentous actin is concentrated in the apices of neuroepithelial cells in SELH embryos during the middle stage of cranial neurulation in the region of Closure 2, as found in embryos with normal neurulation. Some condensations of staining were also seen between neighbouring cells, consistent with the localization of actin microfilaments to adherens junctions (belt desmosomes) that are involved in cell-cell adhesion and possibly intercellular communication (Darnell et al., 1986). The neuroepithelial cells are also grossly normal in shape and size. Because actin distribution appears to be normal, we conclude that the cytoskeleton as a whole is grossly normal. However, other approaches would be required to test for a more subtle defect within the cytoskeleton. In conclusion, the defect in SELH originates very early in cranial neural fold development. The morphology of SELH midbrain folds is obviously abnormal by the 3-5 somite stage, well before the time at which normal neural tube closure would occur in this region. At all stages of cranial neurulation observed, there is a consistent abnormality in the shape of the neuroepithelium. 115 Chapter 4: GENETIC ANALYSIS OF THE CAUSE OF EXENCEPHALY IN SELHA/Bc MICE Introduction The previous major genetic analysis of the exencephaly-liability in SELH mice (Juriloff et al. 1989) involved a cross to the most closely related strain available, ICR/Be. This was done purposefully, as nothing was previously known about the number of loci involved. In the event that there were many, it was thought that a cross to a related but normal strain would minimize this difference and give the number of major loci involved, rather than all genes including any modifier loci for which SELH and ICR/Be could share common alleles. The genetic model that best fit the observed data was multigenic inheritance of two to three additive loci with a developmental threshold. A cross of SELHA mice to SWV/Bc, a normal and unrelated strain, was made in order to examine whether maternal genotype influences the incidence of exencephaly in progeny (Gunn et al., 1992). No evidence of genetic maternal effects were found in this cross. The reciprocal F l and BC1 crosses made in this study also provided the opportunity to test the conclusions of the previous genetic analysis. These data also supported an additive threshold model of inheritance of the exencephaly-causing genes, but a segregating generation was not testcrossed and so it was not possible to rule out epistatic, multiple recessive locus models. In the study described here, SELHA mice were crossed to LM/Bc, a strain with a normal pattern of neural tube closure (including presence of the most common initiation site at Closure 2 at the prosencephalon/mesencephalon boundary; Juriloff et al., 1991; Golden and Chernoff, 1993) and no historical relationship to SELH mice. The purpose of this cross was to uncover all the major genes in SELH mice that contribute to exencephaly-liability, including any that SELH and ICR/Be share in common and were therefore not identified in the previous study. 116 This cross (to LM/Bc) was the foundation for the study to identify the map positions of the exencephaly-causing genes in SELH(A) mice (see Chapter 5), and it is helpful to know the mode of inheritance and number of causative loci segregating in this cross. For the mapping study, the unrelated LM/Bc strain was appropriate to be crossed to SELHA mice instead of the related ICR/Be strain because more loci are likely to differ between SELHA and LM/Bc if the cause of exencephaly is polygenic, and there is a greater likelihood of genetic markers being informative between unrelated strains. Materials and Methods Animal Maintenance and Breeding All mice were maintained and timed pregnancies were obtained as described in Chapter 2 {General Materials and Methods). Mouse stocks The SELHA/Bc, SELHR/Bc and LM/Bc strains of mice have all been described in Chapter 2 (General Materials and Methods). Experimental design Breeding scheme: Two SELHA sires, sibs A5354 and A5356, were crossed to LM/Bc dams and resulting F l mice were intercrossed to produce F2 animals segregating for the exencephaly liability loci. Sire A5354 was only crossed to one LM/Bc dam and produced two litters before he became sick and was retired. Sire A5356 was mated to 4 LM/Bc dams and produced a total of 20 litters of F l animals. 117 F2 animals were generated from the three F l pairs from SELHA sire A5354 and 6 F l pairs from SELHA sire A5356. Sixty litters of F2 animals were produced, containing a total of 331 males and 322 females (not significantly different from Mendelian 1:1 segregation; X 2 , p>0.9). F2 females were raised to 6 weeks of age, at which time they were scored for ataxia and other gross morphological abnormalities, then discarded. F2 males were also raised to at least 6 weeks of age, generally longer, and were also observed for development of ataxia or other abnormalities. A further cross, between SELHA sire A6081 and 4 LM/Bc dams, was made in order to collect additional exencephalic D14 F2 embryos, the purpose of which will be described later (in the mapping study). Eleven F l litters were collected from SELHA sire A6081, and all surviving F l females (47) were mated to their male sibs to collect F2 litters around D14 (23/47; 7 D12, 6 D13, 9 D15 and 1 D17). Classical genetic crosses: In order to study the genetic difference between SELHA and LM/Bc in liability to exencephaly, several classical genetic crosses were made and embryos were collected and scored for exencephaly and other obvious external malformations. The incidence of exencephaly was generally observed in D14 litters (overall, 88% of litters were D14; see results) from timed pregnancies, but in some cases it was not feasible to check plugs every morning, or a plug was missed, and the litters were collected as close to D14 as possible, based on how long a female had been with a sire and on how "big" she was. Palpation of the females was not done as it was thought that this could possibly damage the embryos. As exencephaly can be observed any time after neural tube closure is complete, and as no fetal mortality (between D14 and D18) of exencephalic SELH embryos was seen in a previous study (Macdonald et al., 1989), the data 118 from litters collected between D10 and D18 were kept. In addition, most resorbed embryos die prior to D10 (and those that died later were noted), and it is generally possible to score dead embryos for exencephaly 1-2 days after death occurred. A total of 13 SELHA sires were testcrossed to SELHR dams, and 97-139 embryos (mean= 107) were collected for each sire. SELH litters of any size were included in the data as small litters (<5 embryos) were not rare and their inclusion did not significantly affect the incidence of exencephaly (X 2; p>0.9). Of the 150 SELHA x SELHR litters collected, 130 (86.7%) were collected on D14, 1 on D10, 3 on D l l , 6 on D12, 2 on D13, 3 on D15, and 5 on D16. Litters from SELHA sires x SELHR dams were collected throughout the time that F2 sires were being testcrossed in order to keep track of any potential fluctuations in the incidence of exencephaly during the 22 months during which F2 sires were testcrossed. Ten LM/Bc sires were testcrossed to SELHR dams to determine the incidence of exencephaly in the F l generation, with 102-126 embryos (mean= 109) examined per sire. All litter sizes were included as litters smaller than 5 embryos were not rare. Of the 121 litters of F l embryos collected, 98 (81.0%) were collected on D14 and 23 (1 D10, 1 D l l , 7 D12, 2 D13, 6 D15, 3 D16, 2 D17 and 1 D18) were not. Ten F l sires were testcrossed to SELHR dams and 102-132 embryos (mean= 111) per sire were examined to determine the frequency of exencephaly in the BC1 generation. Of the 93 litters examined, 83 (89.3%) were collected on D14 and 10 (7 D12, 1 D13 and 2 Dl5) of BC1 embryos (SELHR.F1) were not. The reciprocal backcross was also examined by crossing ten SELHR sires to F l dams, to check for the presence of a genetic maternal effect. Only 41-108 embryos (mean= 61) were collected per sire, from 48 litters of which 3 (1 D12, 1 D13 and 1 D15) were not collected on D14. In order to test whether the difference in exencephaly frequency between reciprocal backcrosses is significant while allowing for the large differences 119 in litter size (and therefore potential litter effects), the mean-litter Freeman-Tukey arcsine transformed (Mosteller and Youtz, 1961) values for each sire were calculated and used in a one way analysis of variance (F-test). As the larger sample sizes in the backcross of F l sires to SELHR dams makes these data more reliable than the reciprocal backcross, its exencephaly frequency was used for the BC1 generation in the genetic analysis. The same 10 F l sires crossed to SELHR dams were also crossed to F l dams to determine the incidence of exencephaly in the F2 generation, with 24-89 embryos (mean= 54) collected per sire. Of the 37 litters of F2 embryos examined, 32 (86.5%) were collected on D14 and 5 (1 D l l , 2 D12, 1 D13 and 1 D16) were not. Subsequently, more F2 embryos were collected from F l parents generated from a later cross of SELHA sire A6081 to 4 LM/Bc dams, made for the purpose of generating F2 embryos for the linkage study (Chapter 5). Twenty-four F l sires were bred to 47 F l dams. A total of 610 embryos were generated (11 -43 per sire, mean= 25.4) from 47 litters, 23 of which (7 D12, 6 D13, 9 D15 and 1 D17) were not collected on D14. The majority of exencephalic F2 embryos were frozen and kept at -20°C. Only the frequency from the first data set was used for the genetic analysis as these data came from the same F l sires that were backcrossed to SELHR and were collected during the same time period as the data for other generations. The incidence of exencephaly in LM/Bc embryos was found by crossing ten LM/Bc sires to LM/Bc dams and examining 101-111 embryos (mean= 106) per sire. Litters of all sizes were included as small litters were not rare. Of the 124 litters of LM/Bc embryos observed, 113 (91.1%) were collected on D14 and 11 (2 D l l , 2 D12, 3 D13, 3 D15 and 1 D16) were not. Of the 109 F2 males testcrossed to SELHR females, three would not breed, one produced two litters and would not continue breeding, and three bred well at the outset but then produced small or ho litters, likely due to old age. This is particularly likely with the sires that produced 120 good litters early on but small or no litters later. As with the parental and F l generations, litters of all sizes were included. Complete testcross data were collected for 102 F2 sires that produced 97-141 embryos (mean= 109). Of the 1,006 testcross litters collected from these 102 F2 sires, 889 (88.4%) were collected on D14, 3 on D10, 5 on D l l , 40 on D12, 39 on D13, 23 on D15, 6 on D16 and 1 on D19. Progeny testing of F2 sires: In order to analyze the number and mode of inheritance of the exencephaly-liability loci that differ between SELHA and LM/Bc mice and to map them, an F2 population segregating for the exencephaly liability loci was generated. As exencephaly is a qualitative trait caused by an underlying liability that is quantitative in nature, genotype cannot be inferred from phenotype (i.e., animals with a high genetic liability may be normal, while others with an intermediate genetic liability may be affected). The genotype of animals segregating for the genes that cause exencephaly is most accurately ascertained by determining the breeding values of individual animals (i.e., determining each animal's liability based on the mean frequency of exencephaly in its progeny). As exencephaly is a lethal trait and mice cannibalize their defective young, it is necessary to collect testcross progeny before birth in order to observe all the members of each litter. As this requires the mother to be killed, breeding values for this type of trait are collected for male animals. In addition, as one sire can be bred to several females at once, it is faster to collect testcross data from male animals. The goal was to collect 100 Dl4 embryos per sire from 100 F2 sires. Although the 95% confidence intervals around a sample size of 100 embryos does not eliminate overlap between the expected genotypic classes of sires, it is sufficient that sires with a high genetic risk to exencephaly will not be misclassified as sires with a low genetic risk, and vice versa. As the 121 approach to mapping the exencephaly-liability loci found in the SELH strain was to type only the sires that produced the 10% highest and 10% lowest frequencies of exencephaly (selective genotyping), it was necessary to testcross a sufficient number of sires that the extreme 10% would consist of sires with high or low genetic liabilities to exencephaly. As two to three genes are thought to be involved in exencephaly-liability, at most 1/16 or 1/64 of surviving F2 sires would be expected to carry all or none of their liability alleles from the SELHA strain. It is not feasible to testcross hundreds of sires, so 100 was selected as a goal that would result in the extreme 10% of sires having most or all of their exencephaly-liability alleles from the appropriate parental strain (i.e., SELHA alleles in high-risk sires and LM/Bc alleles in low-risk sires). Entire litters of F2 males were testcrossed as the animals of each litter should contain an assortment of genotypic classes; this minimizes the chance of inadvertent selection of exencephaly-liability genotypes. Twenty litters of F2 males were testcrossed, and the number of sires in each ranged from 1 to 9 (mean= 5.5). At least one litter was testcrossed from each of the F l breeding pairs. Six of the F2 litters were descended from SELHA sire A5354 while the rest descended from sire A5356. The average age of F2 sires at the outset of breeding was approximately 7.2 (range: 2.5-11) months, and it took an average of 2.1 months of breeding (range: 2 weeks to 6 months) to obtain 100 embryos per sire. Data analysis The frequencies of exencephaly expected in the testcross progeny of F l sires (BC1 embryos) and of F2 sires (pooled) for one autosomal recessive locus with 18.0% penetrance (the observed frequency of exencephaly in SELH embryos) and for 2-4 recessive loci with duplicate epistasis and 18.0% penetrance in the multiple recessive homozygote were calculated and compared to the observed frequencies. Duplicate epistasis is based on the concept that some 122 genes may be represented more than once in a genome (Griffiths et al., 1993; Juriloff, 1995). If duplicate copies have identical function, all copies would need to be mutant in order for an abnormal phenotype to be evident. X 2 tests were used to test the fit of the observed data to expected. The frequencies expected in the F l and F2 generations were also calculated and compared (X 2 tests of goodness of fit) to the observed values. The frequencies of exencephaly observed in the F l , F2, F2-testcross and BC1 generations were compared with the frequencies expected under an additive polygenic threshold model using a method adapted from the approach described by Falconer (1989) and described previously in Juriloff et al. (1989). Under a polygenic threshold model with additive inheritance, the mean of the F l generation on the underlying quantitative scale is expected to lie mid-way between the means of the parental populations, and the mean of the BC1 generation is expected to lie mid-way between the means of the F l and SELH. There should, therefore, be a linear relationship between the genotype (i.e., proportion of genes from SELH) and mean liability to exencephaly in each generation. Dominance or epistasis disrupts the linearity of this relationship. The frequency of exencephaly in SELH embryos was used to generate a scale, measured in units of standard deviation, where the assumed normal distribution of liability to exencephaly is positioned relative to a fixed threshold (see Fig. 3), using the probit transformation (Finney, 1971). The frequencies of exencephaly in BC1, F2-testcross, F l , F2, and LM/Bc embryos were used to locate each of their liability distributions relative to the fixed threshold and to calculate the position of their mean liabilities on the same scale as SELH (Fig. 12). The fit of the calculated means to linear regression was tested using the linear regression function on a Hewlett-Packard 11C calculator. The number of loci that cause exencephaly and differ between SELHA and LM/Bc was estimated by comparing the frequency of SELHA-like F2 sires (with respect to the frequencies 123 5 ° CM O LL CZ - CD T - Z3 O °" DQ 9> T~ CZ LU CO CD E CD CD ,s 11 C Q _ - CD CO > o > CO CO >,.g> -Q 5 jS .£5 CD c CD § E £ $ . .1 ^ ^ "3 h - T _ r2 . CD O "co _] cc •o < £5 <r LU 1 —I CZ LU CO co I ., .Q CO •= CD .3 £ o 2 E o Q. CD O CD c CZ CD :== o m CO > , CZ cz 0 CD •-1= CD 1 " O O . o CN to CO T5 CD C -CZ E CO *-* O 0 0 CZ CD .1 cz cb O ~ J -Q. -ti CO _ -Q CD o o c 8 f t CO o £ 2L>E CZ CO 13 •j= Q. O CO CD - C i— Q CO CD c CD CZ CD != CD CD ro^'° CD 15 Tj cz o £ CD — -CZ = • o c i = CD CD = C. -CZ E CD CD •Q CD *-°£ 2 "cf «e--CZ O «-Q _ - ~ O CD CD O > "D CZ CD $ 2 3 0) CO o CO CO •> CD T3 •o CO XJ CZ ro CO J3 CO I 124 of exencephaly in their offspring, here termed "exencephaly production") to that expected if 1-4 loci are segregating. If the strains differ by one liability gene, a quarter (25-26/102) of the F2 sires should produce SELHA-like frequencies of exencephaly, following Mendelian segregation ratios; 1/16 (6-7/102 F2 sires) for 2 loci; 1/64 (1-2/102 sires) for 3 loci; and 1/256 (0-1/102 sires) for 4 loci. As SELHA sires produced 8.7-28.6% exencephaly, F2 sires producing frequencies in the same range could be considered as SELHA-like. The same proportions of sires are expected to produce SELHA-like frequencies of exencephaly if multiple exencephaly-causing loci with equal and additive effects differ between SELHA and LM/Bc (i.e., only sires carrying SELHA alleles at all liability loci will be SELHA-like in exencephaly production). Alternatively, F2 sires with 95% confidence intervals of the mean overlapping the SELHA confidence interval of 16-20% (n=1404) could also be considered as SELHA-like in exencephaly production. This approach is more conservative than the last, and 2/13 SELHA sires themselves would not be considered as SELHA-like (their 95% confidence intervals do not overlap with 16.0-20.0%). Another approach taken to examine the possible number of exencephaly-liability loci that differ between SELHA and LM/Bc was to look at the distribution of the frequencies of exencephaly in the testcross progeny of F2 sires. If one locus differs, there are expected to be three classes of F2 sires; one quarter of them should carry only LM/Bc alleles at the liability locus and produce frequencies of exencephaly like LM/Bc sires testcrossed to SELHR dams, one half should be heterozygous and produce frequencies of exencephaly like F l sires testcrossed to SELHR dams, and one quarter should carry only SELHA alleles and produce frequencies of exencephaly like SELHA sires testcrossed to SELHR dams. If two loci are involved, then five classes of F2 sires are expected. For three loci, nine classes are expected. The expected frequency of each class can be calculated. In each case, the most frequent class of sires is the 125 Fl-like group and the least frequent classes of sires are those that are either carry all of their liability-alleles from LM/Bc or all of their liability-alleles from SELHA. To search for possible modes within the distribution from testcrossed F2 sires, the mean-litter Freeman-Tukey arcsine-transformed frequency of exencephaly (Mosteller and Youtz, 1961) was calculated for each sire, and the values were arranged in rank order and differentiated using a computer program (J. Pond and D.M. Juriloff) for Stewart's midpoint formula for "smoothing", or differentiating, frequency distributions (Stewart, 1969). Stewart's formula (<J>=28/(3x3+2x2+x1-x.,-2x.2-3x 3)) is used to differentiate a cumulative frequency plot where the cumulative number (or frequency) of individuals is plotted against their quantitative trait value. Transformation of this plot by differentiation allows one to look for modes in the resulting distribution. In the case of a single locus segregating, an F2 generation is expected to consist of a 1:2:1 ratio of SELHA, F l and LM/Bc-like animals, with respect to their genotype at that one locus. By taking a 1:2:1 ratio of the testcross values for SELHA, F l and LM/Bc animals, the expected cumulative frequency plot for a one locus model can be made and differentiated as for the F2 data. The two distributions can be compared in order to determine the fit of the observed data to that expected for a single gene. The Castle-Wright formula (Castle, 1921a, b; Wright, 1934; D.M. Juriloff, personal communication) was also applied to the mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly (Mosteller and Youtz, 1961) of the F l , F2, and parental generations. This formula calculates the number of loci, k, with equal and additive effects on a particular trait that are segregating in an F2 generation, based on genetic variance. If two inbred strains with phenotypic difference, D (the distance between their means for the trait of interest), are crossed, the mean of the F l and F2 generations should be mid-way between the two parental values (at VzD). If the genes underlying a trait have equal and additive effects, and all the "+" genes are 126 in one parental strain and all the "-" (liability) genes in the other, then the effect of substitution of any one allele will move the phenotypic value by D/2k. If one gene is segregating, there are 3 possible genotypes in the F2: +- (Fl-like), and ++ or — (like one of the parental strains), found at a 2:1:1 ratio. As the mean of the F2 and F l generations are the same, the genetic variance in the F2 would be: C 2 G = VA(-V2D)2 + V2(0)2 + V*(+VlD)2. = ZP/8 as one quarter of F2 animals have each parental genotype and their mean phenotypic value is + V 2 D from the mean of the F2 generation, while one half of F2 animals are Fl-like and their mean lies at the mean of the F2 generation. a 2 G works out to D2/16 if two loci are segregating, D2/24 for 3 loci, £>2/32 for 4 loci. In other words, a 2 G = £>2/8A\ The genetic variance in the F2, a 2 G , can be calculated from the observed F2 variance, which includes genetic and non-genetic variance, by subtracting the environmental (parental or F l , since inbred strains are used) variance. The formula can then be solved for k, the number of loci segregating. Statistical Methods X 2 tests of goodness of fit and the Freeman-Tukey arcsine transformation have been described in Chapter 2 (General Materials and Methods, "Statistical Methods"). The 95% confidence intervals of the mean frequencies of exencephaly (f) were obtained from Rohlf and Sokal (1981) or calculated using the formula: / ±1.96V(npq) (Snedecor and Cochran, 1967), where n is the total number of embryos examined, p is frequency of non-exencephalics, and q is frequency of exencephalics. 127 Results No ataxic F2 animals were seen in the 653 animals examined. One animal (F2 408) was particularly small, weighing one quarter that of its male litter mates and was difficult to sex externally (had been classified as female). It also had a shortened snout, limbs and tail. Upon autopsy at 7.5 weeks of age, even internal determination of sex was difficult; it is possible that this individual was a hermaphrodite, but the autopsy damaged some of the tissues that would have been required to confirm this. Several other pre-weaning F2 animals also appeared particularly small but most did not survive to weaning; the one that did (F2 937) weighed one third that of its male litter mates at 3 weeks of age and died within 2-3 days of being weaned, before it could be further examined. The observed incidence of exencephaly was 18.0% in SELH embryos (253/1404; range 8.7-28.6% from individual sires; Fig. 13 and Table 3); 0.1 % in LM/Bc embryos (1/1057; range 0-0.9% from individual sires; Fig. 13 and Table 3); 0.3% in the F l generation (3/1093; range 0-1.0%; Fig. 13 and Table 3); 3.6% in the BC1 generation (40/1114; range 0-8.3%; Fig. 13 and Table 3); and 4.4% in the F2 generation (24/541; range 0-13.3%; Fig. 13 and Table 3). The day of gestation on which litters were collected did not affect the frequency of exencephaly in SELH embryos as the size of all moles in SELH litters was consistent with death prior to D10. The 102 F2 sires testcrossed to SELHR dams produced frequencies of exencephaly ranging from 0% (31 sires) to 15.5% (1 sire; Fig. 13 and Table 3), with an overall incidence of 2.4%. The incidence of exencephaly in the reciprocal backcross was 5.9% (36/608; range 1.4-10.6%). The difference between the reciprocal backcrosses was not significant (F ( 1 1 8 ]= 2.65, < F . o 5 [ i , i 8 ] » 0.25>p>0.10). The frequency of exencephaly in the additional, "second" set of F2 embryos was 3.1% (19/610; range 0-10.3%). The frequency of exencephaly in each set of F2 128 Fig. 13: Distributions of the breeding values (frequencies of exencephaly produced in progeny) of individual sires for the crosses indicated. Approximately 100 embryos were scored per sire, except for SELHR sires (for which the mean was 61). Number of sires 0-5 0 5 0 5 0 F2 SIRES X S E L H R DAMS LM/Bc S IRES X LM/Bc DAMS LM/Bc SIRES X S E L H R DAMS S E L H R SIRES X F1 DAMS F1 SIRES X S E L H R DAMS S E L H A SIRES X S E L H R DAMS 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 % exencephaly 129 CO O "co Q _ cu , o c CD X CD O CO CO m CM CO CD o" CO s> CO I f ) o CO "3- CO CNJ CNJ^  IT) CO CNJ CO O j Q E CD O m o co O) o oo o co CN CO o E 00 CN CT) CO CO CN CN CO Is- CO 0) h- CO I f ) O) CO CO 00 CO CO c Q . E 4b o CO CO I f ) CO CO T— T— CO I f ) Is- CN CN l f ) i f ) CO i f ) CO CO o O < I —l LU CO cd LU CO O CO o CO o CO cn LU CO CC I _l LU CO Cd 111 CO CM CN LL cd LU CO < cu UJ E CO o 130 embryos did not differ significantly (X 2; 0.5>p>0.1). The incidences of exencephaly in the parental strains, BC1, F l , and F2 generations, as well as the testcross of F2 sires, were used to examine the mode of inheritance of the exencephaly-liability loci. The low frequency of exencephaly in the F l generation suggests that the genes causing exencephaly in SELHA mice are not dominant in combination with LM/Bc alleles. The frequencies of exencephaly observed in the various crosses were compared to those expected for 1 autosomal recessive gene with 18.0% penetrance and for 2-4 recessive loci with duplicate epistasis and 18.0% penetrance in the multiple homozygote (Table 4). Although the frequency of exencephaly observed in F l and F2 embryos was consistent with a single autosomal recessive gene with 18.0% penetrance, the frequencies observed in BC1 embryos and testcross embryos from F2 sires are significantly lower than expected under this model (Table 4). The incidence of exencephaly in BC1 embryos only agrees with that expected for 2 epistatic loci with 18.0% penetrance in the double homozygote (Table 4), while the frequency observed in testcross embryos from F2 sires only agrees with that expected for 3 autosomal recessive loci with 18.0% penetrance in the multiple homozygote (Table 4). Thus, the observed data collectively do not appear to fit a single recessive or multiple recessive locus epistatic models of gene action because the correct model should fit the data from all generations. The observed frequencies of exencephaly in the parental, F l , BC1, and F2 generations, and in the testcross of F2 sires were compared to those expected under a multifactorial threshold model with additive gene action. The mean liability values of each cross was derived by probit transformation (Table 5). Linear regression (Fig. 14) of the mean liabilities against the proportion of SELH genes present in each generation gave a correlation coefficient value, R2, of 0.7414. If only the parental, F l and BC1 generations are used, the value of R 2 is higher (0.8699). While these values are consistent with an additive model of gene action and suggest Table 4: Comparison of the observed frequencies of exencephaly and the 1 3 1 frequencies predicted by recessive epistatic models of gene action. % exencephaly expected Genotype observed 1 locus 1 2 loc i 2 3 loc i 2 4 loc i 2 SELHA/Bc 18 — — — — LM/Bc 0.1 — — — — F1 0.3 0 0 0 0 BC1 3.6 9.0* 4.5 2.3* 1.1* F2 3.7 4.5 1.1* 0.3* 0.1* F2 testcross 2.4 9.0* 5.1* 2.3 0.1* 'Autosomal recessive with 18.0% penetrance. 2 Autosomal recessive with duplicate epistasis and 18.0% penetrance in multiple homozygotes. *Observed number is significantly different (X^ P<0.05) from that predicted by the model. Table 5: Comparison of the observed frequencies of exencephaly with the values predicted by the multifactorial threshold model. Genotype % SELHA genes Observed % exencephaly Location of threshold1 Location of mean2 Predicted location of mean 3 Predicted % exencephaly S E L H A 100 18 4.0836 5. OOOO4 5.285 11.5 BC1 75 3.6 3.2009 5.883T4 5.8208 4.1 F2-testcross 75 2.4 3.0226 6.162 4 5.8208 4.1 F1 50 0.3 2.2522 6.83244 6.3567 1.2 F2 50 4.4 3.294 5.79064 5.7906 1.2 LM/Bc 0 0.1 1.9098 7.17484 7.4284 <0.1 Relative to a mean of 5.0000 in units of standard deviation (probits). Relative to a fixed threshold at 4.0836 on probit scale. Relative to a fixed threshold at 4.0836, for regression line with R2=0.7414, slope=-0.0197, Y-intercept=7.2869. Used to calculate linear regression against percentage SELHA genes. C 3 - > iu s c CD CD <» a) » * - Q T J T J C CO CD T-132 c CO ™ OT o CO 1c •5 £ * E o to •*= CD CO CO g CO i -CD CO O $ CD £ CD CO E o CO CD xi o — Q) t . "° -o 3 CD CD : T J CO CD C > " O — CD O CD T J E £ T J 22 "5 CO o ir T J x: c i f * J5 CD ni to 2 2 co cL Q-fD CD *- x: CD — c — c T J O CD <4-x: — CO . _ co j^-T J T-I— d II — c •S c5 ifO CO C. CD c Q-O CO '••= CD CD CD c CD < - C _ l 8* CD 0 3 O .CO o c •E 3 ^ T J =3 • - CD t XI CO CD fO CD £ ^ a ~ E x O CO CD tl CD -ro ffl -2 r (D +J Q . £ = CD « X i O ® CO C o — CD C C X T J S t 5 ai — C O c o> . CO CO o x: ~ a . £ CD o O Q . c o CD o CD CD • o ^ CO c CD O o co C CD . CD 3 = C T X i J D ^ g T J C O CD CO II 5 8 o OQ CD X > £ -O t TJ CD T J C CO LL C T— >5 U_ -H CO < I E o CO CD CO c o to to * - CD CO r j i i . CD CD S? 3 CO i - CO ^ > o CO , _ — CD 7\ ro O -= CD a) co o CO O tJJ c ca CD E o _ i T - LU CO CM LL X LO Q_ O C O LU CO O If) (O 0 fZ CD CM O ) LU CO LO CN o C O o i 00 CD LO X (3 X 133 better fit of the data to this model than recessive epistatic models, the amount of the deviation of the correlation coefficient from 1.0000 suggests that gene action may not be simply additive. The proportion of F2 sires that produced frequencies of exencephaly like SELHA sires when testcrossed to SELHR dams can be used to estimate the number of exencephaly-liability loci segregating in this cross. If only one liability locus were segregating, 1/4 of F2 sires (25-26/102) would be expected to produce SELHA-like frequencies (8.7-28.6%) of exencephaly. In contrast, only 1/16 (6-7/102) would be expected to be SELHA-like if two liability loci were segregating. Even fewer, 1/64 (1-2/102 sires) are expected for 3 loci, and if 4 loci were involved, the proportion expected is so small (1/256) that none are expected in a sample of 102 sires. Approximately 1/26 (4/102) F2 sires fall within the range of SELHA sire values, suggesting the segregation of 2-3 loci. This approach suggests that there is a significant deficit (X 2, p<0.001) in the number of F2 sires expected to produce SELHA-like frequencies of exencephaly if there were only one causative gene. The deficit (approximately 21 sires) is larger than can be accounted for by a number of these animals having been exencephalic and therefore not surviving to be testcrossed as only 4.4% of F2 embryos were exencephalic; a maximum of 4-5 additional F2 sires would have been SELHA-like in exencephaly production. Only about 30% of exencephalics are expected to be male (Juriloff et al., 1989), so fewer than 5 exencephalic male F2 embryos would actually have been expected. This suggests that between four (observed value) and nine (maximum possible, including exencephalics) SELH-like F2 animals out of 107 (102 sires + 5 exencephalics) were obtained. This range extends through the value expected if two liability loci are segregating (6-7/107 animals), suggesting that the cause of exencephaly in SELH mice is most likely due to a combination of two loci. The reasonably good fit of generation mean liabilities to linear regression suggests that the effects of these loci are largely additive (i.e., there is little or no dominance within or between loci; Falconer, 1989). 134 The distribution of the breeding values of testcrossed F2 sires (Fig. 13) is similar to that expected if 2 (Fig. 15) additive exencephaly-liability loci are segregating and is significantly different than the distribution expected if one locus is segregating (Fig. 16). For three loci (not shown), seven classes of F2 sires are expected and four of these classes (accounting for 4 2 / 6 4 of sires) would produce a mean frequency of exencephaly in their progeny of 0-3.6% (i.e., the distribution would be similar to that for two loci, where many sires produce 0% exencephaly and few produce high frequencies). None of the F2 sires has a breeding value of 16.0% or greater (putting it within the range of the 95% confidence interval around the SELHA value). The F2 sire with the highest breeding value was sire 317, with 15.5% exencephaly in testcross progeny. This sire is the only one whose 95% confidence interval overlaps with the one around the SELHA value. Thus looking at the data this way suggests more than 2 loci (but fewer than four) are segregating, but as already mentioned, not all SELHA sires would even be considered SELHA-like under these criteria, suggesting that looking at the proportion of F2 sires with values within the range of values actually produced by SELHA sires is the best approach. In another approach, instead of the number of classes of F2 sires, the distribution of frequencies of exencephaly produced by individual F2 sires was also examined in relation to the patterns expected for various numbers of exencephaly-liability loci segregating. If there is only one codominant liability gene, then the expected frequencies of exencephaly produced by F2 sires should reflect the three expected classes of F2 sires (Fig. 16); LM/Bc-like, Fl-like, and SELHA-like. Under this model, one quarter of the F2 sires are expected to produce about 0.1% exencephaly (0-1%), one half of the F2 sires are expected to produce about 3.6% exencephaly (0-8%), and one quarter of F2 sires are expected to produce about 18.0% exencephaly (9-29%). The observed data do not fit such a distribution pattern: significantly more sires produced low 13 Fig. 15: Expected distributions of the breeding values (frequencies of exencephaly produced in testcross progeny) of F2 sires if 2 equal and additive exencephaly-liability loci are segregating. Each distribution represents one genotypic class of F2 sires, and the numbers above the distributions indicate the number (/4) of liability alleles inherited from the SELHA parental strain. The expected shape of the distribution overall is indicated by the pale, thick line. 2 5 10 15 20 25 Frequency of exencephaly produced by F2 sires in testcross to SELHR Fig. 16: Expected distributions of the breeding values (frequencies of exencephaly produced in testcross progeny) of F2 sires if 1 additive (co-dominant) exencephaly-liability locus is segregating. Each distribution represents one genotypic class of F2 sires, and the numbers above the distributions indicate the number (12) of liability alleles inherited from the SELHA parental strain. The expected shape of the distribution overall is indicated by the pale, thick line. 1 0 5 10 15 20 25 Frequency of exencephaly produced by F2 sires in testcross to SELHR 136 frequencies of exencephaly and fewer produced SELHA-like frequencies than expected if exencephaly is caused by a single codominant gene (compare Figs. 13 and 16). Even if the breeding values are transformed by taking the mean-litter Freeman-Tukey arcsine values (Fig. 17), the observed distribution does not resemble that expected for one locus. The smoothed frequency distribution (Fig. 18), using the mean litter Freeman-Tukey arc-sine transformed frequencies of exencephaly, is also different from the distribution expected for a single codominant gene, as determined by generating a frequency distribution from 10 each of LM/Bc, F l , and SELHA sires testcrossed to SELHR dams, and 10 SELHR sires testcrossed to F l dams (Fig. 19) to give a 1:2:1 ratio of LM/Bc: F l : SELHA animals testcrossed to SELHR. The observed F2 distribution is compacted toward smaller Freeman-Tukey frequencies of exencephaly, and the peaks in the data lie at different X-axis values than do those of the simulated single locus distribution (Figs. 18 and 19). It is possible to distinguish modes in the differentiated distribution for the F2 generation, suggesting that a large number of loci are not involved. The number of exencephaly-liability loci segregating in this cross was also estimated using the Castle-Wright formula (k=D2/8(a2G)). Estimating nongenetic variance from the SELHA sample (a2=18.91) gave an estimate of -4.2 loci. The negative value is the result of the F2 variance (CT2=10.94) being smaller than that of the SELHA parental variance. The F2 distribution and therefore the F2 variance is truncated and compacted at 0% exencephaly because the mean frequency of exencephaly is low. The mean of the SELHA generation is higher and no SELHA sire produced less than 8.7% exencephaly, so the variance is not truncated or compacted. In an attempt to correct for this, the mean parental variance (i.e., the average of the SELHA and LM/Bc variances; a2=9.58) was used as that of LM/Bc (o2=0.24) is also truncated at 0% exencephaly; this gave an estimate of about 25 loci. This number is obviously wrong, Fig. 17: Distribution of the breeding values (mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly) of individual F2 sires. Number of sires 3 0 4 2 5 4 7 9 1 1 1 3 1 5 1 7 1 9 2 1 2 3 2 5 mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly 138 Fig. 18: Frequency distribution of the breeding values (frequencies of exencephaly produced in testcross progeny) of F2 sires. The distribution has been smoothed using Stewart's mid-point formula (Stewart, 1969). Frequency at midpoint (0) 5 10 15 20 25 30 Frequency of exencephaly (Freeman-Tukey arc-sine) Fig. 19: Expected frequency distribution of the breeding values of F2 sires if one exencephaly-liability locus is segregating. The plot was constructed by using Stewart's mid-point formula (Stewart, 1969) to smooth the distribution created by taking a 1:2:1 ratio of the breeding values of LM/Bc, F1, and SELHA animals testcrossed to SELHR. ^ Frequency at midpoint (<$) 4 0 -30 -2 0 -10-5 10 15 20 25 30 Frequency of exencephaly (Freeman-Tukey arc-sine) 140 considering the results of the other methods of examining the data. Finally, estimating nongenetic variance from the F l sires (a2=2.36) gave a more realistic estimate of 3.9 exencephaly-causing genes. Discussion The previous genetic analysis of the cause of exencephaly in SELH/Bc mice involved a cross to the normal but related strain, ICR/Be (Juriloff et al., 1989). The use of a related strain was deliberate, as nothing was previously known about the number of loci involved. This allowed an estimation of the number of major genes contributing to exencephaly liability in SELH mice and their mode of inheritance. In the present study, SELHA mice were crossed to LM/Bc, an unrelated strain with a normal pattern of neural tube closure and virtually no spontaneous production of exencephaly. The purpose of this study was two-fold: to obtain an estimate of the number of loci involved in exencephaly-liability that differ between SELHA and LM/Bc (necessary background information for the mapping study) and to extend and test the conclusions of the previous genetic analysis in a cross to an unrelated normal strain. In addition, the SELH mice used in the previous genetic analysis in 1989 had not reached 20 generations of brother-sister inbreeding, and so it was possible that the SELH stock was still segregating (heterozygous) at one or more of the loci involved in causing exencephaly. Although all the SELH mice tested produced exencephaly in their progeny, three SELH embryos (of the 72 embryos observed at a stage scorable for Closure 2) were seen with evidence of Closure 2 (Fig. 8 in Macdonald et al., 1989). In addition, the frequency of exencephaly produced by SELH sires ranged from 4.0-39.6% (Juriloff et al., 1989). Although there still appears to be some fluctuation in the frequencies of exencephaly produced by SELH sires (see Chapter 5), the range is much less and, in general, lower frequencies (below 30%) are now seen. 141 The large range of exencephaly frequencies could have been a result of residual heterozygosity at some loci. The mice used in the present study had the equivalent of over 21 generations of brother-sister inbreeding and are therefore expected to be homozygous at all loci. If the mice in the previous analysis were heterozygous at any of the exencephaly-liability loci, the estimate of the number of genes in that study would likely have been inaccurate, and the number of genes involved could have increased or decreased upon further inbreeding of SELH, depending which allele (normal or mutant) was fixed at any heterozygous liability loci. There is no single method of analysis that allows testing of all possible genetic models. Instead, individual models must be tested for fit to the observed data, sometimes by more than one method. It is therefore not possible to come out with an exact answer, but rather some models can be eliminated to leave a relatively small number of possibilities. The frequencies of exencephaly observed in the various generations examined are not consistent with epistatic models of inheritance but are compatible with a multifactorial model with additive gene action and a developmental threshold. The deviation of the data from linear regression suggests that gene action may not simply be additive. The position of the F l value above the regression line (toward the LM/Bc parental value) suggests that there may be some dominance of the LM/Bc alleles at the exencephaly-liability loci, although it could also be the result of "hybrid vigour" (heterosis; Falconer, 1989), in which heterozygosity at all "background" loci appears to restore nonspecific fitness or robustness to developmental processes that was lost in inbreeding. For example, F l animals often grow larger, and are fertile earlier and have larger litters than either parental strain. It is possible that heterosis in the F l could "dampen" the effect of the exencephaly-liability alleles they carry, decreasing their risk of being affected and causing their mean to be shifted toward that of the LM/Bc parental generation. 142 Another possibility is that the genetic variance is not the same in each generation as is assumed by the method of transforming the data using probits. If the variance differs in some generations, the proportion of individuals expected to fall over the threshold would change since the shape of the distribution is affected. The smoothed distribution of the frequencies of exencephaly produced by testcrossed F2 sires suggests that there are a small number of exencephaly-liability loci, but more than one, segregating in this cross. The proportion of F2 sires that produced frequencies of exencephaly in testcross progeny within the same range as SELHA sires indicates that 2-3 loci are responsible for the high liability to exencephaly seen in the SELH(A) mouse strain; a significantly lower proportion of F2 sires are SELHA-like in exencephaly production than is expected if one locus were segregating (4/102 observed vs 26/102 expected). Taking into account that exencephalic F2's should carry most or all SELHA alleles at all exencephaly-liability loci, that they would not have survived to be testcrossed, and that the frequency of exencephaly in the F2 generation is 4.4%, between 1/12 and 1/27 F2 animals would be expected to be SELHA-like with respect to liability to exencephaly. This value fits best with that expected for two loci (1/16). It is interesting that the 95% confidence intervals around the mean frequencies of exencephaly produced by some SELHA sires do not overlap with the 95% confidence interval around the average value for SELHA sires; i.e., the variance among SELHA sires is greater than expected. This could be a result of the mild temporal fluctuation in frequency of exencephaly produced by SELHA sires in this study (see Chapter 5); sires testcrossed in the first 10-11 months of data collection produced higher frequencies of exencephaly than those testcrossed in the latter 12-13 months (see Fig. 20, Chapter 5). The cause of this fluctuation is not obvious, although it is interesting that fluctuations have been noted in the incidence of human NTD over time (Campbell et al., 1986). 143 As the distributions of the different genotypic classes of F2 sires are expected to overlap (see Fig. 15), the F2 sires that produced the highest frequencies of exencephaly may not all carry only SELHA alleles at liability loci. If two loci are segregating, a sire with 3 of 4 SELHA alleles at these loci could easily, by chance, produce SELHA-like frequencies of exencephaly. Such a sire would lie at the "high" extreme of his distribution and could produce a higher frequency of exencephaly in testcross progeny than a sire with all his liability alleles from SELHA but that lies at the bottom or "low" extreme of his phenotypic distribution. It is possible, therefore, that at least one of the F2 sires that is SELH-like with respect to exencephaly-production is not genotypically completely SELHA-like at the exencephaly-liability loci. This would result in an overestimate of the number of SELHA-like sires and an underestimate of gene number. In addition, a greater proportion of F2 sires is expected to carry one or two LM/Bc alleles at liability loci, making it more likely that a male with a "low" genotype will be misclassified as a high-risk sire than that a high-risk sire will be misclassified as low-risk. It is therefore best to take a conservative estimate of gene number as 2-3 additive loci. The Castle-Wright formula is designed to give an estimate of the number of genes that underlie a quantitative trait and are segregating in the F2 from a cross between two strains (Castle, 1921a, b; Wright, 1934). The following assumptions must be met: that the genes have equal and additive effects (no dominance or epistasis), that involved loci are unlinked, and that the alleles in the high strain all increase the phenotype while those of the low strain all decrease the phenotype. In addition, it is assumed that each generation shows a normal distribution for the trait of interest. There is an immediate difficulty in meeting these assumptions for the exencephaly trait in SELH mice —a low frequency trait with a developmental threshold, not a quantitative trait at 144 the level of direct observation. The underlying liability to exencephaly in SELH mice is assumed to be due to a quantitative characteristic, but the resulting phenotype is qualitative (affected or normal) and therefore binomially distributed. Although the variance is related to the mean, the distribution of exencephaly production by SELHA sires is not truncated because its mean lies far enough from 0% exencephaly. However, the distributions of the LM/Bc parental strain, the F l generation and F2 sires testcrossed to SELHR dams have means closer to 0% exencephaly and their distributions are truncated, affecting their variances. This is also evident in the BC1 generation, where the variance is smaller than that of the SELH parental strain. The Freeman-Tukey arcsine transformation may be used to stabilize the variance of binomial distributions when it is dependent on the mean (see p. 95), thus converting the distribution into an approximately normal one. However, the effect of extreme outlying values is sometimes too great for the variance to be stabilized. Since calculation of the Castle-Wright formula depends entirely upon variances, the asymmetrical distributions observed in most generations of the cross between SELH and LM/Bc mice make this approach inappropriate for the analysis of traits like exencephaly in SELH mice. This is also suggested by the values obtained for gene number in this study, where even the use of the mean parental variance gave an estimate of about 25 loci. This suggests the calculated variances are smaller than they should be, relative to the distance between the two parental means (i.e., the denominator of the equation is too small). Use of the F l variance provided a more realistic estimate of 3.9 loci, although this number is still larger than that suggested by other methods of analysis. In addition, if there is any dominance, as may be indicated by the position of the F l value in the linear regression, then at least one additional assumption underlying Wright's formula would be violated. Zeng et al. (1990) also found that Wright's method was of little value in estimating the number of loci influencing a quantitative trait as its 145 strict assumptions are regularly violated. They suggest that, because there is no reliable method of estimating gene number, the best means of determining the number of loci may be to map them. As liability to exencephaly in SELH mice seems to fit a multifactorial threshold model of inheritance, it is not surprising that it is difficult to identify the precise number of loci involved. In such a system, genetic, environmental and stochastic events influence an embryo's phenotype, and even when two embryos share a common genotype and environment, the stochastic events that take place during their development will differ (Kurnit et al., 1987). Thus, non-genetic factors that influence neural tube closure may result in one embryo being affected whereas another with the same genotype may be normal. In addition, depending on non-genetic factors, one embryo may require more liability alleles than another to push it over the threshold to be affected. The results obtained in this study are very similar to those obtained when SELH mice were crossed to the related strain, ICR/Be. This suggests that either background effects (as a result of "minor" loci) are minimal or that LM/Bc mice do not carry alleles at such loci that affect the expression of the (major) SELH exencephaly-liability loci. Three crosses of SELH mice to normal strains (ICR/Be, SWV/Bc, LM/Bc) have suggested that the liability to exencephaly in SELH mice is*due to a small number of genes with additive effects. The similarity of the results from crosses of SELH to three different normal strains suggests that the exencephaly-liability in SELH mice is due to mutant alleles, rather than polymorphisms. If the liability to exencephaly were the result of polymorphisms, the three different crosses would be unlikely to give similar results as it is unlikely that ICR/Be, SWV/Bc and LM/Bc would share the same polymorphisms at exencephaly-influencing loci. This cross, between SELH and LM/Bc, was the foundation of a study to map the genes 146 responsible for the liability to exencephaly in SELHA/Bc mice (see Chapter 5). It is helpful to have an approximate idea of the number of loci segregating for the trait in the specific cross made in order to map them. 147 Chapter 5: MAPPING THE SELHA/BC EXENCEPHALY-LIABILITY LOCI Introduction As the exencephaly-liability in SELHA mice appears to be caused by 2-3 loci with additive effects that differ between SELHA and LM/Bc mouse strains (see Chapter 4), it should be possible to map these loci using methods for multigenic traits. The emergence and growth in the number of mapped simple sequence length polymorphisms (SSLPs) in the early 1990's has provided a large number of highly polymorphic, densely spaced genetic markers that are ideal for mapping and linkage analysis. These markers are highly polymorphic between inbred strains of mice and can be typed quickly and easily using the polymerase chain reaction (PCR). Using these markers, it is possible to look in a segregating generation for co-segregation of alleles from a parental strain and high or low liability to a trait. Most analytical methods applied to linkage analyses for complex traits assume a large number of liability loci and/or a normal distribution of the phenotypic values for parental and segregating generations (for example, see Soller et al., 1976; Tanksley et al., 1982; Edwards et al., 1987; Paterson et al., 1988; Lander and Botstein, 1989). The traditional single-marker approach to mapping quantitative trait loci (QTLs) assumes that a QTL lies exactly at a marker locus and looks for a significant difference in the mean phenotypic value of progeny with different genotypes at the marker (Soller et al., 1976; Tanksley et al., 1982; Edwards et al., 1987). Interval mapping (Lander and Botstein, 1989) is based on the fact that a QTL is more likely to lie between two marker loci than at one. This approach uses an adaptation of LOD scores employed in human linkage analysis, based on maximum-likelihood methods, to calculate the most likely effect of a putative QTL at any given location. Alternatives to this approach include the use of multiple regression in place of maximum-likelihood methods (Haley and Knott, 1991). Multiple linear regression can also be combined with conventional interval 148 mapping to test models that assume there are multiple QTLs elsewhere in the genome when looking at each individual location (Jansen, 1992; Jansen, 1993). All of these approaches are directed at normally distributed quantitative variables. Kruglyak and Lander (1995) have developed a nonparametric approach for mapping QTLs, in which individuals of a segregating generation are ranked for phenotype and for genotype independently and the distributions are compared using the Wilcoxon rank-sum rank test. Their approach allows interval mapping through generalization of the Wilcoxon rank-sum statistic to the region between markers, where genotype must be inferred from the genotype at adjacent flanking markers, and by determination of appropriate threshold values for significance (Kruglyak and Lander, 1995). This approach appears to require genotyping of a relatively large population (i.e., the entire segregating generation). Exencephaly in SELHA mice is a low frequency, genetically complex lethal threshold trait. At the phenotypic level, threshold traits are binomial (individuals are either affected or normal), not normally distributed, although the underlying genetic liability to these traits is generally assumed to be normally distributed. Although all SELHA embryos are genetically at risk to developing exencephaly, only about 18% are affected. Thus, in a segregating F2 generation from a cross between the SELHA and LM/Bc strains, an embryo may carry SELHA alleles at all the exencephaly-liability loci and yet not be affected. In addition, as the liability loci appear to act in an additive manner, some embryos that carry many (but not all) of their alleles at these loci from SELHA may be exencephalic. This demonstrates the difficulty in correlating genotype to phenotype for a multigenicand lethal threshold trait. However, if adult F2 animals are testcrossed to SELH mice in order to determine the frequency of exencephalic progeny in their offspring, their breeding values may be used as a more accurate measure of their genetic liability to exencephaly. F2 sires that carry SELHA 149 alleles at all the exencephaly-liability loci are expected to produce the same proportion of exencephalic testcross progeny as SELHA sires, while F2 sires that carry LM/Bc alleles at all the liability loci are expected to produce the same proportion of exencephalic testcross progeny as LM/Bc sires. As liability is measured by the proportion of exencephalic offspring produced by individuals within a population, the distribution of a population with a low phenotypic mean will be truncated at zero, resulting in a skewed distribution. For this reason, methods of linkage analysis that assume a normally distributed variable are not appropriate for this type of trait. It is also more difficult to obtain an accurate estimate of each individuals' genotype for a low frequency trait because there is a strong tendency for the phenotypic distributions of some genotypic classes to overlap. For example, as there is overlap between the genotypic classes of F2 sires that produce 0% exencephaly (see Chapter 4 and Fig. 15), some of these low-risk sires may carry some SELHA alleles at exencephaly-liability loci. In addition, as the proportion of F2 animals that carry all their liability locus alleles from the SELHA strain is small (i.e., 1/16 if there are 2 loci, 1/64 if there are 3 loci), and some of these animals are expected to be exencephalic and not survive to be testcrossed, sires that produce high frequencies of exencephaly are likely to carry some of their liability locus alleles from the LM/Bc strain. As the genotypic class of sires carrying half of their alleles from either parent is the most frequent, and sires within this class may, by chance, produce a large range of exencephaly frequencies in their offspring, some high-risk and some low-risk sires (as well as some exencephalic embryos) may carry only half of their liability alleles from either, parental strain, although in general, high-risk sires and exencephalics are expected to carry mostly SELHA alleles and low-risk sires are expected to carry mostly LM/Bc alleles. The number of loci that cause exencephaly in SELHA mice appears to be fairly small, and the effects of individual loci are therefore expected to be quite large. It should therefore be 150 possible to detect linkage of markers to liability loci by simpler methods than are required for the mapping of loci responsible for a strain difference that is due to a large number of QTLs of relatively small individual effect. The methods developed in this study are based on methods of looking for linkage to one marker at a time and, as the combined effects of genes are important for additive traits, the combined genotypes of potential candidate regions are also examined relative to phenotype. These approaches involve a simple statistical method of analysis which should be adequate to indicate the most likely regions of the SELHA genome to carry exencephaly-liability loci. These regions can be tested empirically by other genetic experiments. Mapping loci to chromosomal regions may indicate potential candidate loci, and is an important step in ultimately identifying the genes that cause exencephaly in SELH embryos and understanding their role in normal and abnormal neural tube closure. Preliminary results from this study were presented in a poster at the 44th Annual Meeting of the American Society of Human Genetics (Montreal, Canada) in October 1994. Materials and Methods Mouse stocks and maintenance All mice were maintained as described in Chapter 2. The SELHA/Bc, SELHR/Bc and LM/Bc strains of mice are all described in Chapter 2. Experimental design Selection of F2 animals to be typed for genetic markers: LM/Bc dams were mated to SELHA sires, and F l mice were raised and intercrossed to produce F2 animals, as described in Chapter 4 (Materials and Methods, "Breeding scheme"). A total of 102 F2 sires were testcrossed to SELHR females, and approximately 100 embryos per 151 sire were scored for exencephaly in order to determine each sire's individual liability to exencephaly (as described in Chapter 4 (Materials and Methods, "Progeny testing")). The ten sires that produced the highest frequencies of exencephaly ("high-risk sires") and ten of the 31 sires that produced no exencephalic progeny ("low-risk sires") were selected to be typed for genetic (SSLP) markers. The 10 low-risk sires (pedigree numbers 321, 385, 549, 598, 614, 841, 855, 869, 908, and 944) were selected as follows: only 1 low-risk sire was selected from any one litter; sires testcrossed during different time-periods were chosen; and sires for which the largest number of testcross embryos were collected were selected over those with a smaller sample size, when permitted by the other criteria. The ten sires that produced the highest percentages of exencephalic progeny were sires #317 (produced 15.5% exencephaly), 911 and 930 (both produced 10.2%), 388 (9.7%), 943 (8.2%), 307 and 323 (both 8.0%), 320 (7.7%), 306 (7.6%), and 310 (6.7%). This set of "high-risk" and "low-risk" F2 sires was used to screen the genome for linkage to the SELHA exencephaly-liability trait. In order to find SSLPs that differ between SELHA and LM/Bc, the two strains were typed for 312 SSLP markers, 125 (40%) of which were informative (see Appendix C). Each of the markers to be typed was selected based on its map position and its degree of certainty, its reported level of polymorphism between common inbred strains of mice (Research Genetics Inc. listing), and the size difference between its known alleles with preference given to those with differences of at least 5 bp. Informative markers covering all the autosomes at intervals of approximately 20 cM or less were identified and 109 markers were selected to type the 10 high-risk and 10 low-risk F2 sires, as well as SELHA, LM/Bc and F l animals. After the data for the high- and low-risk F2 sires had been analysed and the chromosomal regions most likely to carry exencephaly-liability loci had been identified, a set of 31 exencephalic F2 embryos was used to test if these linkages were also apparent in a second data 152 set (i.e., to test the hypotheses generated by the F2 sire data). Exencephalic F2 embryos, like the high-risk F2 sires, are expected to carry most (but not necessarily all) of their liability alleles from SELHA. The 31 exencephalic F2 embryos were typed for markers on the chromosomes that were indicated to carry exencephaly-liability loci in the analysis of high- and low-risk F2 sires. In addition, as the methods of linkage analysis utilized in this study focused on showing deviation of the inheritance of marker alleles from Mendelian segregation patterns, it was necessary to show that there is Mendelian segregation at the "best" markers (one marker on each of the chromosomes for which there was the most support for linkage to an exencephaly-liability locus). Each of these markers was therefore typed in 79-81 F2 embryos (6 entire litters, including exencephalics); with a sample of this size, it should be possible to detect deviations such as: one of the homozygote classes deficient by at least one half or one of the homozygote classes deficient by 40% and the other homozygous class with 40% additional animals. Any marker for which the deviation from expected in this set of 6 litters showed borderline significance (0.10>P>0.05) was typed on a second data set of 75 F2 embryos (6 more entire litters). Re-ranking of high-risk F2 sires: Later in the study, potential effects that might change the membership of the high-risk group of F2 sires were explored. There appeared to be a mild time trend in the frequencies of exencephaly produced by SELHA and F2 sires over the 22 months during which they were testcrossed to SELHR dams. This was examined by plotting the mean-litter Freeman-Tukey arcsine transformed frequency (Mosteller and Youtz, 1961) of exencephaly produced by each sire against the median of the months during which each sire was testcrossed (Figs. 20, 21). As a large number of F2 sires were typed, a random fluctuation of breeding values is expected over 153 time, and the data are expected to produce a linear regression with a slope of zero. This assumption was tested (Fig. 22) using the Series Statistics calculation of linear regression in Harvard Graphics 3.0 for Windows (Software Publishing Company, Santa Clara, CA, USA). Fit to polynomial regression (Fig. 23) was also tested, using the "Stat Pac" statistical package of a Hewlett Packard 41C calculator. If significant, the inverse parabola (Y = 5.90 + 0.75x - 0.03x2) can be plotted (Fig. 23) and its values subtracted from the values for the best fitting curve at each value of x to give a straight line with a slope of zero (i.e., as expected if fluctuation in breeding values is random). By determining the difference between the two curves at each value of x and subtracting these values from each F2 sire data point, the data can be corrected to fit a linear regression with a slope of zero, i.e., the effect of the time-trend can be removed. This approach was applied to the data for the 21 highest-ranking sires for exencephaly production and their ranking was corrected for the temporal effect on production of exencephaly. As the ranking of some of the top ten exencephaly-producing sires was altered, the "new" top ten sires were typed for markers on the chromosomes that are thought to carry exencephaly-liability loci, based on the original analysis, and analysed using the same methods as for the original data set. This additional data set does not replace the original data set but provides an additional (alternative) means of examining the data. Genotyping of F2 animals: DNA was prepared from SELHA, LM/Bc, F l and a total of 179 F2 animals (23 F2 sires and 156 F2 embryos). Most adult DNA was prepared from liver tissue, but as the DNA of some F2 sires was required while they were still breeding (to collect data for another study), some samples were prepared from tail tips. For liver preparations, the liver was removed and either used fresh or frozen immediately and stored at -70°C until use. Approximately half of the liver 154 Fig. 20: Plot of the mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly produced in testcross progeny by SELHA sires over the time-period of testcrossing. Each X-axis value represents one month, starting in Nov. 1992 (at 1) and ending in Aug. 1994 (at 22), and sires are plotted by the median of the months during which their litters were collected. 35 30 25 20 15 10 frequency of exencephaly (mean litter Freeman-Tukey arcsine) n — i — i — i — i — i — i — i — i — n — i — r — i — i — i — i — i — i — T i — i — i — 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 11/92 02/93 05/93 08/93 11/93 02/94 05/94 08/94 Fig. 21: Plot of the mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly produced in testcross progeny by F2 sires over the time-period of testcrossing. Each X-axis value represents one month, starting in Nov. 1992 (at 1) and ending in Aug. 1994 (at 22), and sires are plotted by the median of the months during which their litters were collected. frequency of exencephaly (mean litter Freeman-Tukey arcsine) • t • • • • • • • • * • • • • t • • ' t ••. • . • • 1 I ! s • • .„• i t • • i 1 . • 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 11/92 02/93 05/93 08/93 11/93 02/94 05/94 08/94 : 155 Fig. 22: Plot of the mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly produced in testcross progeny by F2 sires over the time-period of testcrossing, showing the fit to linear regression (R2=0.02). Each X-axis value represents one month, starting in Nov. 1992 (at 1) and ending in Aug. 1994 (at 22), and sires are plotted by the mean month during which their litters were collected. 25 20 15 10 5 0 frequency of exencephaly (mean litter Freeman-Tukey arcsine) • • • - • • • • • • • • • W I - • • • • • Mi • • t • • # t • ' • ^ - • - -•- s • • • • • • • • • • • 1 • i • • • 1 • t ! • • i i i i i i i i i i i i r 0 1 2 3 4 5 6 7 8 9 1011 12 1314 1516 171819 2021 22 11/92 02/93 05/93 08/93 11/93 02/94 05/94 08/94 25 20 15 10 5 Fig. 23: Plot of the mean-litter Freeman-Tukey arcsine transformed frequencies of exencephaly produced in testcross progeny by F2 sires over the time-period of testcrossing. Each X-axis value represents one month, starting in Nov. 1992 (at 1) and ending in Aug. 1994 (at 22), and sires are plotted by the mean month during which their litters were collected. The "U" shaped curve shows the curvilinear regression (Rf=0.11), while the inverse curve indicates the mirror-image curve used to correct the sires' breeding values to fit a linear regression with a slope of zero. frequency of exencephaly (mean litter Freeman-Tukey arcsine) • • • 9 • • • • • • • • • • • • * p • • • • • • I 3 1 I 2 i 3 4 5 6 7 i i 8 9 1 1 1 I 1 1 1 1 1 ! 1 1 1 1011 12 13141516171819 20 21 22 11/92 02/93 05/93 08/93 11/93 02/94 05/94 08/94 156 was used. It was rinsed in 2% PBS (150 mM NaCI, 2 mM Na/K P0 4 pH 7.3, 5 mM KC1), chopped into approximately 1 mm pieces with a fresh scalpel blade, then added to 5 ml of lysis buffer (100 mM NaCI, 10 mM Tris-HCl, pH 8, 25 mM EDTA, pH 8, 1% SDS) containing 100-300 ug/ml of proteinase K in a 15 ml centrifuge tube. The tube was placed in a 50°C water bath for 5-20 hours. Phenol-chloroform extractions were performed, followed by ethanol precipitation of the DNA (Sambrook et al., 1989). The DNA was rinsed with 70% ethanol, then resuspended in tris-EDTA, pH 8.0 (TE8; Sambrook et al., 1989). Whole embryo DNA was prepared in the same way, using 3-5 ml of digestion buffer containing 200 u.g/ml of proteinase K. During the phenol-chloroform extractions of embryo preparations, it was difficult to keep the DNA-containing fraction pure as it was very viscous and tended to pull some of the lower layer along with it. The DNA obtained was therefore not very pure but was of sufficient quality for use in PCR. The final concentrations of DNA were measured on 1/110 dilutions (in water) of each sample in an LKB UV-Spectrophotometer. Fractions of the samples were diluted in water to 100ng/ul, for use in PCR. Tail tip DNA was prepared from 1-1.5 cm of tail tip, chopped into small pieces and placed into 0.5 ml of lysis buffer (as above) containing 200-300 ug/ml of proteinase K in a 1.5 ml eppendorf tube. The tubes were placed in a 50°C water bath for 5-20 hours, and phenol-chloroform extractions and ethanol precipitation of the DNA performed. The DNA was resuspended in 500 u.1 of TE8 and diluted in water to 100ng/u.l for use in PCR. Simple sequence length polymorphisms (SSLPs) were used as genetic markers to map the exencephaly-liability loci. SSLPs were typed using PCR with mouse "MapPairs" primers obtained from Research Genetics Inc. (Huntsville, Alabama, USA). Each PCR reaction was carried out in a 25 u,l volume overlaid with mineral oil in a 650 ul reaction tube. Each reaction contained 100-150 ng of target DNA and 0.14 uM of each (forward and reverse) primer. The 157 rest of the reaction mixture was provided by a "master mix" consisting of dATP, dGTP, dCTP and dTTP (final concentration, 50 U.M each; Perkin-Elmer or Pharmacia), Taq DNA polymerase (final concentration, 0.625 Units; Perkin-Elmer (AmpliTaq) or Gibco-BRL), 1 OX PCR buffer (final concentration, 10 mM Tris-HCl, pH8.3, 50 mM KC1; Perkin-Elmer or Gibco-BRL), and magnesium chloride (final concentration usually 1.5 mM, range 1.5-3.0 mM; Perkin-Elmer or Gibco-BRL). PCR was performed in a Perkin-Elmer 4600 thermocycler, usually under the following conditions: 4.5 min. at 94°C (denaturation), followed by 30 cycles of 1 min. at 94°C (denaturation), 1 min. at 55°C (annealing) and 1 min. at 72°C (extension), followed by 7 min. at 72°C. Some primers required different annealing temperatures, ranging from 50-68°C; when 68° was used, the 30 PCR cycles were changed to 1 min. at 94° and 2 min. at 68°C. One marker, D16Mit9, required 4% DMSO in the PCR mixture for adequate results, and D13Mitl0 required a "hot-start", where the DNA and primers were denatured for several minutes at 94°C before adding master mix and continuing as above. The marker dye bromphenol blue-xylene cyanol FF (5 u.1) was added to the PCR product and one third of the mixture (10 u.1) was usually run electrophoretically on 4% NuSieve (4% NuSieve (FMC Bioproducts), 1% agarose) horizontal gels containing 0.5 ug/ml of ethidium bromide. Gels were run in 1XTAE (Sambrook et al., 1989) at about 130 V for 2-5 hours, then observed and photographed (Polaroid 667 film) under UV light. The products for a few markers ran fairly close together on NuSieve gels and were better visualized on 3.5-5% MetaPhor (FMC Bioproducts) gels, run and visualized as above. The resolution of small DNA fragments (under 300 bp) by NuSieve and MetaPhor gels is similar to that of acrylamide gels, but the agarose gels are generally faster and easier to run. 158 Data analysis Inheritance pattern at marker loci: Analysis of the marker allele information for linkage or association with phenotype involved several approaches. The first was a simple examination at each marker locus of the total number of SELHA and LM/Bc alleles in the high-risk and low-risk F2 sires. A marker linked to an exencephaly-liability locus is expected to have an excess of SELHA (S) alleles in the high-risk sires and an excess of LM/Bc (L) alleles in low-risk sires. The total number of S alleles at each marker was plotted for high-risk and for low-risk F2 sires along each chromosome. A Mendelian segregation pattern was expected for any locus not linked to an exencephaly-liability locus, with half (10 of 20) of the alleles within each group (high- and low-risk F2 sires) being S and the rest, L. The predicted pattern on a chromosome carrying an exencephaly-liability locus was one where the pooled high-risk sires would have a total of more than 50% (10) of their alleles from SELHA (S) and, simultaneously, the low-risk sires have inherited a total of less than 50% of their alleles from SELHA (S). X 2 Analysis: For markers linked to exencephaly-liability loci, the "predicted allele" is "S" in high-risk sires and "L" in low-risk sires. The alternative allele in each is the "unpredicted allele". By coding the alleles this way, the data for high- and low-risk sires were combined. Deviation of the observed ratio of predicted and unpredicted alleles in the combined high- and low-risk F2 sires from the expected 1:1 ratio based on random segregation can be tested for significance by X 2 tests of goodness of fit (Sokal and Rohlf, 1981). The advantage of this approach is that the data from the high- and low-risk F2 sire groups can be combined into one value and its statistical significance assessed. The amount of deviation from random segregation is 159 represented graphically in Figures 28-30 by plotting the X 2 value at each marker locus against the position of the marker on the chromosome (in cM from the centromere). As the X 2 test does not differentiate between deviation in the direction predicted for linkage (e.g., an excess of SELHA alleles in high-risk sires) or the opposite direction (e.g., an excess of LM/Bc alleles in high-risk sires), the X 2 values for markers at which deviation is in the predicted direction are plotted above the zero line on the Y-axis, whereas the values for markers at which deviation is in the opposite direction are plotted below the zero line. As multiple X 2 tests were performed, there is an increased risk of obtaining a false positive result. However, as markers on a chromosome are linked, the X 2 tests were not all independent of each other. Thus, instead of using the number of markers typed (109) as the number of independent tests, the number of linkage groups tested was used. As the size of the mouse genome excluding the X chromosome is approximately 1500 cM (Davisson and Roderick, 1989), and one linkage group consists of 50 cM, approximately 30 linkage groups were tested. Under these conditions (30 X 2 tests with one degree of freedom each), the critical value of X 2 at P=0.05 (using Sidak's multiplicative inequality; Rohlf and Sokal, 1981) is 9.839. Combination approach: As exencephaly in SELHA mice shows best fit to an additive, multigenic model of inheritance, the combined genotype at all liability loci for low-risk sires should have a high proportion of LM/Bc ("L") alleles, while the combined genotype at all liability loci for high-risk sires should have a high proportion of SELHA ("S") alleles. Thus, at markers linked to liability loci, the F2 sires should also carry a majority of their alleles from the expected parental strain. The genotypes of each high-risk and low-risk sire at the best marker (based on X 2 analysis) on the two chromosomes most likely to carry liability loci (10 and 13) were plotted 160 against each other. It was of interest to observe whether there was a non-random combination of genotypes at the two loci. In addition, the most favourable genotypes (i.e., in high-risk sires, the most S alleles) of the two best (adjacent) markers on these chromosomes were also plotted against each other. The combined genotypes of the best markers on chromosomes 10 and 13 were then plotted together (5 genotypic classes, with 0-4 S alleles) against each of the other chromosomes that had X 2 values suggestive of a linked liability locus. This approach was also applied to the data from exencephalic F2 embryos. Exencephalic F2 embryos: The exencephalic F2 embryos were typed only for markers on chromosomes that had been indicated as most likely to carry exencephaly-liability loci in the analysis of the high- and low-risk F2 sire data. The analysis of the patterns of segregation of SELHA and LM/Bc alleles at chromosome 2, 10, and 13 markers in the exencephalic embryos involved the same methods as for the F2 sire data (see above). While all the same markers on these chromosomes were typed in order to present the data by the same method used for the F2 sires, the focus of interest in the exencephalic F2 data set was the markers with the best X 2 values in the F2 sire data. As the exencephalic F2 animals are an independent data set, and as only three chromosomal regions were being analysed, the X 2 analysis of the exencephalic F2 data was not corrected for multiple tests. Testing for Mendelian segregation: As the analysis of the linkage data for exencephalic F2 embryos relied on detection of deviation from Mendelian segregation patterns, it was necessary to verify that there was no detectable segregation distortion in any of the regions examined (i.e., Chromosomes 2, 10, and 161 13) using X 2 tests of goodness of fit to the expected 1:1 ratio. For the chromosome 2 marker, it was necessary to type a second group of embryos, and the information from the two data sets was combined using Fisher's technique for combining probabilities from independent tests of significance (Sokal and Rohlf, 1981). Statistical Methods X 2 tests of goodness of fit and the Freeman-Tukey arcsine transformation were described in Chapter 2 (General Materials and Methods, "Statistical Methods"). Sidak's multiplicative inequality (Rohlf and Sokal, 1981) was used to correct for multiple independent X 2 tests. This approach generates a relatively conservative (R. Dreger, personal communication) critical value at the 5% level of significance for a given number of independent tests, k. If the results of a series of separate significance tests on different sets of data that test the same scientific hypothesis are suggestive, but no one of them is statistically significant, their probabilities (P) may be combined using Fisher's technique for combining probabilities from independent tests of significance (-2ElnP; Sokal and Rohlf, 1981). If the null hypothesis (i.e., that the observed and expected values are the same) is true, this value should be distributed as a X 2 with 2k degrees of freedom, where k is the number of probabilities being combined. Results SSLP markers: The map positions of the 109 SSLP markers for which F2 animals were typed to look for linkage to exencephaly-liability loci are shown in Fig. 24. The map distances given between markers were taken from the Research Genetics Murine MapPairs April 1995 Information 162 Fig. 24: Map of D_Mit_ SSLP markers used for linkage analysis. 1 # 231 1170 4-18 -7 -92 ••132 -83 -94 -102 ••149 -17 11 # 2 -20 -4 -38 +14 10 2 + 80 + 143 148 12 ••37 ••136 +33 5 +99 19 3 4 Te2 I101 46 -22 -28 -133 4-113 t16 -107 ±19 +14 -91 -27 -58 123 13 Nds2 + 24 95 168 ±122 6 • 8 10 13 -3 -117 -10 -39 -193 -76 -78 14 -Nds1 -54 -5 41 77 15 t +5 29 -34 -35 83 -16 -102 -10 201 -75 -79 -62 -68 109 16 ••9 +110 60 +19 17 •19 -10 -88 39 •124 -4 •89 -46 -30 •4 A A -139 -14 •11 •81 •137 18 -20 -36 -40 -4 19 • -68 • -41 •-5 --10 33 158 70 180 164 i — i 7.2 cM 163 Release. The distances in all figures are given in cM from the most proximal marker listed in the Research Genetics Inc. April 1995 release, and the end of each chromosome was taken to be at the most distal marker given in that listing; the distances between the most proximal marker and the centromere and between the most distal marker and the telomere are not well established, making it difficult to generate an SSLP map that is consistent with distances on the physical map. Although some markers are listed on the chromosome consensus maps given in the 1994 Mammalian Genome Supplement which would give the relative distance to the ends of the chromosomes, many of the distances between markers on these maps do not agree with the distances given by Research Genetics. This makes it difficult to combine the information of these two sets of maps. Between the spring of 1992 and April 1995, the number of mapped mouse SSLP markers grew from several hundred (Love et al., 1990; Aitman et al., 1991; Cornall et al. 1991; Hearne et al., 1991; Dietrich et al., 1992) to 6,183 (Research Genetics "Murine MapPairs" April 1995 information release). The selection of SSLP markers to be screened for differences ("informativeness") between the SELHA and LM/Bc mouse strains began in June 1993 and was complete in the spring of 1995. During this time, not only did the number of available SSLP markers grow, but their map positions generally became more certain. In addition to being mapped relative to one another, several SSLP markers were also mapped relative to genes on the consensus linkage map. In the Research Genetics information releases, there are three kinds of markers, based on the certainty of their map positions: "framework" markers, where the odds ratio in favour of its given position is 1000:1; markers for which the odds ratio in favour of its position is between 100 and 1000:1; and markers for which the odds ratio in favour of its position is less than 100:1. Framework markers were chosen except for regions in which there were no informative framework markers. In all but one of these regions, markers with an odds 164 ratio of 100-1000:1 in support of their map position were selected; the map position of marker D16Mit60 was supported by an odds ratio of less than 100:1 but was included as it was the only informative marker within a 20 cM region at the time. All but one marker (D10Mitl80) was chosen by October 1994, based on the Research Genetics July 1994 release. The certainty of the map positions of some markers changed between that and the April 1995 release, and the certainty of the location of some markers was changed from having an odds ratio of more than 100:1 to having an odds ratio of less than 100:1. Based on the April 1995 information release, the map positions of 5 markers used in this study are supported by an odds ratio of less than 100:1. These markers are D2Mitl43 (which was a framework marker on the July 1994 map), D4Mit91, D4Mitl6, D6MitlO, and D16Mit60 (see Fig. 24 for map positions). If the map positions of these markers change as they are verified in the future, it is possible that they will leave gaps of more than 20cM between markers. This could affect the linkage analysis for the exencephaly-liability loci. For this reason, the genotypes of the 20 F2 sires at marker loci are given in Appendix A. It is important to note that the genotypes at most of these markers are consistent with the genotypes at the adjacent markers. For example, if one adjacent marker is SS and the other is LL, the marker in question is SL, or if the adjacent markers are LL and SL, the marker in question is either LL or SL. This suggests that while their map position may not be precise, it is most likely that they do lie between the given flanking markers. In the April 1995 release, the map position of D16Mit60 is still only supported by an odds ratio of less than 100:1. For this marker, there is a double recombination event between it and one of its flanking markers (D16Mitl9) in an F2 sire, but the stated distance between these two markers is 14.8 cM. There are also 2 single recombination events between D16Mit60 and its other flanking marker, D16Mitll0, although they are only thought to be 1.5 cM apart. D16Mit60 was originally typed because D16Mitll0 165 and D16Mitl9 were more than 20 cM apart on the July 1994 Genetics Research Inc. map, but as this distance is now given as 16.3 cM, D16Mit60 is no longer a crucial marker. High-risk and low-risk F2 sire data: The markers near an exencephaly-liability locus are expected to carry a large proportion of SELHA alleles in the high-risk F2 sires and a large proportion of LM/Bc alleles in the low-risk sires. A pattern compatible with this expectation was observed for portions of chromosomes 3, 5, 7, 10, 11, 13, and 17 (Figs. 25-27). The deviation between high- and low-risk sires in number of SELHA alleles was most pronounced on chromosomes 10 and 13. This was also evident in the X 2 analysis (Figs. 28-30). The two most distal markers on chromosome 10 were the only ones with X 2 values over the threshold level of 9.8 (D10Mitl80, X 2= 10.0, and D10Mitl64, X 2= 12.1), and one marker on mid-chromosome 13 had a maximum value just below the threshold (D13Mit39, X 2= 8.1). The next highest X 2 values (4.9) were on chromosomes 3 (at D3Mit22), 5 (at D5Mitl68), and 17 (at D17MitlO). From the first comparison of the number of SELHA alleles inherited in the grouped high-and low-risk F2 sires, chromosome 2 displayed a pattern opposite to that expected: the high-risk sires inherited more LM/Bc alleles and the low-risk sires inherited many of their alleles from SELHA (Fig. 25). This suggested the possibility that SELHA mice have a gene on Chromosome 2 that partially suppresses or compensates for the effect of the exencephaly-liability loci or that LM/Bc mice have a weak exencephaly-liability locus that is not strong enough by itself to cause exencephaly but increases the liability when combined with the SELHA liability-loci. The maximum X 2 value for the combined high- and low-risk F2 sire deviations on chromosome 2 was 4.9. At all chromosome 2 markers (Fig. 28) the deviation of the ratio of "predicted" and "unpredicted" parental alleles from the expected 1:1 ratio was in the opposite direction to that 166 Fig. 25: Summed number of SELHA alleles at markers on chromosomes 1-6 in 10 high-risk F2 sires (squares) and 10 low-risk F2 sires (circles). The expected number, based on Mendelian segregation patterns, is indicated by the solid line at 10 (50%) SELHA alleles. (/> = as < X _1 LU CO 20 16 12 10 8 Chromosome 1 0 10 20 30 40 50 60 70 80 90 100 cM 0 10 20 30 40 50 60 70 80 cM 8 a> "co < I _i LU CO 20 16 12-10 8-Chromosome 4 0 10 20 30 40 50 60 cM 0 10 20 30 40 50 60 70 cM 201 0 10 20 30 40 50 60 70 80 cM 10 20 30 40 50 cM 1 Fig. 26: Summed number of SELHA alleles at markers on chromosomes 7-12 in 10 high-risk F2 sires (squares) and 10 low-risk F2 sires (circles). The expected number, based on Mendelian segregation patterns, is indicated by a solid line at 10 (50%) SELHA alleles. 20 16 W 0) I 12 1 10 _i LU 8 CO Chromosome 7 20 16 12 10 8 Chromosome 8 10 20 30 40 50 60 cM 0 10 20 30 40 50 60 70 cM (A a> a> 75 < I _i LU CO 20 16 12 10 8 Chromosome 9 0 10 20 30 40 50 60 cM 0 10 20 30 40 50 60 70 cM 20 16 </> a> cu i l 12 ^ 10 LU 8 CO Chromosome 11 20-16-12 10 8 Chromosome 12 0 10 20 30 40 50 60 70 80 cM 10 20 30 40 50 cM 20 16 12 10 8 Chromosome 14 0 10 20 30 40 50 60 cM 0 10 20 30 40 50 60 cM 201 16 12 10 8-4 0 Chromosome 15 20 16 12 10| 8 Chromosome 16 10 20 30 40 50 cM 10 20 30 40 50 cM 20 * 1 6 75 12 £ 10 LU 0 + 10 20 30 40 cM Chromosome 19 20 16 12 10 8 Chromosome 18 10 20 30 cM 10 20 30 40 50 cM Fig. 27: Summed number of SELHA alleles at markers on chromosomes 13-19 in 10 high-risk F2 sires (squares) and 10 low-risk F2 sires (circles). The expected number, based on Mendelian segregation patterns, is indicated by the solid line at 10 (50%) SELHA alleles. 169 0 10 20 30 40 50 60 70 80 90 100CM 0 10 20 30 40 50 60 70 80 cM 20 30 70 80 CM 15 10 5 0 5 10 15 Chromosome 6 10 20 30 40 50 cM Fig. 28: X 2 values for deviation of the observed ratio of parental alleles at markers on chromosomes 1-6 from the expected 1:1 ratio in high- and low-risk sires. The values plotted above zero deviate in the expected direction, and those below zero deviate in the opposite direction. The X 2 threshold value for significance is indicated by solid lines at 9.8. 170 15 10-5" (/) CD =J 0 CO > CM X 5-10-15 Chromosome 11 10 20 30 40 50 60 70 80 CM 15 10 5 0 5 10 15 Chromosome 12 10 20 30 40 50 cM Fig. 29: X 2 values for deviation of the observed ratio of parental alleles at markers on chromosomes 7-12 from the expected 1:1 ratio in high- and low-risk sires. The values plotted above zero deviate in the expected direction, and those below zero deviate in the opposite direction. The X 2 threshold value for significance is indicated by solid lines at 9.8. 171 CO >; to CD CO > CO CD co CO CD _3 CO 15 10 5 0 5 10 15 15 10 5 0 51 10 -15 _ ( 15-10-5 -0 5-10-15 10 20 30 40 50 60 cM 10 20 60 cM Chromosome 15 10 20 30 40 50 cM Chromosome 17 10 20 30 40 cM Chromosome 19 15 i 10 5 0 5 10 15 Chromosome 16 10 20 30 40 50 cM 15 10 5 0 5 10 15 Chromosome 18 10 20 30 cM 10 20 30 40 50 cM Fig. 30: xf values for deviation of the observed ratio of parental alleles at markers on chromosomes 13-19 from the expected 1:1 ratio in high- and low-risk sires. The values plotted above zero deviate in the expected direction, and those below zero deviate in the opposite direction. The X 2threshold value for significance is indicated by solid lines at 9.8. 172 expected near a liability locus with "liability" alleles in SELHA. A pattern opposite to expected was also seen on Chromosome 12 (Fig. 29), but it was less striking. The maximum X 2 value at D12Mit37, the most proximal marker, was also 4.9. As exencephaly in SELHA embryos appears to be caused by 2-3 loci with additive action (Chapter 4), it is appropriate to consider combinations of potential liability-loci, not just single chromosomal regions in isolation. When the genotypes at D10Mitl64 and D13Mit39 were considered jointly (Fig. 31), all high-risk F2 sires had at least half their alleles from SELHA and all the low-risk sires had at least half their alleles from LM/Bc. If two loci are associated with exencephaly, it is expected that a "poor" genotype (i.e., LM/Bc alleles in a high-risk sire) at one locus would be compensated for by a "good" genotype (i.e., SELHA alleles in a high-risk sire) at the other. As the loci examined here were preselected, it is not surprising that each group of sires has inherited at least half of their alleles from the expected parental strain. Two of the high-risk sires, including sire 317 which ranked first in exencephaly production, were heterozygous at both D10Mitl64 and D13Mit39, as were two of the low-risk F2 sires. This observation, that two sires from each risk group carry equivalent genotypes on chromosomes 10 and 13, suggests that an additional locus elsewhere in the genome also contributes to exencephaly-liability. It is more likely that a liability locus lies between two markers than at a marker. The genotype at the best pair of adjacent markers on chromosomes 10 and 13 were therefore examined, and the marker with the genotype most closely resembling that predicted for a marker closely linked to a liability locus (i.e., the marker with more S alleles in high-risk sires) was selected as being most likely to represent genotype at the putative liability locus. The best pair of markers on chromosome 10 were D10Mitl80 and D10Mitl64, and the two best pair of markers on chromosome 13 were D13Mitl0 and D13Mit39. The selected chromosome 10 and Fig. 31: Genotype at D1 OMitl 64 plotted against genotype at D13Mit39 for high-risk (stars) and low-risk (triangles) F2 sires, showing the combined genotypes for these two markers. S S SL LL LL SL S S Genotype at D1 OMitl 64 174 13 genotypes from high- and low-risk sires were then examined jointly. The results of this analysis (not shown) were the same as the results for markers D10Mitl64 and D13Mit39 alone; the flanking markers did not contribute additional information. Again, this suggests that an additional liability locus is present elsewhere in the genome. The genotypes at the best chromosome 3, 5, and 17 markers were examined in concert with the combined genotype of D10Mitl64 and D13Mit39 (Figs. 32-34). They are expected to show some clustering of alleles since the loci are preselected. If a locus on any of these additional chromosomes compensates for a deficiency in the number of predicted parental-type alleles on chromosomes 10 and 13, then the 6 high-risk F2 sires that carry 2/4 SELHA alleles at D10Mitl64 plus D13Mit39 should have more of the expected parental-type alleles on the additional chromosome than the other high-risk sires. Although the combined genotype of some of the sires with 2/4 SELHA alleles at D10Mitl64 and D13Mit39 was improved when genotype on chromosome 3, 5 or 17 was also considered, most of the sires were not shifted toward having more than half of their alleles from the predicted parental strain (Figs. 32-34). In addition, a number of high- and low-risk sires continued to carry equivalent genotypes. Thus, this approach provides no additional evidence that any of these chromosomes (3, 5, or 17) carries a liability locus. Considering further the possibility that chromosome 2 carries a liability locus with liability alleles in the LM/Bc strain, the joint genotype at D10Mitl64, D13Mit39, and D2Mit7 (Fig. 35a) in high- and low-risk sires was examined. The observation that a number of high-and low-risk sires (three of each) with 2/4 "S" alleles at D10Mitl64 and D13Mit39 shared a common genotype at D2Mit7 suggests that genotype at D2Mit7 does not strongly affect exencephaly-liability. As the X 2 value at D2Mit92 was the same as that at D2Mit7, genotype at D2Mit92 was also examined for compensation of genotype at D l OMitl 64 and D13Mit39 (Fig. 175 Fig. 32: Genotype at D3Mit22 plotted against number of SELHA alleles at D10Mit164 and D13Mit39 combined, showing the combined genotypes for high-risk (stars) and low-risk (triangles) F2 sires. CM CM CO Q CO CD Q . >. -i—' O c CD CD SS S SL & LL 0 1 2 3 4 # SELHA alleles at D10Mit164 + D13Mit39 Fig. 33: Genotype at D5Mit168 plotted against number of SELHA alleles at D10Mit164 and D13Mit39 combined, showing the combined genotypes for high-risk (stars) and low-risk (triangles) F2 sires. co CD £ SS L O Q * SL CO CD Q. O c CD CD % LL •k 0 1 2 3 4 # SELHA alleles at D10Mit164 + D13Mit39 176 Fig. 34: D17Mit10 genotype plotted against number of SELHA alleles at D10Mit164 and D13Mit39 combined to show combined genotypes for high-risk (stars) and low-risk (triangles) F2 sires. £ S S SL € LL (0 CD CL oc CD CD 0 1 2 3 4 # SELHA alleles at D10Mit164 + D13Mit39 177 Fig. 35. Genotype at a: D2Mit7, or b: D2Mit92, plotted against the number of SELHA alleles at D1 OMitl 64 and D13Mit39 combined to show the combined genotypes for high-risk (stars) and low-risk (triangles) F2 sires. a.-. A "fl* * * # A A r * i 1 1 1 1 : 0 1 2 3 4 # SELHA alleles at D1 OMitl 64 + D13Mit39 CM Q •4—' CO CD C L >s -i—• O c CD C D LL SL S S CVJ C O S LL CVJ Q CO CD Q _ - t — • o c CD O ) SL ss 0 1 2 3 4 # SELHA alleles at D1 OMitl 64 + D13Mit39 178 35b). Fewer high- and low-risk sires (one of each) had overlapping genotypes at these three loci. While this is more consistent with the existence of a compensatory locus on chromosome 2, it also suggests that an additional locus is involved in exencephaly-liability. The PCR products of the high- and low-risk F2 sires at the best SSLP markers (D10Mitl64, D13Mit39, and D2Mit7) are illustrated in Fig. 36. Exencephalic F2 embryos: The analysis of the inheritance patterns at SSLP markers in high- and low-risk F2 sires suggested the most likely locations of liability loci to be on chromosomes 10 and 13. The data suggested that there might also be a protective locus on chromosome 2 in SELHA mice. In order to test the predictions derived from the high- and low-risk F2 sire data, chromosome 2, 10 and 13 markers were typed in 31 exencephalic F2 embryos. Under the additive threshold model, the risk of being exencephalic increases as more SELHA alleles are present at liability loci, but many exencephalic F2 embryos will not be homozygous for SELHA alleles at all of the contributing loci (see Fig. 37). Some exencephalic embryos may, in fact, be heterozygous at all the liability loci. When taken as a group, however, the majority of their alleles on chromosomes 10 and 13 should come from SELHA if these regions contain genes involved in causing exencephaly. If these chromosomes do not contain exencephaly-liability loci, exencephalic F2 embryos should show Mendelian segregation ratios (50% SELHA and 50% LM/Bc alleles) for markers on these chromosomes. In addition, if the hypothesis of a protective locus on chromosome 2 in SELHA is true, exencephalics should carry mostly LM/Bc alleles at markers near this gene. The number of SELHA alleles in exencephalic F2 embryos at the chromosome 2, 10, and 13 markers is shown in Fig. 38. Photographs of the PCR products for D10Mitl64, D13Mit39, and D2Mit7 are shown in Fig. 39. 179 a. D2Mit7 Additional c. D13Mi t39 High-risk F2's Additional r-High-risk F2's Fig. 36. Photographs of the P C R products of the 13 high-risk and 10 low-risk F2 sires, and parental strain and F1 animals, for a : D2Mit7, b: D1 OMitl 80 and c : D13Mit39. P C R was executed under standard conditions (see Materials and Methods) and the products were visualized by electrophoretic separation on 4% NuSieve agarose gels and staining with ethidium bromide. The origin of the gels is oriented toward the bottom of the page (i.e., smaller DNA fragments are closer to the top of the page). 180 CD o) CD zB ** 5 CO T3 CO RA CD cu .!= > co to c •£ •- a> to — 0 to CD < 15 < LU X co —I 0) LU £ "ft 2 o — CD CD W > c CD CD E'.ir > O CO O W CD X -° E i s 5 . 8 o o TJ a> ra u _ c = Q .TJ _Q CD T3 CD O CO — c _ J (D £ ra <u d •5 <P = CD ra E " 3 ra «= £ § $ I* *l .9- c J= CD CO O) c o •B * ra CD 2jE ra -£ = 0) CD SZ O) * ; O o >-CL CL >> CO •C CO CD O M- CO O CD ro £ g.E Sis i - ra Q. sz CD CL «- CD ra g JZ X D_ CD ra M— . o tz cn 'o LI .E 0 o = tz ra CD x o CD C CD = XJ ra CL CD U c X CD ra ._ CD U XJ co ° o a> ^  ra 15 2? <^ I LU ra GO — $ s-« CD O O tZ CQ CD _ , t J) tD ra £ c ° S o co ra co 5! » T J - 0) r CO — = 4_ CD o = < ™ x c < _| a) X LU cn _i c/) P LU Q. CO » O CO CD 'o c CD £ 2 Q) co x: co H <D > W tZ CM 0 LL. cn O v p Q- co o ^ To CD CD —I X LU » CO = 1 1 0 ra o 5s T3 CO 2 Q, C L ^ CD O «= — g H c o c CD CD O CD CO CD x: c CO ra TJ CD •4-* O CD St W . _ . _ o> CO CO o o ra . O ± 5 CO CO CD CD ra sz u XI ra C CO C CO CO O a> o ® CO ra £ < CD LU • -CO TJ i - CD CD O C TJ 3 2 CD X) O CO I I c c CD CD cn o) o o CO Q . CO g t CD ra x CD >« ^ 2 o3 ° -X CO CD CO o ° b S I * ! 00 i S8JJS uojijodojd E *vZ a. a. o CD 181 Fig. 38: Number of alleles inherited from the SELHA parental strain on chromosomes 2,10 and 13 in 31 exencephalic F2 embryos. The solid line at 31 indicates the expected number of alleles (50%) based on Mendelian segregation patterns. 60 "cc < X _l LU CO to Q) _0J "(C X _l LU CO to 0) "CC LU CO 50 40 30 20 10 0 60 50 40 30 20 10 0 60 50 40 30 20 10 0 Chromosome 2 0 10 20 30 40 50 60 70 80 cM Chromosome 10 0 10 20 30 40 50 60 70 cM Chromosome 13 0 10 20 30 40 50 60 cM 182 a. D2Mit7 Fig. 39. Photographs of the P C R products of the 31 exencephalic F2 embryos for a: D2Mit7, b: D1 OMitl 80 and c: D13Mit39. P C R was executed under standard conditions (see Materials and Methods) and the products were visualized by electrophoretic separation on 4% NuSieve agarose gels and staining with ethidium bromide. The origin of the gels is oriented toward the bottom of the page (i.e., smaller DNA fragments are closer to the top of the page). 183 In F2 sires, the chromosome 13 marker with the highest X 2 value was D13Mit39. In the exencephalic F2 embryos, the X 2 value at this marker, 14.5 (Fig. 40), was highly significant (P= 0.00014). However, the chromosome 13 marker with the most divergent pattern in exencephalic embryos was D13MitlO (Figs. 38 and 40). On chromosome 10, D10Mitl64 had 32 SELHA alleles and 30 L M alleles, giving a non-significant X 2 of 0.1 (P=0.75; Figs. 38 and 40). On chromosome 2, two markers (D2Mit7 and D2Mit92) had the same highest X 2 value in F2 sires. In exencephalic F2 embryos, the X 2 value for deviation of the observed pattern of parental-type alleles from the expected 1:1 ratio at D2Mit92 was significant (4.1; P=0.043), while that at D2Mit7 (1.6; P=0.21) was not (Figs. 38 and 40). Thus, the data from exencephalic F2 embryos supported the hypotheses that there are exencephaly-liability loci on chromosomes 13 and 2, but not that there is an additional liability locus on chromosome 10. Exencephalic F2 embryos are expected to carry mostly SELHA alleles at markers near exencephaly-liability loci (Fig. 37). However, a number of exencephalic F2 embryos are expected to carry only carry half of their total number of liability locus alleles from SELHA as this genotypic class of animals is much larger than those in which animals carry more S alleles. Of the 31 exencephalic F2 embryos typed for genetic markers, 6 had fewer than half of their alleles at D10Mitl64 and D13Mit39 from SELHA (Fig. 41a). Alternatively, as a liability locus is more likely to lie between two markers, the genotype with the most SELHA alleles at either D10Mitl64 or D10Mitl80 was plotted against the genotype with the most SELHA alleles at either D13Mit39 or D13Mitl0 for each of the exencephalic embryos. Over half (18/31) of embryos carried at least 3/4 alleles from SELHA, but 4 had only 1/4 S alleles (Fig. 41b). These observations support the hypothesis that there must be an additional liability locus (or loci) elsewhere in the genome. In addition, the one S allele in exencephalic embryos carrying 1/4 S alleles at the chromosome 10 and 13 candidate regions was always on chromosome 13, 184 Fig. 40. X 2 values for deviation of the observed ratio of parental alleles at chromosome 2, 10 and 13 markers in exencephalic F2 embryos from the expected 1:1 ratio. The values for markers at which the pattern of deviation was in the expected direction are plotted above zero and the values at markers in which deviation was in the opposite direction are plotted below zero. The )6 threshold value for significance is indicated by solid lines at 9.8. 20 co CD CC > CM X CO CD cc > CM X co CD CO > 10 0 10 20 20 10 0 10 20 Chromosome 2 1 1 1 1 1 1 1 r 0 10 20 30 40 50 60 70 80 c M Chromosome 10 10 20 30 40 50 60 70 c M Chromosome 13 50 60 c M 185 Fig. 41. Genotype of 31 exencephalic F2 embryos at chromosome 10 and 13 markers plotted against one another to show combined genotype, a: Genotype at D1 OMitl 64 vs genotype at D13Mit39; b: most "favourable" genotype (most SELHA alleles) at D1 OMitl 64 or D1 OMitl 80 vs most "favourable" genotype at D13Mit39 or D13Mit10. CD CO C O CO CD CL >> •4—" o c 0 CD S S SL LL LL SL S S genotype at D1 OMitl 64 CO CD CO CL T — O v . c o CD CO po •4—• •*-> CO ^ CD ^ n co £ S S SL LL dddk LL SL S S best genotype at D1 OMitl 64 or D1 OMitl 80 186 suggesting the effect of the putative liability locus on chromosome 13 might be stronger than the putative locus on chromosome 10, that there may not be a liability locus on chromosome 10, or that the chromosome 10 locus is important for transmission of a high exencephaly-liability by F2 sires but not for being affected oneself. In order to examine whether genotype at the potential suppressor or compensator locus on chromosome 2 compensates for genotype at the putative liability loci on chromosomes 10 and 13 in exencephalic F2 embryos, the combined genotype of D10Mitl64 and D13Mit39 was plotted against genotype at D2Mit7 (Fig. 42a) or D2Mit92 (Fig. 42b) for the exencephalic F2 embryo data. If there is compensation, exencephalic embryos with fewer SELHA alleles at D l OMitl 64 and D13Mit39 are expected to carry more LM/Bc alleles at the putative chromosome 2 locus. One of the six exencephalic F2 embryo that carried only 1/4 S alleles at D l OMitl 64 and D13Mit39 was SS at D2Mit7 (Fig. 42a). At D2Mit92, however, all embryos with fewer than 4/4 S alleles at D10Mitl64 and D13Mit39 were either heterozygous or homozygous LL (Fig. 42b). All of the exencephalic embryos that were homozygous LL at D l OMitl 64 carried at least one L allele at D2Mit92, but only two were LL (see Table 7 and Appendix B for individual genotypes). These observations are consistent with those expected if D2Mit92 is linked to a locus at which S alleles compensate for the exencephaly-liability conferred by S alleles at other loci. As a gene involved in liability to exencephaly is more likely to lie between two markers than at a marker locus, the genotype with the most S alleles of each pair of best chromosome 10 (D10Mitl64 andD10Mitl80) and 13 (D13Mit39 andD13Mitl0) markers was plotted against the genotype with the most L alleles at D2Mit7 or D2Mit92 (Fig. 43). If genotype at a chromosome 2 locus compensates for genotype at chromosome 10 and 13 loci, animals with fewer SELHA alleles at the chromosome 10 and 13 loci are expected to have more LM/Bc 187 Fig. 42. Genotype at a: D2Mit7, or b: D2Mit92, plotted against the number of SELHA alleles at D1 OMitl 64 and D13Mit39 combined to show the combined genotypes for exencephalic F2 embryos. CM Q -*—• co CD Q . O c CD LL SL SS 0 1 2 3 4 # SELHA alleles at D1 OMitl 64 + D13Mit39 CVJ CD LL CM Q 2 SL co CD CL O c CD CD S S 0 1 2 3 4 # SELHA alleles at D1 OMitl 64 + D13Mit39 188 Fig. 43: Most "favourable" genotype (most LM/Bc alleles) at D2Mit7 or D2Mit92 plotted against the most "favourable" genotypes (most SELHA alleles) at D1 OMitl 64 or D1 OMitl 80 and D13Mit39 or D13Mit10, for 31 exencephalic F2 embryos. LL SL o §> ss 0 # S E L H A alleles from the best genotypes of D1 OMitl 64 or D1 OMitl 80 + D13Mit39 orD13Mit10 189 alleles at the chromosome 2 locus. Only one embryo was an SS homozygote at the putative chromosome 2 locus (Fig. 43), and it carried 4/4 S alleles at the putative chromosome 10 and 13 loci (i.e., was genotypically SELHA-like). All the other exencephalic embryos carried at least one L allele at the chromosome 2 locus. This is consistent with the hypothesis that there is a liability locus on chromosome 2 at which LM/Bc mice carry liability alleles. Testing for Mendelian segregation: As the analytical methods used to look for association between phenotype (liability to exencephaly) and inheritance of specific parental alleles at marker loci assume Mendelian segregation of the data and test for deviation from the expected segregation patterns, it was necessary to verify that there is Mendelian segregation of these markers in F2 animals in general. The overall segregation pattern at the pooled 109 SSLP markers typed in high- and low-risk F2 sires fit Mendelian expectation: animals were SS homozygotes at a total of 542 markers, SL heterozygotes at 1099 markers, and LL homozygotes at 539 markers (X 2 goodness of fit to 1:2:1 ratio; 0.975>P>0.90), and 2183 (50.1%) of the 4360 alleles typed were from SELHA. To check for Mendelian segregation at the loci of interest in F2 embryos, six litters of F2 embryos (including any exencephalics) were typed for D2Mit92, D10Mitl80 and D13Mitl39. A total of 81 embryos were typed for D2Mit92 and D13Mit39; only 79 embryos were typed for D10Mitl80 as PCR product was difficult to obtain for 2 embryos from different litters for this marker. The segregation patterns at Dl OMitl 80 (20 SS, 38 SL, 21 LL) and D13Mit39 (20 SS, 47 SL, 14 LL) did not differ from the 1:2:1 ratio expected (X 2; P= 0.93 for D10Mitl80 and P= 0.23 for D13Mit39). The pattern at D2Mit92 (14 SS, 39 SL, 27 LL) was also not different from that expected (X 2; P= 0.12), but as the probability was close to the 10% level and the deviation was in the same direction as that seen in exencephalic embryos (a deficiency of 190 SELHA alleles and an excess of LM/Bc alleles), it was thought best to test a second data set of 6 more litters (75 embryos). The segregation of alleles in this data set (15 SS, 41 SL, 19 LL) did not differ significantly from that expected under Mendelian segregation (X 2; P= 0.58). Combining the tests on the two independent data sets gives a value that is compared to a X 2 with 4 degrees of freedom and is not significant (P= 0.26), indicating that segregation at this marker locus does follow a Mendelian pattern. Re-ranking of high-risk F2 sires: After the analysis of the data from high-risk and low-risk F2 sires and exencephalic F2 embryos was complete, potential effects that would change the membership of the high-risk group of F2 sires were explored. The first of these was the unit of measure of liability. When the raw percentage of exencephaly is used, it does not account for potential litter effects (i.e., the potential lack of independence of exencephaly risk within litters). Mean-litter Freeman-Tukey arcsine values correct for litter effects as long as the distribution of litter size is similar among sires (if a sire consistently produces small litters, he will rank higher by this approach). The second way in which the ranking of the high-risk sires might change is if there were temporal variation affecting penetrance of exencephaly. There appeared to be a mild time trend in the frequencies of exencephaly produced by SELHA sires over the 22 months during which they were testcrossed to SELHR dams: the frequencies of exencephaly that they produced were highest at the beginning this period (fall of 1992), dipped in the spring and summer of 1993, and increased again slightly afterward (Fig. 20, p. 153). Any temporal variation in the penetrance of exencephaly might affect the ranking of the high-risk F2 sires, as intermediate-risk sires testcrossed during a period when penetrance was particularly high (i.e., at the beginning of the study) might produce higher frequencies of exencephaly than high-risk sires testcrossed during 191 a period when penetrance was lower. If temporal variation does exist, "correction" of the F2 sires' breeding values to account for this might change the membership of the group of ten high-risk F2 sires, which could in turn affect the results of the linkage analysis. The overall frequencies of exencephaly produced by F2 sires also showed a mild temporal fluctuation, where the breeding values of sires testcrossed in the late-summer and early fall of 1993 were lower than during other times (Fig. 21, p. 153). The fit of the F2 sire data to linear regression is poor (Fig. 21; R2=0.02) and the slope (-0.08) is not significantly different from zero (ts, P>0.9). The pattern of the sires' breeding values over time (Fig. 21, p. 153) suggested that a polynomial regression line (parabola) might be more appropriate, and did show better fit to the data (R2=0.11; PO.001), using the formula Y = 15.27 - 0.75x + 0.03x2 (Fig. 22, p. 154). The inverse parabola (Y = 5.90 + 0.75x - 0.03x2) was plotted (Fig. 23, p. 154), and its values subtracted from the values for the best fitting curve at each value of x, resulting in a straight line with a slope of zero at Y=10.58. The ranking of the 21 F2 sires that produced the highest frequencies of exencephaly were "corrected" for the possible temporal fluctuation in mean breeding values by subtracting the difference of the two parabolas from the F2 sire data points. This changed their ranking relative to that based on the direct frequencies (%) of exencephaly in their testcross embryos (Table 6). Three F2 sires (307, 306 and 310) dropped from the group of ten highest exencephaly producing sires, while sires 684, 863, and 868 moved into the top ten from original rankings of 11th, 12th, and 16th, respectively. It is interesting that the ranking of the top ten sires based on percentage of exencephalic progeny also differs from the ranking based on uncorrected mean-litter Freeman-Tukey arcsine values, with sire #310 moving from 10th (based on percentage) to 16th and sire #863 moving from 11th (based on percentage) to 10th in the ranking. Mean-litter Freeman-Tukey arcsine values correct for litter effects on frequency of exencephaly, and the difference in ranking using 192 Table 6: Ranking of high-risk F2 sires. The F2 sires that produced the 21 highest frequencies (%) exencephaly re-ranked by mean-litter and corrected (for temporal fluctuations in frequency) mean-litter Freeman-Tukey arcsine frequencies of exencephaly. In each case, the ranking of the top ten sires changes. Sire percentage (rank) mean-litter Freeman-Tukey arcsine (rank) corrected mean litter Freeman-Tukey arcsine (rank) 317 15.5(1) 24.11 (1) 21.40(1) 911 10.2 (2) 19.60 (3) 18.33 (3) 930 10.2 (2) 19.51 (4) 17.82 (4) 388 9.7 (4) 19.71 (2) 18.80 (2) 943 8.2 (5) 17.58 (7) 15.41 (9) 307 8.0 (6) 16.96 (9) 13.65(15) 323 8.0 (6) 17.98 (6) 15.81 (8) 320 7.7 (8) 18.10(5) 15.93 (5) 306 7.6 (9) 17.55 (8) 13.58 (16) 310 6.7 (10) 15.19(16) 14.82 (12) 863 6.7 (11) 16.74(10) 15.83 (7) 684 6.3(12) 16.23(11) 15.86 (6) 308 6.1 (13) 15.50(15) 12.19(20) 910 5.6(14) 15.98(14) 14.71 (13) 529 5.5(15) 14.40(19) 14.03(14) 868 5.5(16) 16.22(12) 15.31 (10) 382 5.4(17) 14.90(18) 13.21 (19) 858 4.9(18) 14.16 (20) 13.55 (17) 945 4.5(19) 13.01 (24) 10.84 (21) 918 4.4 (20) 16.13(13) 13.42(18) 642 4.0(21) 15.00(17) 14.93(11) 193 the uncorrected transformed values suggests that the distribution of exencephalics among litters varies for many of the high ranking sires (they did not produce dramatically different litter sizes). Sires with "extreme" litters, in which many embryos are affected rank higher based on percentage of exencephaly but lower based on mean-litter Freeman-Tukey arcsine values. After the re-ranking, F2 sires 684, 863 and 868 were typed for the chromosome 2, 10 and 13 SSLP markers (see Appendix A). The marker genotypes for the revised top ten high-risk F2 sire group on each of these chromosomes showed a similar pattern to that seen for the original group of high-risk sires: a slight excess of LM/Bc alleles on chromosome 2, and an excess of SELHA alleles on chromosomes 10 and 13 (Figs. 36 and 44). The number of SELHA alleles on chromosome 10 in the re-ranked high-risk sires did not appear to vary from expected (Mendelian segregation) by as much as the original group of high-risk sires (a maximum of 13 SELHA alleles in the former versus 16 in the latter). X 2 analysis of the deviation of the observed ratio of predicted parental-type alleles from that expected if there is random segregation was similar to that seen when the original ranking of high-risk sires was used (Fig. 45). The greatest deviation on chromosome 2 was at D2Mit7 (X2= 6.4), but it was not significant. The greatest deviation on chromosome 10 was at D l OMitl 64, but in contrast to the value obtained using the original ranking of the high-risk F2 sire group, it was suggestive but not significant in this data set (X2= 6.4). Finally, on chromosome 13, the greatest deviation was at both D13Mit39 and D13Mitl93 (X2= 8.1); the same value was observed at D13Mit39 in the data using the original ranking of high-risk sires. Again, this value is suggestive but not significant. The combined genotype at D10Mitl64 and D13Mit39 in the revised group of high-risk F2 sires (Fig. 46) did not support the presence of strong compensation for a "poor" genotype at one locus by a "good" genotype at the other. A number of the re-ranked high-risk sires had 194 Fig. 44. Summed number of SELHA alleles at markers on chromosomes 2,10 and 13 in top 10 ranking F2 sires after correction of breeding values for temporal fluctuations in frequencies of exencephaly (squares). Values for the 10 low-risk sires are also shown (circles). Chromosome 2 2 0 , ° 0 10 20 30 40 50 60 70 80 CM Chromosome 10 201 1 OH 1 1 1 1 1 1 r1 0 10 20 30 40 50 60 70 CM Chromosome 13 20 T 1 0-I , 1 1 , , , 1 0 10 20 30 40 50 60 CM 195 Fig. 45. X 2 values for deviation of the observed ratio of parental alleles at chromosome 2, 10 and 13 markers from the expected 1:1 ratio in re-ranked high-risk F2 sires and low-risk F2 sires. The values for markers at which the pattern of deviation was in the expected direction are plotted above zero and the values at markers at which deviation was in the opposite direction are plotted below zero. The X 2 threshold value for significance is indicated by solid lines at 9.8. 20" co CD CO > CM X co CD CO > CO CD _3 CO > CNJ X 10 0 10 20 20 10 0 10 Ghromosome 2 1 1 1 1 1 1 1 r 0 10 20 30 40 50 60 70 80 cM Chromosome 10 70 cM 60 cM Fig. 46: Genotypes at D1 OMitl 64 and D13Mit39 plotted against each other for the re-ranked high-risk F2 sires. S S S L LL LL S L genotype at D1 OMitl 64 S S 197 genotypes that overlap with those seen in low-risk sires (compare Fig. 46 to Fig. 32). The three candidate chromosomal regions were also examined together (Figs. 47 and 48). As with the original high-risk sire group, some of the sires in the re-ranked group had genotypes that overlap with those seen in low-risk sires (compare Figs. 47 and 48 to Fig. 35). These observations are consistent with those made on the original high-risk sire group and also suggest the existence of an additional liability locus elsewhere in the genome. The results of re-ranking the high-risk F2 sire group to correct for potential temporal fluctuations in exencephaly-production are very similar to those obtained for chromosomes 2, 10, and 13 in the original data set. The re-ranking of these sires does not significantly alter the support for the candidate chromosomal regions suggested by the original analysis, and also suggests that an additional, unidentified locus must be involved in exencephaly-liability. Discussion The results of this study strongly suggest that one of the loci involved in the exencephaly-liability in the SELHA mouse strain lies on the top half of chromosome 13, most likely between markers D13Mitl0 and D13Mit39 (see Fig. 49). There is inconsistent support for an additional liability locus on distal chromosome 10, near the marker D l OMitl 64 (see Fig. 50). However, since breeding values are a more accurate measure of phenotype for a multilocus threshold trait than are direct phenotypes (affected or normal), the data from exencephalic F2 embryos should not be taken strongly to discount distal chromosome 10 as a candidate region. The putative chromosome 13 locus appears to have a stronger effect than the putative chromosome 10 locus because some high-risk F2 sires and exencephalic embryos that have only one SELHA allele among the four alleles in both candidate regions always have the susceptibility allele on chromosome 13. 198 Fig. 47: Genotype at D2Mit7 plotted against the number of S E L H A alleles at D1 OMitl 64 + D13Mit39 combined, showing the overall genotype for re-ranked high-risk F2 sires. r- LL OJ Q S SL CD CL f ss-cn 0 1 ¥ 3 4 # S E L H A alleles at D1 OMitl 64 + D13Mit39 " T " Fig. 48: Genotype at D2Mit92 plotted against the number of S E L H A alleles at D1 OMitl 64 + D13Mit39 combined, showing the overall genotype for re-ranked high-risk F2 sires. CM LL Co L L -OJ ^ S L CD CL tss c CD CD 0 1 2 3 4 # S E L H A alleles at D1 OMitl 64 + D13Mit39 199 There may also be a locus that compensates for exencephaly-liability in SELHA embryos on chromosome 2 near markers D2Mit7 and D2Mit92 (see Fig. 51). If this compensatory locus does exist, it may act in a recessive manner as animals with a high liability to exencephaly tend to be either heterozygous (SL) or homozygous LL at D2Mit7 and D2Mit92. As genotype in the chromosome 2, 10, and 13 candidate regions (given in Table 7) does not fully account for the exencephaly-liability observed in F2 sires and exencephalic F2 embryos, there must be at least one additional liability locus elsewhere in the genome. The data analysis in F2 sires suggests that this locus probably lies on chromosome 3 (near D3Mit22), chromosome 5 (near D5Mitl68) or chromosome 17 (near D17MitlO). The candidate regions for the SELHA exencephaly-liability loci do not contain loci that are known to cause neural tube closure defects, such as p53 (chromosome 11), macs (proximal chromosome 10), ApoB (chromosome 12), Csk (chromosome 9), or those mutated in splotch (Pax3; chromosome 1), curly tail (chromosome 4, with modifier loci on distal chromosome 3 and mid-chromosome 5), loop tail (chromosome 1), extra toes (GH3; proximal chromosome 13) or crooked (chromosome 6) animals. This is not particularly surprising as all of these genes, with the possible exception of the curly tail locus, cause neural tube closure defects as single gene mutations. Moreover, several of these genes cause spina bifida or additional developmental defects, while the NTD in SELHA mice appears to be limited to cranial neural tube closure. Identification of the genes responsible for exencephaly in SELHA mice should, therefore, identify loci that were not previously known to be important in neural tube closure. No other loci have previously been identified that cause NTD in an additive, multigenic manner. The NTD in curly tail mice appears to be influenced by more than one locus, but the curly tail gene is a major locus whose expression is affected by modifier loci on other chromosomes. As most human NTD appear to be multifactorial in etiology and there is no Table 7. Genotype, of all exencephaly-liability loci. F2 animals typed, at the markers determined S= SELHA allele, L= LM/Bc allele. most likely 200 to carry D13Mit39 D10Mit164 D2Mit7 D13Mit39 D1 OMitl 64 D2Mit7 High-risk sires: Exencephalic F2 embryos: 317 SL SL SL 1 SS LL SL 911 ss SL SS 2 SS SS SS 930 ss SL SL 3 SL SL SL 388 SL SS LL 4 SS SL LL 943 LL SS SS 5 SS SL SL 307 SS SL SL 6 SL SL LL 323 SL SS SL 7 SS SS SL 320 SS SS SL 8 SL SL SL' 306 SL SL SL 9 SS SL SL 310 SL SS LL 10 SL SS SL Additional 11 SS SL SL High-risk sires: 863 SS LL LL 12 SL SS SL 684 SL LL LL 13 SL LL SL 868 SL SL SL 14 SL SL LL Low-risk F2 15 SS LL LL sires: 321 LL LL SS 16 SL SL LL 385 LL LL SS 17 SS SL LL 549 LL SL SS 18 SL LL LL 598 LL SL SS 19 SS SL LL 614 SS LL SL 20 SL LL SL 841 LL LL SS 21 SL SS SS 855 LL SS SS 22 SS SS SL 869 LL LL SS 23 SS LL SL 908 SL SL SL 24 SS SS SS 944 SL LL SL 25 SL LL LL 26 SS SS SS 27 LL SL SS 28 SS SL LL 29 SS SS SS 30 SL SL SL 31 SL LL LL 201 evidence for a major liability locus, identification of the SELHA exencephaly-liability loci will suggest candidate loci for NTD susceptibility in humans. Mapping these loci in mice is a first step in this process, and will suggest candidate regions that can be examined for linkage to human NTD-liability based on the homology between the mouse and human genomes. The strongest candidate region appears to lie between D13Mitl0 and D13Mit39, on chromosome 13 (Fig. 49), although it is possible that the putative liability locus lies near but outside of the region flanked by these markers. The region between D13MU10 and D13Mit39 shows homology to human chromosomes 5q and 9q, while the regions immediately adjacent show homology to chromosomes 5 and 6. The homeobox-containing gene Msxl (Hox8) lies 2 cM distal to D13Mitl0 and 4 cM proximal to D13Mit39. This gene encodes a transcription factor that is expressed in the dorsolateral aspects of the cranial neural folds in the neuroepithelium and mesenchyme in the region of NCC emigration (MacKenzie et al., 1992; Liu et al., 1994; Davidson, 1995). A single base pair substitution in the homeodomain of the human MSX2 gene was found in individuals affected with Boston-type craniosynostosis, an autosomal dominant disorder involving premature fusion of the brain sutures (Jabs et al., 1993). Mice carrying the same mutation in a transgene expressed in the developing skull had premature closure of brain sutures and a small patch of ectopic bone above the sagittal suture (Lui et al., 1995). An abstract by Potts and Sadler (1995) reports that injection of antisense oligonucleotides to Msxl and Msx2 into early somite cultured mouse embryos resulted in "abnormal patterning" of the neural tube, although exactly what this means was not explained. It is conceivable that different mutations in this gene might not be dominant and/or might act at other stages of craniofacial development to cause a different phenotype. Msxl is a good candidate gene for an exencephaly-liability locus in SELHA mice because it is expressed in D8-9 embryos in the same region of the cranial neural folds (MacKenzie et al., 1992; Liu et al., 1994) that appeared 202 Fig. 49: Map of chromosome 13, showing the position of S S L P markers, and the candidate region relative to the position of loci that could be involved in neural tube closure. The regions of homology to human chromosomes are also shown. regions of homology in the human genome - 1q 7p | 5q • 9q Xt ch Tpmt Lamb1-2 Msx2 Ntrk2 Dhfr -10 cM D13Mit3 D13Mit117 | D13Mit10 D13Mit39 l I I D13Mit193 • Candidate region D13Mit76 D13Mit78 203 abnormal in histological sections of 3-11 somite SELH embryos (see Chapter 3). Tpmt (thiopurine methyltransferase) also maps to the region of interest on chromosome 13 (at 27 cM). As S-adenosyl-methionine can be the methyl donor for reactions involving Tpmt (Otterness and Weinshilboum, 1987), this enzyme may be involved in methionine metabolism and therefore also affect folate metabolism (Finkelstein, 1990; Scott et al., 1994), both of which are thought to be important in normal neural tube closure (Smithells et al., 1981; Smithells et al., 1983; Coelho etal., 1989; Coelho and Klein, 1990; MRC, 1991; Essien, 1992; Nosel and Klein, 1992; Czeizel and Dudas, 1994; Scott et al., 1994; Mills et al., 1995). Also within the region of interest are Lambl-2, which encodes a laminin 131-chain homolog and Ntrk2, which encodes neurotrophic tyrosine kinase receptor, type 2. Although GU3 (which appears to be mutated in Xt mice), and Dhfr (dihydrofolate reductase) both map to chromosome 13, they lie outside the candidate region: the former is 21 cM proximal to D13Mitl0 and the latter is 14 cM distal to D13Mit39. Dhfr is closer to the marker D13Mitl93, for which the evidence of linkage to a liability locus was lower than that for D13Mitl0 and D13Mit39, except in the analysis of the F2 sire data using the re-ranked high-risk sires. The gene ch (congenital hydrocephalus) maps 12 cM proximal to D13Mitl0 and above D13Mitll7, for which the evidence of linkage was poorer than for D13Mitl0 and D13Mit39. Hydrocephalus is rarely (if ever) seen in SELH mice, although it is possible that different mutations in the same gene could cause exencephaly but not hydrocephalus. As the map position of the SSLP markers relative to the map of genes on any chromosome is likely to be somewhat inaccurate, D13Mitl0 and D13Mit39 might actually lie closer to Dhfr or ch than is given by the chromosome 13 consensus map (Mammalian Genome 1994 Supplement; Justice and Stephenson, 1994), but it seems unlikely that they would be mismapped by more than 12 cM. Based on the map information currently available, Msx2 appears to be the best candidate 204 locus for an exencephaly-liability locus. It is also possible that the exencephaly-causing gene in SELHA mice has not yet been identified or has been identified but its map position is not yet known. The evidence for a candidate region on chromosome 10 is strong in the high- and low-risk F2 sire data, but it is not supported by the exencephalic F2 embryo data. While it is possible that the result in the F2 sires is due to statistical chance, other possibilities should also be considered. For example, if there is an exencephaly-liability locus on chromosome 10, the different results of these two data sets might reflect their biological difference. It is possible that exencephalic F2 embryos have such a high genetic risk to exencephaly as a result of their genotype at other liability loci that their genotype at the putative chromosome 10 is less relevant than in the high-risk sires, who may carry fewer liability alleles at other loci and are therefore not affected. The transmitted genotype at the putative chromosome 10 locus could, however, affect the risk of exencephaly in the progeny of these high-risk sires. An alternative hypothesis is that the liability-locus on chromosome 13 has a stronger effect than the other locus or loci. Under this hypothesis, exencephalic F2's carry a sufficient majority of their chromosome 13 alleles from SELHA (see Appendix A) that they may not need as many SELHA alleles on chromosome 10. In fact, four exencephalic embryos carried only one SELHA allele in the candidate regions (looking at both flanking markers) on chromosomes 10 and 13, and this one SELHA allele was always on chromosome 13. This suggests that the putative chromosome 13 locus may have a stronger effect on phenotype than the putative chromosome 10 locus. High-risk F2 sires have fewer SELHA alleles on chromosome 13 and so they may require more SELHA alleles on chromosome 10 to put them at high risk to having exencephalic offspring, while low-risk sires would be expected to carry mostly LM/Bc alleles at all liability loci. SELHA alleles on chromosome 13 having a stronger effect than SELHA alleles on chromosome 205 10 might also explain why animals with more SELHA alleles on chromosome 13 are generally affected (i.e., exencephalic). A new data set is required to investigate this question as the current data sets were used to develop this hypothesis and therefore cannot be used to test it. The candidate region for the putative chromosome 10 locus (Fig. 50) is most likely to lie near the most distal marker typed, D l OMitl 64. The candidate region shows homology to human chromosome 12, and the region immediately distal shows homology to human chromosome 19q. The only potential candidate genes that appear to lie within this region are Gli (at 67 cM) and Steel (SI, at 55 cM). Gli is related to GH3 (mutated in Xt mice). It is thought to encode a transcription factor and is expressed in the ventral neural plate paraxial cephalic mesenchyme (Hui et al., 1994). SI encodes a haematopoietic growth factor that is the ligand for the c-kit receptor tyrosine kinase, and SI is expressed in midline floorplate cells of the prosencephalon of D9.5 mouse embryos (Matsui et al., 1990). The macs gene (21 cM), which encodes a protein kinase C substrate, produces exencephaly in 25% of mice homozygous for a null allele, also maps to chromosome 10 but is located much further proximal of the candidate region. It is also possible that the putative liability locus on this chromosome encodes a gene that either has not yet been identified or is unmapped. It may seem odd to suggest the existence of an exencephaly-suppressor locus in SELHA, a strain with a high liability to spontaneous exencephaly, but it should be noted that all SELHA embryos close their cranial neural tube in an abnormal manner, but only about 18% of embryos fail to complete this process. SELHA mice might be expected to have a more robust Closure 3 mechanism than average, which would enable it to compensate for lack of Closure 2 in most embryos. A more robust Closure 3 could either begin earlier, giving it more time to complete fusion of the prosencephalon and mesencephalon, or proceed more quickly than Closure 3 in other strains. As the SELH/Bc mouse stock was developed by selecting for animals that 206 Fig. 50: Map of chromosome 10, showing the position of S S L P markers, and the candidate region relative to the position of loci that could be involved in neural tube closure. The regions of homology to human chromosomes are also shown. regions of homology in the human genome -10 cM 6q a 10q 22q 21q 19p 12 • - 19q l 1 2 Macs SI Gli D10IW7 D10Mit3 D1 OMitl 58 D10Mit42 D10IVM66 D10MK70 D1 OMitl 80 D1 OMitl 64 Candidate region 207 produced both exencephaly (a lethal defect) and good weaned litter size, it is conceivable that such a suppressor locus would have been selected along with the exencephaly-liability loci because without the suppressor most or all SELH embryos might have been exencephalic. There does appear to be some normal genetic variation in the timing of Closure 3; it occurs before Closure 2 in Jcl:ICR (Sakai, 1989) and CD-I (Waterman, 1976) embryos, while its initiation appears to be delayed in ICR/Be embryos (Juriloff et al., 1991). If SELHA embryos do carry an exencephaly-suppressor locus that acts on the timing or progress of Closure 3, such a gene would be expected to be expressed in the rostral prosencephalon of embryos during neural fold elevation and fusion in this region (D8.5-9 of gestation). The candidate region for this potential suppressor locus on chromosome 2 (Fig. 51) lies near D2Mit92 (at 41 cM from the centromere), and probably between it and D2Mit7 (at 28 cM). The region around these two markers shows homology to regions of human chromosomes 9q, 2q, 11, and possibly 15q. Of the genes that map within this region, Notchl (formerly Tan-1 and Motch, at 16 cM), Dlxl and Dlx2 (at 46 cM), Crapbll (at 53 cM), and Pax-6 (at 58 cM) are expressed in the prosencephalon of D8.5-9 embryos. Other loci in this region that might be expressed in the head during early development are: Bmil (at 12 cM), a proto-oncogene which, when "knocked-out" by a targeted deletion, causes ataxia and decreased cell densities in the cerebellum, as well as other abnormalities (van der Lugt et al., 1994); Rxra (at 17 cM); Spna2 (brain alpha spectrin-2; at 17 cM); Pbx3 (a proto-oncogene homeobox gene; at 23 cM); Scnla, Scn2a, Scn3a (brain sodium channel alpha-subunits; at 36 cM); and Scn7a (at 38 cM). Confirmation that exencephaly-liability loci lie on chromosomes 13, 10 and 2, and further refinement of their map positions will involve the establishment of reciprocal congenic strains (D.M. Juriloff, personal communication). The plan is to breed the regions containing the putative liability loci from SELHA onto the LM/Bc strain background to determine whether they 208 Fig. 51: Map of chromosome 2, showing the position of S S L P markers, and the candidate region relative to the position of loci that could be involved in neural tube closure. The regions of homology to human chromosomes are also shown. regions of homology in the human genome 10p 2q 9q 11 15 2 Bmi1 Notch 1 Spna2, Rxra Pbx3 Scnla, Sen2a, Scn3a Scn7a DM, Dlx2 Crapb2 un (Pax1) 20 -10 cM D2Mit80 D2Mit7 D2Mit92 Candidate region D2Mit133 D2Mit107 D2Mit143 D2Mit148 209 affect the liability to exencephaly. The genetic model suggested by the mapping data is slightly different from that suggested by the genetic analysis of exencephaly-liability in the cross between SELHA and LM/Bc. While the latter supported an additive model involving 2-3 liability loci in the SELHA strain, the mapping study suggests the hypothesis that at least 2 liability loci and one suppressor locus may affect the incidence of exencephaly in SELHA embryos. LM/Bc alleles would be expected to act like liability alleles, but only when at least some SELHA alleles are present at liability loci elsewhere in the genome. However, F l animals carry one LM/Bc allele on chromosome 2 and are heterozygous at the putative chromosome 10 and 13 loci, but have a very low incidence of exencephaly (0.3% of embryos). This suggests that other loci may be involved in exencephaly liability in the cross between SELHA and LM/Bc. If all F2 sires were to be typed for genetic markers, it would be difficult to tell for many of them (i.e., all the intermediate-risk sires) what the expected underlying genotype at the exencephaly-liability loci would be. By selecting only the top and bottom 10% of sires with respect to exencephaly-production, this problem is circumvented. Based on the 95% confidence intervals around the breeding values, a high-risk sire should not be misclassified as a low-risk sire, nor a low-risk sire as a high-risk one. The animals whose genotype can most easily be inferred from their phenotype contribute the most linkage information (Lander and Botstein, 1989). Therefore, the search of the genome in this study looked for a correlation between genotype at marker loci and breeding values of the "extreme" (high- and low-risk) F2 sires, rather than with phenotype in exencephalic embryos. Another way in which using the high- and low-risk F2 sires is more powerful than the use of exencephalic F2 embryos is that segregation at marker loci can be compared to Mendelian expectation in both the high- and low-risk sires; two deviations, in opposite directions, are 210 required at the same time, and deviation from expectation will therefore be less likely to occur by chance. At a marker near a liability locus, not only are the high-risk sires expected to inherit mostly SELHA alleles, but the low-risk sires are expected to carry mostly LM/Bc alleles. The exencephalic embryos cannot be compared to normal embryos as the latter group will include embryos with a high genetic risk to exencephaly that succeeded in closing their neural tubes (like approximately 80% of SELHA embryos). The exencephalic embryos provided a means of testing the candidate regions selected from the analysis of the high- and low-risk F2 sires. If the putative liability locus on chromosome 10 is real, it would have been missed by only examining exencephalic F2 embryos, while if will not be surprising if chromosome 10 does not carry a liability locus, as this was suggested by the analysis of the exencephalic F2 data. If only the exencephalic F2 embryos had been typed, there would not have been any means to check for false results (positive or negative) without collecting an additional data set. Finally, it should be remembered that it was not known going into the study whether any F2 embryos would be exencephalic. The re-ranking of the high-risk sires to correct for the apparent temporal fluctuation in the frequencies of exencephaly they produced provides an alternate method of examining the data. It is interesting that the ranking of sires based on (uncorrected) mean-litter Freeman-Tukey arcsine frequencies is also different from the ranking based on percentage exencephaly (Table 6), although not as dramatically as the corrected values. However, even the substitution of three of the high-risk sires, based on their re-ranking, does not change the analysis for chromosomes 2, 10, and 13 appreciably: the results still suggest the existence of loci involved in liability to exencephaly on these chromosomes, and the odds ratios in support of linkage on these chromosomes remain higher than those for any of the other chromosomes in the original analysis. This suggests that sires classified as high-risk do carry many liability alleles even if 211 the ranking is inaccurate to some degree. The ten sires typed may only be a sample of 10 of the 15-20 F2 sires that actually carry the highest liabilities to exencephaly, but inexact ranking of the sires is not a major concern with respect to the power of detecting linked liability loci. The approaches used to identify candidate regions most likely to be involved in exencephaly-liability fall into two categories. The first is similar to the traditional "single locus search" method of analysis for linkage analysis of complex traits (Soller et al., 1976; Tanksley et al., 1982; Edwards et al., 1987; Dupuis et al., 1995). In a single locus search, one locus is looked for at a time, but it is assumed that multiple loci will be detected if each has a strong enough effect and/or the sample size is sufficiently large (Dupuis et al., 1995). The pattern of inheritance at marker loci in high- and low-risk F2 sires was used to identify regions that are most likely to be involved in causing exencephaly in SELHA mice by examining the genotype at SSLP markers covering the autosomes and testing for significant deviation of the ratio of predicted and unpredicted genotypes (based on breeding value-phenotype) from the expected 1:1 ratio by X 2 tests. The "combination" approach used to look at the combined genotype of more than one potential candidate region at a time is similar to the "conditional search" described by Dupuis et al. (1995), where analysis of the same data set in which a preliminary analysis (such as a single locus search) has been performed and has detected linkage to some loci is continued to search for linkage to additional loci. In this study, the preliminary search identified regions on chromosomes 13 and 10 that are likely to be linked to exencephaly-liability loci, and other regions on chromosomes 2, 3, 5, and 17 for which there was marginal evidence of involvement. If one locus has a lesser effect than the other(s), it would be possible, for example, for most animals to carry fewer SELHA alleles at this locus but more SELHA alleles at the stronger locus or loci and still have a high genetic liability to exencephaly. In this case, the weaker locus 212 would not show as "favourable" a genotype as the stronger loci, but it would be expected to carry more SELHA alleles in high-risk sires that carry fewer SELHA alleles at the stronger loci. If the putative chromosome 13 and 10 loci are considered to have the strongest effects based on their higher probabilities of carrying liability alleles, then by looking at their combined genotype along with the genotype of the best marker on each of the other chromosomes, it should be possible to identify weaker liability loci based on whether the genotype at any of these markers compensates for a less favourable genotype relative to breeding value on chromosomes 13 and 10 in high- and low-risk F2 sires. One problem with the "conditional search" method is that the exact location of the trait loci identified by the single locus search (here, the putative loci on chromosomes 13 and 10) can only be estimated, based on the probability of linkage to the marker loci. In other words, D13Mit39 and D10Mitl64 were selected as the markers that best represent the genotype at the putative linked liability loci, but unless these markers lie within the liability loci, there will be recombination events between the liability loci and these markers. A recombination event could result in a sire carrying a less favourable genotype (i.e., a high-risk sire carrying more LM/Bc alleles) at the liability locus than at the marker. A third liability locus with a weaker effect would need to carry a more favourable genotype (i.e., more SELHA alleles) in this sire than would be required if the chromosome 13 and 10 genotype were more favourable. A recombination event could also result in a sire carrying a more favourable genotype on chromosomes 13 and 10 than is represented by his genotype at the marker loci. This sire would not require as favourable a genotype at a third, weaker locus to compensate as would be expected based on marker genotype. In order to decrease the effect that recombination between a marker and liability locus would have on the analysis, the combination approach applied to the data in this study also used the most favourable genotype from the best adjacent pair of markers 213 in addition to the genotype at the best chromosome 13 and 10 markers. The use of a conditional search has little value when there is epistasis between different loci. For example, if the putative exencephaly-suppressor locus on chromosome 2 acts in a recessive manner, then being heterozygous at this locus would have the same phenotypic effect (in embryos) as being homozygous for LM/Bc alleles but this is not reflected in the combination approach used. Animals heterozygous at the chromosome 2 marker would need to be plotted with the LL homozygous animals to reflect gene action, and carrying at least one LM/Bc allele at this locus would be sufficient to increase the liability of an animal carrying at least some SELHA alleles at the putative chromosome 13 and 10 loci. Application of the "combination" approach to these data suggested that, if chromosome 13, 10, and 2 do all carry exencephaly-liability loci, they are unlikely to be the only loci involved as genotype in the candidate regions on these chromosomes do not sufficiently account for the phenotype of all F2 sires and embryos. It also suggested that one liability locus on either chromosome 3, 5, or 17 is not sufficient in combination with the putative chromosome 13 and 10 loci to account for the phenotype of all the animals included in this study. Although it is widely available, the "QTL-Mapmaker" computer program, developed by E.S. Lander and D. Botstein (see Lander and Botstein, 1989), was not used to analyze these data because it uses a parametric approach and these data do not meet its assumptions. The QTL-Mapmaker program appears to calculate maximum-likelihood estimates based on LOD scores that involve variances, assuming normal distribution of the data (Lander and Botstein, 1989). As was shown in attempting to use Wright's formula to determine how many exencephaly-liability loci are segregating in the cross between SELHA and LM/Bc mice (Chapter 4), the variances of low-frequency threshold traits such as exencephaly in SELHA mice are truncated. Any calculations involving these variances will therefore be inaccurate. For example, Wright's 214 formula estimated 25 exencephaly-liability loci when the mean parental variance was used, while the data support a much lower number. The approaches that were applied to these data to look for linkage between genetic markers and liability loci do not rely on the data being normally distributed and were, in fact, much simpler than the non-parametric methods being proposed as alternatives to the original QTL-Mapmaker program (for example, Kruglyak and Lander, 1995). Nevertheless, the methods used here were still powerful enough to detect linkage to at least one exencephaly-liability locus. The methods of analysis developed here have been used to indicate the probability of linkage between an exencephaly-liability locus and a genetic marker. These methods have primarily been used as a qualitative means of identifying the chromosomal regions that are most likely to carry liability loci. Further analysis of these data is planned as additional methods are developed (Diana Juriloff, personal communication). The QTL-Mapmaker computer program detects linkage when the "LOD" score exceeds a predetermined threshold value that is based on the genome size and marker spacing (Lander and Botstein, 1989). The X 2 values calculated in this study are also compared to a threshold value and are used to indicate the regions with the strongest probability of carrying liability loci. The regions with the highest values were examined by additional methods to further study their potential contribution to exencephaly-liability. The chromosome for which there was the strongest support for a linked liability locus in the original high- and low-risk F2 sire data analysis was chromosome 10. Although subsequent analysis of exencephalic F2 embryo data did not appear to support this conclusion, the use of an alternate method of analysis would not have changed the strength of the original F2 sire data set. In other words, chromosome 10 should have been the strongest candidate region, even if QTL-Mapmaker or other methods had been used to analyze the data. In addition, because the liability alleles in the chromosome 2 215 region are in the normal strain, it is likely that the putative compensatory locus on chromosome 2 would have been missed had the marker allele information simply been entered into a computer program. The combination of approaches used in this study also provides more insight into the mode of gene action than a simple indication of probable map position would give. Using this information, it was possible to propose a genetic model for the liability to exencephaly in SELHA mice that can be tested in future studies. 216 Chapter 6: L A C K OF CLOSURE 2 CAUSES EXENCEPHALY IN SELH/Bc MICE Introduction It is thought that the failure of the mesencephalic neural folds to elevate and initiate contact and fusion at the site of Closure 2 and the high liability to exencephaly in SELH mice both reflect the same underlying defect in neural fold mechanics. It has not been possible previously to reject the alternative hypothesis, that the absence of Closure 2 and the presence of a high incidence of exencephaly in SELH embryos and newborns occur together in this strain by chance. As a part of the genetic analysis of exencephaly in SELH/Bc mice and the mapping of the liability genes that differ between the SELH and LM/Bc strains, a large F2 population was generated and 102 F2 sires were testcrossed to SELHR dams in order to determine their individual genetic liability to exencephaly, based on the frequency of exencephaly observed in their progeny. F2 sires with a high liability to exencephaly (i.e., those that produced high frequencies of exencephaly in testcross progeny) are expected to inherit most or all of their liability-loci alleles from the SELH strain while low-risk F2 sires (those that produced no exencephalic testcross progeny) are expected to inherit most or all of their liability-locus alleles from the LM/Bc strain. This panel of F2 sires provides a powerful means of looking for a relationship (or lack thereof) between lack of Closure 2 and liability to exencephaly, using genetic correlation. This analytical method is based on the fact that if two traits are caused by the same genes, they will always be transmitted together. If they are caused by different genes, they will be transmitted in all possible combinations. If absence of Closure 2 and high liability to exencephaly are caused by the same genes, then F2 sires that produced high frequencies of exencephaly in testcrosses to SELHR dams should also produce testcross embryos that lack Closure 2; F2 sires that produced no 217 exencephalic testcross progeny are expected to produce testcross embryos with a normal pattern of neural tube closure (including Closure 2). This study has been published (Gunn et al., 1995) and much of its presentation here, particularly the results section, has been taken from the published version. I collected the embryos and scored them for stage of cranial neural tube closure. Diana Juriloff, Muriel Harris and I were all involved in analysing the data and writing the published manuscript. Materials and Methods Mouse maintenance and breeding All mice were maintained and timed pregnancies were obtained as described in Chapter 2 (General Materials and Methods). Embryos for scoring pattern of cranial neural tube closure were collected between D8/20h and D9/llh. They were collected intact in their deciduae into Bouin's fixative and transferred to 70% ethanol after at least 24 hours in fixative. Embryos were dissected out of their decidua under 70% ethanol and scored for somite count and stage of cranial neural fold elevation and/or closure. The classification of stages of closure generated by K.B. Macdonald (Macdonald et al., 1989) were followed. These stages, also generally described in the discussion of neurulation in the introduction of this thesis (p. 3-8), are as follows: folds evident: the cranial neural folds are evident as a symmetrical pair of biconvex bulges; folds bent: the cranial flexure has developed, there has been lateral growth and expansion of the neural folds, and the rostral and rostro-lateral edges of the folds have begun elevation; 218 prosencephalon folds beginning elevation: the neural folds have become concave in shape and begun elevation across the neural groove, reaching as much as half the distance from being unelevated to making contact in the mid-line; prosencephalon folds completing elevation: the folds are nearing the mid-line prior to making contact at Closure 2; initial contact at Closure 2: the cranial neural folds make contact across the mid-line at the prosencephalon/mesencephalon boundary (this includes embryos where the zone of contact at Closure 2 has extended to include some portion of the mid-prosencephalon); Closure 3 begun without Closure 2: a stage unique to SELH and SELH-like (in exencephaly-liability) embryos, where fusion has begun at the most rostral edge of the neural tube and is extending caudally through the prosencephalon without extending past the site of Closure 2 into the mesencephalon; fused to the mid-mesencephalon with ANP open: fusion at Closure 2 has occurred and extended caudal into the mid-mesencephalon and rostral into the mid-prosencephalon, leaving an open region of rostral prosencephalon (ANP: anterior neuropore); fused to the mid-mesencephalon with ANP closed: the rostral neural tube is fused from its most rostral aspect caudal to the mid-mesencephalon, due either to extension of fusion from Closure 3 only or to extension in both directions from the site of Closure 2 as well as some fusion of the rostral prosencephalon by extension of Closure 3; fused to the apex with ANP open: the cranial neural tube is fused from the mid-prosencephalon to the apex (the most dorsal part of the cranial neural tube and mesencephalon, nearing the caudal mesencephalon) and the ANP is still visible as an opening over the rostral prosencephalon; 219 fused to the apex with ANP closed: fusion of the cranial neural tube is complete from its most rostral aspect to the apex; fused to the rhombencephalon with ANP open: the cranial neural tube is fused from near the rostral prosencephalon to the rostral rhombic lip, a small region of the rostral prosencephalon (the ANP) remaining open; fused to the rhombencephalon with ANP closed: fusion of the cranial neural tube is complete from its most rostral aspect to the rostral rhombic lip; fusion complete: fusion of the rhombencephalon is complete and the entire cranial neural tube is closed. Additional observations were made at each stage of closure on the progress of Closure 4 over the rhombencephalon; these observations were important for the study of fusion in the prospective cerebellar region (Chapter 7). Mouse stocks The SELHA/Bc, SELHR/Bc and LM/Bc strains of mice have all been described in Chapter 2. Experimental design The generation of the F2 sires and the scoring of their testcross embryos for exencephaly is described in the Materials and Methods of Chapter 4. Litters for scoring stage of neural tube closure were generated by timed matings of SELHR dams x SELHA sires (SELH embryos), LM/Bc sires x LM/Bc dams (LM/Bc embryos), LM/Bc sires x SELHA and SELHR dams (Fl embryos), and of 13 F2 sires x SELHA and SELHR dams (testcross embryos). Six of these F2 sires (sires 385, 517, 525, 636, 841 and 950) 220 were among the 31 low-risk sires that produced no exencephaly in testcrosses to SELHR; 6 F2 sires were the highest-risk sires available and ranked 2nd, 3rd, 4th, 5th, 11th and 12th in exencephaly production (sires 911 (10.2%), 930 (10.2%), 388 (9.7%), 943 (8.2%), 863 (6.7%) and 684 (6.3%)); and one sire (858; 4.9%) was intermediate-to-high for exencephaly-liability (ranked 18th in exencephaly production). Both SELHA and SELHR dams were used in crosses to LM/Bc and F2 sires; the study began using SELHA dams, but later the SELHA breeding colony was not generating many animals and SELHA females were not breeding well. Litters were first collected from SELHR dams x SELHA sires to verify that Closure 2 is absent in SELHR/SELHA embryos. Once this was shown to be the case, some SELHR females were used to collect litters from some LM/Bc and some high- and low-risk F2 sires. Litters from 12 SELHA and 16 SELHR dams mated to LM/Bc sires were examined. All litters collected from F2 sires 385 (n=10), 517 (n=9), 525 (n=8) and 841 (n=10) were from SELHA dams. Only 1 litter of the 10 collected from F2 636 was from an SELHR dam, and all 13 litters collected from F2 950 were from SELHR dams. All litters collected from F2 sires 911 (n=l 1), 858 (n=10), 388 (n=7), 863 (n=l 1) and 684 (n=9) were from SELHA dams. All 12 litters from F2 930 were from SELHR dams, and three of the 15 litters from F2 930 were from SELHA. Data analysis For each strain and for each category of F2 sire, the number of embryos with each combination of somite count and neural tube closure stage was tabulated graphically in the manner used by Macdonald et al. (1989). Somite count was used as the indicator of developmental age. Within each somite count, the proportion of embryos in each neural tube closure stage was calculated. These distributions were compared graphically among the strains 221 and classes of F2 sires. The number of embryos in each closure stage was compared between the strains and between high- and low-risk F2 sires within somite classes by X 2 tests of independence. In a second approach, only the embryos in stages where the presence or absence of formation of the anterior neuropore (ANP; signifying Closure 2) could be evaluated were considered (see Fig. 52 and Table 8). Frequencies of embryos lacking Closure 2 detected by this approach were compared among the strains and classes of F2 sires. The statistical significance of the difference between classes of F2 sires was tested by a X 2 test of independence. The probability of obtaining the observed relationship between lack of Closure 2 and production of exencephaly in testcross progeny by high- and low-risk F2 sires was calculated using Fisher's exact test (Sokal and Rohlf, 1981). For each individual F2 sire, the frequency of embryos lacking Closure 2 was plotted against the frequency of exencephaly produced in previous testcross litters. Spearman's coefficient of rank correlation (Sokal and Rohlf, 1981, and "Statistical Methods" in Chapter 2) for these two variables was applied to the seven F2 sires that produced exencephaly. The sires tested were selected based on their values for one of the traits (production of exencephaly in offspring) and therefore may not show a strictly linear relationship between the two traits. Statistical Methods The X 2 test of independence and Spearman's coefficient of rank correlation are described in Chapter 2 {General Materials and Methods, "Statistical Methods"). 222 Results Patterns of neural tube closure The numbers of embryos that were examined for stage of neural tube closure and somite count were as follows: 394 SELH, 321 LM/Bc, 273 F l , 649 from high-risk F2 sires, 613 from low risk sires, and 89 from the intermediate-risk sire. In any given litter, there was variation among embryos in somite count and in stage of neural tube closure. As Closure 2 is a relatively brief stage (Macdonald et al., 1989), only some embryos will be passing through it even in litters collected at the most appropriate gestational age. Therefore, not all embryos will be in a stage scorable for Closure 2. The pattern of neural tube closure in SELH embryos followed the pattern described previously (Macdonald et al., 1989). Embryos were often observed at a stage where the only fusion in the cranial neural tube was in the rostral prosencephalon (see Figs. 52 and 53), and the majority of 16-17 somite embryos (81%, 21/26) were at this stage (C3B, Closure 3 begun without Closure 2). No SELH embryos were seen initiating contact and fusion at the site of Closure 2 at any somite count (IC2, Fig. 53). The pattern of neural tube closure in LM/Bc was as previously described (Juriloff et al., 1991) and similar to that seen in other normal strains (Macdonald et al., 1989). Embryos were often seen with initiation of fusion between the apposed neural folds at the prosencephalon/mesencephalon boundary (Closure 2, stage IC2, Fig. 53), and this stage was most common in 13-15 somite embryos; 29% (5/17) of 13-somite embryos, 50% (18/36) of 14 somite embryos, and 9% (3/32) of 15 somite embryos were at this stage. No LM/Bc embryos were seen undergoing Closure 3 without Closure 2 (C3B, Fig. 53). The pattern of neural tube closure in F l embryos was very similar to that of the LM/Bc strain (Fig. 53). The distribution of embryos through closure stages was not different between Fig. 52. Diagrammatic side-views of normal and SELH-like embryos during first stages of cranial neural tube closure, showing initiation and extension of closure sites 1, 2, and 3, and showing the region of Closure 4. In normal embryos, fusion at Closure 2 creates the anterior neuropore (ANP). In SELH-like embryos, absence of Closure 2 results in one continuous rostral opening with no ANP. P=prosencephalon; M=mesencephalon; R=rhombencephalon. 224 soAjqwa jo % soAiqwe jo % soAjqwe JO % 225 LM/Bc and F l embryos at any somite stage (X 2; all p>0.25). A single F l embryo with 21 somites had only the rostral prosencephalon fused, a closure pattern consistent with incipient exencephaly. The low frequency of this observation is consistent with the 0.3% (3/1074) exencephaly seen in F l embryos (see Chapter 4). Comparison of the patterns of neural tube closure among the two parental strains and F l (Fig. 53) demonstrated delay of completion of neural tube closure in SELH, as previously noted (Macdonald et al., 1989). At the 18-19 somite stage, 77% (36/47) of LM/Bc embryos and 69% (9/13) of F l embryos had completed fusion of the entire cranial neural tube (FC; see asterisks in Fig. 53). In contrast, only 8% (3/37) of SELH embryos with 18-19 somites had completed fusion, a significant delay compared with the F l and LM/Bc (X2=41.1, pO.OOl). The pattern of neural tube closure in embryos from high-risk F2 sires and low-risk F2 sires differed from each other (Fig. 54). The pattern for the embryos of high-risk sires tended to resemble the SELH pattern, having embryos undergoing Closure 3 without Closure 2 (C3B in Fig. 54); for example, 31% (20/64) of 16-17 somite embryos were at this stage. As in SELH, completion of fusion (FC) was delayed; for example, only 33% (8/24) of 18-19 somite embryos had reached this stage. Unlike SELH, however, a small proportion of embryos had initiated fusion at Closure 2 (e.g., 14% (18/132) of 14-15 somite embryos). In contrast to the embryos from high-risk F2 sires, the pattern for testcross embryos from low-risk F2 sires tended to resemble that of LM/Bc and F l embryos (Fig. 54). Among 14-15 somite embryos, 34% (35/104) were observed with initiation at Closure 2 (IC2), and 68% (17/125) of 18-19 somite embryos had completed fusion (FC; see asterisks in Fig. 54). Only a very small minority had the SELH-like stage of Closure 3 begun with Closure 2 (C3B); for example, 3% (2/61) of 16-17 somite embryos. The distribution of embryos among stages of closure differed significantly (X 2; p<0.05) " O CD E | CO ^ CD LU c 2 CO 4= CD cd to CO £ 11 w o CM — o E * o co «t= V CO - C O O ) CO CD co 75 2 8 co ,„ CD CO 4-1 O c CD to a 0 ZJ CO 00 CD 1 s o o O 55 ^ ( 0 2 1 o E o I * CO CD CD O £ CD CO CO CD O CO CO O CM" CD 1— 3 CO o Q L 0 o c 0 _ Q E o J Z co o JZ! E CD 0 CD CC -4—« 9 ab £ o CO cn ^ o ^ ra 0 . 1 ca -a. § i < 5 i s Q ° 0 CO ZJ E 0 H— o C o "t o C L O i _ C L 0 JZ co co CO o CO c CE O L L o «» 0 *± a. ca cSi 2 O T 3 — 0 . - CO c zs L L O 0 ca Z! 0 c 7a § _>• 0 Q -£ 2 1 S 75 x a a> 0 2 ° 1 1 § ca o C L ~ E 0 5 O _ Q ° o LD C O £ cvi z> CO 0 ZJ \l si O +-• o ca 0 3 I £ 0 c CO 1 -§ o ~ M— C i t g 4 - " co ca 11 CD 00 CD 0) 4_, CO ca o £ O ca 0 0 > ±i 0 E 0 5 , o B ca co co ca > 0 0 g TTJ C L 0 O c 0 _ 0 C L E o o co co I J2 E o -g o E ca o f Y CD 0 o co 0 . CO >5 + 00 0 CO 0 ca o TJ c CO 0 4—" CO ca co" CO _ca 0 £ 8 IS 1 £ O CO ca 0 c 0 CD CO ^ o © cT-Q - Q T J E 2 0 ca C L cn LU . . O CM CL 0 • - ZJ C CO O O ca O > 4 -0 CJ 0 ° C D *t £ 5 1 i CD C D 0 0 0 J Q E ZJ C E o o 0 _ Q o C 0 CD is ^  7D * 75_ £ E =3 o g c ° O Q CO LL 2 0 . . - C o r 2 2 6 co+i- -soA jqws p % 3 b \°° i Hd Vd lAIIAId aeo 201 30d aad ad 0 CD 03 -t—1 CO 0 i— Z3 CO CJ o 0 n Zi 4—I To i— Zi 0 o o o CD o CD O o OJ soAjqiuep % 227 between high- and low-risk F2 sires for each of the 12+13, 14+15, 16+17 and 18+19 somite classes. Small sample size within the 20+21 and 22+23 somite classes may account for the lack of statistical significance there; all embryos from low-risk sires had completed cranial neural tube closure by the 20 somite stage whereas many embryos with 20-23 somites from high-risk sires had not. Presence of an anterior neuropore Another way of summarizing the data, focused on whether or not the embryos had undergone Closure 2, was applied. Within several of the closure stages shown in Figs. 53 and 54, there were two types of embryos: one type with an open section of the neural tube at the mid-prosencephalon but fusion in the region of Closure 2, the other with fusion extending throughout the prosencephalon. For this discussion, this open section over the prosencephalon is referred to as the anterior neuropore (ANP). Initiation of Closure 2 during normal neurulation splits the open cranial neural tube into two regions and creates the ANP (Fig. 52) and a more caudal opening, referred to as the mesencephalic neuropore in this discussion. The ANP is the region of the neural tube left open rostral to the site of Closure 2, and it is normally closed by rostral extension of Closure 2 and caudal extension of Closure 3. Embryos that initiate fusion at Closure 2 have an open ANP through several subsequent stages of neural tube closure, as fusion proceeds caudally from Closure 2 through the mesencephalic neuropore. If the step of initiation of fusion at Closure 2 is omitted as in SELH embryos, no separate ANP is formed, there is one continuous rostral opening of the neural tube (Fig. 52), and fusion through the prosencephalon and mesencephalon proceeds caudally from the most rostral aspect of the neural tube (Closure 3). SELH-like embryos are not expected to have an ANP (as defined here) at any stage of closure. The presence of an ANP can therefore be used to indicate that Closure 2 has 228 taken place. In addition, rostral closure complete to the usual zone of initiation of Closure 2 can be used to detect the absence of Closure 2. The embryos that could be scored in this way were compared between high- and low-risk F2 sires, F l , SELH, and LM/Bc (Table 8). Among embryos scorable for the presence of Closure 2, all 58 SELH embryos lacked Closure 2, all 55 LM/Bc embryos had Closure 2, and all but 1/53 (2%) F l embryos, like the normal LM/Bc strain, had Closure 2. The 6 high-risk F2 sires produced 67/103 (65%) testcross embryos lacking Closure 2; the intermediate-risk F2 sire produced 8/20 (40%) testcross embryos without Closure 2; and the 6 low-risk F2 sires produced 5/76 (7%) testcross embryos with no Closure 2. The difference between high- and low-risk F2 sires was significant (X 2; p<0.001). That is, the F2 sires that transmitted a high liability to exencephaly also transmitted a high liability to lack of Closure 2, and the F2 sires that transmitted a low exencephaly-liability also transmitted a low liability to lack of Closure 2. Within the high-risk and low-risk F2 sire groups, the individual sire values were consistent (Table 8). The frequency of exencephaly in the testcross progeny of F2 sires was strongly related to the frequency of failure of Closure 2 (Fig. 55). The probability of observing lack of Closure 2 in less than 15% of the testcross progeny of all 6 low-risk sires and in more than 40% of the testcross embryos from all 7 high-risk sires (including the intermediate-risk sire) by chance was 0.00058 (Fisher's exact test). Among sires that produced exencephaly, the absence of Closure 2 was positively correlated with the percent exencephaly (rs=0.88; p<0.05). DISCUSSION Previously, it was known that SELH mice have a high frequency of exencephaly, that the exencephaly is due to failure of elevation of the mesencephalic neural folds, and that they also have an unusual mechanism of cranial neural tube closure (Macdonald et al., 1989). All SELH 229 Table 8. Numbers of D8-D9 embryos scored for stage of neural tube closure in S E L H , LM/Bc and F1, and from testcrosses of high-risk3, intermediate-riskb1 and low-risk0 F2 sires. No. of embryos not scoreable for presence of Closure 2 No. of embryos scoreable for presence of Closure 2 Genotype No. too early No. too late Present Absent SELH strain 267 69 0 58 LM/Bc strain 128 138 55 0 F1 (SELH.LM) 162 58 52 1 Testcross from: High-risk F2 sires (id) #911 89 4 1 16 #930 67 42 3 13 #388 56 8 5 8 #943 73 37 13 11 #863 76 24 6 12 #684 67 3 8 6 total 428 118 36 67 Intermediate-risk F2 sire (id) #858 63 6 12 8 Low-risk F2 sires (id) #385 57 36 8 1 #517 54 8 13 1 #525 73 9 10 1 #636 79 6 14 0 #841 97 4 8 0 #950 52 62 18 2 total 412 125 71 5 High-risk F2 sires transmitted high liability to exencephaly (produced 6.3-10.2%). bIntermediate-risk F2 sire transmitted an intermediate liability to exencephaly (4.9%). cLow-risk F2 sires transmitted low liability to exencephaly (produced 0%). Fig. 55. Relationship, for individual F2 sires, between the genetic liability to exencephaly they transmitted and the proportion of testcross embryos with an SELH-like mechanism of cranial neural tube closure (lack of Closure 2) they produced. Two sires produced 0% for both traits scored. % testcross embryos lacking Closure 2 i "T 1 1 1 1 1 1 1 I I I I 0 1 2 3 4 5 6 7 8 9 10 11 12 % e x e n c e p h a l y p r o d u c e d i n t e s t c r o s s 231 embryos fail to elevate the mesencephalic folds at the normal time to make contact and fuse at the prosencephalic/mesencephalic boundary (Closure 2). The mesencephalic folds of most SELH embryos do elevate and fuse later from the leading edge of a contact and fusion zone that begins at the most rostral end of the neural folds. The interpretation has been that the absence of Closure 2 in all SELH embryos is the defect that creates liability to exencephaly. However, an alternate hypothesis was that absence of Closure 2 is a polymorphic variant among inbred strains of mice and that total failure of elevation of the mesencephalic folds, leading to exencephaly in some embryos, is a separate trait coincidentally present in SELH mice. This alternative hypothesis could not be rejected by the previous experimental approaches. Using a genetic approach, this study demonstrates that the genes that cause exencephaly in SELH mice are the same as those that cause the absence of Closure 2. Segregating sires that transmit a high liability to exencephaly also transmit an SELH-like pattern of closure, with a large proportion of embryos detectably lacking Closure 2 and a clear delay in completion of neural tube closure at later somite stages. Segregating sires that transmit a low liability to exencephaly also transmit a normal pattern of closure, with many embryos detectably undergoing Closure 2 and a large proportion of embryos having completed neural tube closure at the normal time (the 18-19 somite stage). Furthermore, among segregating sires there is a strong correlation between the proportion of their progeny lacking Closure 2 and the proportion having exencephaly later. This result means that the cause of exencephaly is the same as the cause of absence of Closure 2, and that the fundamental defect in SELH embryos leads to both exencephaly and lack of elevation of the mesencephalic folds at the normal time. The distributions of embryos across closure categories (Figs. 53 and 54) are not completely the same from high-risk and low-risk F2 sires as from the SELH and LM/Bc strains, respectively. The high-risk sires produced some LM/Bc-like embryos and the low-risk sires 232 produced some SELH-like embryos. This is expected based on Mendelian segregation of the exencephaly-liability loci (discussed in depth in Chapter 5). In an F2, if two or three loci are responsible for the trait, only 1/16 or 1/64 of F2 sires, respectively, are expected to be genetically the same as each parental strain. Therefore, not all low-risk F2 sires are expected to carry only LM/Bc alleles and not all high-risk F2 sires are expected to carry only SELH alleles at the exencephaly-liability loci. In addition, sires transmitting a low but non-zero liability to exencephaly may, by chance, produce no exencephalic progeny (i.e., the 95% confidence intervals around their expected values overlap with 0%). It was therefore expected that the low-risk F2 sires would produce some testcross progeny with an SELH-like pattern of closure and that the high-risk sires would produce some testcross embryos initiating fusion at Closure 2. The closure mechanism in SELH embryos demonstrates the developmental redundancy of the closure initiation sites; to some extent, they can compensate for one another. It is possible that Closure 3 may be particularly robust in SELH embryos. As exencephaly is lethal and cranial neural tube closure is completed by Closure 3 in SELH embryos, natural selection would have favoured fixation of genes conferring a robust Closure 3 during the early generation of the SELH stock. A robust Closure 3 may either begin earlier in SELH embryos than in most normal strains, or be more efficient (proceed more quickly through the prosencephalon and mesencephalon). An alternative possibility is that the neural folds of SELH embryos retain their ability to fuse later into development, giving the slower mechanism of fusion by extension of Closure 3 more time to complete closure over the mesencephalon. This study also demonstrates the power of Mendelian segregation studies as an analytical tool in tracing paths of causality in complex developmental errors. It is a simple yet powerful way of showing that the mutant genes causing lack of Closure 2 are the same as those that cause 233 exencephaly and so that lack of Closure 2 leads to exencephaly. The F2 sires used in this study were taken from a panel F2 sires that was already being generated and testcrossed as part of other, concurrent studies; the only extra work necessary for this study involved additional testcrosses to collect D8-9 embryos and examination of these embryos to score them for stage of neural tube closure. Observations made on these embryos also provided some information on the mechanism of closure in the caudal mesencephalon in SELH embryos (see Chapter 7). It would be difficult to show the relationship between lack of Closure 2 and liability to exencephaly by other means. The ideal method would be to identify the specific exencephaly-liability mutations in SELH mice, cause the same mutations in a normal strain, and to look for lack of Closure 2 and high liability to exencephaly in the mutant animals. An alternative method would be to create a congenic strain, where the exencephaly-liability loci from SELH are bred onto a normal strain background and to examine the mice for absence of Closure 2 along with exencephaly-liability. This would require knowledge of the map position of the liability loci and several generations of breeding to accomplish. The latter method and the Mendelian segregation studies do not rule out the possibility that different, closely linked genes are responsible for each trait. Intuitively, however, it seems probable that absence of Closure 2 would affect subsequent closure of the neural tube. It would be an unlikely coincidence for these traits to be caused by different but closely linked loci. The demonstration that absence of Closure 2 and liability to exencephaly probably share a common genetic cause validates measuring phenotype (liability to exencephaly) by scoring production of exencephaly in testcross progeny. Indeed, this is a more accurate approach than scoring for lack of Closure 2 directly as litters are not synchronized in their development, making it impossible to collect litters in which all the embryos are in this brief stage of development. 234 Chapter 7: STUDIES ON THE CAUSE OF CLEFT CEREBELLUM-ATAXIA Introduction Absence of Closure 2 in SELH mice is thought to lead to abnormal cerebellar development in a proportion of non-exencephalic embryos. About 5-10% of weaned SELH mice are ataxic (unable to coordinate voluntary muscular movements) and have a midline cleft of the cerebellum. The cerebellum develops from the caudal mesencephalon and rostral rhombic lips (Bonnevie and Brodal, 1946; Hallonet et al., 1990) and is thought to be the last region of the cranial neural tube to fuse in SELH embryos, as the advancing edge of Closure 3 meets Closure 4 (Juriloff et al., 1993). Histology of D10 and D l l embryos is consistent with the hypothesis that, in some SELH embryos with extremely delayed neural fold elevation and fusion, only the surface epithelium in the prospective cerebellar region retains the ability to fuse in the midline; the neuroepithelium does not make contact across the neural groove (Harris et al., 1994). It appears, from observation of the caudal limits of the opening in the cranial neural tube of exencephalic SELH embryos, that it is possible for Closure 4 to extend rostral into the mesencephalon (Harris et al., 1994). Fusion over the rhombencephalon by Closure 4 differs from fusion of the rest of the cranial neural tube, as Closure 4 appears to involve elongation of a membrane rather than apposition and fusion of the neural folds across the midline (Geelen and Langman, 1977; Copp et al., 1990; Golden and Chernoff, 1993). Initial contact in the rhombencephalon is between surface ectoderm, followed by fusion of a thin layer of neuroepithelium (Geelen and Langman 1977, 1979). It is possible that in some SELH embryos with severely delayed neural fold elevation and fusion, Closure 4 completes fusion of the rhombencephalon before Closure 3 has completed fusion to the rostral rhombic lip. Rostral continuation of Closure 4, involving the rostral elongation of the membranous covering of the 23 5 rhombencephalon, might then result in fusion of surface epithelium but not neuroepithelium in the prospective cerebellar region. The neuroepithelium may never overcome its inadequate degree of elevation to make contact in the midline, or at least not within the time period during which fusion between neuroepithelial cells is possible. Alternatively, the