UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Cell cycle related effects on the radiation survival responses of human tumor cells Hill, Andrew Arthur 1998

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

Item Metadata

Download

Media
831-ubc_1999-388956.pdf [ 9.32MB ]
Metadata
JSON: 831-1.0085449.json
JSON-LD: 831-1.0085449-ld.json
RDF/XML (Pretty): 831-1.0085449-rdf.xml
RDF/JSON: 831-1.0085449-rdf.json
Turtle: 831-1.0085449-turtle.txt
N-Triples: 831-1.0085449-rdf-ntriples.txt
Original Record: 831-1.0085449-source.json
Full Text
831-1.0085449-fulltext.txt
Citation
831-1.0085449.ris

Full Text

C E L L C Y C L E R E L A T E D EFFECTS ON THE RADIATION S U R V I V A L RESPONSES OF H U M A N TUMOR CELLS by A N D R E W ARTHUR FULL B.Sc.(Eng.) Queen's University, 1991 M.A. Sc. University of Toronto, 1993 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Physics) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A August, 1998 ©Andrew Arthur Hill, 1998 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. 1 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. Department of P H-ySlC^ A v o The University of British Columbia Vancouver, Canada Date <b t>e^  Wb DE-6 (2/88) ABSTRACT Cell cycle-related phenomena have profound effects on the ability of cells to survive exposure to ionizing radiation. These phenomena are important because they shed light on the underlying mechanisms of the cell cycle, and because they provide avenues to improve the efficacy of clinical radiotherapy. Radiobiologists have developed a partial picture of how and why radio sensitivity varies during the cell cycle. Nevertheless, our understanding of the role of cell cycle effects on the radio sensitivity of human cells is still far from complete. In this thesis, cell populations which were synchronized at specific points in the mitotic cycle were used to explore cell cycle-related effects on the radiosensitivity of human tumor cells in vitro. The survival of synchronized cells after irradiation was measured at low doses which are relevant to radiation therapy, using the cell-sorter assay, which utilizes a cell-counting flow cytometer to reduce the uncertainties associated with traditional survival assays. The radiosensitivity of three human tumor cell lines varied significantly over the course of the cell cycle, in a manner which was in general agreement with earlier studies by others who examined the responses of rodent, and some human, cell types, using similar synchronization techniques. However, discrepancies were found with other studies which used alternative synchronization methods. The widely-used linear quadratic model of cell survival was tested in synchronized cell populations. Consistent, significant deviations from the model were found. A ii mathematical model of synchronized cell radio sensitivity was developed to explore these deviations. Departures from the linear-quadratic model in two cell lines could be adequately explained by cell cycle-related heterogeneity in experimental cell populations. In a third cell line, however, the deviations from the linear-quadratic model were not attributable to cell heterogeneity alone. One of the three human tumor cell lines examined here underwent a prolonged arrest in the Gi phase of the cell cycle after irradiation. The arrest was characterized by following the progression of synchronized cells after irradiation at different times in Gi phase. The characteristics of the arrest were consistent with a "checkpoint" in late Gi phase where radiation-damaged cells stopped cycling for an extended period. iii TABLE OF CONTENTS A B S T R A C T ii T A B L E OF CONTENTS iv LIST OF FIGURES vi LIST OF TABLES ix LIST OF ABBREVIATIONS x A C K N O W L E D G M E N T S xi Chapter 1: Introduction 1 1.1 Effects of ionizing radiation on living cells 1 1.1.1 Radiation damage to D N A is critical to cell killing 3 1.2 The normal cell cycle 5 1.2.1 Sequence of events in the normal cell cycle 5 1.2.2 Molecular control of the cell cycle 6 1.3 Models of the cellular radiation survival response 11 1.3.1 Description of survival models 11 1.3.2 Why examine deviations from the LQ model? 16 1.3.3 Known departures from the LQ model 17 1.4 Cell-cycle related responses to ionizing radiation; apoptosis 19 1.4.1 G 2 block 19 1.4.2 G i arrest 20 1.4.3 Apoptosis 23 1.5 Radiosensitivity varies over the course of the cell cycle in rodent, HeLa, and some other human cell types 24 1.5.1 Previous investigations of cell-age related variation in radiosensitivity 24 1.5.2 Explanations for cell cycle related variation in radiosensitivity 27 1.6 Relevance of intrinsic tumor cell radiosensitivity to radiation therapy 28 1.6.1 Accurate knowledge of low dose intrinsic radiosensitivity in tumor cells may be important for designing radiotherapy regimes 28 1.6.2 Cell-cycle related variation in radiosensitivity may be an important factor in radiotherapy 29 1.7 Objectives of this thesis 32 Chapter 2: Methods 33 2.1 Cell Lines..., 33 2.2 Routine cell culture 33 2.3 Selection of synchronous cells by mitotic rolloff 34 2.4 Monitoring of cell synchrony 35 2.4.1 Measurement of mitotic index 37 2.4.2 Measurement of D N A content by flow cytometry 38 2.4.2.1 DAPI staining for D N A content 38 2.4.2.2 BrdU-FITC/DAPI staining for S-phase fraction and D N A content 39 2.4.3 Quantitation of cell cycle distribution by D N A histogram fitting 39 2.5 Irradiation of cells 39 iv 2.6 Measurement of cell survival 40 2.6.1 Single dose survival profiles (Conventional clonogenic assay) 40 2.6.2 Survival curves (FACS presort assay) 42 2.7 Calculation of cell survival 43 2.8 Fitting of measured survival curves 44 2.9 Mathematical models of synchronized cell radiation survival 45 2.10 Measurement of apoptosis 53 2.11 Western blotting for p53 53 2.12 Transfection of A549 cells to abrogate p53 function 55 Chapter 3: Results 58 3.1 Characteristics of mitotically selected cells 58 3.2 Kinetics of mitotically selected cells 58 3.3 Radiosensitivity variations during the cell cycle in HT-29, A549, and U l 66 3.4 Substructure in the survival responses of synchronized HT-29, A549, and U l 71 3.5 Radiation induced kinetic responses of HT-29, A549 and U l 106 3.6 Mathematical model predictions of substructure in the survival responses of synchronous and asynchronous cell populations 126 3.6.1 Characterization of the model 126 3.6.2 Modelling of the single dose survival responses 128 3.6.3 Modelling of substructure in survival curves 138 Chapter 4: Discussion 168 4.1 Three human tumor cell lines exhibited common variations in single-dose radiosensitivity after mitotic selection. These variations were related to age-specific radiosensitivity by a mathematical model 168 4.2 The survival responses of asynchronous and synchronized A549 and HT-29, but not U l cells, are consistent with the presence of age-specific subpopulations with different LQ radiosensitivities. Alternative explanations for U l substructure are considered.. 178 4.3 A549, but not HT-29 or U l , undergoes a prolonged G i arrest after irradiation which does not necessarily play an important independent role in determining clonogenic survival 192 4.4 A549, HT-29, and U l cells have lower linear quadratic a/(3 ratios at clinically relevant doses than have commonly been reported for other tumor cells and normal tissues 199 Chapter 5: Conclusions 204 References , 206 Appendix 1: Effects of binucleate cells on classification of G i , S, and G2/M phase cells220 v LIST OF FIGURES Figure 1.1: The cell cycle 8 Figure 1.2: Gi and G2 arrests 9 Figure 2.1: Schematic of mitotic shakeoff apparatus 36 Figure 2.2: Top view of irradiation sample holder 41 Figure 2.3: Three step process to specify model parameters 49 Figure 2.4: Schematic view of the cell cycle model 52 Figure 2.5: Schematic of the BCMGSNeo plasmid vector 57 Figure 3.1: Examples of mitotically selected U l cells 61 Figure 3.2: D N A histograms of synchronized A549 cells 62 Figure 3.3: D N A histograms of synchronized HT-29 cells 63 Figure 3.4: D N A histograms of synchronized U l cells 64 Figure 3.5: A549 cell cycle distribution after mitotic selection 66 Figure 3.6: HT-29 cell cycle distribution after mitotic selection 67 Figure 3.7: U l cell cycle distribution after mitotic selection 68 Figure 3.8: Cell age related variation in radiosensitivity in A549 71 Figure 3.9: Cell age related variation in radiosensitivity in HT-29 72 Figure 3.10: Cell age related variation in radiosensitivity in U l 73 Figure 3.11: Survival curve substructure in asynchronous A549 cells 77 Figure 3.12: Survival curve substructure in Oh mitotic A549 cells 78 Figure 3.13: Survival curve substructure in 3h Gi phase A549 cells 79 Figure 3.14: Survival curve substructure in lOh Gi/S phase A549 cells 80 Figure 3.15: Survival curve substructure in 15h S /G2 phase A549 cells 81 Figure 3.16: Alpha ratios for synchronized A549 cells 84 Figure 3.17: Beta ratios for synchronized A549 cells 85 Figure 3.18: a/p ratios for A549 cells 86 Figure 3.19: Survival curve substructure in asynchronous HT-29 cells 87 Figure 3.20: Survival curve substructure in Oh mitotic phase HT-29 cells 88 Figure 3.21: Survival curve substructure in 3h Gi phase HT-29 cells 89 Figure 3.22: Survival curve substructure in 8h Gi/S HT-29 cells 90 Figure 3.23: Survival curve substructure in 15h Gi/S HT-29 cells 91 Figure 3.24: Alpha ratios for synchronized HT-29 cells 94 Figure 3.25: Beta ratios for synchronized HT-29 cells 95 Figure 3.26: ot/p ratios for A549 cells 96 Figure 3.27: Survival curve substructure in asynchronous U l cells 97 Figure 3.28: Survival curve substructure in Oh mitotic U l cells 98 Figure 3.29: Survival curve substructure in 3h Gi phase U l cells 99 Figure 3.30: Survival curve substructure in 17h Gj/S phase U l cells 100 Figure 3.31: Survival curve substructure in 25h S /G2 phase U l cells 101 Figure 3.32: Alpha ratios for synchronized U l cells 104 Figure 3.33: Beta ratios for synchronized U l cells 105 Figure 3.34: a/p ratios for synchronized U l cells 106 Figure 3.35: Kinetics of synchronized A549 cells after irradiation at t=3h 112 vi Figure 3.36: Kinetics of synchronized A549 cells after irradiation at t=6h 113 Figure 3.37: Kinetics of synchronized A549 cells after irradiation at t=l l h 114 Figure 3.38: Kinetics of synchronized A549 cells after irradiation at t-16h 115 Figure 3.39: Kinetics of synchronized HT-29 cells after irradiation at t=3h 116 Figure 3.40: Kinetics of synchronized U l cells after irradiation at t=3h 117 Figure 3.41: Kinetics of synchronized U l cells after irradiation at t=8h 118 Figure 3.42: Kinetics of synchronized U l cells after irradiation at t=18h 119 Figure 3.43: Kinetics of synchronized U l cells after irradiation at t=22h 120 Figure 3.44: D N A histograms of A549 and A549BPV cells, 16h after irradiation 121 Figure 3.45: Kinetics and survival of mitotically selected A549BPV transfected cells... 122 Figure 3.46: Kinetics of synchronized A549BPV cells after irradiation 123 Figure 3.47: Induction of p53 by irradiation in A549, HT-29, U l , and A549BPV cells 124 Figure 3.48: D N A fragmentation in A549, HT-29, U l 125 Figure 3.49: Progression of model HT-29 cells after mitotic selection 129 Figure 3.50 Comparision of model with A549 cells 130 Figure 3.51: Comparision of model with HT-29 cells 131 Figure 3.52: Comparision of model with U l cells 132 Figure 3.53: Comparision of distribution of model A549 cells with measured cells 133 Figure 3.54: Comparision of distribution of model HT-29 cells with measured cells .... 134 Figure 3.55: Comparision of distribution of model U l cells with measured cells 135 Figure 3.56: Model A549 and HT-29 single dose survival profiles 139 Figure 3.57: Schematic of A549 model 140 Figure 3.58: Schematic of HT-29 model 141 Figure 3.59: Comparison of model U l with smoothed model 142 Figure 3.60: Model and measured survival curves in asynchronous A549 cells 145 Figure 3.61: Model and measured survival curves in A549 (0 h mitotic cells) 146 Figure 3.62: Model and measured.survival curves in A549 (3 h Gi phase cells) 147 Figure 3.63: Model and measured survival curves in A549 (lOh Gi/S phase cells) 148 Figure 3.64: Model and measured survival curves in A549 (15h S/G2 phase cells) 149 Figure 3.65: A549 model and measured phi/ Pio 150 Figure 3.66: Schematic of A549 model 151 Figure 3.67: Model and measured survival curves in asynchronous HT-29 cells 152 Figure 3.68: Model and measured survival curves in HT-29 (Oh mitotic cells) 153 Figure 3.69: Model and measured survival curves in HT-29 (3h Gi phase cells) 154 Figure 3.70: Model and measured survival curves in HT-29 (8 h Gi/S phase cells) 155 Figure 3.71: Model and measured survival curves in HT-29 (15 h S/G2 phase cells).... 156 Figure 3.72: HT-29 model and measured Phi/ Pio 157 Figure 3.73: Schematic of HT-29 model 158 Figure 3.74: Model and measured survival curves in asynchronous U l cells 159 Figure 3.75: Model and measured survival curves in U l (Oh mitotic cells) 160 Figure 3.76: Model and measured survival curves in U l (3 h Gi phase cells) 161 Figure 3.77: Model and measured survival curves in U l (17 h Gi/S phase cells) 162 Figure 3.78: Model and measured survival curves in U l (25 h S/G2 phase cells) 163 Figure 3.79: U l model and measured Phi/ Pio 164 Figure 3.80: Schematic of U l model 165 vii Figure 3.81: Measured substructure and model variance in survival in HT-29 166 Figure 3.82: Measured substructure and model variance in survival in U l 167 Figure 4.1: Fits to U l 3 h G. phase survival response 189 viii LIST OF TABLES Table 3.1: Mitotic index of cells prepared by mitotic shakeoff 58 Table 3.2: Values of a fitted from A549 survival curves 82 Table 3.3: Values of P fitted from A549 survival curves 82 Table 3.4: Linear quadratic fits to the full range of A549 survival curves 83 Table 3.5: a/p ratios for full, low, and high dose range fits in A549 cells 83 Table 3.6: Linear-quadratic-cubic (LQC) fits in A549 cells 83 Table 3.7: Values of a fitted from HT-29 survival curves 92 Table 3.8: Values of p fitted from HT-29 survival curves 92 Table 3.9: Linear quadratic fits to the full range of HT-29 survival curves 93 Table 3.10: a/p ratios for full, low, and high dose range fits in HT-29 cells 93 Table 3.11: Linear-quadratic-cubic (LQC) fits in HT-29 cells 93 Table 3.12: Values of a fitted from U l survival curves 102 Table 3.13: Values of P fitted from U l survival curves 102 Table 3.14: Linear quadratic fits to the full range of U l survival curves 103 Table 3.15: a/p ratios for full, low, and high dose range fits in U l cells 103 Table 3.16: Linear-quadratic-cubic (LQC) fits in U l cells 103 Table 3.17: Kinetic parameters of the mathematical model 127 Table 3.18 Radiosensitivity parameters of the mathematical model 127 Table 4.1: Possible 'feeder effects' on U l survival 187 Table 4.2: Fits to U l 3 h G i phase survival response 191 ix LIST OF ABBREVIATIONS 5-FU 5-fluorouracil AT Ataxia-telengectasia ATM Ataxia-telengectasia-mutated BrdU Bromodeoxyuridine CDK Cyclin-dependent kinase CHART Continuous hyperfractionated accelerated radiotherapy DAB Diaminob enzidine DAPI 4', 6-diamidino-2-phenylindole dihydrochloride hydrate DNA Deoxyribonucleic acid ECF Enhanced chemofluorescence EDTA Ethylenediaminetetraacetic acid FACS Fluorescence activated cell sorting FBS Fetal bovine serum FdUrd 5-fluoro-2'-deoxyuridine g Acceleration due to gravity, «9.81 m/s2 Gy Gray (=1 J/kg) HEPES N-2-hydroxyethylpiperazine-N' -2-ethanesulfonic acid HU Hydroxyurea kbp Kilo-basepair LET Linear energy transfer LPL Lethal-potentially lethal LQ Linear quadratic LQC Linear-quadratic-cubic MPF Mitosis-promoting factor NP-40 Non-idet P-40 PBS Phosphate buffered saline PCNA Proliferating cell nuclear antigen PLDR Potentially lethal damage repair PMSF Phenylmethylsulfonyl fluoride Rb Retinoblastoma RMR Repair-misrepair RS3 Repair saturation model (3-parameter) SDS Sodium dodecyl sulphate SEM Standard error of the mean SF2 Surviving fraction at 2 Gy Tris Tris(hydroxymethyl)aminomethane TX-100 Triton X-100 x ACKNOWLEDGMENTS First I would like to thank my supervisor, Dr. Lloyd Skarsgard, for all the assistance that he provided over the course of my thesis work. I greatly appreciate the opportunity he gave me; I hope I have learned something from Lloyd about the persistence that is required to be able to ask the right questions and get the best answers. I would also like to thank the members of my supervisory committee, Dr. R.E. Durand, Dr. A.I. Minchinton, and Dr. R. Johnson for contributing their time, and also providing numerous valuable suggestions which improved this thesis at every level. Without the help of numerous others, it would have been simply impossible to complete the work which is described in this thesis. Deanna Acheson made a lot of medium, and contributed most of the excellent technical assistance which I benefited from during my time in the lab. Fung Wan and Art Sy made it possible to complete some very time-consuming experiments by contributing their help. Adriana Vinczan counted a mind-boggling number of colonies. Denise McDougal sorted an even larger number of cells. And Gary DeJong also provided key FACS time and help in the early stages of my thesis work. I very much appreciate the assistance of numerous other people, who all contributed thoughts, insights and lab help which were invaluable, including Dr. Brad Wouters, Dr. Peter Johnston, Dr. Peggy Olive, Hans Adomat, and Susan MacPhail. I am also very grateful to Dr. Graeme Dougherty for his great generosity in taking the time to teach an engineering physicist some molecular biology. xi I am indebted to the Natural Sciences and Engineering Council of Canada for 2 years of financial support during this thesis, and also to the BC Cancer Research Centre for financial assistance. Finally, and above all, Fd like to thank my parents, Ardeth and Gilbert Hill, who provided the help and encouragement which really made this thesis possible. xn Experience does not ever err, it is only your judgment that errs in promising itself results which are not caused by your experiments. Leonardo da Vinci, c. 1510 xiii Chapter 1: Introduction 1.1 Effects of ionizing radiation on living cells The effects of ionizing radiation on living cells are commonly classified in broad terms as physical, chemical, or biological; each of these effects occurs over a characteristic time scale (see e.g. Chapman 1980, Mozumder 1985). A very brief description of these effects follows here. More detailed comments, especially on radiation-induced cell killing, will follow later in this Chapter. Physical effects Ionizing radiations (either photons or charged particles) directly excite and ionize molecules within the cell, creating various radical species. These interactions may occur directly in critical biological target molecules, or they may occur in other molecular components of the cell, including (commonly) water molecules. Direct interaction of ionizing radiation with biological target molecules such as deoxyribonucleic acid (DNA) can occur on a picosecond time scale. Chemical effects Interaction of radiation with water molecules in the cell produces short-lived radicals (such as OH* and H*) which can diffuse over nanometer distances and then react with biological target molecules over a time scale measured in microseconds (Johns 1983). Such "indirect" reactions contribute significantly to the molecular damage produced by radiation. 1 Chapter I: Introduction Once radicals have been produced in cellular targets by either direct or indirect action, they undergo a "metionic reaction" with surrounding molecules (Alper 1979) which determines the nature of the resulting chemical damage. The metionic reaction may restore the target molecule to its original state, or it may alter the target molecule so as to render it non-functional or dysfunctional. This latter occurrence is known as the "fixation" of radiation damage. The nature of the metionic reaction is dependent on the intracellular chemical environment at the time of irradiation. For example, the presence of molecular oxygen in the cell tends to favour fixation of radiation damage, which is a part of the well-known "oxygen effect" (Alper 1979). Once radiation damage to cellular components has been fixed, cellular enzymatic processes will repair or misrepair some of the damage over a time scale of minutes to hours. There are numerous pathways of cellular repair which have varying degrees of fidelity; some of the repair processes that occur in D N A following exposure to ionizing radiation have been studied in detail (Leadon 1996). Biological effects The myriad biological effects of radiation on cells become evident over a time scale of hours to years. On this time scale, cells which have unrestituted radiation damage may display a wide spectrum of radiation responses, including temporary disruptions of the cell cycle, permanent loss of reproductive capacity, or rapid cellular disintegration (e.g. apoptosis). Actively dividing cells with non-lethal levels of damage that continue to reproduce may pass on genetic mutations arising from radiation-damaged D N A to their progeny. 2 Chapter 1: Introduction 1.1.1 Radiation damage to DNA is critical to cell killing DNA, as the repository of a cell's genetic blueprint, is absolutely central to the normal functioning of all cells. Transcription and translation of genes to produce proteins, the successful replication of the genome, and the partition of the chromosomes at mitosis are all critical events in the life cycle of any actively dividing cell, which depend on the integrity of a cell's genomic DNA. For these reasons, it is a natural hypothesis that ionizing radiation should exert its effects on cells through the damage it imposes on D N A molecules. Numerous experiments have been carried out over the last 40 years which provide strong evidence for the hypothesis that damage to nuclear D N A and/or the supporting nuclear matrix is the critical lesion that leads to radiation-induced cell death. For example, microbeam irradiation of the cell nucleus is orders of magnitude more effective at producing cell killing and growth inhibition than irradiation of the surrounding cytoplasm (Munro 1970), and localized radiation damage from radionuclides which are incorporated directly into D N A is vastly more effective at killing rodent cells than radiation damage from radionuclides which are incorporated into other cellular components (Burki et al. 1968). Furthermore, correlations have been shown between radiation-induced cell killing and chromosome aberrations (Dewey et al. 1962), and the relative biological effectiveness of several different ionizing radiations for inducing chromosome aberrations is highly correlated with their effectiveness for cell killing (Skarsgard et al. 1967). 3 Chapter 1: Introduction More evidence for the importance of D N A as a radiation target comes from studies of the autosomal recessive genetic disorder ataxia-telengectasia (AT). AT homozygotes are highly radiosensitive, and also prone to cancer, abnormal immune gene recombination, and genetic instability. Some (but not all) AT cell lines have specific defects in D N A adduct excision repair. Taken together, this evidence suggests that the radiation-induced cell killing is modulated by pathways which involve D N A metabolism (Meyn 1995). Detailed studies of the biochemical characteristics of radiation-induced lesions in D N A molecules have revealed the nature of some of these lesions, and the mechanisms by which these lesions may cause cell death. It is currently believed that ionizing radiation tracks produce "local multiply damaged sites" in D N A molecules (Ward 1985). These sites may contain several different types of biochemical lesions, including D N A single strand breaks, double strand breaks, or various types of base damage, which will be more or less repairable. Al l of these types of damage, if left unrestituted, or if misrepaired, would be expected to be deleterious to the cell. For example, studies of chromosome aberrations have documented how acentric or dicentric chromosomes resulting from radiation-induced double-strand breaks to D N A can permanently stall the process of mitosis or lead to the production of non-viable daughter cells. Thus, a direct causative link can be postulated between radiation-induced D N A damage and cell killing. 4 o Chapter I: Introduction 1.2 The normal cell cycle 1.2.1 Sequence of events in the normal cell cycle Early investigators who observed dividing cells under the microscope could only see visible changes in the morphology of the cell during the various stages of mitosis. During these stages (prophase, metaphase, anaphase, and telophase), the massive restructuring of the cell and segregation of the chromosomes into two daughter cells were obvious. Thus, these dramatic events were the first parts of the cell division cycle to be described. During the remainder of the cell cycle, the relatively long period between mitoses, no obvious activity took place that could be observed by optical microscope; hence, this period was initially given the vague label "interphase". Howard and Pelc (1953) were the first to elucidate the events which occur during interphase, by measuring the incorporation of 3 2P and 35S into the root meristem of the bean plant Vicia faba. Their original delineation of these events remains today as the standard description of the cell cycle. After mitosis is complete, newly divided cells enter a first "gap" phase, Gi, which was originally described as the time interval between telophase and the start of DNA synthesis. After the end of Gi, cells begin to replicate their DNA. The period of DNA synthesis, during which cells will take up radioactive DNA precursors, is known as S phase (for "synthesis"). At the end of S phase, cells have doubled their DNA content. Following S phase, there is a second "gap" phase, G 2 , during which no DNA synthesis takes place. Finally, at the end of G 2 phase, cells enter mitosis, divide, and the daughter cells may repeat the cycle of Gi, S, and G 2 phases (see figure 1.1). For human tumor 5 Chapter 1: Introduction cells in vitro, durations of the stages of the cell cycle vary considerably (Gi phase especially); but some "typical" times are 10 h for G i , 6 h for S, and 4 h for G 2 . 1.2.2 Molecular control of the cell cycle Transitions between phases of the cell cycle (M, G i , S, and G2) are thought to be controlled by proteins known as cyclin-dependent kinases (CDK's) and cyclins (Murray et al. 1989). In order to be "active", a C D K must be complexed with an associated cyclin. At specific stages of the cell cycle (figure 1.1), the cyclin-CDKs phosphorylate other substrates which are parts of molecular pathways that are crucial for proper completion of the cycle. Many of these pathways are still not fully understood. The activity of cyclins and CDK's is normally regulated at the level of expression (i.e. the amount of a cyclin/CDK protein may oscillate during the cell cycle) and at the level of activation (i.e. cyclin/CDK complexes are activated by phosphorylation only at specific points in the cycle) (Sherr 1994). Befitting their putative role as master cell cycle controllers, the cyclins are implicated in cellular responses to proliferative signals, as well as cellular transformation and oncogenesis (Hunter et al. 1994). Under normal growth conditions, the major task of the cyclin control system is to ensure that the processes of D N A replication (S phase) are properly coordinated with those of cell division (M phase). This guarantees that at the end of each cycle, two identical daughter cells will be produced with the correct ploidy. Thus, G i and G 2 phases might be viewed as interchangeable "gaps" which are distinguished only by the type of cyclin-CDK complex which is active in each phase (Nurse 1994). In the following paragraphs, a few of the better understood pathways of the cyclin cell cycle control system in mammalian cells will be discussed. 6 Chapter I: Introduction Gi Phase and the Gi/S transition Cyclins D, E, and A are implicated in progression through Gi phase and triggering of D N A replication at the start of S phase. Good evidence for the importance of these cyclins has emerged from studies linking cyclin D and E overexpression to transformation and acceleration of Gi phase, and abnormal cyclin A function to oncogenesis (Hunter et al. 1994). Currently, it is thought that both the cyclin D-CDK4/6 and cyclin E-CDK2 complexes (which are most active in Gi and at the Gi/S transition, respectively) may cooperate in phosphorylation of the retinoblastoma (Rb) protein. Phosphorylation of Rb dissociates Rb-E2F complexes that are present during Gi phase, and releases the E2F transcription factor, which may activate genes like thymidine kinase and dihydrofolate reductase (Maity et al. 1994) that are important for D N A replication (see figure 1.2B). At the Gi/S transition, Cyclin E is degraded, and Cyclin A enters into complexes with CDK2. One functional link between the cyclins and S phase may involve proliferating cell nuclear antigen (PCNA). PCNA, which is an accessory protein of D N A polymerase 5, forms complexes with cyclins D, E, and A (Sherr 1994). In yeast, some intriguing investigations have suggested that protein complexes which are known to bind origins of replication may be directly phosphorylated by activated CDKs (Heichman et al. 1994). Evidence for such an interaction would provide another direct link between the cyclin control system and the initiation of S phase. There are numerous small molecules which can inhibit cyclin-CDK activity, depending on the stoichiometry of their binding to C D K complexes. These 7 Chapter 1: Introduction Figure 1.1: The cell cycle. Gl , S, G2 and M phases are shown. M phase is subdivided into prophase, metaphase, anaphase, and telophase. Points at which various cyclin-CDK complexes are thought to be active in triggering cell-cycle events are indicated: cyclin D during Gi, cyclin E at the Gi/S border, and cyclin B at the G 2/M transition. 8 Chapter I: Introduction (A) G2/M TRANSITION T14 Y15 Cyclin B-CDK1@(p) (inactive) • T161 Cyclin B-CDK1 @ (active) G2/M Transition CDC25C Cyclin H-CDK7 CHK1? DNA Damage (B) G1/S TRANSITION p21 <«- p53 A- DNA Damage Cyclin D-CDK4(P, ? Cyclin E-CDK2(p) Rb-E2F (inactive) E2F (active) G1/S Transition • Rb(P Figure 1.2: G2/M and Gi transitions. (A) The G2/M transition is triggered by activation of cyclin B-CDK1. In G 2 arrest, D N A damage may act through a pathway involving the CHK1 and CDC25C kinases to hold Cyclin B-CDK1 in a hyperphosphorylated, inactive state, preventing entry to M phase. (B) The Gi/S transition involves the release of E2F by the action of cyclin D/E complexes. In radiation-induced Gi arrest, upregulation of p53 activates p21, which inhibits the activity of Cyclins D and/or E, preventing the normal dissociation of the Rb-E2F complexes which are involved in triggering the start of S phase. 9 Chapter 1: Introduction molecules appear to play key roles in regulating cyclin activity in Gi phase in response to growth, differentiation, and D N A damage signals; they include pi5, pi6, p21, and p27 (for a good review, see Hunter et al. 1994). Briefly, pi5 and pi6 are structurally similar proteins which specifically bind and inhibit CDK4 and CDK6 (the Gi phase CDK's); it has been hypothesized that the pl6 gene, which is mutated in most tumor cell lines, is in fact the tumor suppressor gene on chromosome 9 (called M L M ) which is implicated in some familial melanomas. The p21 protein binds and inhibits several different cyclin-CDK complexes which are important in Gi and S phase (including cyclin D-CDK4, cyclin E -CDK2, and cyclin A-CDK2), as well as inhibiting PCNA. Unlike pi5 and pi6, however, there is currently no evidence that p21 is mutated in human tumors. The p27 protein shares significant amino-acid sequence similarity with p21, and also inhibits CDK4 and CDK2 complexes; low levels of p27 have been associated with tumor progression in human breast cancer (Catzavelos et al. 1997). It is important to note that the function of C D K "inhibitors" like p21 is probably not limited to simple inhibition. For example, there is evidence that p21 also acts as an assembly factor and targeting agent for cyclin-CDK's, and as noted above, the nature of p21 action is dependent on the stoichiometry of its binding with C D K complexes (LaBaer et al. 1997). G2/M Transition Cyclin B protein levels rise through late S and G 2 phase. During interphase, the yeast Cyclin B-CDK1 (CDK1 is also known as CDC2) complex is kept in an inactive state by phosphorylation on tyrosine-15 and threonine-14. In yeast, Weel /Mikl protein kinases carry out this phosphorylation, hence maintaining the inhibition of the CDK1 complex. At 10 Chapter I: Introduction the G 2 / M transition, the activation of the cyclin B-CDK1 complex (also known as the mitosis-promoting factor, or MPF) appears to be effected by further transfer of phosphate groups: CDC25C dephosphorylates tyrosine-15 and threonine-14 of the CDK1 complex, while cyclin H-CDK7 kinase phosphorylates threonine-161 of CDK1. The active cyclin B-CDK1 complex can then phosphorylate critical targets (e.g. histones and nuclear lamins) that are essential for M phase events like chromatin condensation and degradation of the nuclear envelope (Maity et al. 1994) (figure 1.2A). MPF is rapidly degraded at anaphase by a ubiquitin pathway: mutations in a short conserved sequence at the N -terminus of the cyclin B protein, which is termed the "destruction box", stabilize cyclin B against mitosis-specific proteolysis (King et al. 1994). 1.3 Models of the cellular radiation survival response 1.3.1 Description of survival models In this thesis, cell survival is defined as the fraction of normally viable cells in a given population which remain clonogenic after receiving a dose of ionizing radiation. Clonogenicity is defined as the unlimited capacity to divide and produce daughter cells. In practical terms survival is calculated from plating efficiencies, which are defined as the fraction of the total number of cells in any sample which can form colonies. The survival after a radiation dose X, Sx, is given by: s ( 1 .1 ) X PE0 ' 11 Chapter 1: Introduction where PEX is the plating efficiency after a dose X, and PE0 is the plating efficiency in an untreated control sample. In this way survival is normalized to a value of 1 for unirradiated samples. Cell survival is often plotted on a semi-log plot as a function of radiation dose; the resulting curve is simply called a survival curve. Why measure cell survival? Primarily because cell killing (the destruction of clonogenic capacity, used interchangeably here with the term "cell inactivation") is the most important route by which ionizing radiation exerts its acute effects in living organisms. For example, it is cell killing in rapidly proliferating tissues such as the bone marrow and the digestive tract that produces acute radiation sickness, and it is radiation induced cell killing that is responsible for the regression of cancerous tumors after radiotherapy. Of course, radiation also produces non-lethal alterations to DNA, yielding transformations and mutations which can lead to carcinogenesis and genetic effects. These are long-term (or "late") effects and are another part of the harm associated with environmental exposure to radiation. For investigators interested in measuring cell killing by ionizing radiation, it is fortunate that there are excellent survival assays designed for cells in vitro. The first measurements of mammalian cell survival curves in vitro were carried out by Puck and Marcus in the 1950's (Puck et al. 1956). Survival was calculated as the fraction of initially viable single cells in a sample which formed colonies of greater than 50 cells over a growth period of about 2 weeks. This clonogenic assay has since become ensconced as the standard way to measure cell survival. Other assays of cellular viability based on different endpoints, such as membrane integrity, have been developed, but in the strict sense, these are not survival assays according to the standard definition which is adopted in this thesis. 12 Chapter 1: Introduction The earliest measurements of survival in viruses and bacteria showed that the survival curves of these organisms were exponential in form (i.e. they were well fitted by straight lines on a semi-log plot of survival against dose). Many of these early survival measurements are now somewhat difficult to unearth, but Lea (1946) provides references to a-particle survival measurements in bacteria dating back to at least 1912, and shows exponential curves for the y-ray inactivation of viruses dating from as early as 1940. In the 1950's, starting with the work of Puck and Marcus, mammalian cells were found to have "shouldered" survival curves (i.e. they were concave-downward on a semi-log plot of survival against dose). Thus new survival models were needed to account for the observed responses of mammalian cells. The following is a brief survey of some of the most commonly used survival models which have been developed in the last 50 years. In the absence of detailed information about the mechanisms of radiation induced cell killing, many early models were based primarily on "target" theory, as it was pioneered by Lea (1946). In these models it was postulated that each cell contained one or more unspecified critical "targets". If all the targets were hit some critical number of times by ionizing radiation, the cell would be inactivated. Numbers of hits per target per unit dose were based on Poisson statistics. The multi-target plus single-hit model is a generalization of this type of model (Bender et al. 1962), and is described by the equation: S(D)=e-^D[\-(\-e-^D)n], (1.2) where /L and X2 are inactivation coefficients for single hit and multiple-target processes, D is the radiation dose, and n is the number of targets involved in the multiple target process (each target requiring at least one hit). This expression was widely used in the past, but has more recently been supplanted by other models. 13 Chapter 1: Introduction Later models incorporated a more sophisticated description of radiation damage which appeared to be consistent with the known molecular structure of DNA, and observations linking chromosome aberrations to cell killing. Based on experimental data, these models postulated that D N A double strand breaks were lethal lesions which could be produced by a single ionizing particle track, or by the interaction of sublethal damage from two independent particle tracks which were suitably close in space and time. Examples of this type of model include the linear quadratic (LQ) model (Chadwick et al. 1973, Kellerer et al. 1972). The LQ model is described by the equivalent expressions S(D)=e~aD-pDl O-3) -\nS(D) o n (1.4) or = ct + BD v ' D y where a and B are inactivation coefficients linked to single and double track actions, respectively. A simple extension of the LQ model is the linear-quadratic-cubic (LQC) model, which is given by S(D) = e'aD-f,Dl-^ 0-5) -\nS(D) n 2 (1.6) or ^-L = a + 8D + yD . v ' D Another model of this type is the Neary model of chromosome aberrations (Neary 1965), given by S(D) = e - 'Mi-« l l J > [Hi -« - < * 1 -*" i , > 2 ]> ) (1.7) where A}, X2, and X3 are coefficients related to the number of targets, the probability of interaction between damaged targets, and the probability of lesions occurring in any given target by a single or double track mechanism. 14 Chapter 1: Introduction In another class of models, the emphasis was not on the physical description of damage deposition in DNA, but rather on the kinetics of cellular repair of the damage. In these models it is typically postulated that damage, which is induced by ionizing radiation tracks, is repaired and/or misrepaired by a repair mechanism which obeys some explicit kinetic rules. One prominent example of this type of model is the lethal-potentially lethal (LPL) damage model (Curtis 1986), which for high dose rates and short irradiation times can be expressed as S(D) = e - ( " ^ ) D [1 + — — (1 - e~ent' )]E, ( 1 8 ) (SPL I &2PL ) where rjL and rjpi are the coefficients of production of lethal and potentially lethal lesions, S = S P J S 2 P L , and sPL and S2PL^XQ the rates of correct repair and binary misrepair respectively. Another example of this type of model is the repair-misrepair (RMR) model (Tobias et al. 1980), which postulates in its two-parameter form that S(D) = e-">[l + ^ r, ( L 9 ) e where Sis a lesion-induction term, and the s coefficient is a ratio of correct self-repair to lethal quadratic misrepair. A third example of this type of model is the repair saturation model (Goodhead 1985). In this four-parameter model, it is hypothesized that lesions are induced proportionately to radiation dose (i.e. by one-track action) with proportionality constant a (lesions/Gy), a proportion p of those lesions are lethal, and cellular repair is a saturable process with rate constant k which is dependent on the number of lesions, and the number of available repair enzymes. The initial number of repair enzymes is c0, and this number is depleted by repair. The dependence of survival on dose D in this model is: 15 Chapter I: Introduction S(D) = exp -p(aD-co) \-\exp(k(co-aD)) au (1 1 0 ) 1.3.2 Why examine deviations from the LQ model? As can be seen from the preceding section, there are a large number of disparate survival models based on different mechanistic schemes which purport to describe the creation of radiation-induced damage in cells, and the cells' reactions to that damage. Two obstacles have made it impossible to establish the validity of any one model of cell killing by showing that one model fits experimental data better than any other. The first is experimental: statistical limitations in the traditional measurement of cellular survival create large uncertainties in experimental survival determinations (Boag 1975). As a result, it has been difficult to exclude models of cell killing on the basis of poor fit to experimental data. The second problem is much more fundamental: it can be shown that models which postulate completely different mechanisms of radiation induced cell killing may, with appropriate parameter values, predict very similar radiation survival under some conditions. Thus an equally good fit to experimental data may be achieved by multiple incompatible models. To emphasize this fact, it has been shown that the first few terms of infinite series McLaurin expansions of the LQ, Neary, LPL, and R M R models can be matched exactly, implying that these models can be made indistinguishable when survival is acceptably close to unity by the appropriate choices of parameter values (Tobias 1985). Given these facts, why should one compare experimental data to any survival model? One reason is logical: while it is not possible to prove the validity of any survival 16 Chapter 1: Introduction model by fitting to experimental data, the presence of consistent deviations of experimental survival from model predictions, outside those expected on the basis of sampling statistics, would allow one to reject a model, and such a rejection would force the creation of newer, better models. Of all the models described in Section 1.3.1, the LQ model has been the most extensively used over the past 20 years to quantify and predict radiation survival in numerous situations. These include the quantification of the survival of asynchronous or synchronized tumor cells in vitro (e.g. Biade et al. 1997, Fertil et al. 1980, Gillespie et al 1975, McGinn et al. 1994), the modelling of early and late responses to radiotherapy in vivo (e.g. Barendsen 1982, Dileto et al. 1996, Douglas 1982, Douglas et al. 1976, Douglas et al. 1979, Hopewell et al. 1991, Thames et al. 1982, Travis et al. 1987), and the theoretical prediction of the effects of alternative radiation therapy regimes which are now in clinical trials (Horiot et al. 1992, Williams et al. 1985). Given this ubiquitous usage, deviations from the L Q model in any system where it is currently applied would be of interest from both radiobiological and clinical standpoints. 1.3.3 Known departures from the LQ model There are cases where deviations from the LQ model in specific systems have been investigated in detail. These deviations can be divided into two distinct types: greater-than-LQ-resistance to radiation as dose is increased, which has been theoretically predicted on the basis of population heterogeneity, and less-than-LQ-resistance to very low doses (< 1-2 Gy), known as the hypersensitive response. 17 Chapter 1: Introduction The hypersensitive response A hypersensitivity to killing by ionizing radiation at very low doses below 1 Gy has been observed in some types of both rodent (Marples et al. 1993) and human (Wouters et al. 1996) cells. In these cell lines, the entire cell population appears to be more sensitive to ionizing radiation at low doses below 2 Gy than would be predicted based on fitting to 2 or more decades of cell kill by the LQ model. Above the hypersensitive region, the cells appear to acquire a more radioresistant response. It has been hypothesized that such cells require some threshold dose to induce molecular radioresistance mechanisms, such as upregulation of D N A repair pathways (Joiner et al. 1996, Lambin et al. 1996, Marples et al. 1997). Below this threshold dose, the cells do not efficiently respond to radiation damage, and are therefore exceptionally sensitive to low doses (Wouters et al. 1997). The LQ model does not account for such a mechanism. Putative population substructure At doses above ~2 Gy, deviations from the LQ model have been found in the form of greater-than-LQ-resistance to ionizing radiation, which is conveniently observed as a downward curvature in a plot of the negative logarithm of survival divided by dose against dose: -\n(S)/D vs. D. If cells followed the LQ model, such a plot should be purely linear, with y-intercept a and slope P (see equation (1.4)). In this thesis, deviations from the L Q model of this type will be referred to as "substructure". Substructure has been observed in the acute responses of rodent cells (Skarsgard et al. 1991, Skwarchuk et al. 1993) and human tumor cells (Skarsgard et al. 1996) in vitro, as well as indirectly in the fractionated 18 Chapter I: Introduction radiation responses of rodent cells in vivo (Douglas et al. 1979, Vegesna et al. 1985). When noted, the response has been attributed either to all cells following a non-LQ survival response (Douglas et al. 1979), or to the presence of multiple populations of cells with heterogeneous radiation responses in the irradiated cell samples (Skarsgard et al. 1997). The latter "population substructure" hypothesis is supported by observations of reduced substructure in partially synchronized rodent cells (Skwarchuk et al. 1993), and by theoretical calculations (Schultheiss et al. 1987) demonstrating that population heterogeneity leads to deviations from the LQ model which are qualitatively similar to those that were observed in fractionated radiation experiments. However, this substructure has not been directly correlated with population heterogeneity in human tumor cells. 1.4 Cell-cycle related responses to ionizing radiation; apoptosis 1.4.1 G 2 block Blocks in G 2 phase after exposure to ionizing radiation were measured by many early workers in radiation biology (Terasima et al. 1963). It was hypothesized that the G 2 block provided time for a cell to repair critical damage before proceeding to mitosis; this idea of a "pause for repair" appears to date back to Lea (1946). The repair hypothesis has been reinforced by experiments which showed that long G 2 blocks in human cell lines were correlated with radiation resistance (see refs. in Maity et al. 1994), and that drugs which abrogated G 2 block sensitized human carcinoma cells to radiation (e.g. Wang et al. 1996). The G 2 block is dependent on the dose and the age of cells at the time of 19 Chapter 1: Introduction irradiation: the closer HeLa cells are to the end of G 2 phase, the longer they will block in G 2 after irradiation (Terasima et al. 1963). Radiation appears to induce G 2 block by inhibition of of the cyclin B-CDK1 complex at multiple levels. In some irradiated cells which are blocked in G 2 , cyclin B rnRNA levels are reduced, possibly due to destabilization of the message (Bernhard et al. 1994), and cyclin B protein levels are also held low, while cyclin A protein levels remain abnormally high (Muschel et al. 1992, Muschel et al. 1993). These reduced levels of cyclin B are implicated causally in the G 2 block, since artificial induction of cyclin B1 can shorten the duration of the block in some cell types. However, in addition to changes in cyclin B protein levels, the phosphorylation status (and thus activity) of the cyclin B -CDK1 complex may also be modulated during G 2 block. Recent findings suggest, for example, that the CDC25C kinase which activates CDK1 may itself be inhibited during G 2 block by radiation-induced activation of CHK1, which deactivates CDC25C via phosphorylation (figure 1.2A). A recent review (Hwang et al. 1998) provides more details of these mechanisms. 1.4.2 G i arrest Some cell types appear to respond to ionizing radiation exposure by arresting in the Gi phase of the cell cycle. This was first observed in human fibroblasts 30 years ago (Little 1968). The Gi arrest in response to ionizing radiation is currently believed to be induced by a p53-dependent (Kastan et al. 1992) upregulation of p21, a C D K inhibitor which inhibits the transition from Gi to S phase (el-Deiry et al. 1993, Harper et al. 1993), 20 Chapter I: Introduction although other proteins such as gadd45 may play a role (Rudoltz et al. 1996). Inhibition of cyclin D and/or E by p21 prevents the release of E2F from the Rb-E2F complex, hence preventing the transcription of genes which are necessary for S phase (figure 1.2B). Viral oncoproteins can interfere with the normal regulation of the Gi arrest checkpoint. For example, papillomavirus E6 can bind to and inactivate p53 (preventing radiation-induced Gi arrest), and papillomavirus E7 binds to the retinoblastoma protein Rb, causing the release of E2F (and facilitating the Gi/S transition) (DeWeese et al. 1997, Kessis et al. 1993). Gi arrests appear to be more prolonged than G 2 blocks in some cell lines, with investigators reporting radiation-induced Gi arrests lasting 48 hours or more. However, the accuracy of observations of the Gi arrest in asynchronous cells have been questioned because of potential confounding effects caused by the progression of non-arrested cells (see e.g. Nagasawa et al. 1995). Furthermore, it has been argued that the capacity of p53 to mediate Gi arrests in established tumor cell lines is much less than in untransformed cells, based on investigations which suggested that tumor cell lines did not exhibit prolonged Gi arrests (Li et al. 1995, Nagasawa et al. 1998). It is possible that the lack of a Gi arrest in some p53-competent tumor cell lines is due to deficiencies in the action of other proteins; for example, at least one other protein, p33 m G 1 , interacts with p53 in human cell lines and is necessary for p53-mediated growth inhibition (Garkavtsev et al. 1996, Garkavtsev et al. 1998). Based on extrapolation from earlier evidence suggesting that the G 2 block provided time for cells to repair D N A damage before entering mitosis, many investigators have hypothesized that the Gi arrest serves a similar purpose, allowing for cellular repair 21 Chapter 1: Introduction before the initiation of D N A replication in S phase. Such a theory is consistent with evidence that potentially lethal damage repair (PLDR) stops once cells enter S phase (Iliakis et al. 1983). However, proof of this "time for repair" hypothesis has remained elusive. Many experiments have examined the effects of abrogating p53 function on Gi arrest and radiation response; the overwhelming majority of these studies have found that elimination of normal p53 function reduces or eliminates the Gi arrest after ionizing radiation exposure (Chang et al. 1997, Griffiths et al. 1997, Gupta et al. 1996, Haas-Kogan et al. 1996, Mcllwrath et al. 1994, Pellegata et al. 1996, Siles et al. 1996, Slichenmyer et al. 1993, Yount et al. 1996), and makes various cell types radioresistant (Chang et al. 1997, Delia et al. 1997, Griffiths et al. 1997, Gupta et al. 1996, Haas-Kogan et al. 1996, Lee et al. 1993, Lowe et al. 1993, Mcllwrath et al. 1994, Siles et al. 1996, Tsang et al. 1995, Yount et al. 1996). A few studies suggest no effect of p53 modulation on radiosensitivity, or a radiosensitization after p53 abrogation (Pellegata et al. 1996, Slichenmyer et al. 1993). Most investigators have concluded that in cell lines which undergo radiation-induced apoptosis, active p53 promotes this mode of cell death and therefore tends to make cells radiosensitive (see e.g. Lowe et al. 1993). Of course, p53 is a known modulator of numerous cellular pathways involved in cell cycle control and damage repair, so there is little direct evidence that any of the radiosensitizing effects of p53 in the above mentioned experiments are specifically due to a cell cycle arrest in Gi phase. It has been suggested that inducible processes like the Gi arrest do not, of themselves, play a major role in the radiation response of some human tumor cells (Schwartz 1994). However, some investigators have argued that the Gi arrest itself, independent of apoptosis, makes cells radiosensitive; a prolonged or permanent arrest in 22 Chapter I: Introduction Gi would clearly tend to lower the survival of cells as it is measured in a clonogenic assay. Indeed, these investigators have observed that some Gi arrest-competent cells are more radiosensitive than isogenic p53/Gi arrest-deficient cells, specifically in Gi phase, and that this effect is independent of apoptosis (Haas-Kogan et al. 1996, Yount et al. 1996). Others have argued that the transient G i delay that is observed in other human cells is unlikely to have much effect on radiation survival (Olive et al. 1996). In summary, two very basic questions about the Gi arrest are currently the subject of debate: (1) Do prolonged Gi arrests occur in a significant number of human tumor cell lines? (2) What is the impact of such arrests, if they occur, on human tumor cell radiosensitivity? 1.4.3 Apoptosis Apoptosis is a process by which cells intentionally self-destruct (Steller 1995); the process was given its name by Kerr and Wylie (1972) (for a review of early work see Kerr et al. 1980). Ionizing radiation can induce apoptosis in numerous cell types (Meyn et al. 1993, Mirkovic et al. 1994, Stephens et al. 1991, Stephens et al. 1993, Story et al. 1994, Thompson 1995). The early markers of apoptotic cell death (which can include inter-nucleosomal D N A fragmentation and chromatin condensation) can sometimes be clearly differentiated from the indicators of necrotic cell death (e.g. loss of membrane integrity), but categorization of apoptotic and non-apoptotic cells in practice can be difficult (Olive et al. 1997). The molecular mechanisms of apoptosis are currently the subject of intense investigation, and a complete description of what is known about these mechanisms is 23 Chapter 1: Introduction beyond the scope of this thesis. In radiobiology, it is only recently that recognition of apoptosis as an important mode of cell death in some cell types has become widespread (Blank et al. 1997). There has been intense debate about the potential roles of apoptosis in tumor cell biology: it has been suggested that deregulation of apoptosis confers a growth advantage to some tumor cells (Graeber et al. 1996), and apoptosis is a significant mode of tumor cell killing by radiation; hence the induction of apoptosis in tumor cells may be a useful therapeutic strategy (Milas et al. 1994). There has also been much interest in the role of apoptosis in determining the dose-dependence of radiation survival (Ling et al. 1994, Ling et al. 1994, Olive et al. 1997). Current evidence would suggest that the induction of apoptosis by ionizing radiation follows approximately the same dose-dependence as clonogenic cell killing (Langley et al. 1995). The dependence of apoptosis on position in the cell cycle appears to be cell-type dependent (e.g. Langley et al. 1995, Olive et al. 1996); cell-age dependent changes in the apoptotic fraction of cells are not clearly correlated with changes in clonogenic survival. 1.5 Radiosensitivity varies over the course of the cell cycle in rodent, HeLa, and some other human cell types. 1.5.1 Previous investigations of cell-age related variation in radiosensitivity A number of important studies done in the 1960's provided some of the first measurements of "cell-age dependent" variation in radiosensitivity through the cell cycle, also known as the "age response" (these studies include Dewey et al. 1962, Djordjevic et 24 Chapter 1: Introduction al. 1967, Erikson et al. 1963, Hahn et al. 1966, Sinclair et al. 1963, Sinclair et al. 1965, Sinclair et al. 1966, Terasima et al. 1961, Terasima et al. 1963, Terasima et al. 1963, Vos et al. 1966, Whitmore et al. 1965). The studies of Tolmach's and Sinclair's laboratories especially are widely cited today as benchmarks in standard texts (see for example, chapter 6 in Hall 1994). The major findings of these early studies, which were almost all carried out in rodent or HeLa human cells, can be briefly summarized as follows (see the review by Sinclair 1968): 1. In Chinese hamster cell lines with a short Gi phase, radiosensitivity was greatest during mitosis, slightly less during Gi , and then decreased until the end of S phase. Radiosensitivity in G 2 increased again until mitosis. 2. In HeLa and some hamster cell lines with a long Gi phase, radiosensitivity followed the same pattern as above, but there was usually a resistant peak sometime around mid-Gi phase, followed by a increase in radiosensitivity towards S-phase. However, studies in mouse L cells and human kidney cells provided some exceptions to this rule. 3. In most cell lines radiosensitivity was minimal in later S phase. 4. The radiosensitivity of synchronous cells varied with time even when D N A synthesis was inhibited. Since the 1960's there have been numerous examinations of the age response of rodent cells in vitro or in vivo, using various cell synchronization techniques. For example, Grdina (1980) and Keng (1984) used centrifugal elutriation to examine the age response of rodent cells, with results that were interpreted to be generally in agreement with earlier investigations. Chapman (1970) showed that some Chinese hamster cells 25 Chapter I: Introduction resuming growth after plateau phase exhibited a peak of radioresistance during the first Gi phase after release from growth inhibition. It is striking that only two of the studies cited above (Erikson et al. 1963, Vos et al. 1966) measured the radiosensitivity of human cells other than the HeLa cell line. More recently, there have been some studies of the cell-age dependence of radiation survival in human tumor cells other than HeLa. Blakely (1985) found the age response of mitotically selected human T - l cells to be qualitatively similar to that of HeLa cells for low L E T radiation: resistant in early Gi phase and later S/G2 phase, and sensitive in mitosis and near the Gi/S border. At higher LET the age response of T - l cells was relatively flat (Blakely et al. 1984). Drewinko (1977) found somewhat similar results: LoVo human adenocarcinoma cells synchronized by tritiated thymidine block were relatively radioresistant in late Gi, and radiosensitive in early S and early Gi phase. In contrast, Petterson (1977) determined the age response of mitotically selected human NHIK 3025 cells to be markedly different from the HeLa age response: they were resistant in Gi phase, but uniformly sensitive through S phase, and even more sensitive in G 2 / M phase. Five other investigations of the age response of human tumor cells have also reported varied results (Biade et al. 1997, McGinn et al. 1994, Quiet et al. 1991, Tang et al. 1994, West et al. 1988); these studies will be discussed in more detail in Chapter 4. 26 Chapter 1: Introduction 1.5.2 Explanations for cell cycle related variation in radiosensitivity The factors which might determine cell-age dependent variation in radiosensitivity can be broadly organized into two categories: initial radiation damage, and D N A damage recognition and repair pathways. The first variable that has the potential to explain cell-cycle dependent variation in radiosensitivity is the yield of initial damage from ionizing radiation. Damage yield could be modulated by cell-age dependent changes in D N A conformation (Dewey et al. 1972) or intracellular levels of radioprotectors such as sulfhydryls (e.g. Blakely et al. 1988, Han et al. 1976). Unfortunately, assays of D N A damage in cell-age specific cells are often confounded by age-dependent differences in D N A mobility that are unrelated to radiation damage (Iliakis et al. 1991). There is some limited evidence that different levels of initial damage after X-irradiation in radiosensitive and radioresistant variants of one rodent cell line may correlate with radiosensitivity (but only in G2 phase) (Wlodek et al. 1988). On the other hand, there is also evidence from other cell lines that the yield of initial D N A damage from ionizing radiation does not change significantly throughout the cell cycle (e.g. Olive et al. 1990). Once D N A damage is present, the ability of a cell to recognize and repair that damage could be cell-age dependent if, for example, there were cell-cycle dependent fluctuations in the activity of important elements of damage sensing and repairing pathways like D N A dependent protein kinase (DNA-PK) or the R A D proteins (e.g. Chen et al. 1997, Lee et al. 1997). Some investigators have provided evidence that fluctuation in the level of PLDR is correlated with variation in radiation survival in synchronized rodent cells (fliakis et al. 1983, Wlodek et al. 1988). However, clear correlations between 27 Chapter 1: Introduction D N A repair capability and clonogenic survival in age-specific cells have not always been achieved (see e.g. Bussink et al. 1996). Alternatively, there could be cell-age dependent variations in the activity of other molecules which are hypothesized to play a role in deciding whether damaged cells continue to proliferate, stop cycling, or undergo apoptosis. Two candidates for such molecules are the ataxia-telengectasia gene product (ATM), and p53 (Meyn 1995). 1.6 Relevance of intrinsic tumor cell radiosensitivity to radiation therapy 1.6.1 Accurate knowledge of low dose intrinsic radiosensitivity in tumor cells may be important for designing radiotherapy regimes The outcome of clinical radiotherapy is determined by a host of factors, including normal tissue responses, the intrinsic radiosensitivity of tumor cells, the growth rate of tumors, and the rate of tumor cell loss (Hall 1994). Although the issue is somewhat controversial, studies suggest that the radiosensitivity of cell lines in vitro is correlated with positive outcomes in radiotherapy (Fertil et al. 1985, Schwartz et al. 1996). Our current understanding of radiosensitivity at clinically relevant low doses (< 2 Gy) is limited for several reasons, including the lack of a consistent response among different cell types. This inconsistency may be exaggerated by statistical limitations on the precision of survival measurements when the surviving fraction is close to 1. Nevertheless, some of the measurements which have been made suggest the presence of effects at low doses, such as hypersensitivity and heterogeneity in the responses of mixed cell populations (Section 1.3.3), which could have clinical significance. For these reasons, high-precision 28 Chapter 1: Introduction measurement of low-dose radiation survival in human tumor cells has the potential to supply valuable information in the ongoing effort to improve radiotheraputic outcomes. 1.6.2 Cell-cycle related variation in radiosensitivity may be an important factor in radiotherapy There are numerous reasons why variation in radiosensitivity through the cell cycle could impact the outcome of radiotherapy in the clinic. Three examples are given here, related to cell-cycle redistribution, combined radiotherapy and chemotherapy (fluorodeoxyuridine), and the relative responses to radiation of normal and tumor tissue. Redistribution in fractionated radiotherapy Redistribution of tumor cells throughout the cell cycle has been hypothesized to contribute to the sensitivity of rapidly cycling tumor cells during the course of a fractionated radiotherapy regime (Hall 1994, Withers 1975), based largely on measurements of cell-age related changes in radiosensitivity in rodent and HeLa cells (see e.g. Denekamp 1986). According to this hypothesis, an initial dose of radiation will preferentially kill rapidly cycling tumor cells that are at sensitive stages of the cell cycle, leaving behind tumor cells at relatively resistant stages of the cycle. These resistant cells would then redistribute into more sensitive stages of the cycle in the interval before the next fraction of radiation. The net effect of such redistribution would be to preferentially sensitize rapidly cycling tumor cells compared to slowly- or non-cycling normal tissue cells, which would not redistribute. Obviously the clinical importance of the redistribution 29 Chapter 1: Introduction effect can only be predicted if the degree of radiosensitivity variation as a function of cell age is known in varied types of human tumor cells. Mechanism of action of fluorodeoxyuridine Fluorodeoxyuridine is a widely used chemotheraputic agent in the treatment of various neoplasms (reviewed in McGinn et al. 1993). It is believed that this halogenated pyrimidine radiosensitizes cells at least partially by synchronization of cells in early S phase (McGinn et al. 1993, McGinn et al. 1994). This hypothesis depends on early S-phase being a relatively radiosensitive part of the cell cycle; preliminary investigations of the radiosensitivity of S-phase HT-29 cells have not clearly confirmed this theory, and further investigation of cell-cycle related variation in radiosensitivity in these cells has been suggested (McGinn et al. 1994). Role of cellular heterogeneity in biasing comparisons of the a/B ratio in early and late responding tissues. A central goal of fractionated radiotherapy is to minimize effects on normal tissue while maximizing effects on tumor tissue. It has been proposed that giving a course of radiotherapy in a larger number of smaller fractions over the same time period (e.g. daily treatments of 2x 1 Gy instead of 1 x2 Gy over a course of ~4-6 weeks) would reduce detrimental effects in later-responding normal tissues more than it would spare tumor cells. This would provide a theraputic advantage because it would allow the total dose imparted to the tumor to be increased without corresponding escalation of unwanted late 30 Chapter 1: Introduction tissue responses. Such a strategy is known as hyperfractionation. Important evidence in favour of the hyperfractionation proposal has emerged from the measurement, in various tissues, of the total doses required for a defined effect level as a function of the dose given per fraction. Measurements of this type are often summarized in a "F e plot", which is a plot of (inverse total dose) versus (dose per fraction) for a constant effect level. When early and late-responding tissues are compared on such plots, isoeffect curves for late responding tissues have a greater slope than that for early responding tissues. In terms of a LQ survival model, this implies that late responding tissues have a smaller a/p ratio than early responding tissues (Thames et al. 1982). The corollary of this observation is that a reduction in the dose per fraction will reduce unwanted late-tissue responses more than it will reduce tumor cell kill, allowing the total dose to be increased with a concomitant theraputic gain. Thus hyperfractionation should bring a net benefit to the patient who is undergoing radiotherapy. One criticism of the just-described F e plot analysis has emerged which has the potential to weaken this argument for hyperfractionation. Schultheiss (1987) has shown theoretically that the use of a linear quadratic survival model to fit the heterogeneous radiation response of a mixed cell population will lead to biased overestimates of the a/p ratio. If relatively rapidly cycling early responding tissues have more cellular heterogeneity than more slowly cycling late-responding tissues, this implies that the typical F e plot analysis could overestimate the difference in the a/p ratios between early- and late-responding tissues, and therefore this argument for a therapeutic advantage from hyperfractionation would be overstated. Furthermore, the data used to generate F e plots, 31 Chapter 1: Introduction and hence a/p ratios, has been largely limited to doses per fraction well above the clinically relevant 2 Gy/fraction range in an overwhelming majority of cases (see e.g. tables 1-3 in Williams et al. 1985); this would tend to exacerbate the bias in the a/p ratios which is introduced by heterogeneity in cellular radiation responses. Given this possibility, it appears that measurements of cell-age related heterogeneity in the radiation response of human tumor cells may contribute to our analysis of possible benefits or drawbacks of alternative radiotherapy schemes. 1.7 Objectives of this thesis The objectives of this thesis were threefold: 1. To measure the cell cycle related changes in the radiosensitivity of three human tumor cell lines, and compare them to the patterns of cell-age variation that have been measured previously in other cell lines. 2. To determine if the survival responses of asynchronous human tumor cell lines can be explained by the presence of multiple cell-age specific subpopulations with simple LQ survival responses. 3. To establish the characteristics of Gi arrest in human tumor cell lines, and examine the role that the Gi arrest response plays in determining the radiation survival responses of these cells. 32 Chapter 2: Methods 2.1 Cell Lines Human tumor cell lines HT-29 (colon adenocarcinoma), DU-145 (prostate carcinoma, isolated from metastasis to brain), A549 (lung carcinoma), MCF7 (breast adenocarcinoma, isolated from pleural effusion), and HT-144 (malignant melanoma, isolated from metastasis to subcutaneous tissue) were obtained from American Type Culture Collection (Rockville, MD). Cell line Ul (melanoma) was obtained from Dr. J.B. Mitchell. MCF-7 , HT-144, and DU-145 were examined in preliminary experiments only. We did not carry out complete studies with MCF-7 because we were not able to achieve an adequate degree of synchronization in mitotically selected MCF-7 populations. The A549 cell line was between passage numbers 73-247. The HT-29 cell line was between passage numbers 125-209. The total passage number of the Ul cell line was unknown, but all experiments with this cell line were carried out between passage numbers 2-34 of our frozen stock cells. 2.2 Routine cell culture All cell lines were maintained as monolayer cultures in disposable plastic tissue . culture flasks (Falcon or Nunc). HT-29, DU 145, and A549 were grown in McCoy's 5A medium (Gibco) containing 2.2 g/L sodium bicarbonate as a buffer and 10% (v/v) fetal bovine serum (FBS) (Gibco). Ul and HT-144 were grown in D-MEM medium (Gibco), containing the same amounts of sodium bicarbonate and fetal bovine serum. Cells were 33 Chapter 2: Methods subcultured twice a week by trypsinization and reseeding at reduced density. Trypsinzation consisted of a 3 minute incubation with either 0.05% trypsin containing 0.53 mM E D T A (for HT-29, D U 145, and A549) or 0.1% trypsin (for U l ) . Cell cultures were checked periodically for mycoplasma contamination. 2.3 Selection of synchronous cells by mitotic rolloff Populations of synchronized mitotic cells were prepared by using a modification of the standard mitotic shakeoff procedure (Terasima et al. 1961). 25-50 x 106 cells were seeded into each of 1-4 1750 cm2 plastic roller bottles (Falcon), in 250 mL of prewarmed growth medium per bottle. Cell suspensions were gassed for 40 seconds with 5% CO2, 95% air, and bottles were closed by adding derotator caps. Bottles were then fitted into a roller bottle table (Talandic Research Corp., Pasadena, CA) which rotated the bottles at selected speeds. Bottles were rotated at approximately 0.2 R P M (speed setting A-14) for 2-6 hours after seeding, during which time the cells attached to the surface of the bottles. Once attachment was deemed to be complete, the rotation speed was increased to approximately 1 R P M (speed setting A-3.1) for a 30-48 h growth period before harvesting of mitotic cells. During this time, the cultures grew to approximately 80% confluence. To harvest mitotic cells, the bottles were spun at approximately 200 R P M (speed setting C-57) for 5 minutes. Then the medium in the bottles was pumped off with a 4-head peristaltic pump (Cole Parmer, Chicago) into a pre-chilled glass collection bottle on ice. The pump tubing was rinsed by pumping approximately 30 mL fresh growth medium at 37°C into the bottles momentarily, then immediately flushing this medium into the 34 Chapter 2: Methods collection bottle. Fresh growth medium was then pumped into the bottles, at 125 mL of medium per roller bottle (see figure 2.1). Cell suspensions from each harvest were immediately transferred from the collection bottle to 200 mL centrifuge bottles (Falcon) and centrifuged at 120g for 15 minutes (at 4°C). After centrifugation the supernatant was poured off, and cell pellets were resuspended at a concentration of approximately 5 x 105 cell/mL in suspension-modified growth medium (Sigma-Aldrich, Oakville, ON) containing 10% FBS and 20 mM N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid (HEPES) as a buffer, at 4°C. These concentrated cell suspensions were held on ice until sufficient cells had been harvested for an experiment. Cells were harvested at 40 minute intervals as described in the previous paragraph. At least the first 3 harvested populations in all experiments were discarded, as they were found to include large numbers of non-mitotic cells and significant amounts of cellular debris. Cells from all later harvests were pooled together on ice. In typical experiments, 4-12 consecutive harvests were required to collect sufficient numbers of mitotic cells. 2.4 Monitoring of cell synchrony The synchrony of experimental cell populations was routinely monitored by the measurement of mitotic index and cellular DNA content. The ability of these methods to quantify cell synchrony depends on the position of cells in the cell cycle. Direct counting of the fraction of cells in a specific phase of the cycle is a good indicator of synchrony for phases which are relatively short, like mitosis. DNA content increases as cells progress through S phase, so it is a good indicator of synchrony during this phase, as long as it can 35 Chapter 2: Methods cold collection bottle Figure 2.1: Schematic of mitotic shakeoff apparatus. Cell monolayers were grown in plastic roller bottles for 30-48h prior to rolloff. To collect mitotic cells, bottles were spun at 200 RPM for 5 minutes, and the growth medium was then pumped into a cold collection bottle. Collected mitotic cells were immediately centrifuged, resuspended at higher density, and held on ice. After each rolloff, bottles were refilled with fresh warm growth medium from the feeding bottle. 36 Chapter 2: Methods be measured with acceptable resolution. On the other hand, DNA content is constant during G i and G2 phases. This means that flow cytometric measurements of DNA content, especially during the relatively long G i phase, cannot by themselves fully describe the degree of cell synchrony in an experimental population. 2.4.1 Measurement of mitotic index 5 x 105 to 2 x 106 freshly harvested mitotic cells in cold medium were centrifuged for 7 minutes at 120g. Cell pellets were then resuspended in ice-cold 0.025 M sodium citrate at a concentration of approximately 5 x 105 cell/mL and held on ice for 15 minutes. An equal volume of Carnoy's fixative (25% acetic acid, 75% ethanol) was added dropwise to the cell suspension, cells were centrifuged 10 minutes at 250g, and then resuspended in Carnoy's fixative. At this point, cells were either stored at -20°C or immediately centrifuged for 10 minutes at 250g and resuspended in 0.1-1 mL Carnoy's fixative. This concentrated cell suspension was dropped onto a clean microscope slide and air dried for 30 minutes. Slides were then stained with Giemsa stain (Sigma) for 20 minutes, rinsed in distilled water, and air dried. Stained slides were sealed with Permount. In some cases, cell suspensions were applied to microscope slides using a CytoSpin centrifuge for 10 minutes at 250g. To determine the mitotic index, at least 100 (typically 200-500) stained cells on a slide were identified as mitotic or non-mitotic; the mitotic index was calculated as the percentage of the total number of observed cells which were in mitosis. Mitotic cells in 37 Chapter 2: Methods prophase, metaphase, anaphase, or telophase were differentiated from non-mitotic cells based on the clear presence of condensed chromosomes and other morphological changes characteristic of mitotic cells (e.g. dissolution of the nuclear membrane). 2.4.2 Measurement of DNA content by flow cytometry 2.4.2.1 DAPI staining for DNA content For routine measurements of D N A content, 5x105 t o l x l O 6 cells were centrifuged for 7 minutes at 120g, and then resuspended in 0.25 mL Saline G M (1.1 g/L glucose, 8.0 g/L NaCl, 0.4 g/L KC1, 0.39 g/L Na 2HPO 4-12H 20, 0.15 g/L KFf 2 P0 4 , containing 0.5 mM ethylenediaminetetraacetic acid (EDTA) (Crissman et al. 1990)). To this cell suspension, 0.75 mL ethanol (at -20°C) was added dropwise, while vortexing. This procedure fixed the cells. Fixed cells were then stored at -20°C until the time of flow cytometric analysis. Immediately prior to flow cytometric analysis, the fixed cells were centrifuged for 7 minutes at 120g, the supernatant was aspirated away, and the cells were resuspended in 1 ug/mL DAPI (4', 6-diamidino-2-phenylindole dihydrochloride hydrate), dissolved in phosphate buffered saline (PBS) containing 0.1% TX-100; DAPI binds quantitatively to double stranded DNA. Flow cytometric analysis was carried out on a Coulter Elite flow cytometer with U V laser excitation at 351-364 nm; the emission maximum of DAPI was at 454 nm. Twenty thousand events were collected for D N A histograms; events were gated to remove fluorescent signals from cellular debris and cell aggregates. 38 Chapter 2: Methods 2.4.2.2 BrdU-FITC/DAPI staining for S-phase fraction and DNA content In some experiments, cells were pulse labelled with bromodeoxyuridine (BrdU) and analysed by dual-parameter flow cytometry for BrdU incorporation and D N A content. Preparation of these cells followed the FITC-anti-BrdU antibody manufacturer's suggested protocol (Becton Dickinson source book section 3.80.1). 2.4.3 Quantitation of cell cycle distribution by DNA histogram fitting Single parameter D N A histograms acquired by flow cytometry were analysed using the Winlist and Modfit analysis programs (Verity Software House, Topsham, ME) . The percentage of cells in Gi, S, and G 2 phases was calculated by fitting of D N A histograms with Gaussian (for Gi and G 2 phases) or polygonal (for S phase) functions. 2.5 Irradiation of cells All irradiations were carried out using a 250 kV x-ray machine with 2 mm A l and 0.5 mm Cu filtration (Phillips). Filtration of the x-ray beam removed low energy x-ray photons and strongly attenuated the characteristic x-rays resulting from photoelectric interactions in the anode. Experimental populations of cells were irradiated while in suspension, contained in 5 mL polypropylene or polystyrene test tubes (Falcon). Tubes were filled with at least 4.0 mL of cell suspension for irradiations. During irradiation, test tubes were held in a water-filled perspex jig mounted on a fixed length aluminum spacer (figure 2.2). This apparatus accurately positioned the jig relative to the x-ray beam. Water at 37°C was pumped through the jig during irradiation to maintain the temperature 39 Chapter 2: Methods of the cell suspensions, using a temperature-regulating recirculating water bath (Neslab). The x-ray dose was monitored by a dose integrator (TRIUMF biomedical) which was calibrated using an electrometer (Victoreen model 500, with 0.6 cm3 probe). The electrometer was in turn calibrated against standard dosimeters maintained at the BC Cancer Agency physics department. 2.6 Measurement of cell survival 2.6.1 Single dose survival profiles (Conventional clonogenic assay) A conventional clonogenic assay was used to measure the radiation survival of cells in the single dose studies which are described in Section 3.1. In these experiments., synchronized mitotic cells were harvested as described in Section 2.3. Pooled mitotic populations, kept at 4°C, were seeded into 20-30 10 cm diameter tissue culture dishes (Falcon) filled with 15 mL prewarmed growth medium at a density of 3-6 x 105 cells per dish. Al l dishes were seeded at the same time, which was designated t=0 h. At regular intervals from 0.5-40 h after this time, single dishes were harvested by trypsinization, resuspended in growth medium, and the cell suspension was divided into two samples. The first sample was processed for D N A content analysis (Section 2.4.2.1). The second sample was centrifuged at 120g for 7 minutes, and diluted to 6-10 x 103 cells/mL in growth medium. From this diluted sample 300-500 cells were transferred into each of three 5 mL polypropylene test tubes (Falcon), prefilled with 4.5 mL growth medium. 40 Chapter 2: Methods perspex sample holder Figure 2.2: Top view of irradiation sample holder. Sample test tubes were held in the perspex sample holder, which was filled with 37°C water from a recirculating water bath. The sample holder fitted into locating holes on the aluminum spacer bar, which was fixed to the x-ray head. In the diagram, the central 6 test-tube holes contain samples. 41 Chapter 2: Methods These control tubes were mock irradiated. From the same diluted sample 3000-5000 cells were transferred into each one of a matching set of three 5 mL test tubes and irradiated with a single dose of 5-8 Gy, as described in Section 2.5. Immediately after irradiation., cells from all 6 tubes were plated by pouring the contents of each tube into a 10 cm diameter tissue culture dish and rinsing the tube twice into the dish with warm growth medium. Concentrated feeder cells which had been irradiated with 80 Gy of Cs-137 gamma rays were then added to each dish (7 x 104 feeder cells/dish). These clonogenically dead feeder cells promoted the growth of surviving cell colonies and made the total number of cells (feeders+sample cells) in each dish approximately equal. Dishes were swirled gently to distribute cells evenly over the growth surface, and then incubated at 37°C in a humidified 5% C 0 2 atmosphere for 9-14 days to allow colonies to develop. At the end of this growth period the cells were fixed and colonies visualized by staining the dishes with a 6 g/L solution of Malachite green (Fisher Scientific) in distilled water. Colonies of greater than 50 cells were counted for determination of cell survival. 2.6.2 Survival curves (FACS presort assay) For measurement of cell survival dose response curves ("survival curves") the cell sorter assay was used (Durand 1984, Durand 1986). For this assay, experimental populations of cells at a density of approximately 50-75% confluence in 180 cm 2 culture flasks were harvested by trypsinization, centrifuged 7 minutes at 120g and then resuspended at approximately 5 x 105 cells/mL in growth medium containing 20 mM HEPES as a pH buffer. This cell suspension was held at 37°C for the duration of a 42 Chapter 2: Methods survival curve measurement (approximately 1 hour). Aliquots of the cell suspension were sorted on a Becton Dickinson FACS 440 cell sorter (FACS). An appropriate number of cells was sorted into 5 mL polypropylene tubes (Falcon) which were prefilled with growth medium at 37°C, using the FACS perpendicular and forward light scatter signals to detect cells, without the use of a cell stain. Typically, 500-80000 cells were sorted into each tube. Tubes containing sorted cells were then irradiated as described in Section 2.5. Alter irradiation the cell suspensions were added to tissue culture dishes along with feeder cells and the dishes were incubated and stained for colony counting as described in Section 2.6.1. 2.7 Calculation of cell survival Cell survival was calculated as follows. For each sample (corresponding to one tissue culture dish) the plating efficiency P was defined as: where Nc was the number of surviving colonies, and Np was the number of cells plated in that dish. Then the survival S for any sample was given by where Po was the average plating efficiency of samples which received zero dose. Survival values for many equivalently treated samples were averaged to give the mean survival S: (2.1) (2.2) 43 Chapter 2: Methods S = ^ —. (2-3) n The standard error of the mean (SEM) of S was calculated using the standard formula: Z^.-" 5 ) 2 (2.4) 5£M = l -a . «(« -1) 2.8 Fitting of measured survival curves Theoretical equations were fitted to measured survival curves using the Levenberg-Marquardt nonlinear least squares fitting algorithm (Press et al. 1992) to minimize the x 2 value between observed survival values and equivalent equation-derived values Sj for all of the m survival values in a single survival curve, weighted by the standard error cr, of each observed survival value: i=l o-, This algorithm was used either as implemented in the Microcal Origin 3.5 plotting package (Microcal, Northhampton, MA), or alternatively in a set of Fortan 77 programs which were written by Brad Wouters and modified by Andrew Hill for specific use in fitting multiple parts of survival curves. The fitting routines in these Fortran programs were validated on standard datasets by direct comparison of fits against identical fits from the Microcal Origin fitting module. 44 Chapter 2: Methods 2.9 Mathematical models of synchronized cell radiation survival A mathematical model of cell kinetics and survival was based on the state-vector model of Thames (Thames et al. 1977). The derivation of this model will be described briefly here; for full details see the reference above. In the Thames model of cell kinetics the entire cell cycle is divided into an arbitrary number of compartments, K. Of these K compartments, KGi are allocated to G i phase, KS are allocated to S phase, and KG2/M are allocated to G2 /M phase, such that K = K O l + K S + KG2/M • (2-6) A number of cells «, are distributed into the /'th compartment according to the situation which the user wishes to model, so that the total number of cells N is given by N = fjni. (2.7) At time / the cell population is described by the state vector V,, made up of the s: nK (2.8) The progression of cells is followed in time steps of arbitrary length At. After each time step, the cells in the /'th compartment have defined probabilities ah bh and c, of advancing zero, one, or two compartments forward in the cycle, respectively, where or,.+ +c, = 1. (2.9) Thus at any time t+At, the number of cells in each compartment is given by: 45 Chapter 2: Methods nx it + At) = axnx (t) + 2bKnK (t) + 2cK_xnK_x (t) n2 (t + At) = a2n2 (t) + bxnx (J) + 2cKnK (t) n3(t + At) = a3n3 (/) + b2n2 (t) + cxnx (t) (2 n4 (t + At) = a4n4 (t) + b3n3 (?) + c2n2 (t) ns (t + At) = a5n5 (t) + b4n4 (t) + c3n3 (t) etc.... Note that in the calculation of rti and ri2, the transition probabilities from compartments K and K-l are multiplied by the daughter factor (in the current model, 2.0) in order to account for the doubling of cells as they progress from the last compartment of mitosis to the first compartment of G i phase. In a typical invocation of the model, the values of KGi, Ks, and KG2/M were chosen by the user, subject to mild constraints described by Thames (Thames et al. 1977). The initial distribution of cells among the K compartments, the durations of each phase, the variances of these durations, and a value for At were also supplied by the user. Once the values of these parameters were chosen, the values of ah bh and c, were calculated using equations which are described fully by Thames. The model determined the state vector V of /?, values after each time step At by repeated application of equations (2.10). The model displays the required characteristic that regardless of the initial distribution of cells throughout the cell cycle, as / —>• oo, the state vector V approaches the same exponential-type distribution vector which is characteristic of asynchronous cells (Thames et al. 1977). The model also explicitly differentiates between biological age and time. At a given time / after a cell begins the cycle, the cell's biological age is specified by the number of the compartment /' in which it resides, not by the time /. The original Thames model was designed to predict only the kinetics of cycling cell populations. In order to extend the model to predict radiation survival, it was necessary to 4 6 Chapter 2: Methods assign specific values to the LQ survival parameters a, and /?, for each of the K compartments in the cell cycle. Then, the survival S of any particular distribution of cells after a radiation dose of D is given by: f>,exp(-a,Z)-/?,/J>2) ( S(D) = ^ . 1=1 In the models described here, the entire cell cycle was divided into 4 radiosensitivity regions (see below), which did not necessarily correspond to the Gi, S, or G2/M cell cycle stages. Within each of the radiosensitivity regions, the values of a, and /?, were constant. At borders between regions, the values of a, and /?, changed abruptly in a stepwise fashion. The objective of this modelling was to investigate the potential role of cell-cycle related variation in radiosensitivity on the survival responses of asynchronous cells. To do this, the kinetics and radiation survival responses of asynchronous and synchronized cells in a simple but realistic model were compared to the equivalent observed responses. The approach taken was to create a simple model with the minimum number of adjustable parameters, and to then select the values of those adjustable parameters by matching model predictions of kinetics and radiation responses against measurements of the same quantities. Although, as described immediately below, several adjustable parameters in the model had to be assigned values which could not be known with absolute certainty, it will be seen in Chapter 3 and 4 that the conclusions which are drawn from the kinetic and radiation survival predictions of the model are not highly dependent on the exact values of these adjustable parameters. 47 Chapter 2: Methods Adjustable parameters of the model The three step process which was used to fix the adjustable parameters in the model is outlined in figure 2.3. The adjustable parameters in the final model were: 1. The initial mitotic index of synchronized cells at the time of mitotic selection, and the durations of GI , S, and G 2 / M phases, with associated variances in duration. 2. The a, and B, parameters of the LQ model at 4 radiosensitivity stages in the cell cycle. 3. The relative positions of the borders between the radiosensitivity stages in the model, i.e. the compartment numbers i in the model at which radiosensitivity changed from one level to another. Initial mitotic index and cell cycle stage durations The initial mitotic index and the durations and variances of the Gi , S, and G 2 / M phases were simply set by varying these parameters and comparing the model predictions of fractions of cells in Gi , S, and G 2 / M during the first cycle after mitotic selection to measurements of the same fractions (step 1 in figure 2.3). The combination of mitotic index and phase durations and variances which gave the lowest root-mean-square deviation from the measurements were chosen for use in the final model. 48 Chapter 2 : Methods 1 'Si TQ2/MI Adjust T 0 1 , cQ„ " i s , CTS CTG2/M- and initial mitotic index in model to best fit measured %G1, %S, and %G2/M over the first cycle after mitotic selection. i measured time after mitotic selection (h) 2 Set a , p for each of the 4 radiosensitivity regions (A, B, C, and D) according to fits to measured survival curves of synchronized cells in M, G1, G1/S, and S/G2. 1 ro > W \ B ^ ^ A C V ^ dose (Gy) 3 Choose the location of borders between the radiosensitivity regions A, B, C, and D to best fit measured single-dose surivival profiles for synchronized cells in the first cycle after mitotic selection. 1 measured model time after mitotic selection/ model compartment The model is now completely specified. Figure 2.3: Three step process to specify model parameters. 49 Chapter 2: Methods Parameters of the LQ model throughout the cell cycle (a, and fit) The a, and /?, parameters (J=\,2,3,...,K) throughout the cell cycle were set as follows. To form the simplest model which could incorporate the basic observation that radiosensitivity was different when measured at 4 different times after mitotic selection (see Chapter 3), it was assumed that cells in any compartment / of the cell cycle exhibited one of 4 discrete levels of radiosensitivity. Each of these levels (labelled as levels A-D) was characterized by a single pair of the LQ parameters a and /?. These values of a and ft were assigned to all cell cycle compartments within a given stage A, B, C, or D. The a and f3 parameters of levels A-D were determined from LQ fits to the measured survival curves of Gi , Gi/S, S/G 2 and M cells, respectively (step 2 in figure 2.3). Positions of boundaries between radiosensitivity stages in the model The positions of the boundaries in the model between each of the four radiosensitivity regions (A, B, C, and D) were determined by comparing the model predictions of single-dose cell survival as a function of time after mitotic selection against the measured single-dose cell survival during the first cycle after mitotic selection. Boundary positions were varied in the model and the corresponding model predictions of single-dose survival were compared to the measurements. It was found during this process that the differences in the survival measured in single dose experiments (using the standard clonogenic assay) and survival predicted by the model (based on survival curves measured using the FACS survival assay as described above) made direct matching of model survival predictions to measured single dose survival impractical for determining 50 Chapter 2: Methods optimum boundaries between radiosensitivity regions in the model. Although the trends in radiosensitivity through the cell cycle were the same when measured by conventional and FACS assays, the absolute survival levels were different enough that it was clearly invalid to attempt to directly match model-predicted single dose survival to single-dose survival measurements for the purposes of the boundary-positioning process. So, for modelling purposes only, the measured single dose survival profiles were "normalized" by multiplying by an appropriate constant factor (equivalent to a change in plating efficiency) so that the survival at 3 h after mitotic selection in single dose survival profiles matched the corresponding survival according to the survival curve measurements. Then the boundary positions which minimized the root-mean-square deviation of the model predictions of single dose survival from the "normalized" matching measurements were chosen for use in the model (step 3 in figure 2.3). The need for normalization of the single dose survival profiles simply implied that the zero dose plating efficiency was consistently higher in the FACS survival assay than in the conventional clonogenic assay. This did not seem unreasonable, since cell handling in the two assays was, by necessity, different, and selection by gating of cells during FACS sorting could have excluded non-viable cells which were necessarily included in the conventional clonogenic assay samples. The model was implemented by a Fortran 77 program which accepted as input an initial cell distribution and the kinetic and survival parameters of a cell population. The program returned cell distributions and survival values for the cell population at time steps At after the starting point. The overall scheme of the model is summarized in figure 2.4. 51 Chapter 2: Methods cell cycle phases GI S G2 M 7 CM II I I I ! w 1 CM II • I l l A B C D 0 0 (aA,pA) K , P B ) (ac,Pc) (aD,pD) radiosensitivity phases Figure 2.4: Schematic view of the cell cycle model. Cells progressed from the start of the cycle (at left) through Gi, S, G 2 , and M phase, at which time they doubled by mitosis (at right), and then re-entered Gi phase. In the model, the cycle was divided into K indexed compartments (k=l, k=2, etc.), each of which contained nu cells. In parallel with G i ; S, G 2 , and M phases, the cycle was divided into 4 radiosensitivity stages (A, B, C, and D). All compartments which were in a given radiosensitivity stage had the same radiation response, characterized by the LQ parameters a and B for that stage. 52 Chapter 2: Methods 2.10 Measurement of apoptosis In order to examine the mode of cell death in human tumor cell cultures, a simple flow cytometric assay of D N A fragmentation, which is one of the hallmarks of apoptosis in many cell lines, was used (Gorczyca et al. 1993). Control or irradiated cell monolayers were harvested by trypsinization, and pooled with detached, floating cells from the same cultures. These cells were then fixed in 75% ethanol according to the protocol described in Section 2.4.2.1. After fixation, the cells were stored at -20°C until the time of flow cytometric analysis. Immediately prior to analysis, cells were centrifuged, resuspended in 0.1%) TX-100 and incubated for 1 hr at room temperature to allow small D N A fragments in apoptotic cells to diffuse through the permeabilized cell membrane. At the end of this incubation, cells were centrifuged and stained with DAPI as described in Section 2.4.2.1. In analysis of the flow cytometric data, apoptotic cells from which D N A fragments had diffused were observed in a "sub-Gi" peak of events with less than a Gi complement of DNA. The fraction of the total events which appeared in this sub-Gi peak was used as a measure of apoptosis. 2.11 Western blotting for p53 p53 protein was detected by standard Western blotting techniques (Laemmli 1970). Cells were harvested by trypsinization, resuspended in lysis buffer (50 mM Tris (pH 8.0), 150 mM NaCl, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (PMSF), 1% v/v NP-40), and lysates were stored at -70°C. Protein concentration in all lysates was 53 Chapter 2: Methods determined by the Lowry assay (Sigma protein assay kit P5656). For electrophoresis, aliquots of lysate containing equal amounts of total protein were mixed 1:1 with loading buffer (per 50 mL: 0.3 8g Tris-base (pH 6.8), 5 mL glycerol, 1.15g sodium dodecyl sulphate (SDS), 45 mL double-distilled H 20, 0.0025 g bromophenol blue), and separated on a 12% acrylamide gel by electrophoresing at 130V for approximately 45 minutes. When separation was complete, protein was transferred onto nitrocellulose membrane (BioRad, Hercules, CA) by electrophoretic blotting (70V for 2 hours). Membranes were then blocked in 5% non-fat milk in PBS for 1 h, and probed with primary anti-p53 antibody (PAb 1801, Calbiochem, La Jolla, CA), followed by secondary peroxidase-conjugated anti-mouse IgG antibody (Sigma-Aldrich). Protein was detected by incubation with diaminobenzidine (DAB) (Sigma-Aldrich) for approximately 5 minutes, followed by rinsing with PBS to stop the development of precipitate. Protein levels were quantitated by scanning of developed membranes with a laser densitometer (Molecular Dynamics, Sunnyvale, CA). In some experiments an enhanced chemofluorescent (ECF) protein detection technique was used. In this method, the electrophoresis was the same as described above, but the ECF kit protocol (Amersham, ECF kit RPN5780) was followed for protein blotting and immunodetection. In the ECF protocol, p53 was quantified directly on the membrane by fluorescence using a Storm phosphoimager (Molecular Dynamics). 54 Chapter 2: Methods 2.12 Transfection of A549 cells to abrogate p53 function Of the three cell lines which are characterized in this thesis, only A549 is known to express wild type p53 protein (Noble et al. 1992). HT-29 cells express a mutant form of p53 (Rodrigues et al. 1990) and the status of p53 in U l cells is apparently not known. To investigate the potential role of the Gi delay in the radiosensitivity of A549 cells, we aimed to abrogate p53 function in these cells. p53 function is required for the activation of the p21 cyclin dependent kinase inhibitor which initiates the Gi delay in response to D N A damage (Kuerbitz et al. 1992). Hence, A549 cells were transfected with the BCMGSNeo plasmid expression vector (see figure 2.5), which was obtained from Dr. G. Dougherty. This vector is based on the BMGNeo plasmid vector (Karasuyama et al. 1988). The BCMGSNeo vector (14.5 kbp) contains 69% of the bovine papillomavirus-1 (BPV-1) genome; this D N A fragment is known as the transforming fragment (Lowy et al. 1980). The transforming fragment includes the BPV-E6 gene (Chen et al. 1982), which encodes a protein which has been shown to enhance the degradation of p53 in human epithelial cells in vitro (Band et al. 1993). A549 cells were grown to 50% confluence in 5 cm diameter tissue culture dishes (Falcon), and then transfected with BCMGSNeo by electroporation. Three days after electroporation, 0.5 mg/mL G418 antibiotic was added to the cultures as a selection agent. Following 14 days of growth in G418, several surviving colonies were picked off the dishes and grown up for individual examination. Cells from each clone were assayed for Gi delay by flow cytometric measurement of D N A content after irradiation. Cells from one clone which appeared to have lost the ability to block in Gi phase after irradiation were then selected for further investigations of p53 status, radiosensitivity and cell kinetics 55 Chapter 2: Methods after irradiation. The cell line established from this clone was designated as A549BPV. 56 Chapter 2: Methods Figure 2.5: Schematic of the BCMGSNeo plasmid vector (14.5 kbp). Plasmid contains Neomycin (G418) resistance gene and Ampicillin resistance genes. The C M V promoter drives transcription of D N A inserted into the stuffer region (no insert was used in the experiments described here), and the poly-A addition signal is included for efficient termination of insert transcription. The 69% B P V fragment contains the E6 gene. The human (3-globin fragment enhances the transforming effects of the B P V gene fragment (DiMaio et al. 1982). 57 Chapter 3: Results 3.1 Characteristics of mitotically selected cells The mitotic index of mitotically selected populations of cells was determined by staining of cells immediately after mitotic rolloff with Giemsa stain, and counting of mitotic figures. Populations of mitotic cells were routinely 85-95% pure. Typical values of the mitotic index in 3 cell lines are summarized in table 3.1. Figure 3.1 shows 3 examples of mitotically selected U l cells in telophase and pro/metaphase. Table 3.1: Mitotic index of cells prepared 3y mitotic shakeoff. Cell Line Mitotic Index Range HT-29 A549 U1 94 93 87 (92-97) (90-98) (79-94) Typical values and ranges of mitotic index are shown. 3.2 Kinetics of mitotically selected cells After mitotic selection, cells progressed through Gi , S, and G 2 phases. It was found that a fraction of the cells (approximately 25%) did not fully complete cytokinesis immediately after mitotic selection, presumably due to some disturbance of cell division by the rolloff collection procedure. This effect has been noted by others using the mitotic rolloff procedure (see e.g. the technical guide Protocols 1: Automated mitotic selection into monolayer cultures, Talandic Research Corporation, Pasadena, CA, USA). These cells maintained two nuclei contained in a single joined cytoplasm (figure 3.1 A, 3. IB). In flow cytometric analysis of fixed cell samples, binucleate cells could not be unambiguously 58 Chapter 3: Results distinguished from singlet cells with the equivalent total D N A content, and therefore could not all be excluded from analysis by the typical forward scatter/time of flight gate used to exclude doublet cells from D N A content analysis. Thus binucleate cells appear on the D N A histograms which are discussed below as cells with twice the corresponding singlet cell D N A content. It should be noted here that in cell sorter assays of survival (results in Section 3.4), where live cells were sorted on the basis of light scatter only, binucleate cells could be at least partially distinguished from singlet cells. The binucleates appeared as a sub-population with slightly greater forward and perpendicular light scatter, presumably due to their unique size and/or shape. Sorting gates were set up so as to exclude the greatest possible number of binucleate cells in the cell sorter assay of survival; periodic observation of the sorted populations under the microscope confirmed that sorted synchronized populations in interphase were enriched in singlet cells. Figures 3.2 to 3.4 show histograms of D N A content in synchronized populations of HT-29, A549, and U l cells at intervals after mitotic selection. A l l three cell lines displayed similar behaviour over the course of the first cycle after mitotic selection. In figures 3.2 to 3.4, the peak labeled Gi contained singlet Gi cells. The peak labeled G 2 contained either singlet G 2 cells or binucleate Gi cells. Consequently, as the synchronized population moved into S phase (~10 h in A549 cells), cells appeared at 2 positions in the D N A histograms: between the Gi and G 2 peaks (singlet cells) and beyond the G 2 peak (binucleate cells which persisted through S phase). The binucleate cell population with > G 2 D N A content in all 3 cell lines disappeared from all synchronized cell populations during the first 40 h after mitotic selection. The zero-dose plating efficiencies of cells in 59 Chapter 3: Results various experiments were examined in order to determine the viability of the binucleate cells. For example, it was observed that for A549 and HT-29 the zero-dose plating efficiency of asynchronous sorted cell populations (containing no binucleates) was typically 70-90%. This was about 10% higher than the zero-dose plating efficiency of mitotically selected sorted populations (containing a reduced fraction of binucleates after sorting; see previous paragraph) which was in turn typically 5-10% higher than the zero-dose plating efficiency in unsorted mitotically selected cells (containing the largest fraction of binucleate cells). Hence, populations with up to ~25% binucleate cells had plating efficiencies ~20% lower than populations without binucleate cells. This suggested that the binucleate cells were non-viable, but since there were unavoidable differences in cell handling between sorted/unsorted and synchronized/asynchronous cell populations, it was not possible to exclude the possibility that some or all of the binucleate cells were viable. The percentages of Gi , S, and G 2 cells in synchronized populations of HT-29, A549, and U l cells were calculated by fitting the D N A histograms taken at intervals after mitotic selection (including the histograms shown in figure 3.2-3.4). Results are shown in figures 3.5-3.7. Calculation of the fractions of Gi , S, and G2/M phase cells was carried out by fitting D N A histograms as described in Section 2.4.3. The presence of the binucleate cells complicated the attribution of cells in the G 2 D N A peak, because of the potential mixing of singlet G 2 cells and binucleate Gi cells in this peak. To account for the binucleate Gi cells, at times from 0 h after mitotic selection 60 Chapter 3: Results M l Figure 3.1: Examples of mitotically selected U l cells. A ,B: typical mitotically selected cells in telophase or early GI phase. Some cells of this type that did not complete cytokinesis persisted in the synchronized populations as "binucleate" cells. C: typical pro/metaphase cell with clearly condensed chromosomes. 61 Chapter 3: Results Figure 3.2: DNA histograms of synchronized A549 cells. Cells were synchronized at t=0h by mitotic selection and samples were collected for DNA analysis at times afterward as indicated on the figure. 62 Chapter 3: Results Figure 3.3: D N A histograms of synchronized HT-29 cells. Cells were synchronized at t=0h by mitotic selection and samples were collected for D N A analysis at times afterward as indicated on the figure. 63 Chapter 3: Results rfi\A'r;rrir-> Figure 3.4: DNA histograms of synchronized Ul cells. Cells were synchronized at t=0h by mitotic selection and samples were collected for DNA analysis at times afterward as indicated on the figure. 64 Chapter 3: Results to the approximate duration of (Gi+S) phase, all cells in the G 2 peak were counted as Gi cells. However, at times corresponding to the duration of (Gi+S) phase, synchronized G 2 cells would start to appear in the G 2 peak. Therefore, from this time forward, which was, in practice, clearly defined as the first time when there was a significant increase in the fraction of cells in the G 2 peak, all cells in the G 2 peak were counted as G 2 cells. For a discussion of the rationale for this accounting of cells and the potential uncertainties involved, see Appendix 1. The approximate lengths of the cell cycle in each of the cell lines can be estimated directly from figures 3.5-3.7. The length of the cell cycle was approximately equal to the time after mitotic selection at which the rate of transition from G 2 / M phase into the second post-mitotic Gi phase is greatest. This time was about 20 h in A549, 20 h in HT-29, and 32 h in U l . These times agree well with measured cell doubling times for these cell lines which were typically 18-24 h in A549 and HT-29, and 30-40 h in U l . 65 Chapter 3: Results A549 time after mitotic selection (h) Figure 3.5: A549 cell cycle distribution after mitotic selection. Cells were selected at t=0 and samples taken at various times after selection. 66 Chapter 3: Results Figure 3.6: HT-29 cell cycle distribution after mitotic selection. Cells were selected at t=0 and samples taken at various times after selection. 67 Chapter 3: Results U1 tirrB after mitotic selection (h) Figure 3.7: Ul cell cycle distribution after mitotic selection. Cells were selected at t=0 and samples taken at various times after selection. 68 Chapter 3: Results 3.3 Radiosensitivity variations during the cell cycle in HT-29, A549, and Ul. In order to determine the overall variation in radiation sensitivity through the cell cycle of these three cell lines, mitotically selected populations of each cell type were irradiated with a single dose at intervals after mitotic selection, according to the method described in Section 2.6.1. The single dose which was given to each cell line (5 Gy for A549, 6 Gy for HT-29, 8 Gy for U l ) was chosen to give a survival of approximately 0.1 in each of the cell lines, on the basis of earlier experiments done by others (Skarsgard et al. 1996). Figures 3.8-3.10 show the cell survival as a function of time after mitotic selection in synchronized populations of A549, HT-29, and U l cells. Data in these figures represent the averaged results of 2 experiments (HT-29 and U l ) or 3 experiments (A549). The average number of individual survival determinations which were made for each point on these single dose survival responses was 8 (range 3-9) for A549, 6 (range 3-6) for HT-29, and 5 (range 3-6) for U l . The approximate durations of G i , S, and G2/M phases are indicated in the figures by the labeled horizontal bars. The vertical arrows shown on the figures mark some important kinetic reference points, determined from the kinetic data in figures 3.5 to 3.7. The thick arrows represent the times of maximum rate of increase (up-arrow) or decrease (down-arrow) in the Gi-phase fraction. The thin arrows represent the times of maximum rate of increase (up-arrow) or decrease (down-arrow) in S-phase fraction. Al l three cell lines showed common elements in the pattern of radiosensitivity variation after mitotic selection. At roughly comparable survival levels (around S=0.1) A549 showed the smallest magnitude of variations during interphase (a survival ratio of 1.6 between most resistant and most sensitive points within interphase). HT-29 and U l 69 Chapter 3: Results exhibited larger variations in radiosensitivity (ratios of 2.5 and 3.5 between most resistant and most sensitive points within interphase for HT-29 and U l respectively). In all cell lines the ratios of survival at the most resistant and sensitive points in interphase were significantly greater than one, with p<0.001 by the T-test. Therefore, there were significant variations in radiation survival over the course of interphase after single doses of 5, 6, and 8 Gy in A549, HT-29, and U l respectively. In all the cells, mitosis (t=0 h) was more sensitive than any time during interphase. Early G\ phase (t=2-4 h) was relatively radioresistant. Around the time when the synchronized populations made the transit from G i to S phase, there was a period of maximum interphase radiosensitivity (8-10 h in A549 and HT-29, 14-16 h in U l ) . Following this, cells became more radioresistant, with a local peak of radioresistance which occurred when the synchronized population was primarily in S/G2 phase (about 15 h in A549 and HT-29, 25 h in U l ) . In all cell lines, survival at the peak of radioresistance in early G i phase was at least as great as survival at the peak of the S/G 2 peak of radioresistance. 70 Chapter 3: Results 0.3 0.2 H> 0.07 0.06— 0.05" 0.04-0.03 0.02 ~i—i—i—i—i—i—i—i—i—i—r • i A549 i—i—i—I—i—i—i—i—I—i—r i—i—i—i—i—i—r 1—i—i—i—I—i—i—i—i—|—r "i—i—i—i—l—i—i—i—r 0 10 15 20 25 30 35 40 time after mitotic selection (h) Figure 3.8: Cell age related variation in radiosensitivity in A549. Cells were synchronized by mitotic selection at t=0h and irradiated with 5 Gy at intervals afterwards. Horizontal bars indicate approximate durations of Gl , S, and G2/M phases. Thick arrows represent times of maximum rate of increase (up arrow) or decrease (down arrow) in Gl phase fraction (see fig. 3.5). Thin arrows represent the same times for S phase fraction. 71 Chapter 3: Results Figure 3.9: Cell age related variation in radiosensitivity in HT-29. Cells were synchronized by mitotic selection at t=0h and irradiated with 6 Gy at intervals afterwards. Horizontal bars indicate approximate durations of GI , S, and G2/M phases. Thick arrows represent times of maximum rate of increase (up arrow) or decrease (down arrow) in GI phase fraction (see fig. 3.6). Thin arrows represent the same times for S phase fraction. 72 Chapter 3: Results U1 t i m e a f t e r m i t o t i c s e l e c t i o n (h ) Figure 3.10: Cell age related variation in radiosensitivity in U1. Cells were synchronized by mitotic selection at t=0h and irradiated with 8 Gy at intervals afterwards. Horizontal bars indicate approximate durations of GI , S, and G2/M phases. Thick arrows represent times of maximum rate of increase (up arrow) or decrease (down arrow) in GI phase fraction (see fig. 3.7). Thin arrows represent the same times for S phase fraction. 73 Chapter 3: Results 3.4 Substructure in the survival responses of synchronized HT-29, A549, and U l . Survival responses were measured for A549, HT-29 and U l cells for doses of 0-12 Gy, corresponding to approximately the first 3 decades of cell kill in these cells. For each cell line, asynchronous and synchronized responses were measured. The synchronized responses were measured in mitotically selected cells at 4 times after mitotic selection. In each cell line, the chosen times (0 h, 3 h, 8/10/17 h, and 15/15/25 h after mitotic selection) corresponded to the points when most synchronized cells were in mitosis, mid-Gi phase, Gi/S phase, or late S/G2 phase. Each survival response was fitted with linear quadratic curves: a fit to the low dose range (up to and including 4 Gy) and a fit to the high dose range (above 4 Gy). Both high and low dose range fits were forced through S=l at D=0 Gy. The survival responses are shown in figures 3.11-3.15 (A549), figures 3.19-3.23 (HT-29), and figures 3.27-3.31 (Ul). In each of the figures, the top panel is a plot of survival against radiation dose, and the bottom panel is the linearized survival plot of -ln(S)/D against dose. Tables 3.2-3.4, 3.7-3.9, and 3.12-3.14 give the parameters of L Q fits to synchronized and asynchronous populations of A549, HT-29, and U l on the low, high, and full dose ranges. Tables 3.5, 3.10, and 3.15 give the a/(3 ratios of the linear quadratic fits for each cell population.. For clarity, figures 3.16-3.17, 3.24-3.25, and 3.32-3.33 plot the othi/aio and phi/Pio ratios and figures 3.18, 3.26, and 3.34 plot the ratios of a/p for the low, high and full dose ranges in A549, HT-29, and U l . To provide another independent measure of deviations from the LQ model over the entire dose range 0-12 Gy, all survival curves were also fitted with the (LQC) model (see equation 1.5). In this model, the fit parameters were not dependent on the arbitrary choice of high and low dose 74 Chapter 3: Results ranges, as the LQ fits were. Tables 3.6, 3.11, and 3.16 give the parameters of linear-quadratic-cubic (LQC) fits to the survival curves. In HT-29 and A549 cells, all the synchronized and asynchronous survival responses exhibited LQ substructure (with one exception, synchronized 3 h Gi HT-29 cells). Parameters of the LQ fits to the high dose range were different from those of the low dose range at the level of 95% confidence limits. On the high dose range, a values were always larger and P values always smaller than the a and p values derived from low dose range fits. In addition, the full dose range LQ "Q values" (in table 3.4 and 3.9) were less than 0.001 for all A549 and HT-29 cell populations. The Q value represents the probability of getting a worse fit to the data by chance with a single LQ function, based on the value of x2 (also in tables 3.4 and 3.9), and assuming the survival values are normally distributed (Press et al. 1992). HT-29 cells, synchronized and sampled 3 h after mitotic selection in a radioresistant stage of Gi, exhibited different survival substructure than other HT-29 and A549 cell populations. In these cells, there was no significant difference between a and P values derived from LQ fits to the low and high dose ranges, and the Q value for a single LQ fit to the data was greater than 0.5, indicating a "reasonable" single LQ fit to this survival curve. LQC fits to A549 and HT-29 cells indicated that although 3/10 cell populations had y values that were not significantly less than zero, 9/10 cell populations did had negative y values (tables 3.6, 3.11), where non-zero y values indicated deviations from the LQ model. I n U l cells, results of L Q fitting of survival curves were fundamentally different. Values of a for the high dose range were not different from those for the low dose range for 3 h Gi, 17 h Gi/S, and 25 h S/G2 cells (noting that a was constrained to be > 0 in the 75 Chapter 3: Results fitting procedure). There were two cases where was significantly different from ai„: in asynchronous cells OCM was less than ai0 and in 0 h mitotic cells ahi was greater than ai0. The values of ph; for the high dose range were greater than p]o on the low dose range for all synchronized populations. LQC fits to Ul cell populations indicated that in 4 of 4 populations in which 95% confidence limits could be determined, y was significantly greater than zero; in 0 h mitotic Ul cells, the y value was positive, but the confidence limit calculation in the fitting software failed (table 3.16). 76 Chapter 3: Results Figure 3.11: Survival curve substructure in asynchronous A549 cells. Asynchronous cells were irradiated with doses from 0-12 Gy, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 77 Chapter 3: Results A549 Oh mitotic 1 mrp 1 ' 1 ' 1 • 1 1 r Figure 3.12: Survival curve substructure in Oh mitotic A549 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy immediately after selection, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 78 Chapter 3: Results A549 3h G1 -0.2 t-I I I I I I I I I I I I 2 4 6 8 10 12 Dose (Gy) Figure 3.13: Survival curve substructure in 3h G i phase A549 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at 3h after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 79 Chapter 3: Results A549 10h G1/S 1 J E 1 I 1 I 1 1 ' 1 ' 1 ' =| Figure 3.14: Survival curve substructure in lOh Gi/S phase A549 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at lOh after selection, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 80 Figure 3.15: Survival curve substructure in 15h S/G 2 phase A549 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at 15h after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 81 Chapter 3: Results Table 3.2: Values of a fitted from A549 survival curves Cycle Time #of curves Ctio 95% CL Ct h i 95% CL asynchronous 3 0.1025 0.0968-0.1082 0.2533 0.2484-0.2585** M 4 0.4194 0.4055-0.4337 0.5111 0.4972-0.5264** 3h d 4 0.107 0.1008-0.1133 0.1829 0.1782-0.1877** 10h d /S 5 0.1284 0.1226-0.1343 0.2114 0.2081-0.2147** 15h S/G2 4 0.1074 0.1015-0.1135 0.1605 0.1576-0.1635** The number of curves represents the number of survival responses which were averaged into the dataset which was used for parameter fitting. o,0 is fitted from the dose range from 0-4 Gy. a h l is fitted from the dose range 4-12 Gy. 95% confidence limits are shown for each parameter. ** indicates that the 95% confidence limits for a H do not overlap with the 95% confidence limits for ai 0. Table 3.3: Values of p fitted from A549 survival curves Cycle Time #of curves Pio 95% CL Phi 95% CL asynchronous 3 0.0576 0.0549-0.0602 0.0258 0.0253-0.0263** M 4 0.0325 0.0267-0.0386 0.0058 0.0039-0.0081** 3h d 4 0.0433 0.0409-0.0458 0.0248 0.0242-0.0254** 10h d /S 5 0.0495 0.0475-0.0515 0.0308 0.0304-0.0313** 15h S/G2 4 0.043 0.0411-0.0449 0.0304 0.03-0.0308** The number of curves represents the number of survival responses which were averaged into the dataset which was used for parameter fitting. pto is fitted from the dose range from 0-4 Gy. phi is fitted from the dose range 4-12 Gy. 95% confidence limits are shown for each parameter. ** indicates that the 95% confidence limits for pw do not overlap with the 95% confidence limits for pto. 82 Chapter 3: Results Table 3.4: Linear quadratic fits to the full range of A549 survival curves (0-12 Gy) Cycle Time #of curves a P x2 n Q asynchronous 3 0.1496 0.0362 190.58 21 O.001 M 4 0.4609 0.0136 105.74 29 <0.001 3h G, 4 0.1381 0.0304 79 21 O.001 10h G^S 5 0.1696 0.0364 103.66 21 O.001 15h S/G2 4 0.1421 0.0328 101.85 21 <0.001 Tabulated are the number of survival responses which were averaged together in the fitted dataset, the LQ parameters a and p, the goodness-of-fit parameter x2, the number of datapoints in the fitted average survival response n, and Q, which is the probability of obtaining a worse fit to the data by chance, based on the % value. Table 3.5: a/p ratios for full, ow, and high dose range fits in A549 ce lis. Cycle Time a/p (0-12 Gy) 95% CL a/p (0-4 Gy) 95% CL a/p (4-12 Gy) 95% CL asynchronous 4.133 3.959-4.312 1.780 1.608-1.971 9.818 9.445-10.217 M 33.890 27.981-41.743 12.905 10.505-16.243 88.121 61.383-134.974 3h d 4.543 4.329-4.785 2.471 2.201-2.770 7.375 7.016-7.756 10h d / S 4.659 4.518-4.808 2.594 2.381-2.827 6.864 6.649-7.063 15h S/G2 4.332 4.183-4.472 2.498 2.261-2.762 5.280 5.117-5.450 The table shows the a/p ratio determined from LQ fits to the full range of survival data (0-12 Gy), the low dose range (0-4 Gy), or the high dose range (4-12 Gy), with 95 % confidence limits on each value. Table 3.6: Linear-quadratic-cubic (LQC) fits in A549 cells. Cycle Time a 95% CL P 95% CL Y 95% CL asynchronous 0.10237 0.0725-0.1413 0.06304 0.0599-0.0683 -0.00211 -0.002-0.0016" M 0.4331 0.3457-0.5692 0.02834 0.0167-0.0751 -0.00137 -.002-0.0118 3hGi 0.1079 0.0831-0.1364 0.0457 0.0423-0.0504 -0.0014 -0.002-0.0008" 10h Gi/S 0.1338 0.1136-0.1569 0.0512 0.0484-0.0546 -0.0013 -0.002-0.0009" 15h S /G 2 0.1114 0.0873-0.1393 0.0442 0.0408-0.0485 -0.0009 -0.001-0.0004" Shown are values of the parameters a, p, and y in the LQC model, with 95% confidence limits. ** indicates significantly less than zero. 83 Chapter 3: Results A549 a h i / ratio 1 1 1 1 1 1 i i " < 1 • • • i •J LL 1 1 i u . J . • -4-1 L + J L + J L + J L + J L asyndirrjnous M 3hG1 10hG1/S 15hS/G2 Figure 3.16: Alpha ratios for synchronized A549 cells. Shown is the ratio of the a values (±95% confidence limits) derived from fitting a linear quadratic function to the low (0-4Gy) and high (4-12 Gy) ranges of the survival curves of synchronized A549 cells. If the survival curves were homogenous LQ responses, the a ratio would equal 1, indicated by the solid horizontal line. 84 Chapter 3: Results A549ph./p | o ratio asynchronous M 3hG1 10hG1/S 15hS/G2 Figure 3.17: Beta ratios for synchronized A549 cells. Shown is the ratio of the P values (+95% confidence limits) derived from fitting a linear quadratic function to the low (0-4Gy) and high (4-12 Gy) ranges of the survival curves of synchronized A549 cells. If the survival curves were homogenous LQ responses, the p ratio would equal 1, indicated by the solid horizontal line. 85 Chapter 3: Results Figure 3.18: a/p ratios for A549 cells. Figure shows the a/p ratios (±95% confidence limits) calculated for asynchronous and synchronized A549 cells using LQ fits to the low (0-4 Gy), high (4-12 Gy), and full (0-12 Gy) dose ranges. 86 Chapter 3: Results HT29 Asynchronous 4 6 8 Dose (Gy) 10 12 Figure 3.19: Survival curve substructure in asynchronous HT-29 cells. Exponentially growing cells were irradiated with doses from 0-12 Gy, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 87 Chapter 3: Results HT29 Mitotic Dose (Gy) Figure 3.20: Survival curve substructure in mitotic HT-29 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy immediately after selection, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 88 Chapter 3: Results HT29 3h G1 4 6 8 Dose (Gy) Figure 3.21: Survival curve substructure in 3h Gi phase HT-29 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at 3h after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 89 Figure 3.22: Survival curve substructure in 8h Gi/S phase HT-29 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at 8h after selection, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 90 Chapter 3: Results 0.6 r Figure 3.23: Survival curve substructure in 15h S/G 2 phase HT-29 cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at 15h after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 91 Chapter 3: Results Table 3.7: Values of a fitted from HT-29 survival curves. Cycle Time #of curves Ct|o 95% CL a h i 95% CL asynchronous 7 0.0381 0.0349-0.0412 0.1323 0.1298-0.1348** M 5 0.2669 0.2491-0.2853 0.4736 0.4588-0.4904** 3h d 3 0.0273 0.023-0.0317 0.021 0.017-0.025 8h G i / S 5 0.0077 0.0013-0.0142 0.121 0.1139-0.1285** 15h S / G 2 3 0.0135 0.0091-0.0178 0.0477 0.0415-0.0542** The number of curves represents the number of survival responses which were averaged into the dataset which was used for parameter fitting. a,0 is fitted from the dose range from 0-4 Gy. a h i is fitted from the dose range 4-12 Gy. 95% confidence limits are shown for each parameter. ** indicates that the 95% confidence limits for <xhi do not overlap with the 95% confidence limits for cq0. Table 3.8: Values of (3 fitted from FT ^-29 survival curves. Cycle Time #of curves Pio 95% CL Phi 95% CL asynchronous 7 0.0475 0.0462-0.0489 0.0259 0.0256-0.0263** M 5 0.111 0.1018-0.1209 0.0275 0.0259-0.0294** 3h G^ 3 0.0374 0.0358-0.039 0.04 0.0394-0.0406 8h G i / S 5 0.0683 0.0653-0.0713 0.0438 0.043-0.0448** 15h S / G 2 3 0.0395 0.0379-0.041 0.0302 0.0293-0.0313** The number of curves represents the number of survival responses which were averaged into the dataset which was used for parameter fitting. pi0 is fitted from the dose range from 0-4 Gy. pw is fitted from the dose range 4-12 Gy. 95% confidence limits are shown for each parameter. ** indicates that the 95% confidence limits for pw do not overlap with the 95% confidence limits for pto . 92 Chapter 3: Results Table. 3.9: Linear quadratic fits to the full range of HT-29 survival curves (0-12 Gy) Cycle Time #of curves a P x2 n Q asynchronous 7 0.0709 0.0334 341.43 27 <0.001 M 5 0.4054 0.0357 208.32 31 O.001 3h d 3 0.0227 0.0397 10.45 18 >0.5 8h G i /S 5 0.0386 0.0542 85.16 27 <0.001 15h S / G 2 3 0.0293 0.0333 49.7 27 <0.01 Tabulated are the number of survival responses which were averaged together in the fitted dataset, the LQ parameters a and b, the goodness-of-fit parameter %, the number of datapoints in the fitted average survival response n, and Q, which is the probability of obtaining a worse fit to the data by chance, based on the % value. Table 3.10: a/p ratios for fu 11, low, and high dose range fits in HT-29 cells Cycle Time a/p (0-12 Gy) 95% CL a/p (0-4 Gy) 95% CL a/p (4-12 Gy) 95% CL asynchronous 2.123 2.038-2.205 0.802 0.714-0.892 5.108 4.935-5.266 M 11.356 10.358-12.364 2.405 2.060-2.803 17.222 15.605-18.934 3h G, 0.572 0.486-0.662 0.730 0.590-0.885 0.525 0.419-0.635 8h G,/S 0.712 0.608-0.821 0.113 0.018-0.217 2.763 2.542-2.988 15hS/G2 0.880 0.746-1.031 0.342 0.222-0.470 1.579 1.326-1.850 The table shows the a/p ratio determined from LQ fits to the full range of survival data (0-12 Gy), the low dose range (0-4 Gy), or the high dose range (4-12 Gy), with 95 % confidence limits on each value. Table 3.11: Linear-quadratic-cubic (LQC) fits in HT-29 cells. Cycle Time 95% CL 95% CL 95% CL asynchronous 0.03688 0.02618-0.04794 0.05174 0.05025-0.0534 -0.00158 -0.00173-0.00141" M 0.34937 0.223776-0.61558 0.06527 0.05275-0.24719 -0.00226 -0.02554-0.10245 3h d 0.02415 0.01361-0.0351 0.03893 0.03716-0.04092 0.00007 -0.00014-0.00032 8h G,/S 0.00136 -0.0207-0.02592 0.07641 0.07311-0.08121 -0.002 -0.00232-0.00146" 15h S/G2 0.00964 -0.01342-0.03496 0.04371 0.03904-0.05027 -0.00099 -0.00151-0.00007 Shown are values of the parameters a, p, and y in the LQC model, with 95% confidence limits. " indicates significantly less than zero. 93 Chapter 3: Results HT-29 a h i / a l 0 ratio 20 15 U i 10 u 5U 0 asyrx^ronous M 3hG1 8hG1/S 15hS/G2 Figure 3.24: Alpha ratios for synchronized HT-29 cells. Shown is the ratio of the a values (±95% confidence limits) derived from fitting a linear quadratic function to the low (0-4Gy) and high (4-12 Gy) ranges of the survival curves of synchronized HT-29 cells. If the survival curves were homogenous LQ responses, the a ratio would equal 1, indicated by the solid horizontal line. 94 Chapter 3: Results 2 0 1.8 U 1.6 U 1.4 L 1.2 o fe 1.0 CO. a 0.8 06 0.4 0.2 0.0 HT-29 P h . / P | o ratio i 1 r asynchronous M 3hG1 8hG1/S 15hS/G2 Figure 3.25: Beta ratios for synchronized HT-29 cells. Shown is the ratio of the P (±95% confidence limits) values derived from fitting a linear quadratic function to the low (0-4Gy) and high (4-12 Gy) ranges of the survival curves of synchronized HT-29 cells. If the survival curves were homogenous LQ responses, the P ratio would equal 1, indicated by the solid horizontal line. 95 Chapter 3: Results HT-29 a/p ratios 20 15 U 1 101-~8 5U 0 ^ 04GyLQfit 4-12 G/LQ fit ^ 0-T2Gyl_Qfit 'A k \ W ^ asynchronous M 3hG1 8hG1/S 15hS/G2 Figure 3.26: a/p ratios for HT-29 cells. Figure shows the a/p ratios calculated for (±95% confidence limits) asynchronous and synchronized HT-29 cells using LQ fits to the low (0-4 Gy), high (4-12 Gy), and full (0-12 Gy) dose ranges. 96 Chapter 3: Results U1 Asynchronous 0.01 h L 05 03 -I , 1 , , . , , r 0 1 2 3 4 0.001 0.8 h 0.6 f 0.4 - J Q •0.2 -i l i i i l i l i i 0 2 4 6 8 10 12 Dose (Gy) Figure 3.27: Survival curve substructure in asynchronous U l cells. Exponentially growing cells were irradiated with doses from 0-12 Gy, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 97 Chapter 3: Results U1 Oh rritotic 0.0 h 42 ' 1 1 1 1 1 1 1 1 ' 1 ' I 0 2 4 6 8 10 12 Dose (Gy) Figure 3.28: Survival curve substructure in mitotic U l cells. Mitotically selected cells were irradiated with doses from 0-12 Gy immediately after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 98 0 2 4 6 8 10 12 Do 39 (Gy) Figure 3.29: Survival curve substructure in 3h G i phase Ul cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at t=3h after selection, and survival assayed by the cell sorter assay. Data points are shown +SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 99 Figure 3.30: Survival curve substructure in 17h Gi/S phase U l cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at t=17h after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 100 Figure 3.31: Survival curve substructure in 25h S/G2 phase U l cells. Mitotically selected cells were irradiated with doses from 0-12 Gy at t=25h after selection, and survival assayed by the cell sorter assay. Data points are shown ±SEM. Solid line: fit to data from 0-4 Gy; dotted line: fit to data above 4 Gy. 101 Chapter 3: Results Table 3.12: Values of a fitted from U1 survival curves Cycle Time #of curves Olio 95% CL a h i 95% CL asynchronous 4 0.0289 0.0245-0.0334 0 0-0.0082** M 4 0.2964 0.2748-0.3188 0.7713 0.7560-0.7889** 3h d 6 0 0-0.0181 0 0-0.0181 17h G^S 4 0.0092 0.0037-0.0146 0 0-0.0051 25h S/G2 4 0.0064 0-0.014 0 0-0.0085 The number of curves represents the number of survival responses which were averaged into the dataset which was used for parameter fitting. a,0 is fitted from the dose range from 0-4 Gy. am is fitted from the dose range 4-12 Gy. 95% confidence limits are shown for each parameter. ** indicates that the 95% confidence limits for do not overlap with the 95% confidence limits for cu0. Table 3.13: Val ues of (3 fitted from Ul survival curves Cycle Time #of Pio 95% CL Phi 95% CL curves asynchronous 4 0.0235 0.0218-0.0253 0.0378 0.0368-0.0389** M 4 0.1902 0.1785-0.2030 0.001 0-0.003** 3h d 6 0.0215 0.0205-0.0225 0.0335 0.0331-0.0338** 17h d /S 4 0.041 0.0388-0.0433 0.0505 0.0499-0.0512** 25h S/G2 4 0.0315 0.0288-0.0344 0.0429 0.0422-0.0437** The number of curves represents the number of survival responses which were averaged into the dataset which was used for parameter fitting. pto is fitted from the dose range from 0-4 Gy. phl is fitted from the dose range 4-12 Gy. 95% confidence limits are shown for each parameter. ** indicates that the 95% confidence limits for phl do not overlap with the 95% confidence limits for Pi„. 102 Chapter 3: Results Table 3.14: Linear quadratic fits to the full range of U l survival curves (0-12 Gy) Cycle Time #of curves a 0 x2 n Q asynchronous 4 0.0001 0.0367 115.5 21 <0.001 M 4 0.65805 0.01585 390.3 21 <0.001 3h d 6 0 0.0326 1407 21 <0.001 17h Gi/S 4 0 0.0498 140 21 <0.001 25h S / G 2 4 0 0.042 218.7 21 O.001 Tabulated are the number of survival responses which were averaged together in the fitted dataset, the LQ parameters a and p, the goodness-of-fit parameter %2, the number of datapoints in the fitted average survival response n, and Q, which is the probability of obtaining a worse fit to the data by chance, based on the x 2 value. Table 3.15: a/ft ratios for full, low, and high dose range fits in U l cells Cycle Time a/p (0-12 Gy) 95% CL a/p (0-4 Gy) 95% CL a/p (4-12 Gy) 95% CL asynchronous 0.003 0-0.223 1.230 0.968-1.532 0.000 0-0.223 M 41.517 0.863-" 1.558 1.354-1.786 771.300 2 5 2 - " 3h G, 0.000 0-0.555 0.000 0-0.883 0.000 0-0.547 17h G,/S 0.000 0-0.102 0.224 0.085-0.376 0.000 0-0.102 25h S/G 2 0.000 0-0.202 0.203 0-0.486 0.000 0-0.201 The table shows the a/p ratio determined from LQ fits to the full range of survival data (0-12 Gy), the low dose range (0-4 Gy), or the high dose range (4-12 Gy), with 95 % confidence limits on each value. Table 3.16: Linear-quadratic-cubic (LQC) fits in U l cells Cycle Time a 95% CL P 95% CL T 95% CL asynchronous 0.03757 0.0077-0.0703 0.01218 0.0046-0.0223 0.00284 0.0018-0.0050" M 0.43736 * 0.12112 * -0.00767 * 3h G, 0.01136 -0.0061-0.0302 0.00638 0.0029-0.0104 0.00323 0.0028-0.0039" 17h G,/S -0.00387 -0.0411-0.0422 0.04514 0.0399-0.0542 0.00067 0.0001-0.0019" 25h S/G 2 0.02086 -0.0144-0.0632 0.01595 0.0022-0.0040 0.00283 0.0022-0.0040" Shown are values of the parameters a, p, and y in the LQC model, with 95% confidence limits. " indicates significantly greater than zero. 103 Chapter 3: Results U1 a h i /a l 0 ratio 2U 2 a 0 _L X asyrxtircnous M 3hG1 17hG1/S 25hS/G2 Figure 3.32: Alpha ratios for synchronized U l cells. Shown is the ratio of the a values (±95% confidence limits) derived from fitting a linear quadratic function to the low (0-4Gy) and high (4-12 Gy) ranges of the survival curves of synchronized U l cells. If the survival curves were homogenous LQ responses, the a ratio would equal 1, indicated by the solid horizontal line. 104 Chapter 3: Results 2 0 -1.8-1.6-1.4-1.2-g | 5 1.0-ea ca * 0.8-0.6-0.4-0.2J 0.0-U1 p h i / p I o ratio i 1 r —=f=— asyncrronous M 3hG1 17hG1/S 25hS/G2 Figure 3.33: Beta ratios for synchronized U l cells. Shown is the ratio of the P values (±95% confidence limits) derived from fitting a linear quadratic function to the low (0-4Gy) and high (4-12 Gy) ranges of the survival curves of synchronized U l cells. If the survival curves were homogenous LQ responses, the P ratio would equal 1, indicated by the solid horizontal line. 105 Chapter 3: Results U1 a/p ratios 800 400 o ca 1.5 1.0 0.5 0.0 04GyLQfit 4-12 Gy LQ fit 0-12 Gy LQfit X asynchronous 3hG1 17hG1/S 25hS/G2 Figure 3.34: a/p ratios for U l cells. Figure shows the a/p ratios (+95% confidence limits) calculated for asynchronous and synchronized HT-29 cells using LQ fits to the low (0-4 Gy), high (4-12 Gy), and full (0-12 Gy) dose ranges. 106 Chapter 3: Results 3.5 Radiation induced kinetic responses of HT-29, A549 and U l . Figures 3.35-3.38 show the kinetic responses of mitotically selected synchronized A549 cells after they were irradiated with 0 or 5 Gy at 3 h, 6 h, 11 h, or 16 h post-mitotic selection. It should be noted that measurements of the cell cycle progression of irradiated cells are typically subject to somewhat greater experimental uncertainty than, for example, the survival measurements in Sections 3.3 and 3.4. Nevertheless, clear differences between the behaviour of irradiated cells and unirradiated controls could be observed. A549 cells irradiated at 3 or 6 h post selection, in Gi phase, exhibited a prolonged arrest in Gi phase which blocked about 80% of the cells in Gi phase for at least 38 h. A549 cells irradiated at 16 h post selection (in later S/G2 phase) showed a radiation-induced G2 block of approximately 8 h, with little effect on the exit of cells from S-phase between 20 and 26 h. Surprisingly, A549 cells irradiated at 11 h post selection (in late Gi/early S phase) showed no arrest in Gi phase (exit of irradiated cells from Gi between 11 and 22 h was normal), and little evidence of a G2 block (peak levels of G 2 cells around 24 h were somewhat higher in irradiated samples, but by 32-38 h levels of G 2 cells in the irradiated and unirradiated cells were similar. There may be some evidence of a post-mitotic delay in Gi phase, since levels of Gi cells after mitosis in the irradiated samples peaked higher and later than in unirradiated samples (around 32 h), while the percentage of S phase cells in the post-mitotic cycle were lower in the irradiated samples (see 28-38 h). However, this is not clear cut. It was noted, as expected, that processing (trypsinizing, irradiating with 0 Gy, and replating) the samples did introduce a delay in the cycling of cells. For this reason, the exit from Gi phase occurred later in A549 cells which were mock-irradiated at 6h (figure 3.36) 107 Chapter 3: Results than in cells that were mock-irradiated near the end of G i phase at 1 l h (figure 3.37). Nevertheless, each paired series of irradiated and control samples was treated identically except for the dose given. So comparisons of kinetics within each of the figures 3.35-3.38 are not skewed by this sampling effect. Figure 3.39 illustrates the progression of mitotically selected synchronized HT-29 cells after they were irradiated with 0 or 5 Gy at 3 h post-mitotic selection, in mid-Gi phase. Note the lack of any delay in exit from G i phase in the irradiated cells, and the presence of a G 2 block in the irradiated cells, evidenced by the continued increase in G 2 cells after 17 h and the lack of an increase in G i cells after 12 h in the irradiated samples. Figures 3.40-3.43 show the kinetic responses of mitotically selected synchronized U l cells after they were irradiated with 0 or 5 Gy at 3 h, 8 h, 18 h, or 22 h post-mitotic selection. There was no prolonged G i arrest in U l cells irradiated at 3 or 8 h after mitotic selection; exit from G i in these irradiated samples paralleled the exit of unirradiated samples. In synchronized U l cells irradiated at all times after mitotic selection, there was evidence for a G 2 block of 2-4 h (see the delayed increase in post-mitotic G i cells in the irradiated samples vs. controls around 24 to 32 h); the duration of the G 2 block appeared marginally greater when synchronized U l cells were irradiated later in the cell cycle (compare the relative delayed rise in post-mitotic G i cells between 24 and 32 h in samples irradiated at 18 h and 22 h vs. samples irradiated at 3 h and 8 h). As noted above with A549 cells, there was some variation in the timing of U l cell kinetics between populations which were sampled at different times in G i phase. For example, the exit from G i may have occurred later in mock-irradiated controls that were sampled at 3 h (figure 3.40) compared to mock-irradiated controls that were sampled at 18 h (figure 3.42). This 108 Chapter 3: Results would be an expected side-effect of the cell sampling process, but does not affect comparisons between the paired control and irradiated time courses within each of the figures. The A549BPV cell line was examined for the presence of a functional G i arrest. Figure 3.44 shows D N A distributions of A549 and A549BPV cells taken 16 h after irradiation with 0 or 5 Gy, in a preliminary experiment. Examination of cell cycle distribution at this single time after irradiation has been used as a measure of G i arrest competence (see e.g. Kastan et al. 1992). In figure 3.44, control samples of A549 and A549BPV exhibited very similar D N A histograms (compare left hand panels). However, there was a pronounced decrease in S phase cells 16 h after irradiation with 5 Gy in A549 that was not present in equivalently treated A549BPV cells (compare right hand panels). There was also a larger fraction of G 2 cells in A549BPV compared to A549. These observations were consistent with an abrogation of the G i arrest in the transfected A549BPV cells, which was the expected result. A549BPV cells were synchronized by mitotic selection and the kinetics and single-dose radiation survival of these cells were monitored for 40 h after mitotic selection. A549BPV cells cycled at very close to the same rate as A549 cells, but exhibited a more resistant radiation response, especially in G i phase (figure 3.45; compare to figures 3.5 and 3.8 for A549 cells). To confirm the abrogation of G i arrest in A549BPV, synchronized populations of A549BPV were irradiated with 0 or 5 Gy 3 h after mitotic selection, and the cell cycle progression of these populations was followed for the subsequent 20 h. Surprisingly, irradiated A549BPV did appear to briefly delay in Gj phase after irradiation, as is indicated in figure 3.46 by the retarded exit of irradiated cells from G i relative to the control cells. Thus, the 109 Chapter 3: Results results of this experiment with synchronized cells were not entirely consistent with the preliminary experiment (figure 3.44), and suggested that the G i checkpoint had not been fully disabled in the transfected A549BPV cell line. To complement these measurements of cell kinetics, the induction of the p53 protein after a dose of 5 Gy was measured by Western blotting as a function of time after irradiation in A549, HT-29, U l , and A549BPV (figure 3.47). A549 upregulated levels of the p53 protein by approximately 6-fold within 2 hours of irradiation, while HT-29 and U l cells did not significantly upregulate the protein. Contrary to expectations, A549BPV also upregulated p53 in a similar fashion to A549. This, and the observation of a G\ arrest in A549BPV (figure 3.46) implied that the transfection of A549BPV to abrogate p53 function had not been successful, or was unstable. For this reason, no further studies with A549BPV were undertaken. The mode of cell death after irradiation in A549, HT-29, and U l cells was examined by irradiating cultures of these cells with equitoxic doses (A549: 5.5 Gy, HT-29: 7 Gy, and U l : 7.8 Gy) and following the irradiated cells by observation under the microscope and measurement of D N A fragmentation up to approximately 160 h after irradiation. In A549 cells, the growth of irradiated cultures was slowed compared to control cultures, but some growth continued from 0-160 h after irradiation. Irradiated cells appeared to stay attached to the culture surface. Very few cells detached from irradiated A549 monolayers, and when floating cells did appear it was when the cultures had reached confluence. In HT-29 cells, many irradiated cells detached from the monolayer by 48-72 h after irradiation (unlike control cultures) and irradiated cultures were clearly less confluent 110 Chapter 3: Results than unirradiated controls by 160 h post-irradiation. Especially after 96 h post-irradiation, there were massive numbers of floating cells which did not appear in control HT-29 cultures. The pattern of cell death in U l cells was similar to HT-29 cells over the post-irradiation period. Very large numbers of cells detached from the irradiated U l monolayers, especially after 96 h. D N A fragmentation was quantitated by flow cytometry of fixed samples from the above cultures, after permeabilization of the cell membrane to allow small D N A fragments to diffuse out of the cells (Section 2.10). This experiment was done twice; trends in both experiments were similar. The second experiment showed less D N A fragmentation in all cell lines; the author believes this was due to assay variability. The results of the first experiment are shown in figure 3.48. In both HT-29 and U l cells, a population of cells with less than G i D N A content (presumably due to D N A fragmentation) appeared around the same time as large numbers of floating cells became evident in irradiated cultures. A549 cultures did not contain as many of these sub-Gi cells. Ill Chapter 3: Results +0 Gy (—o—) or 5 Gy (—><—) A549 0 10 20 30 time after mitotic selection (h) Figure 3.35: Kinetics of synchronized A549 cells after irradiation at t=3h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. This experiment was done twice; the results of one representative experiment are shown. 112 Chapter 3: Results Figure 3.36: Kinetics of synchronized A549 cells after irradiation at t=6h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. The results of a single experiment are shown. 113 Chapter 3: Results Figure 3.37: Kinetics of synchronized A549 cells after irradiation at t=l l h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. The results of a single experiment are shown. 114 Chapter 3: Results +0 Gy (• 100 o 80 r-60 h 40 h 20 h 0 0 -) or 5 Gy (—x—) A549 T >— G1 S >•• G2 o o x- - _ _ v X X 10 20 30 time after mitotic selection (h) Figure 3.38: Kinetics of synchronized A549 cells after irradiation at t=16h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. The results of a single experiment are shown. 115 Chapter 3: Results +0 Gy ( - o - ) or 5 Gy ( - x - ) HT-29 time after mitotic selection (h) Figure 3.39: Kinetics of synchronized HT-29 cells after irradiation at t=3h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. 116 Chapter 3: Results Figure 3.40: Kinetics of synchronized U l cells after irradiation at t=3h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. 117 Chapter 3: Results Figure 3.41: Kinetics of synchronized U l cells after irradiation at t=8h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. 118 Chapter 3: Results +0 Gy (—o—) or 5 Gy ( — x — ) 100 1—r U1 T 1 1 1 1 —o—G1 - - 0 - - S - o . . . G2 80 60 40U 20U 0 i L -i i i L 0 10 20 time after mitotic selection (h) 30 Figure 3.42: Kinetics of synchronized U l cells after irradiation at t=18h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. 119 Chapter 3: Results +0 Gy- (— o—) or5Gy (— x — ) 100 U1 80 60 "33 O "6 40U 20 0 0 —r t i i r T r _L -i 1 1 r —o—G1 - - 0 - - S • o.. G2 10 20 time after mitotic selection (h) 30 Figure 3.43: Kinetics of synchronized U l cells after irradiation at t=22h after mitotic selection. Open circles: mock irradiated cells; crosses: cells irradiated with 5 Gy. 120 Chapter 3: Results Figure 3.44 D N A histograms of A549 and A549BPV cells, 16h after irradiation. Samples were irradiated with 0 or 5 Gy and then allowed to grow normally for 16 h. At this time cells were fixed and stained for D N A analysis. Arrows emphasize the lack of a decrease in S phase fraction, and the increase in G2 phase fraction 16h after irradiation in the A549BPV cells, compared to the A549 cells. 121 Chapter 3: Results A549BFV 100 I i i i i | i i i i | i i i i | i i i i | i i i i | i i i i | i i i i | i i i i > 0.07 • j> 0.06 -13 0.05 -003 ' 1 1 1 1 1 1 1 1 1 1 ' 1 1 ' 1 1 1 1 1 1 1 1 1 ' 1 1 1 1 1 1 ' 1 1 ' 1 1 ' ' ' ' 0 5 10 15 20 25 30 35 40 time after mitotic selection (h) Figure 3 . 4 5 : Kinetics and survival of mitotically selected A 5 4 9 B P V transfected cells. Compare responses to Figure 3.5 and Figure 3.8 which show the equivalent kinetic and survival data for A 5 4 9 cells. Top panel is the result of a single experiment; in bottom panel, survival values are the average of 3 determinations( +SEM). 122 Chapter 3: Results Figure 3.46: Kinetics of synchronized A549BPV cells after irradiation. Cells were synchronized by mitotic selection at t=0h, irradiated with 0 or 5 Gy at t=3h, and then sampli were taken for D N A content analysis at staged times during the next 20h. The result of a single experiment is shown. Compare with Figure 3.35 for wild-type A549 cells. 123 Chapter 3: Results 8 7h 6h O 5 O S _ 0) > 4 O CD CO 03 CD 3r-C 2 0 A549, HT-29, U1, A549BPV p53 induction T — i — i — i — i — i — i — i — i — i — r —•— A549 - ©- HT-29 - - -A - U1 - V- A549BPV J I I I I I L 0 1 2 3 4 5 6 7 8 9 10 time after irradiation (h) Figure 3.47: Induction of p53 by irradiation in A549, HT-29, U l , and A549BPV cells. The level of p53 was measured by Western blotting in irradiated samples of each cell line and compared to unirradiated controls. Relative protein levels were quantified by enhanced chemifluorescence and are expressed as fold increases over control levels. For A549, U l , and A549BPV cells, points are the mean of 2-4 determinations ±SEM; for HT-29, the result of 1 typical experiment is shown. 124 Chapter 3: Results A Figure 3.48: D N A fragmentation in irradiated cells. Figure shows D N A histograms of (A) A549; (B) HT-29; (C) U l . Cells with less than G i D N A content, which are presumably dead, and may be apoptotic, are indicated by the arrows in (B) and (C). The appearance of these sub-Gi cells in (B) and (C) coincided temporally with the appearance of large numbers of floating cells in the irradiated cultures. A549 cells did not appear to undergo the D N A fragmentation to the same extent as HT-29 and U l . 125 Chapter 3: Results 3.6 Mathematical model predictions of substructure in the survival responses of synchronous and asynchronous cell populations. 3.6.1 Characterization of the model Modelling was carried out to examine the relationship between the radiation survival responses of mitotically selected synchronized cells and asynchronous cells and the underlying intrinsic radiation sensitivity of age-specific cells. The general scheme of mathematical models was described in Section 2.9. Tables 3.17 and 3.18 show the values of adjustable parameters in the models which were used for all simulations. Figure 3.49 is a typical simulation of the kinetics of HT-29 cells after mitotic selection. The plot is a graphical representation of the number of cells in each cell-cycle compartment of the model as a function of time after mitotic selection. At t=0 h, 85% of the cells are in compartment 15 (the last stage of mitosis). From t=0 h to approximately t=15-20 h most cells complete mitosis, progress into G i phase (starting in compartment #1), and advance thorough the remainder of G i , S, and G 2 phases, gradually losing synchrony as they progress. Around t=T 5-20 h significant numbers of cells begin to enter the first post-selection mitosis, and then re-enter G i phase after division is complete. 126 Chapter 3: Results Table 3.17: Kinetic parameters of the mathematical model Cell Line t(h) a(h) S t(h) a(h) G 2 t(h) a(h) Ml (%) A549 HT-29 U1 10.7 3.0 7.7 1.9 21.0 4.0 6.4 3.0 9.4 2.5 5.0 2.0 5.2 0.5 3.7 1.0 7.0 1.0 95.0 85.0 80.0 Shown are the duration and standard deviations of duration, in hours, for each phase of the cell cycle, and the mitotic index (Ml) in percent. These values were the ones that provided the best fit between the model kinetic predictions of %G, , %S, and %G 2 /M cells in the first complete cycle after mitotic selection and the corresponding measurements of the same quantities. Table 3 18 Radiosensitivity parameters of the mathematical model Cell Line region A k a p region B k a p region C k a p region D k a p A549 HT-29 U1 1-6 0.1381 0.0300 1-4 0.0227 0.0397 1-10 0.0000 0.0326 7-9 0.1284 0.0495 5-8 0.0077 0.0683 11-18 0.0000 0.0580 10-17 0.1421 0.0330 9-14 0.0293 0.0333 19-31 0.0000 0.0420 18-20 0.3583 0.0800 15 0.0022 0.3987 32-33 0.1570 0.3079 For each radiosensitivity region (A, B, C, and D) the table shows k, the range of model compartments covered by a region, and a and p, the LQ parameters of the survival response of all cells in a region. These values were determined by best-fitting to the single dose survival responses of A549, HT-29, and U1 cells, as described in Section 2.9. From the data in figure 3.49, the percentage of model HT-29 cells in Gi, S, and G2/M phase can be calculated as a function of time after mitotic selection. This was done for HT-29, and repeated for models of A549 and U l cells in the first cycle after mitotic selection (figures 3.50, 3.51 and 3.52, top panels). A good correspondence was achieved between model and measured cell kinetics in the mitotically selected cells: the root-mean-square deviation of the model Gi, S, and G2/M fractions vs. the measured fractions in the first cycle after mitotic selection was approximately 4.1%, 6.1%, and 5.9% for A549, HT-29, and U l cells respectively. A simulation of asynchronous cells was generated by allowing an initially synchronized population to progress through at least 20 cell cycles until it reached a stable, exponentially growing, cell cycle distribution. Asynchronous model distributions were in generally good agreement compared to measured asynchronous distributions for A549, HT-29, and U l (figures 3.53, 3.54, and 3.55). 127 Chapter 3: Results 3.6.2 Modelling of the single dose survival responses Linear quadratic radiation responses were assigned to each compartment of models as detailed in Table 3.18, and the predicted single dose radiation responses of mitotically selected cells were generated and compared to normalized measured single dose responses in the first cycle after mitotic selection. The rationale for normalizing the measured single dose responses was discussed in Section 2.9. The normalizing factors (the fractions by which observed single dose survivals were multiplied before comparison with models) were 1.74, 1.45, and 1.35 for A549, HT-29, and U l respectively. The effect of the normalization on survival was equivalent to the effect of a change in zero-dose plating efficiency. In A549 cells, the model single dose response appears to follow the observed response well, although the model tends to underestimate survival during the first 8 h after mitotic selection, in G i phase (figure 3.50, bottom panel). In HT-29 cells, the model single dose survival is in good agreement with experiment throughout the first cycle (figure 3.51, bottom panel). In U l cells, the model reproduces the general trends in the measured survival response fairly well, but tends to underestimate the magnitude of changes in radiation survival between 15 and 25 h after mitotic selection, when most cells are concentrated in later G i and S phase (figure 3.52, bottom panel). These initial model simulations raised two questions about the kinetics and survival of mitotically selected synchronized cells which were addressed by some additional modelling work. These questions related to (1) the counterintuitive single dose survival, responses around the end of the first cycle after mitotic selection in HT-29 cells, and (2) the quality of resolution of the observed single dose response during G i phase in mitotically selected U l cells. 128 Chapter 3: Results Model HT-29 Figure 3.49: Progression of model HT-29 cells after mitotic selection. Cells start in the last compartment of the model (mitosis) at t=0. At each time step after this, cells progress and disperse through the cell cycle. The average cycle time in the example above is 20.8h. The gradual time-dependent dispersion of the initial mitotic population is evident in the widening of the synchronized population distribution. 129 Chapter 3: Results A549 0.C8 I i i i i I i i i i I i i i i I i i • • I 0 5 10 15 20 time after mitotic selection (h) Figure 3.50: Comparison of model with A549 cells . Top panel shows the percentage of cells in each phase of the cell cycle as a function of time after mitotic selection. Bottom panel shows the measured and model-calculated cell survival after a dose of 5 Gy. Agreement in kinetics (top panel) between model and measured values is good; radiation survival agreement is not as good for reasons discussed in the text. 130 Chapter 3: Results HT-29 o CO > to A A • -A • • A- • -I • • A • • » -A i - • i • • A - -A - • i A i I 1 1 >- T — r 0.3 0.2 0.1 0.C9 0.C6 0.07 normalized HT-29 -\ model x -i i_ 0 10 15 20 t i m e a f t e r mi to t i c s e l e c t i o n (h) Figure 3.51: Comparison of model with HT-29 cells. Top panel shows the percentage of cells in each phase of the cell cycle as a function of time after mitotic selection. Bottom panel shows the measured and model-calculated cell survival after a dose of 6 Gy. Agreement in kinetics (top panel) between model and measured values is good; radiation survival agreement is not as good for reasons discussed in the text. 131 Chapter 3: Results Figure 3.52: Comparison of model with U l cells . Top panel shows the percentage of cells in each phase of the cell cycle as a function of time after mitotic selection. Bottom panel shows the measured and model-calculated cell survival after a dose of 8 Gy. Agreement in kinetics (top panel) between model and measured values is good; radiation survival agreement is not as good for reasons discussed in the text. 132 Chapter 3: Results Model A549 asynchronous distribution 1 1 1 1 1 model A549 asynch (n=3) Figure 3.53: Comparison of distribution of model A549 cells with measured cells. The percentages of G i , S, and G2/M cells were measured by D N A content analysis in asynchronous A549 cells and compared to model predictions of the same percentages. 133 Chapter 3: Results Model HT-29 asynchronous distribution CO 30 h (D O o 2 0 h model HT-29 asynch (n=3) Figure 3.54: Comparison of distribution of model HT-29 cells with measured cells. The percentages of Gi , S, and G 2 / M cells were measured by D N A content analysis in asynchronous HT-29 cells and compared to model predictions of the same percentages. 134 Chapter 3: Results Model U1 asynchronous distribution model u1 asynch (n=4) Figure 3.55: Comparison of distribution of model U l cells with measured cells. The percentages of G i , S, and G2/M cells were measured by D N A content analysis in asynchronous U l cells and compared to model predictions of the same percentages. 135 Chapter 3: Results 1. Single dose survival trends at the end of the first cycle in A549 and HT-29 It was observed that in the measured HT-29 single dose survival responses, there was a monotonic decrease in survival between 16 and 24 h after mitotic selection (figure 3.56), even though the synchronized cells presumably passed through the resistant early G i phase during this time. On the contrary, in A549, there was some suggestion of the expected increase in measured single dose survival between 18 and 24 h after mitotic selection (figure 3.56) when these cells would be expected to be traversing the resistant early G i phase. To understand the difference between the two cell lines over this time period, the model single dose survival responses were extended past the end of the first cycle after mitotic selection and the distributions of synchronized cells in the models of HT-29 and A549 were calculated for the time periods around 16-24 h. The model single dose survivals of A549 and HT-29 did generally agree with the measured trends in survival over the 16-24 h time periods (figure 3.56): the models showed an increase in survival during this time period in A549, and a decrease over the same time period in HT-29. The model distributions of the synchronized cells over this time period gave a clue as to the reason for the different survival trends in the two cell lines. In HT-29, the synchronized cells at 16 h were distributed mostly over the radioresistant region C (figure 3.58) resulting in relatively high cell survival at this time. At 20 h, the HT-29 cells were centred on the radiosensitive region D, leading to a decrease in cell survival that was partially offset by the presence of radioresistant cells in region A, which were relatively greater in number due to cell division (figure 3.58). Between 20 and 24 h, the number of radiosensitive HT-29 cells in region D decreased, but this was offset by the entry of the 136 Chapter 3: Results leading edge of the synchronized cells into the radiosensitive region B, which led to further overall decreases in survival. A549 cells at 18 h after mitotic selection were centred over the later part of the cycle, with a substantial fraction of cells in the radiosensitive D region, which resulted in a local minimum in survival at this time (figure 3.57) (the "dip" in the A549 cell distributions in the first part of G2/M phase is due to change of transition probabilities at the S/G2 border, and is an expected feature of the model). At 21 h, there were slightly more A549 cells in the D region, but this was offset by the production of newly divided resistant cells in the A region, leading to a small increase in overall survival. Between 21 and 24 h, the trailing edge of the A549 cells left the D region, more cells appeared in the radioresistant A region, while only small numbers of cells at the leading edge of the population reached the more radiosensitive B region. Thus the overall trend between 21 and 24 h was a small increase in single dose survival. 2. Resolution of the single dose response in Ul cells The model of U l cells (which had the longest G i phase of all the cell lines) postulated a discrete increase in radiosensitivity at the border between regions A and B in G i phase. This single-compartment increase took place over approximately 3% of the total duration of the cycle (between compartments 10 and 11 in the model; see figure 3.80). It was of interest to know what the effect of postulating a different distribution of radiosensitivity would be on the model-predicted single dose survival response in G i phase. A "smooth" model was created where the radiosensitivity of U l cells increased gradually over an interval in G i which was approximately 30% of the total cycle length (10 of 33 compartments), with the point of maximum radiosensitivity maintained at the centre 137 Chapter 3: Results of region B. When the measured single dose survival response of synchronized U l cells was compared to the prediction of this smooth model and the original model, it was observed that the two models could probably not be distinguished based on the measured data (figure 3.59). Thus, desynchronization of the U l cells in Gi phase appeared to prevent the resolution of the fine structure of the underlying radiosensitivity changes during this phase. 3.6.3 Modelling of substructure in survival curves In order to examine the expected degree of substructure in the survival curves of mitotically selected populations, the models were used to generate survival curves for asynchronous and mitotically selected A549, HT-29, and U l cells at 4 times after mitotic selection. Asynchronous model simulations were carried out by allowing initially synchronized cells to advance at least 20 full cell cycles, at which time they had reached a stable, exponentially growing distribution. The model survival curves are shown in figures 3.60-3.64 (A549), 3.67-3.71 (HT-29), and 3.74-3.78 (Ul), plotted against the corresponding measured curves. Single LQ fits (table 3.4, 3.9, 3.14) to the measured curves are also included for comparison on these figures. As a measure of substructure in these survival curves, the ratio of the LQ P values on the low (0-4 Gy) and high (4-12 Gy) dose ranges (Ph;/Pi0) was calculated for each model simulation as well as for the measured survival curves. This ratio is equal to 1 for a homogenous LQ survival curve with no substructure, and is theoretically less than 1 for a heterogeneous cell population that displays substructure. 138 Chapter 3: Results 0.07 • • • 1 1 • • • 1 • • • • 1 • • • • 1 • 1 • • 1 • • • • 1 • • • • 1 • • • • I 0 5 10 15 20 25 30 35 40 time after mitotic selection (h) Figure 3.56: Model A549 and HT-29 single dose survival profiles over the full 40h time course. 139 Chapter 3: Results Model A549 model ccfrpartment 1 5 10 15 20 i i i i | i i i i | i i i i | i i i i | Figure 3.57: Schematic of A549 model. Cell distributions among the model compartments are shown at 18, 21, and 24 h after mitotic selection. For reference, the limits of G l , S, and G2/M phases and the A, B, C, and D radiosensitivity regions in the model are indicated by bars at the bottom of the figure. 140 Chapter 3: Results Model HT-29 Figure 3.58: Schematic of HT-29 model. Cell distributions among the model compartments are shown at 16, 20, and 24 h after mitotic selection. For reference, the limits of GI , S, and G2/M phases and the A, B, C, and D radiosensitivity regions in the model are indicated by bars at the bottom of the figure. 141 Chapter 3: Results U1 0.2 0.1 0.09 0.08 I-TB 0 0 7 I" • | 0.06 h ™ 0.05 0.04 0.03 o.o3 -m t 0.02 '—1—•—•—L 0 rrxmalized U1 model smoothed model j i_ j_ 0.3 -0.2 '-co. 0.05 h-O 8 5 10 15 tirre after rritotic selection (h) 1 1 1 I 1 1 T i i i l i 0-0-0 * - V 7 - t ? - 9 o - o - o - o - o - s 1 I I I I I - • — a - o - p -A—smoothed a —v— smoothed p 20 o - o - O - o - 6 - s =>o -9 -<y -o -9 -Q-^ -o -o -o Q Q Q ^^ -A -& -A -a -& -a -a -n -a -A -a -a -A -a -& -a -a -a - f l - a -A -a -a -a^ I • . . . I • . . . i . . . • I . . . . I . . . . I 10 15 20 25 model coirpartirent (1-33) 30 Figure 3.59: Comparison of model U l with "smoothed" model. Bottom panel: plot of the LQ a and P parameters in the model and smoothed model. In U l , ct=0 throughout interphase in both models; P changes abruptly between radiosensitivity regions in the model (at compartments 11,19, and 32), while it varies gradually in the smoothed model. Top panel: Single dose survival (at 8 Gy) is plotted during G i and early S (t=0 to 20 h after mitotic selection) in U l (normalized), and the two models. The rates of change of survival during G i phase in the two models appear to be experimentally indistinguishable. 142 Chapter 3: Results In A549 cells, the model predicts some substructure in all asynchronous and synchronized populations except 3 h Gi phase cells (figure 3.65). The model prediction of ph/Piofor 0 h mitotic cells is close to that which is measured, and is less than for any other cell population. However, the A549 model generally predicts less substructure than was measured (model predictions of Pi,i/Pi0 were closer to 1 than the measured values of Phi/Pio)- The model ranking of substructure in the different populations (from most substructure to least) was: 0 h mitotic cells, 10 h Gi/S phase cells, asynchronous cells, 15 h S/G2 phase cells, 3 h Gi phase cells.. The corresponding ranking by the measured pw/Pio values was: 0 h mitotic cells, asynchronous cells, 3 h Gi phase cells, 10 h Gj/S phase cells, 15 h S/G 2 phase cells. The degree of substructure in the model was determined by the distribution of cells in each of the radiosensitivity regions for each of the synchronized populations. Figure 3.66 shows the cell distributions in the model at 0, 3, 10, and 15 h after mitotic selection. In HT-29 cells, the model predicts some substructure in all cell populations, synchronized and asynchronous (figure 3.72). In the HT-29 model, the degree of substructure is greatest in 0 h mitotic cells (i.e. Pm/Pio close to 0) and least in 3 h Gi phase cells (i.e. Phi/Pio close to 1), and is intermediate in 8 h, 15 h, and asynchronous cells. This is in approximate agreement with the measured Ph;/Pi0 values, which indicate a similar rank ordering of substructure: most in 0 h mitotic cells, least in 3 h Gi cells, and intermediate values in 8 h, 15 h, and asynchronous cells. Figure 3.73 shows the cell distributions in the HT-29 model at 0, 3, 8, and 15 h after mitotic selection. In U l cells, there is a dramatic discrepancy between model and measured Phi/Pio values (except for 0 h mitotic cells) (figure 3.79). While all the model Phi/pio values are 143 Chapter 3: Results less than one, as expected, all of the measured Phi/Pio values are greater than one. For the one exception, 0 h mitotic cells, the model predicts a Phi/Pio much lower than any other population, in accordance with the measured Phi/Pio for 0 h mitotic cells. During interphase, the 17 h G\/S cells had the lowest degree of measured substructure (i.e. the measured Pw/Pio was closer to 1 for these cells than for any other synchronized cells in interphase). Figure 3.80 shows the cell distributions in the U l model at 0, 3, 17, and 25 h after mitotic selection. Theoretically, if the model was accurate and the measured substructure was due to cell-age related variation in radiosensitivity alone, then the measured ratio Phi/Pio for each synchronous and asynchronous population should be correlated with the heterogeneity in the survival response of the corresponding cell population in the model. One measure of survival heterogeneity in the model was the variance of the survival S(D) (equation (2.11)). This variance was calculated for D=4 Gy (an arbitrary choice) using standard statistical methods. Mitotic cells had much larger variance in model survival than any other synchronized or asynchronous population, as expected. In HT-29, the expected correlation between measured Ph/Pio and model survival variance was found. This correlation was negative, and weakly significant with or without including the mitotic cells, (R=-0.87, p=0.06 with mitotic cells; R=-0.76, p=0.24 without mitotic cells) (figure 3.81). In A549, where the fluctuations in model survival were smaller than in HT-29, no significant correlation was found (data not shown). In U l cells, there was, as in HT-29, a weakly significant negative correlation (R=-0.89, p=0.04) when mitotic cells were included, but a non-significant positive correlation when these cells were excluded from the analysis (figure 3.81). 144 Chapter 3: Results Figure 3.60: Model and measured survival curves in asynchronous A549 cells. Also shown is a single LQ fit to the survival data, for comparison. 145 Chapter 3: Results A549 Oh M phase treasured model single LQfit 0.0 -0.2 _l_ 6 8 Cose(G/) 10 12 Figure 3.61: Model and measured survival curves in A549 (0 h mitotic cells). Also shown a single LQ fit to the survival data, for comparison. 146 Chapter 3: Results Figure 3.62: Model and measured survival curves in A549 cells (3h G\ phase cells). Also shown is a single LQ fit to the survival data, for comparison. 147 Chapter 3: Results A549 10hG1/S phase 0.001 0.8 h 0.6 h -0.2 h 0 2 4 6 8 10 12 Coee(Gy) Figure 3.63: Model and measured survival curves in A549 (lOh Gi/S phase cells). Also shown is a single LQ fit to the survival data, for comparison. 148 Chapter 3: Results A549 15h S/G2 0.001 — I — I — I — I — I — I 1—I 1—I V 0.8 h 0.6 h -0.2. L 1 I i l i I i ' • i • (I 2 4 6 8 10 12 D08e(Gy) Figure 3.64: Model and measured survival curves in A549 (15h S/G 2 phase cells). Also shown is a single LQ fit to the survival data, for comparison. 149 Chapter 3: Results A549 1.2 1.0 0.8 o '•+—' rc o CO. CO. 0.6U 0.4 U 0.2 U 0.0 T5hS/G3T> async Q3hG1 8^ 10h G1/S >15hS/G2 >10h G1/S |3hG1 (j|async O M m e a s u r e d model Figure 3.65: A549 model and measured Phi/ Pio- Comparison of the ratio of ph;/ P i 0 calculated from fitting the low (0-4 Gy) and high (4-12 Gy) dose ranges in model and measured data for A549 cells. 150 Chapter 3: Results Model A549 model ccrrpartment 5 10 15 T 1 1 1—1 1 1 — I — i — •>—r 20 1 t=3h t=0h CD O / I / » / > / > I » / I / » t=10h t=15h N * / \ / U.--i^-.-<". . • • • • - I - • • \ - • • • • • I • • • • \ G2/M A B C D Figure 3.66: Schematic of A549 model. Cell distributions among the model compartments are shown at 0, 3, 10, and 15 h after mitotic selection. For reference, the limits of Gi, S, and G2/M phases and the A, B, C, and D radiosensitivity regions in the model are indicated by bars at the bottom of the figure 151 Chapter 3: Results 0.61--0.2 r-I i I i i i i i i i i i I 0 2 4 6 8 10 12 Dose (Gy) Figure 3.67: Model and measured survival curves in asynchronous HT-29 cells. Also shown is a single LQ fit to the survival data, for comparison. 152 Chapter 3: Results HT-29 Oh M phase measured model single LQfit 0.0 -0.2 1 J i l i I i I i_ 2 4 6 8 10 12 Dose (Gy) Figure 3.68: Model and measured survival curves in HT-29 (Oh mitotic cells). Also shown is a single LQ fit to the survival data, for comparison. 153 Chapter 3: Results 0.6 h 0.4 CO, c 0.2 0.0 -0.2+-4 6 8 Dos9(Gy) 10 12 Figure 3.69: Model and measured survival curves in HT-29 (3h G i phase cells). Also shown is a single LQ fit to the survival data, for comparison. 154 Chapter 3: Results HT-29 8h G1/S phase 5 3 CO 0.01 0.001 0.8 0.6 -0.2 measured model single LQfit H 1 1 1 1 1 1 1 y 2 4 6 8 DDse(Gy) 10 12 Figure 3.70: Model and measured survival curves in HT-29 (8 h Gi/S phase cells). Also shown is a single LQ fit to the survival data, for comparison. 155 Chapter 3: Results HT-29 15h S/G2 phase measured model single LQfit 2 CO 0.01 0.001 0.8 0.6 -0.2 L H 1 1- H 1 1 1 1 1 H 2 4 6 8 10 12 Dose (Gy) Figure 3.71: Model and measured survival curves in HT-29 (15 h S/G 2 phase cells). Also shown is a single LQ fit to the survival data, for comparison. 156 Chapter 3: Results HT-29 g '-I—» 03 o CO. measured model Figure 3.72: HT-29 model and measured Phi/ pi0- Comparison of the ratio of ph;/ Pi 0 calculated from fitting the low (0-4 Gy) and high (4-12 Gy) dose ranges in model and measured data for HT-29 cells. 157 Chapter 3: Results Model HT-29 model oompartment 1 5 10 15 i — i — i — i — | — i — i — i — i — | — i — i — i — i — | Figure 3.73: Schematic of HT-29 model. Cell distributions among the model compartments are shown at 0, 3, 8, and 15 h after mitotic selection. For reference, the limits of Gi, S, and G2/M phases and the A, B, C, and D radiosensitivity regions in the model are indicated by bars at the bottom of the figure. 158 Chapter 3: Results U1 Asynchronous 03 I , , . , . , . r 0 1 2 3 4 0.001 1 1 1 1 H-H ' 1 1 1 ^ 0.8 -0.6 f -0.2 h i i i i i • i i i i 0 2 4 6 8 10 12 Cose(Gy) Figure 3.74: Model and measured survival curves in asynchronous Ul cells. Also shown is single LQ fit to the survival data, for comparison. 159 Chapter 3: Results U1 Oh M phase o.oocra h — i 1—i 1—i 1—i—| 1 1 1 .0 2 ' 1 1 1 1 1 1 1 1 ' 1 1 I 0 2 4 6 8 10 12 Do 33 (Gy) Figure 3.75: Model and measured survival curves in U l (0 h mitotic cells). Also shown is single LQ fit to the survival data, for comparison. 160 Chapter 3: Results 0.6 h 161 Chapter 3: Results U1 17hG1/S phase 03 3 0.001 0.01 0.CO01 0.8 0.6 0.44-•0.2 0 1 2 3 4 H 1 1 1 1 1 1 1 1 1 H J i l_ -1 i l_ 0 2 4 6 8 10 12 Doee(Gy) Figure 3.77: Model and measured survival curves in U l (17 h Gi/S phase cells). Also shown is a single L Q fit to the survival data, for comparison. 162 Chapter 3: Results Figure 3.78: Model and measured survival curves in U l (25 h S/G2 phase cells). Also shown is a single LQ fit to the survival data, for comparison. 163 Chapter 3: Results 2.0' 1.8-1.6-1.4-1.2-g o c a 0.8-I c a 0.6-0.4-0.2-0.0-U1 asynchrcnous 3hG1 )23r\S/G2 M7hG1/S l-M-8 G2 asynchrcnous 17hG1/S o M measured model Figure 3.79: Ul model and measured Ph;/ Pio- Comparison of the ratio of Phi/ P i 0 calculated from fitting the low (0-4 Gy) and high (4-12 Gy) dose ranges in model and measured data for Ul cells. 164 Chapter 3: Results Model U1 model compartment 1 5 10 15 20 25 30 l 1 1 • I 1 1 1 1 l ' 1 1 1 l 1 1 1 1 I 1 1 1 1 l 1 1 1 1 I 1 1 1 t=0h = 3 t=3h / \ t=17h t=25h V 1 r i - r t - T ^ t T -l-l-t- • • • 1 • • • • 1 • s G2/M II^ IHIIIHIIII I B c 11 Figure 3.80: Schematic of Ul model. Cell distributions among the model compartments are shown at 0, 3, 8, and 15 h after mitotic selection. For reference, the limits of GI, S, and G2/M phases and the A, B, C, and D radiosensitivity regions in the model are indicated by bars at the bottom of the figure. 165 Chapter 3: Results HT-29 i — 1 — i — 1 — r I i i i i i i i i i i i I 0.0 0.2 0.4 0.6 0.8 1.0 1.2 measured phj/r3 Io ratio Figure 3.81: Measured substructure and model variance in survival in HT-29. The data suggest that cell populations with greater heterogeneity (greater variances in survival) also have more substructure (lower bVPio ratios). However, the correlation is fairly weak (P=0.06, for all the data points; P=0.24 when mitotic cells are excluded). 166 Chapter 3: Results U1 —r 4.0x10" 2 m CD 8 c CO ro 2.0x10"2 > "ro > ' f c Z J CD 0 "D O E 0.0 \-X repression (all data points) R =-0.88781 P=0.04434 async Ftec/ession (exdudng mitotic cells): R =0.33298 P=0.66702 _L J . X 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 beta ratio Figure 3.82: Measured substructure and model variance in survival in U l . When mitotic cells are included, there appears to be a significant correlation (P=0.04) between model variance (heterogeneity) and the (3hi/f31o ratio. However, there is no significant correlation when mitotic cells are excluded. 167 Chapter 4: Discussion 4.1 Three human tumor cell lines exhibited common variations in single-dose radiosensitivity after mitotic selection. These variations were related to age-specific radiosensitivity by a mathematical model. Summary of findings Cell-cycle related variations in radiosensitivity were measured by giving approximately equitoxic single doses of ionizing radiation to synchronized cells of three human tumor cell lines from different disease sites: A549 (lung), HT-29 (colon), U l (skin). It was found that all cells shared important traits: they were most radiosensitive in mitosis, radioresistant in early Gi phase and in later S/G2 phase, and locally radiosensitive at times close to, or just before, the Gi/S transition. At the level of approximately 10% survival, the ratio of cell survival at the most and least radioresistant points in interphase was approximately 1.6, 2.5, and 3.5 in A549, HT-29, and U l respectively. The differences in radiosensitivity of cells at the most resistant and most sensitive points in interphase as described by this ratio were significant in all cell lines (i.e. the ratios were greater than one, p < 0.001 by a two-tailed t-test). In all three cell lines, there was an increase in the radiosensitivity of cells in Gi phase, before the start of D N A synthesis at the Gi/S border. This was established by the observation that in all three cell lines, a local maximum rate of decrease in the single dose survival occurred before the maximum rate of increase in the S phase fraction, or the maximum rate of decrease of the Gi phase fraction, as determined by D N A content 168 Chapter 4: Discussion measurements (figures 3.5 to 3.7). The interphase minimum in single dose cell survival for A549 and HT-29 occurred at the same time as the maximum rate of transition into S, and out of Gi phase. This implied that there was a region of maximum radiosensitivity centred at the Gi/S transition point in A549 and HT-29. Intriguingly, in U l cells, which had a longer Gi phase than A549 or HT-29, the interphase minimum in single dose cell survival was located before the time of maximum rate of G]/S transition, and at the time of maximum Gi/S transition, the rate of increase in the single dose survival profile of U l was close to a local maximum. This implied that in U l cells, the region of maximum interphase radiosensitivity around the Gi/S transition was not centred at the Gi/S transition, but before the Gi/S transition and the start of D N A synthesis. The mathematical model of synchronized cell progression supported these observations about the timing of radiosensitivity changes in the cell cycles of the three cell lines. Importantly, the model incorporated a realistic picture of cell progression and desynchronization. It was shown that the observed rates of change in single dose radiation survival during the first cycle after mitotic selection were greatly influenced by cell desynchronization. For example (figure 3.80), measured rates of variation in single dose survival during the first 20 h after mitotic selection in U l cells could be reasonably reconciled with either a rapid step-wise change in radiosensitivity (occurring over approximately 3% of the cell cycle), or a much slower equal-magnitude gradual variation in radiosensitivity (occurring over approximately 30% of the cell cycle). The finite age-width of synchronized cell distributions made it unlikely that these two types of radiosensitivity variation could be unambiguously discriminated by analysis of the single dose survival responses. Regardless of the exact nature of the time course of 169 Chapter 4: Discussion radiosensitivity variation during Gi and S phase, the best correspondence between model and measured single dose survival trends was found by assuming a radiosensitive region approximately centred on the Gi/S transition in A549 and HT-29, and a radiosensitive region centred in Gi, before the Gi/S transition, in U l . This finding in the model was in agreement with analysis of the measured single dose survival responses, as described in the previous paragraph. Other than this suggestion of a subtle difference between A549 and HT-29 on one hand, and U l on the other, the measured single dose survival responses and the mathematical model of those responses provided evidence for important similarities between the cell-age dependent radiation responses of all three cell lines. In all three lines, as noted above, mitosis and points near the Gi/S transition were radiosensitive, and early Gi phase and later S/G2 phase were radioresistant. The mathematical model usefully explained an interesting feature of the HT-29 single dose response. In this response, it was observed that radiation survival decreased monotonically from about 16 to 24 hours after mitotic selection (figure 3.9). During this time interval, the model HT-29 cells advanced from later S/G 2 at around 16 h, through mitosis and early Gi phase, to the Gi/S border at approximately 24 h (figure 3.58). Since early Gi phase was radioresistant, it was initially expected that the cells should have shown a local peak in radiation survival between 16 and 24 h, as they did between 0 and 8 h in the first cycle after mitotic selection. The model showed that the reason no such peak was present during the 16-24 h period was twofold: 1. The leading edge of the synchronized population entered mitosis around 16 h (figure 3.58), which produced a decrease in survival. This decrease in survival was not nearly as 170 Chapter 4: Discussion dramatic as seen in freshly selected mitotic cells (0 h after mitotic selection) because the maximum percentage of cells in mitosis was much reduced by desynchronization, and because of the production at the end of the cell cycle of 2 radiation resistant Gi daughter cells for every 1 radiation sensitive mitotic cell which completed mitosis. 2. From about 20-24 h the major change in the distribution of model cells was an entry of the leading edge of the desynchronized population into the radiosensitive region around the Gi/S transition (figure 3.58), which produced further reductions in radiation survival to the minimum in survival at 24 h. It is interesting that unlike HT-29, A549 cells showed some evidence of an increase in radiation survival as cells entered the second post-selection Gi phase (between 18 and 24 h; see figure 3.8). The mathematical model suggested that this increase was observable in A549, but not HT-29 because the temporal width of the radioresistant part of Gi phase was greater in A549 than HT-29. In the model A549, the resistant part of Gi was long enough relative to the dispersion width of the synchronized populations that the synchronized cells developed observable resistance between 18 and 24 h, before this effect was overwhelmed by entry into the radiosensitive region around the Gi/S transition (see figures 3.56 and 3.57). Comparison to results of other investigators The present findings of radiosensitivity variation during the cell cycle in A549, HT-29, and U l are in general agreement with several of the earlier studies in HeLa cells which were discussed in Section 1.5.1. The finding of a resistant portion of the cell cycle in early 171 Chapter 4: Discussion Gi phase is, like in HeLa cells, in contrast to the relative radiosensitivity in Gi phase which was found in numerous studies of rodent cells with short Gi phases, discussed in Section 1.5.1. There have been three recent investigations of cell-cycle related variations in radiosensitivity in human tumor cells (Biade et al. 1997, McGinn et al. 1994, Tang et al. 1994), which were published while the present work was underway. These studies provide comparative measurements of radiosensitivity in synchronized HT-29 cells. The author is unaware of any extant measurements of radiosensitivity in synchronized A549 or U l cells. In the study of Biade (1997), HT-29 cells were synchronized by mitotic selection, and had a doubling time of about 20 h, with the durations of Gi and S phases estimated to be about 10 h and 6.5-9.5 h respectively. The initial mitotic index of synchronized populations was approximately 90%. Al l these figures are very similar to those determined in the present work. Using a single dose of 5 Gy at intervals of about 3 h, corresponding to a survival level of approximately 0.1 for interphase cells (very similar to the current work, which used a single dose of 6 Gy to give a mean interphase survival of about 0.1) Biade found the HT-29 cells were "relatively resistant in late S phase" and "exhibited minimal variation throughout Gi phase and did not appear to be radiosensitive in G 2 phase". Agreement of these findings with the present work is mixed; in this thesis the radiosensitivity of HT-29 cells was found to vary distinctly in Gi phase, and late S phase was not as radioresistant relative to early Gi phase as was found by Biade. However, on close examination it appears that some of the disagreement between Biade 172 Chapter 4: Discussion and the current study might be explained by the timing and number of survival determinations over the first 20 h after mitotic selection. McGinn (1994) synchronized HT-29 cells using ( N 2 0 + mitotic selection), (colcemid + mitotic selection), and mimosine (a metabolic inhibitor which blocks at the Gi/S border). Radiation survival curves (6-7 doses between 0 and 12.5 Gy) were measured at 3-4 times after synchronization. It was concluded that there was "no significant difference between any of the [synchronized] populations at the level of 10% survival", although analysis of LQ a and P values fitted from measured survival curves indicated "a trend toward increasing sensitivity as cells entered S phase...[and]...increased resistance in populations enriched in late S-phase when compared to Gi populations". Among these findings, only the observation of increasing sensitivity as cells enter S phase is supported by the present work. The finding of no significant age-related differences in survival might be explained by the relatively limited number of experiments (in some cases n=2 determinations) and the time resolution of the study; survival was sampled 3-4 times over the course of a cell cycle, compared to approximately 10-20 times in the present work. Tang (1994) measured the radiosensitivity of HT-29 populations enriched in Gi , Gi/early S, mid to late S, and G 2 / M phase cells by centrifugal elutriation. Using mean inactivation dose as a measure of radioresistance, Tang concluded that in HT-29 cells "radiosensitivity...is [not] strongly dependent on cell age". There were no significant differences in mean inactivation dose through the cell cycle, but "elutriated Gi/early S and mid to late S-phase cells [tended to be] more resistant to radiation than cells in Gi and G 2 / M phase". These findings are partially contradictory to the present results, which 173 Chapter 4: Discussion suggest in particular that Gi/early S cells are more sensitive to radiation than Gi cells. The discrepancy could be due to the fact that elutriated Gi cells are probably not as age-specific as mitotically selected cells allowed to progress into Gi , due to natural variation in the size of age-specific cells (Grdina et al. 1984). Elutriated Gi populations might contain enough later Gi/S (sensitive) cells to make them radiosensitive compared to elutriated Gi/early S populations, which conversely might contain enough later S /G2 (resistant) cells to make them more radioresistant. Such mixed populations could account for the discrepancies between the present work and that of Tang. It may be indicative that in addition to Tang's report, three other recent studies in other human tumor cells found elutriated Gi-phase cell populations were more radiosensitive than later S or G 2 cells: West (1988) with WiDr (human colon adenocarcinoma), Quiet (1991) with JSQ-3 and SCC-61 (human squamous cell carcinomas), and Olive (1997) with TK6 (human lymphoblasts). Implications of the present findings The finding of radiosensitivity increases during Gi phase in A549, HT-29, and U l cells, before the initiation of S phase, is intriguing. This has been recognized before in HeLa cells (Terasima et al. 1963), and may shed light on the mechanisms responsible for cell-age variation in radiosensitivity. Depending on what one considers to be the "baseline" radiation response, it indicates that biochemical events which occur in Gi phase before the start of D N A synthesis make later Gi phase cells radiosensitive, or conversely, make early Gi phase cells relatively radioresistant. 174 Chapter 4: Discussion According to current thinking about intrinsic radiation sensitivity, several potential mechanisms for G i radiosensitivity variation could be envisioned. First, there may be cell-age dependent variations in the yield of initial D N A damage from ionizing radiation, caused by, for example, changing levels of radioprotective molecules such as sulfhydryls during G i phase (see e.g. Han et al. 1976). However, there is some evidence that this is not the case (e.g. Olive et al. 1990). Second, the cell's ability to restitute damage could be cell-age dependent during G i , due to fluctuations in the levels of key molecules involved in damage recognition and repair like the DNA-PK, A T M , or R A D proteins. There is only limited information about such molecular fluctuations (e.g. Chen et al. 1997, Lee et al. 1997) and current assays of D N A repair capacity have not given clear conclusions on this point. A third possible explanation would be that as G i cells approach S-phase, the progressive reduction of time available for repair before the initiation of D N A synthesis leads to increased unrestituted damage at the start of D N A replication, failure of normal D N A replication, and eventual cell death. This is consistent with a commonly held view that G i arrest protects cells from the potentially fatal effects of D N A replication from a damaged template (see discussion in Gupta et al. 1996). To this hypothesis, two features of the observed U l age response seem relevant. First, in U l cells, which have the longest G i phase of the three cell lines examined here (around 21+4 h; see table 3.17), and did not arrest after irradiation in G i phase, radiosensitivity increased during the interval from 4-8 h after mitotic selection. Based on the mathematical model of U l cells, this time interval (4-8 h after mitotic selection) was at least 5-9 h before the start of D N A synthesis for the fastest-cycling of these cells, which would have commenced S phase no earlier than ~13 h after mitotic selection (based on the 175 Chapter 4: Discussion mathematical model of U l cells; see figure 3.52). Reported rates of D N A double strand break repair vary, but figures for typical human tumor cells suggest that the largest fraction (60-100%) of double strand breaks is repaired by 1 h after irradiation (see e.g. Whitaker et al. 1995), and in addition there is some evidence that in more radioresistant cell lines (like U l ) , double-strand break rejoining is faster and/or more complete (Giaccia et al. 1992). Therefore, it seems somewhat unlikely that DSB repair would have been cut short by the start of D N A synthesis in U l cells which were 5-9 h away from entering S phase. If repair was complete for all cells which were irradiated during this interval in Gi phase, then the differential "fixation" of D N A damage at the Gi/S boundary could not have had any effect on the radiosensitivity of these cells; yet, the cells in the later part of the interval were slightly more radiosensitive. Second, the measured U l age response indicated that the period of maximum radiosensitivity in Gi or S phase occurred in later Gi phase, before the onset of D N A synthesis at the Gi/S border (see figure 3.10). Correspondingly, the mathematical model of the U l age response predicted that the period of radiosensitivity in mid- to late-Gi phase ended before the Gi/S boundary (see figure 3.80). If cessation of D N A repair at the Gi/S border was the dominant determinant of radiosensitivity in Gi phase, then one would expect that the point of maximum radiosensitivity would occur at the Gi/S border, when the cells had the minimum possible time to repair before the start of D N A replication. Taken together, these two findings suggest that radiosensitivity increases in mid-to late-Gi phase in U l cells were not primarily attributable to cessation of repair at the Gi/S boundary. One alternative hypothesis which is consistent with these observations 176 Chapter 4: Discussion would be that D N A repair was relatively depressed in later Gi phase before the start of D N A synthesis, as opposed to a cutoff of repair at the Gi/S border. However, there is currently no obvious mechanism which might account for such an effect. The confirmation of a radiosensitive region which encompasses the Gi/S phase border is relevant to issues such as the mechanism of 5-fluorouracil (5-FU) radio sensitization (Lawrence et al. 1996, McGinn et al. 1994). Findings reported here suggest that the cell cycle redistribution into early S phase which occurs after 5-FU administration (McGinn et al. 1993) would make cells treated with 5-FU radiosensitive, because early S phase was found to be a relatively radiosensitive interval in the normal cell cycle. This could contribute to the observed radiosensitizing effects of 5-FU. Compared to previous studies, the main contribution of the present work in respect to cell-age variation in radiosensitivity has been to make measurements of radiation survival in synchronized human tumor cells using a minimally disruptive synchronization technique with good resolution in Gi phase (mitotic selection), and to make enough measurements over the course of the cell cycle to provide a clear picture of cell-age related changes in radiosensitivity. In addition, the present work has quantified the effects of desynchronization on the radiation sensitivity of synchronized cells by the use of a mathematical model, an approach which was not taken in any of other studies discussed in this Section. In summary, the results shown here suggest that the variations in radiosensitivity in three different human cell lines over the course of the cell cycle are fundamentally very similar with respect to the transitions between Gi , S, G 2 , and M phases. Evidence was 177 Chapter 4: Discussion shown that radiosensitivity increases in mid- to late-Gi phase, before the initiation of D N A synthesis. 4.2 The survival responses of asynchronous and synchronized A549 and HT-29, but not Ul cells, are consistent with the presence of age-specific subpopulations with different LQ radiosensitivities. Alternative explanations for Ul substructure are considered. In this Section, findings in A549 and HT-29 will initially be discussed, followed by a separate discussion of findings with U l cells. Summary of flndings(A549 and HT-29) Deviations from the single population LQ model were examined in synchronous and asynchronous populations of both cell lines. The values of the LQ parameters a and P fitted using survival at higher doses (4-12 Gy, corresponding roughly to surviving fractions of 0.4-0.001) were different from those same parameters fitted from survival at lower doses (0-4 Gy, corresponding roughly to surviving fractions of 1-0.3). In the low dose range in A549 and HT-29, a was always less, and P always greater, than in the higher dose range, with one exception. The only population to show a homogeneous LQ response was synchronized HT-29 cells at 3 h after mitotic selection; for these cells ai0 was not significantly different from ahi, and Pi0 was not significantly different from phi (to two significant figures) at the level of 95% confidence limits. The observed deviations 178 Chapter 4: Discussion from the LQ model were not dependent on the particular dose range used to evaluate the high and low dose fits to survival curves; this was established by fitting survival curves with an alternative linear-quadratic-cubic (LQC) model. The LQC model predicted negative values of the cubic parameter for 9/10 of the A549 and HT-29 populations, with the single exception, as above, of HT-29 cells at 3 h after mitotic selection (note that 3/9 negative y values were not significantly less than zero at the upper 95% confidence limit, but these cases were the two mitotic populations, which had exceptionally high heterogeneity, and 15 h S/G2 HT-29 cells, which also had relatively little substructure as determined by the LQ Phi/Pio ratio). The LQC model tended to predict values of a and P which were closer to the low dose range fits than the high dose range fits, because the cubic term in the LQC model provided the degree of freedom required to fit the deviations from the low-dose range fit which occurred at higher doses. Except for mitotic populations, all synchronized populations in A549 and HT-29 had less substructure than asynchronous populations, when substructure was quantified by the ratio Phi/Pio- In HT-29 populations, which generally displayed more substructure than A549, the relative degree of substructure in the measured survival curves of different synchronized and asynchronous populations was weakly correlated with the variance in radiation survival predicted by the model for the same populations (figure 3.81). When all asynchronous and synchronized populations were considered, the model predicted ranges of Phi/Pio ratios from 0.16-1.0 (A549) and 0.03-0.98 (HT-29). These overlapped with measured ranges of phi/pio from 0.18-0.71 (A549) and 0.25-1.07 (HT-29) (figures 3.65, 3.72). The mathematical model of cell survival tended to underpredict the degree of substructure in asynchronous and synchronized A549 and HT-29 (especially A549), when 179 Chapter 4: Discussion the model ratios of Phi/Pio were compared to measured ratios in all cell populations. Underprediction of substructure in the model was expected because the radiation responses of cells in the model were based on fits to measured survival curves which would be affected by heterogeneity. Thus, at radioresistant points in the cell cycle, the model would underestimate radioresistance, and at radiosensitive points in the cell cycle, the model would underestimate radiosensitivity. The net result would be to underestimate the magnitude of fluctuations in survival during the cell cycle, and hence underpredict substructure in the survival curve. However, even with this underprediction, the degree of substructure predicted by the model was large enough to account for the observed values of Phi/Pio, as described above. It was felt that further attempts to improve the agreement of model with experiment by altering the model radiation responses (which would be difficult to rigorously justify from measured survival responses) would not contribute significantly to the understanding of the heterogeneous radiation responses of A549 and HT-29 cells. Comparison to results of other investigators (A549 and HT-29) As noted in Section 1.3.3, in the vast majority of investigations involving radiation survival measurement with use of the LQ model, deviations from the LQ model have either not been investigated, or, in a few cases, observed and concluded to be negligible compared to experimental uncertainty. However, substructure has been observed in asynchronous rodent cells and in asynchronous human tumor cells (Skarsgard et al. 1991, Skarsgard et al. 1996). In previous investigations of the cell lines used in this thesis, substructure was observed in asynchronous populations of A549, HT-29, and U l 180 Chapter 4: Discussion (Skarsgard et al. 1996); the same types of substructure were confirmed in asynchronous cells in this thesis. The only other investigations of substructure in synchronous cells were done in V79 rodent cells. These studies concluded that substructure in rodent cells was likely due to subpopulations of different radiosensitivity (Skarsgard et al. 1993, Skwarchuk et al. 1993). Implications of the present findings (A549 and HT-29) The best described and established deviation from the LQ model in asynchronous human tumor cells is characterized by concave downward -ln(S)/D plots, or equivalently, ratios Phi/Pio < 1. The results shown here demonstrate that these deviations can be fully accounted for in a model which assumes that age-specific cell populations obey the LQ model. In other words, the observed deviations from the LQ model in asynchronous populations could be attributed to the variation in radiosensitivity over the course of the cell cycle in the human tumor cells studied here. The best evidence for this was that the range of Phi/Pio predicted by a realistic model of heterogeneous LQ populations was as large as the measured range of phi/Pio in A549 and HT-29 (figures 3.65 and 3.72). Additional (but weaker) evidence was that in synchronized and asynchronous populations of HT-29, which had more heterogeneity in radiation response than A549, there was a (non-significant) correlation between the variance in model survival (an estimate of heterogeneity) and the measured Phi/Pio ratio (a measure of substructure) (figure 3.81). If the substructure was due to cell-age heterogeneity, such a correlation should exist. There are several possible explanations for the observed correlation not being statistically 181 Chapter 4: Discussion significant. One plausible explanation would be that although the model predicted single-dose survival reasonably well, the model-predicted survival variance might have differed from the true survival variance in any of the synchronized or asynchronous populations of HT-29 cells. For example, the heterogeneity (and survival variance) of synchronized HT-29 cells would be expected to increase rather rapidly over an interval between 8 h and 15 h after mitotic selection (HT-29 cells cross the B-C radiosensitivity border during this time period; see figure 3.73). Therefore, the true heterogeneity of experimental 8 h G i / S phase HT-29 cells may have been greater than the model-predicted heterogeneity at 8 h. Shifting the 8 h time point to higher variance could improve the correlation in figure 3.81. The experiments described here have extended the observation of substructure to mitotically selected synchronized populations of A549 and HT-29 human tumor cells. Significant deviations from the LQ model were measured in these synchronized populations, and the mathematical models of the synchronized cells established that even in these populations, which are commonly viewed as "homogeneous", heterogeneity in radiosensitivity was present, and had a magnitude which could account for the observed deviations from the LQ model. The only synchronized population which did not show significant survival substructure was HT-29 cells, 3 h after mitotic selection. In these cells it appeared that the combination of relatively high synchrony and high radioresistance combined to eliminate measurable substructure from the survival curve. These findings are consistent with the hypothesis that the radiation survival of age-specific homogeneous A549 and HT-29 cells obeys the LQ model; they show that it is not necessary to invoke non-LQ models of survival to quantitatively explain the clearest deviation from the LQ model in asynchronous or synchronized A549 and HT-29 human 182 Chapter 4: Discussion tumor cells. Because most human cell lines investigated so far have shown the same type of substructure (with the important exception of U l ; discussed below), these conclusions about radiation survival modelling will be relevant to many other types of human tumor cells. However, it is important to note that there is at least one other significant deviation from the LQ model which has been measured in some human tumor cell lines, although not as consistently as the substructure described here. This is the hypersensitive response at doses below 1 Gy (Marples et al. 1993). No combination of plausible LQ subpopulations is likely to explain this response (Wouters et al. 1996). However, the hypersensitive response was not convincingly observed in the experiments described in this thesis. It is important to note that, based on the data shown in this thesis, we cannot rule out the hypothesis that age-specific cells do not follow the LQ model, i.e. a better fit to experimental data could be obtained by an alternative model (e.g. the RMR, LPL, or Neary model, to name a few). In fact, the results described here highlight the extreme difficulty of unambiguously rejecting the LQ model based on even careful survival measurements which are inevitably affected by heterogeneity in radiation response. By quantitatively accounting for the observed substructure in A549 and HT-29, we have provided support for the LQ model by removing one of the potential reasons for requiring alternative models of radiation survival for these cells. 183 Chapter 4: Discussion Summary of findings (Ul) In U l cells, the LQ parameters a and P determined from survival curves on high and low dose ranges were different. The LQ a parameters cthi and cti0 were significantly different in only 2/5 populations (asynchronous, and 0 h mitotic cells); but the values of a i 0 were influenced by a limit of oti0 > 0 imposed by the fitting algorithm. In a better indicator of substructure, the values of P M were greater than P i 0 in 4/5 populations (all except 0 h mitotic cells). These indicators of substructure were confirmed by fitting the same survival curves with the LQC model; in this model, 4/4 populations in which 95% confidence limits could be ascertained had y values significantly greater than zero (asynchronous, 3 h Gi cells, 17 h Gi/S cells, and 25 h S/G2 cells). Mitotic U l cells were the only population that had a negative y value, presumably due to the presence of very strong population heterogeneity in these mitotically selected cells. The mathematical model of U l cells, which incorporated cell-age related variation in cell survival, predicted (as expected) that asynchronous and synchronized U l cells should display the same type of substructure as A549 and HT-29 cells, with Phi/Pio ratios less than 1. In direct contradiction to this, all populations of asynchronous and synchronized U l cells had Phi/Pio ratios significantly greater than 1 at the level of 95% confidence limits. It was interesting that in U l cell populations, which all had Phi/Pio > 1 (figure 3.79), asynchronous cells had the highest ratio Phi/Pio, although there was no clear relationship between Ph;/pi 0 and the population heterogeneity predicted by the model of U l cells, if mitotic cells were excluded from the analysis (figure 3.82). This was different from the case in HT-29 cells; for these cells, even if mitotic populations were excluded 184 Chapter 4: Discussion from analysis, there was a (non-statistically significant) trend toward smaller values of Phi/Pio for populations which had greater model-predicted heterogeneity, and asynchronous cells had the lowest Pw/Pio ratios (figure 3.81). One way to interpret this is that in HT-29, the deviations from a homogeneous LQ response were due to heterogeneity among subpopulations which followed the LQ model, so that the asynchronous cells (which according to the model were the most heterogeneous populations, except for mitotic cells) had the lowest Pw/Pio ratio (furthest from P h ; / P i 0 = l ) among the non-mitotic populations. In contrast, for non-mitotic U l cells, the Phi/Pio ratios were more consistent with age-specific populations exhibiting a non-LQ response (with Phi/Pio ratios > 1), and increasing heterogeneity tending to have little consistent effect on the Phi/Pio ratio. Comparison to results of other investigators Substructure in U l cells of the same type as found in this thesis has been reported in asynchronous cells previously (Skarsgard et al. 1996). The experiments described here are the first report of the presence of substructure with the same properties in synchronized U l cells at 3 time points (3 h Gi , 17 h Gi/S, and 25 h S/G2) spread throughout interphase (figures 3.29-3.31). Implications of the present findings Three potential explanations for the "reverse" substructure which was measured in U l cells (both asynchronous and synchronous) can be discarded with some confidence: 1. Experimental error (random or systematic) 185 Chapter 4: Discussion Random experimental errors should have been averaged out by the extensive data averaging done for the survival responses. In U l , 4-6 individual survival curves were averaged for each cell population; thus each survival determination is the average of 24-36 single determinations per dose point (and typically 3 times this number for zero dose control samples). If the unique U l response was due to systematic error, it should have been observed in the A549 and HT-29 responses, which were measured using the same technique. 2. LQ population heterogeneity As was suggested by the mathematical model, if age-specific U l cells had LQ survival responses, then population heterogeneity cannot explain the observation of the "reverse" substructure, with Phi > Pio, and a "concave upward" -ln(S)/D plot. It is important to note that this can be rigorously proven if the distribution of LQ parameters is normal (Schultheiss et al. 1987). While it is easy to imagine non-normal distributions of a and P in a heterogeneous population, the form of these distributions in U l is unlikely to be much different than in A549 and HT-29, based on the single dose survival responses, and A549 and HT-29 did not exhibit the U l substructure. 3. Feeder effect In a clonogenic assay, cells which are seeded at low density sometimes have a lower probability of survival than identical cells which are seeded at much higher density. This "feeder effect" is presumably due to the secretion of growth promoting factors by cells, whether clonogenic or not. It could be hypothesized that differences in the total number of cells per sample (i.e. tissue culture dish) skewed the measurements of U l survival at higher dose compared to survival at lower dose, because in the experiments 186 Chapter 4: Discussion described here, high dose dishes contained a greater number of experimental cells than low dose dishes (at higher doses more cells were required to form an adequate number of colonies). Four potential cell-density related effects on survival are summarized in table 4.1. However, it is important to note that such cell-density effects seem unlikely, because a large constant number of heavily irradiated non-clonogenic feeder cells (7x104 per tissue culture dish) was added to each dish at the time of plating. Therefore, the total cell number per dish never varied by more than a factor of ~1.6 for all U l samples in the cell sorter assay; for all doses below about 7-9 Gy, the cell number per dish was constant to within approximately 10%. Hence, if a "feeder effect" skewed survival measurements in interphase U l cells, it would have had to have been caused by differences in cell density of no more than -10-60% (60% only for the highest 2-3 dose points). Table 4.1: Possible "feeder effects" on U l survival Case Possible change in PE Potential mechanism Effect on survival at higher dose/density vs. lower dose/density Could this explain the shape of the U1 response (assuming LQ survival)? 1 increased PE0, increased PEX "traditional" feeder effect skewed upwards no II decreased PEo.decreased PE„ nutrient deprivation in denser cultures? skewed downwards theoretically III increased PE„only improved repair in denser cultures? skewed upwards no IV decreased PE„only inhibition of repair in denser cultures? skewed downwards theoretically The table considers 4 possible cell density-related effects on cell survival. The effects are classified according to their impact on PE0 and PE„, where S(D) = PEo/PEx. The hypothesis being tested in the table is that U1 cells followed the LQ model, but "feeder effects" skewed the survival measurements away from a LQ response. Cases I and III appear to be incompatible with the measured U1 responses; cases II and IV could theoretically explain the observed responses, but it is important to note that cell density in survival samples never varied by more than 10-60% in the U1 survival assays (see text). It is interesting to consider the effects of heterogeneity on the measured radiation response of U l cells. These cells, as described above, displayed a trend to greater-than-187 Chapter 4: Discussion L Q sensitivity at higher doses, evidenced by "concave upward" -\n(S)ID plots. Regardless of the "true" survival model which might apply to perfectly homogeneous U l cells, heterogeneity will always tend to increase the radioresistance of cells at higher doses (lower survival levels) because the most resistant cells die last. Hence, the "true" survival model for an idealized homogeneous U l cell population would have even larger deviations from the L Q model (trends toward greater-than-LQ-sensitivity at higher doses) than were observed in the synchronized and asynchronous populations of U l cells, which were inevitably somewhat heterogeneous. This is the opposite of the case in A549 and HT-29 cells, where "correction" of measured survival responses for heterogeneity suggests a "true" homogeneous response which is closer to an ideal LQ response, not farther away. Given these facts, it seems necessary to reject the LQ model as an adequate description of the survival response of even homogeneous U l cells over the first 2-3 decades of cell kill. There are numerous alternative models to consider. At a basic mathematical level, any model in which survival could be expressed as an infinite series exponential expression in the dose D (i.e. e'"0'^0 ~yD ") with terms in the exponent of order greater than D2 could adequately account for the "reverse" substructure seen in U l , even including the effects of heterogeneity in cell populations. Such models include the Neary, LPL, and R M R models, for example (see equations 1.7, 1.8, and 1.9). To test these models, they were fit to the survival curve of 3 h G\ phase U l cells. This fitting revealed problems with matching any of these models to the measured U l survival . response. When the R M R and L P L models were fitted to the 3 h G i U l cell survival data 188 Chapter 4: Discussion -0.2 \-0 2 4 6 8 10 12 Dose (Gy) Figure 4.1: Fits to U l 3 h Gi phase survival response. Shown are fits to the survival response from 6 models: LPL, RMR, LQC, RS3 (repair saturation), and the Neary model. The L Q C and RS3 models give the best fit to the data; see table 4.2 189 Chapter 4: Discussion with properly constrained parameter values (i.e. for these models, parameter values > 0), the fits were not better than a LQ model fit (table 4.2). In the case of the Neary model, similar difficulties occurred. Although very good fits to U l survival responses were possible with the Neary model, they required negative values of X\ and X2, but the theory used to derive the model implies that Xi and X2 are positive values. Two other models were more successful at reproducing the 3 h Gi phase U l survival response. These were the LQC model (equations 1.5 and 1.6) and the repair saturation model (equation 1.10). Biophysical interpretation of the L Q C model is difficult because this model can be interpreted as an extension of a generalized interacting-damage model like the LQ model, or as a 3-term infinite series approximation to many other models. In the repair saturation model, it is hypothesized that lethal lesions are induced proportionately to radiation dose (i.e. by one-track action) and cellular repair is a saturable process which is dependent on the number of lesions, and the number of available repair enzyme molecules. For better comparison with other 2 and 3 parameter models, this model was reduced to 3 parameters by fixingp=\ (this implies all lesions acted on by the saturable repair system are lethal if unrestituted, see Goodhead (1985)). Both the L Q C and the 3-parameter repair saturation model fitted the Gi-phase U l survival response better than a single LQ, RMR, LPL, or Neary model, without requiring parameter values that were obviously incompatible with theory (figure 4.1; table 4.2). These two models could also adequately fit the responses of A549 and HT-29 cells. 190 Chapter 4: Discussion Table 4.2 Fits to 111 3 h G i phase survival response Model Best Fit Reduced % Free parameters LQC 4 3 RS3 9 3 LQ 79 2 Neary 91 3 RMR 99 2 LPL 99 3 Notes: 1. Points at D=0 and 0.05 Gy were excluded because the fitting program could not correctly evaluate the repair saturation equation at these low doses. 2. There were n=19 data points in the fitted data set. 3. Better fits could be obtained with the Neary model, but only by allowing Xy > X 2 , which is incompatible with Neary's theory. 4. There were 3 free parameters in the LPL fit because 2 of the 5 LPL parameters had to be fixed to allow convergence of the fit. 5. RS3 model is the repair saturation model with p=1; implies all lesions acted on by the saturable repair system are lethal if unrepaired. 6. The reduced % value is a measure of goodness-of-fit; smaller % values indicate better fits. Final summary of findings (all cell lines) To summarize the present findings in all three cell lines: 1. Analysis of heterogeneity in A549 and HT-29 cells was consistent with homogeneous populations of these cells following the LQ model. In other words, deviations from the LQ model in asynchronous and synchronized A549 and HT-29 could be reasonably attributed to cell-age related heterogeneity alone. This does not imply that these cells must obey the LQ model (see point (3) below). 2. Analysis of heterogeneity in U l cells strongly suggested that U l cells do not obey the LQ model. This was most clearly indicated by the fact that heterogeneous LQ cell populations cannot display "concave upward" -\n(S)/D responses. Reinforcing this conclusion, it was observed that, regardless of the survival model which U l cells obey, heterogeneity in these cells would have tended to "bend" the -ln(S)/D responses back toward a straight-line LQ response, effectively reducing the measured Phi/Pio ratio. Hence, the deviations from the LQ model in the U l cell populations in this thesis, which were 191 Chapter 4: Discussion quantified by the Pw/Pio ratio, are probably underestimates of the true deviations of U l cells from the LQ model. 3. The repair saturation model and the LQC model (but not the LQ, RMR, LPL, or Neary models) could fit the response of mitotically selected GI phase U l cells reasonably well over 2-3 decades of cell kill. These two models could also adequately fit the responses of A549 and HT-29 cells. Thus, one could hypothesize, for example, that all three cell lines obeyed the repair saturation model, but that differences in susceptibility to damage of various types caused saturation to occur with markedly different kinetics in U l , compared to A549 or HT-29 cells. The implication that U l cells employ a repair system which is saturable in the first 2-3 decades of cell kill might be tested by measuring the velocity of D N A repair as a function of dose in these cells. 4.3 A549, but not HT-29 or U l , undergoes a prolonged Gi arrest after irradiation which does not necessarily play an important independent role in determining clonogenic survival Summary of findings Synchronized A549 cells irradiated with 5 Gy at times up to and including 6 h after mitotic selection, while most cells were in Gi phase, exhibited a prolonged Gi arrest. Approximately 80% of the Gi cells were inhibited from entering S phase for at least 36 h after irradiation. Synchronized A549 cells irradiated at 11 and 16 h, when most cells were 192 Chapter 4: Discussion near the Gi/S border and in S phase, respectively, displayed no inhibition of progression through S phase, but did exhibit a block in G 2 phase. Gi arrest was not observed in synchronized HT-29 or U l cells. These cells did, however, block in G 2 phase; the G 2 block appeared to be strongest for cells which were closest to G 2 / M phase when irradiated. In numerous previous studies, the presence of wild-type p53 protein, and its upregulation following irradiation, has been correlated with the presence of a Gi arrest. The results reported here are in agreement with these previous findings. The presence of Gi arrest in A549 and its absence in HT-29 and U l was correlated with the regulation of p53 protein over the first 10 h after irradiation. Specifically, A549 (previously reported to express wild-type p53 (Noble et al. 1992)) upregulated this protein approximately 6-fold within 2 hours of irradiation with 5 Gy, while HT-29 (reported to express a mutant p53 (Rodrigues et al. 1990)) and U l cells did not significantly upregulate the protein (figure 3.47) Irradiated A549 cells in monolayers largely stayed attached to the growth surface for around 160 h after irradiation. This seemed consistent with the prolonged Gi arrest in these cells. The mode of cell death in both HT-29 and U l cells was dramatically different; it was characterized by large-scale detachment of irradiated cells from monolayers around 96 h after irradiation, and the development of D N A fragmentation in irradiated cells around the same time (figure 3.48). Comparison to results of other investigators Others have quantified G\ arrest by examination of fractions of Gi and S cells at various times after irradiation of asynchronous cells (e.g. Kastan et al. 1992). Populations 193 Chapter 4: Discussion synchronized by elutriation or starvation have been examined for the time course of exit from Go/Gi after irradiation (e.g. Little 1968, Pellegata et al. 1996). The closest comparison to the work reported here is provided by two recent reports of prolonged Gi arrest in mitotically selected synchronized human glioblastoma cells irradiated in early Gi phase (Haas-Kogan et al. 1996, Yount et al. 1996). Another recent investigation suggested that Gi-arrest deficient variants of the HCT116 human tumor cell line produced by deletion of the p21 CDK-inhibitor gene had the approximately the same radiation sensitivity in vitro as Gi-arrest proficient normal cells (Wouters et al. 1997). Implications of the present findings The present findings are consistent with the hypothesis that the Gi arrest in A549 cells is a mechanism of cell death, not a delay which provides time for D N A repair. In the experiments described here, Gi-arrested A549 cells stayed in Gi phase for at least 36 h after irradiation with 5 Gy. This implies that the prolonged Gi arrest inhibited cell proliferation to such an extent that clonogenic survival of arrested cells must have been reduced by the arrest, not enhanced. For example, based on typical doubling times, the observed 36 h pause in proliferation in A549 would have led to at least a ~2-fold reduction in the size of colonies formed by temporarily arrested A549 cells compared to equivalent unarrested cells when colonies were assessed for cell survival 10-14 days after irradiation. Repair may take place during the time that cells are arrested, but even if cells recovered from the arrest sometime after 36 h, they would have had to subsequently proliferate faster than unblocked cells in order to exhibit the enhanced survival which they showed relative to later Gi/S phase cells in clonogenic assays (figure 3.8). This possibility 194 Chapter 4: Discussion cannot be ruled out, but circumstantial evidence, in the form of heavy "backgrounds" of non-proliferated A549 cells in tissue culture dishes stained for colony assessment 10-14 days after irradiation (these "backgrounds" did not appear in HT-29 or U l cell samples; data not shown) suggested that many of the arrested cells never escaped from the Gi arrest even over this longer time period. Indirectly supporting this idea, it was found that in A549 cells at 3-6 h after mitotic selection, the fraction of cells which arrested in Gi phase after 5 Gy (about 80%; figures 3.35, 3.36) was approximately the same as the fraction of cells which were killed (about 85%; figure 3.8). The evidence just cited suggests that Gi-arrested A549 cells were permanently inactivated. However, definitive proof that Gi-arrested A549 cells were clonogenically dead would require further observations of arrested cells beyond 36 h after irradiation, which was the limit of experiments reported here. If, for the purposes of discussion, we accept the reasonable theory that the arrested cells were killed, then the hypothesis that the arrest provided time for D N A repair before the start of S phase in A549 cells must be discarded. One should then address the question of the relative contribution of the Gi arrest to clonogenic cell killing. From the point of view of potential genetic or chemical manipulation of the Gi checkpoint to enhance radiation killing of cancer cells, the most important and basic question is this: is the Gi arrest a redundant pathway of cell death? In other words, would the cells which are inactivated via the Gi arrest die by other mechanisms if the arrest was not active? The data presented in this thesis cannot answer this question unequivocally, but they do provide some clues. 195 Chapter 4: Discussion The finding that Gi phase A549 cells between 3 and 6 h after mitotic selection which did arrest in Gi phase were actually more radioresistant than cells at any other phase of the cell cycle (figure 3.8) does not, of itself, allow us to make any firm conclusions except that the Gi arrest, in combination with other cell killing mechanisms which may operate on Gi phase cells, does not provide a large enough radiosensitizing effect to make Gi phase radiosensitive compared to the remainder of the cell cycle. The observed age response of A549 could be consistent with the Gi arrest acting as a redundant cell killing mechanism. For example, there is no evidence that some or all of the arrested cells would not have died by other mechanisms if the arrest was selectively abrogated. On the other hand, one could reasonably posit that the arrest was a non-redundant cell killing pathway which inactivated a specific subpopulation of Gi cells that would not have died otherwise. At first glance, this view would appear to contradict the evidence that Gi phase was more radioresistant than the rest of the cell cycle in A549, but it is crucial to remember that fluctuations in the activity of other cell killing pathways like delayed apoptosis or abortive cell division could account for the observed age response of A549 even if the Gi arrest was an independent, non-redundant cell inactivation pathway. In further examining the cell killing role of the Gi arrest in A549, it may be useful to consider the single dose age responses of HT-29 and U l cells, which did not arrest in Gi phase after irradiation. In these two cell lines, and in some other cell types which do not arrest in Gi phase, such as HeLa (Terasima et al. 1963), the radiosensitivity variations over the course of Gi phase in mitotically selected populations were qualitatively similar to those in A549. Al l these cell types are relatively radioresistant in early-mid Gi phase, and become radiosensitive later in Gi phase. This shows that there are mechanisms of 196 Chapter 4: Discussion radioresistance in early-mid G i phase which are completely independent of the G i arrest in several human tumor cell lines, and naturally suggests that these mechanisms operate in A549, just as they must in HT-29 and U l cells. If that is the case, then it suggests (but does not prove) that the G i arrest in A549 may not exert a large non-redundant radiosensitizing effect in early-mid G i phase because it did not fully reverse the arrest-independent mechanisms of radioresistance which probably operated in this part of the A549 cell cycle. Essentially, this argument that the G i arrest is not a significant non-redundant mode of cell killing is simply based on the qualitative similarity between the age responses of arresting (A549) and non-arresting (HT-29 and U l ) cells during G i phase. How can this argument be reconciled with the numerous previous studies that have confirmed that wild-type p53 producing, Gi-arrest competent cells are more radiosensitive than otherwise identical cells with mutant p53 which do not arrest (e.g. Chang et al. 1997, Delia et al. 1997, Griffiths et al. 1997, Gupta et al. 1996, Haas-Kogan et al. 1996, Lee et al. 1993, Lowe et al. 1993, Mcllwrath et al. 1994, Siles et al. 1996, Tsang et al. 1995, Yount et al. 1996)? A simple explanation could be that normal p53 function, which was required for the G i arrest in the cell lines observed here, confers radiosensitivity to wild-type p53 expressing cell lines through mechanisms which are unrelated to the G i arrest. For example, p53 plays roles in modulating D N A repair (see refs. in Bae et al. 1995) and apoptosis (e.g Lowe et al. 1993); these roles do not have a well-described relationship to the cell cycle. Consistent with this notion, A549 (with wild-type p53) was more radiosensitive at all points in interphase than HT-29 (mutant p53) or U l (which had a mutant-type p53 response; figure 3.47). This can be seen by computing a commonly used radiosensitivity measure, the surviving fraction at 2 Gy (SF2). The SF2 is simply 197 Chapter 4: Discussion calculated from the oci0 and p\0 values in tables 3.2-3.3, 3.7-3.8, and 3.12-3.13, using the L Q equation (1.5). The computed ranges of SF2 in asynchronous and synchronized interphase cells (excluding mitotic cells) are 0.63-0.68 (A549), 0.75-0.83 (HT-29), and 0.83-0.92 (Ul) . Thus, one could envisage, for example, that the loss of normal p53 function in HT-29 and U l , compared to A549, contributed to making these cell lines relatively radioresistant through cell cycle-independent mechanisms, in addition to abrogating Gi arrest function in these cells. To conclude, we can speculate on a role for Gi arrest cell inactivation which would be relevant to cell lines which undergo a suitably prolonged Gi arrest, and is consistent with the data in this thesis, as well as observations by others. In this purely hypothetical scheme, radiation would produce similar yields of D N A damage in cells at all stages of the cell cycle. Cellular responses to this damage would include different, cell-age dependent mechanisms of death, including (for example) delayed apoptosis, abortive cell division, or permanent Gi arrest. The Gi arrest would be a prolonged or permanent cessation of cell division induced only in Gi phase by some of the same upstream molecular signals that lead to other death pathways. The dose response of Gi arrest induction would be similar, but not necessarily identical to, the dose response of other death pathways for Gi cells, hence the arrest would be at least partially a redundant mode of cell killing. Depending on the relationships between these dose responses, some or all of the cells which are killed by Gi arrest would still die by other pathways if the Gi checkpoint was selectively turned off. In this scheme, selective activation of the Gi arrest would make Gi cells more radiosensitive, not more radioresistant, but the increase in radiosensitivity due to the activation of the Gi checkpoint would vary between zero and an 198 Chapter 4: Discussion upper limit which would depend on the dose response of Gi arrest relative to the dose responses of other modes of cell death. 4.4 A549, HT-29, and Ul cells have lower linear quadratic a/p ratios at clinically relevant doses than have commonly been reported for other tumor cells and normal tissues. Summary of findings LQ fits to cell survival in asynchronous cells over a clinically relevant, low dose range (0-4 Gy) determined that the ratio of LQ fit parameters a/p was 1.78 (95% confidence limits 1.608-1.971) in A549, 0.802 (0.714-0.892) in HT-29, and 1.230 (0.968-1.532) in U l . When all populations, synchronized and asynchronous, were included, the a/p values ranged from 1.78 to 12.9 (in A549), 0.113 to 2.405 (in HT-29), and 0 to 1.558 (in U l ) . Comparison with investigations by others These values of the a/p ratio are much lower than most previous measurements of the a/p ratio in tumor, or normal, tissues. For example, a survey of a/p values measured in 48 studies of various tumor tissues in vivo found values ranging from 0.9 to 47.9, with a median value of 11.2 (excluding 1 value which was not exactly specified) (Williams et al. 1985). A more meaningful comparison might be with the dataset of Fertil and Malaise, 199 Chapter 4: Discussion who summarized a and P in 59 cell lines in vitro (including, it should be noted, 12 measurements in HeLa S3 cells) (Fertil et al. 1981). In these 59 measurements, excluding 3 cell lines with P values less than or equal to zero, the range of a/p was 0-231, with a median of 7.9. Compared to the values in these large datasets, the a/p ratios in the cell lines investigated here are low. In terms of percentile ranking, the a/p ratios measured for asynchronous cells are in the bottom 18% of values in the Fertil and Malaise dataset, and the bottom 5% of the values in the Williams dataset. A probable explanation for the fact that the current a/p ratios are lower than most of the values summarized by, for example, Fertil and Malaise, is that the current values were determined from datasets with relatively more survival determinations at low doses. For example, a check of 3 of the sources of a/p ratios (Fertil et al. 1980, Weininger et al. 1978, Wells et al. 1977) which together provided 18 of the 59 values in the Fertil and Malaise report (Fertil et al. 1981) showed that in these sources, there were 1-3 data points at or below 2 Gy, compared to 11 survival determinations at or below 2 Gy in most of the survival responses described in this thesis. Since we expect that values of a/p will be biased upward by inevitable heterogeneity in cell populations, and that as a result, survival determinations at higher doses (lower survivals) will tend to overpredict a/p for the more sensitive subpopulations which are important to survival in the clinically relevant 0-2 Gy dose range, this seems to be a case where "more" is really "better": more survival determinations at low doses should give a more accurate estimate of a/p. 200 Chapter 4: Discussion Implications of the present findings It is commonly asserted that the a term in the linear quadratic equation dominates cell killing at clinically relevant doses (see e.g.Biade et al. 1997, Haas-Kogan et al. 1996), and a/{3 ratios on the order of 10 Gy or greater for tumor cells are commonly cited (for example, Hall 1994, Horiot et al. 1992). The results presented in this thesis would suggest that such values significantly overestimate the a/p ratio of tumor cells in the clinically relevant dose range. The acceptance of lower a/p ratios could have an impact on the assessment of biophysical models of radiation survival. For example, theories of radiation survival have been postulated based on saturable repair mechanisms (Goodhead 1985). In these theories, the curvature of the cell survival curve (which is inversely related to the ratio a/p) is due to the progressive saturation of cellular repair pathways in the presence of increasing numbers of radiation-induced lesions. In the context of such theories, the relatively low a/p ratios reported in this thesis would imply that cellular repair mechanisms begin to saturate at very low doses, on the order of a/p « 1 Gy. To test this implication, direct measurement of residual D N A damage after doses < 1 Gy would be required. Recent advances in the detection of low levels of D N A damage suggest that such a test may soon be feasible (Le et al. 1998). As a descriptor of the "shape" of cell survival curves, the a/p ratio of tumor cell populations has been hypothesized to be an important indicator of the clinical response of these cells relative to normal tissue. For example, based on many previous measurements of a/p ratio in normal and tumor tissues, it has been postulated that normal tissues which exhibit delayed toxic responses after radiotherapy have smaller a/p ratios than early-201 Chapter 4: Discussion responding tumor and normal tissues (Thames et al. 1982). Hence, the use of smaller radiation fractions (less than 2 Gy) might have greater sparing effects on late-responding normal tissues than on tumor cells. This hypothesis has been used as one of the justifications for alternate fractionation schemes which are currently being evaluated in clinical trials like hyperfractionation (Horiot et al. 1992) and CHART (continuous hyperfractionated accelerated radiotherapy) (Denekamp 1986). Indeed, results to date with CHART have suggested that this regimen does reduce late morbidity compared to conventional fractionation at roughly equivalent levels of tumor control (Dische et al. 1997). Are acute radiation survival responses predictive of the response of tumor or normal cells to fractionated radiotherapy? If not, then the observation in this thesis of low a/p ratios in human tumor cells has little clinical importance. There is evidence that acute survival of tumor cells in vitro does correlate with clinical response (e.g. Fertil et al. 1985). But from the viewpoint of measuring a/p ratios (i.e. the shape of the dose-response relationship) in tumor or normal tissue, it seems reasonable that fractionated radiation experiments in vivo may be more relevant to fractionated radiotherapy than acute in vitro measurements like the ones described in this thesis. Nevertheless, it is striking that the vast majority of the in vivo studies which measured the uniformly "high" a/p ratios in the review of Williams (1985), used no dose fractions less than 2 Gy, the typical clinical fraction. At the very least, the in vitro experiments described here strongly suggest that cellular heterogeneity will make the a/p ratios derived at higher doses overestimates of the effective a/p ratio at lower clinical doses. This impact of heterogeneity has been explored on theoretical grounds by others (Schultheiss et al. 1987). 202 Chapter 4: Discussion Given these facts, one might argue that the relatively low a/p ratios which were determined in this thesis for human tumor cells could throw some doubt on one of the rationales for hyperfractionation in radiotherapy. But, it is important-to remember that most of the a/p ratios for normal tissues which have been previously measured are, like currently accepted a/p values for tumor cells, probably biased upwards. So, it is quite possible that if normal tissue a/p ratios were measured based on survival responses with many more determinations at clinically relevant doses, these ratios, like the tumor cell a/p ratios found in this thesis, would be found to be less than the currently accepted values. Indeed, this idea is supported by studies in pig skin which have suggested that the a/p ratio measured from low dose-per-fraction effects is less than that fitted from high dose-per fraction measurements (Hopewell et al. 1991). Hence, it is reasonable to postulate that more complete measurement of survival at low doses in both tumor and normal tissues might not change the current view that late-responding normal tissues have lower a/p ratios than tumor and early responding tissues. This would be consistent with the observed reduction of late morbidity in the CHART trials (Dische et al. 1997). 203 Chapter 5: Conclusions The major conclusions of this thesis can be described in terms of the three objectives which were described in Section 1.7. 1. The cell-age related variations in three different human tumor cell lines were found to have important similarities: extreme sensitivity in mitosis, radioresistance in early Gi phase and later S/G2 phase, and an interval of sensitivity in late Gi/early S phase. The magnitude of these changes varied between the three cell lines, but the nature of the variations appeared to be essentially the same in all. These findings agreed generally with several previous studies in HeLa cells, but they differed from the results of some recent studies in other human cell types. It is argued here that most of the discrepancies with these recent studies can be attributed to the different measurement techniques and synchronization techniques used by these studies. 2. The survival responses of asynchronous and synchronized populations of two human tumor cell lines (A549 and HT-29) were consistent with the presence of multiple subpopulations of cells at different stages of the cell cycle which individually had simple L Q survival responses. In a third cell line (Ul), the dose dependence of survival could not be explained by cell-age specific subpopulations which followed the LQ theory; hence, in this cell line, there was evidence that cell survival did not obey the LQ model. 3. A prolonged Gi arrest was observed in A549, but not HT-29 or U l cells. The arrest appeared to be p53-dependent. The arrest in A549 was measured by following the progression of synchronized cell populations, and was shown to be consistent with a cycle-specific "checkpoint" at a period in late Gi phase. Correlation of the Gi arrest, or 204 Chapter 5: Conclusions lack of arrest, with survival measurements in all three cell lines suggested that in A549, the arrest was a mode of cell death, not recovery. Quantification of any independent effects of the Gi arrest on radiation survival requires more experimentation; however, the fact that during Gi phase, the age response of A549 cells (which arrested) was qualitatively similar to that of HT-29 and Ul cells (which did not arrest), suggested that the arrest could be a partially redundant mode of cell killing. Hence, the Gi arrest may not play a large independent role in modulating radiosensitivity. 205 References Alper T. Cellular Radiobiology. Cambridge: Cambridge University Press; 1979. Bae I, Fan S, Bhatia K, Kohn KW, Fornace AJ, O'Connor PM. Relationships between G l arrest and stability of the p53 and p21cipl/wafl proteins following gamma irradiation of human lymphoma cells. Cancer Res 1995;55(June l):2387-2393. Band V , Dalai S, Delmolino L, Androphy EJ. Enhanced degradation of p53 protein in HPV-6 and BPV-1 E6-immortalized human mammary epithelial cells. EMBO J 1993;12(5):1847-1852. Barendsen GW. Dose fractionation, dose rate and iso-effect relationships for normal tissue responses. Int JRad One Biol Phys 1982;8(11): 1981-1997. Bender M A , Gooch PC. The kinetics of X-ray survival of mammalian cells in vitro. Int J Rad Biol 1962;5:133-145. Bernhard EJ, Maity A, Muschel RJ, McKenna WG. Increased expression of cyclin B l mRNA coincides with diminished G2-phase arrest in irradiated HeLa cells treated with staurosporine or caffeine. RadRes 1994;140:393-400. Biade S, Stobbe CC, Chapman JD. The intrinsic radiosensitivity of some human tumor cells throughout their cell cycles. RadRes 1997;147:416-421. Blakely EA, Chang PY, Lommel L. Cell-cycle dependent recovery from heavy-ion damage in Gl-phase cells. RadRes 1985;104:S145-S157. Blakely EA, Ngo FQH, Curtis SB, Tobias CA. Heavy-ion radiobiology: cellular studies. In: Lett JT, editor. Advances in Radiation Biology. Orlando: Academic Press; 1984. p. 295-389. Blakely EA, Roots RJ, Chang PY, Lommel L, Craise L M , Goodwin E H , Yee E, Dodgen DP, Blakely WF. Cell-cycle dependence of X-ray oxygen effect: role of endogenous glutathione. NCI Monographs 1988;6:217-223. Blank KR, Rudoltz MS, Kao GD, Muschel RJ, McKenna WG. The molecular regulation of apoptosis and implications for radiation oncology. Int J Rad Biol 1997;71(5):455-66. Boag JW. The statistical treatment of cell survival data. In: Alper T, editor. Cell survival after low doses of radiation: theoretical and clincal implications: John Wiley and Sons Ltd.; 1975. p. 40-53. 206 References Burki HJ, Okada S. A comparison of the killing of cultured mammalian cells induced by decay of incorporated tritiated molecules at -196 degrees C. Biophys J 1968;8(4):445-456. Bussink J, Tofilon PJ, Brock WA. Repair of chromosome and D N A breaks versus cell survival in Chinese hamster cells. Int J Rad Biol 1996;70(l):23-32. Catzavelos C, Bhattacharya N , Ung Y C , Wilson JA, Roncari L, Sandhu C, Shaw P, Yeger H, Morava-Protzner I, Kapusta L, Franssen E, Pritchard KI, Slingerland JM. Decreased levels of the cell-cycle inhibitor p27Kipl protein: prognostic implications in primary breast cancer. Nature Med 1997;3(2):227-30. Chadwick K H , Leenhouts HP. A molecular theory of cell survival. Phys Med Biol 1973;18(l):78-87. Chang E H , Jang YJ , Hao Z, Murphy G, Rait A, Fee WE, Jr., Sussman HH, Ryan P, Chiang Y , Pirollo KF. Restoration of the GI checkpoint and the apoptotic pathway mediated by wild-type p53 sensitizes squamous cell carcinoma of the head and neck to radiotherapy. Archives of Otolaryngology — Head & Neck Surgery 1997;123(5):507-12. Chapman JD. Biophysical models of mammalian cell inactivation by radiation. In: Meyn E, Withers HR, editors. Radiation biology in cancer research. New York: Raven Press; 1980. p. 21-32. Chapman JD, Todd P, Sturrock J. X-ray survival of cultured Chinese hamster cells resuming growth after plateau phase. Rad Res 1970;42:590-600. Chen E Y , Howley P M , Levinson AD, Seeburg PH. The primary structure and genetic organization of the bovine papillomavirus type 1 genome. Nature 1982;299:529-534. Chen F, Nastasi A, Shen Z, Brenneman M , Crissman H, Chen DJ. Cell cycle-dependent protein expression of mammalian homologs of yeast D N A double-strand break repair genes Rad51 and Rad52. Mut Res 1997;384(3):205-11. Crissman HA, Tobey RA. Specific staining of D N A with the fluorescent antibiotics, mithramycin, chromomycin, and olivomycin. In: Darzynkiewicz Z, Crissman HA, editors. Methods in cell biology. San Diego: Academic Press; 1990. p. 97-103. Curtis SB. Lethal and potentially lethal lesions induced by radiation - a unified repair model. Rad Res 1986;106:252-270. 207 References Delia D, Goi K, Mizutani S, Yamada T, Aiello A, Fontanella E, Lamorte G, Iwata S, Ishioka C, Krajewski S, Reed JC, Pierotti M A . Dissociation between cell cycle arrest and apoptosis can occur in Li-Fraumeni cells heterozygous for p53 gene mutations. Oncogene 1997;14(18):2137-47. Denekamp J. Cell kinetics and radiation biology. Int J Rad Biol 1986;49(2):357-380. DeWeese TL, Walsh JC, Dillehay LE , Kessis TD, Hedrick L, Cho KR, Nelson WG. Human papillomavirus E6 and E7 oncoproteins alter cell cycle progression but not radiosensitivity of carcinoma cells treated with low-dose-rate radiation. Int J Rad One Biol Phys 1997;37(1): 145-54. Dewey WC, Humphrey R M . Relative radiosensitivity of different phases in the life cycle of L-P59 mouse fibroblasts and ascites tumor cells. Rad Res 1962;16:503-530. Dewey WC, Noel JS, Dettor C M . Changes in radiosensitivity and dispersion of chromatin during the cell cycle of synchronous Chinese hamster cells. Rad Res 1972;52:373-394. Dileto CL, Travis EL . Fibroblast radiosensitivity in vitro and lung fibrosis in vivo: comparison between a fibrosis-prone and fibrosis-resistant mouse strain. RadRes 1996;146:61-67. DiMaio D, Treisman T, Maniatis T. Bovine papillomavirus vector that propagates as a plasmid in both mouse and bacterial cells. Proc Natl Acad Sci USA 1982;79:4030-4034. Dische S, Saunders M , Barrett A, Harvey A, Gibson D, Parmar M . A randomised multicentre trial of CHART versus conventional radiotherapy in head and neck cancer. Radiother Oncol 1997;44(2): 123-36. Djordjevic B, Tolmach LJ. X-ray sensitivity of HeLa S3 cells in the G2 phase: comparison of two methods of synchronization. Biophys J 1967;7:77-94. Douglas B G . Implications of the quadratic cell survival curve and human skin radiation "tolerance doses" on fractionation and superfractionation dose selection. Int J Rad One Biol Phys 1982;8:1135-1142. Douglas B G , Fowler JF. The effect of multiple small doses of X rays on skin reactions in the mouse and a basic interpretation. RadRes 1976;66:401-426. Douglas BG, Henkelman R M , Lam G K Y , Fowler JF, Eaves CJ. Practical and theoretical considerations in the use of the mouse foot system to derive epithelial stem cell survival parameters. RadRes 1979;77:453-471. 208 References Drewinko B, Yang L Y , Rohsdahl M M . Radiation response of cultured human carcinoembryonic antigen-producing colon adenocarcinoma cells. Int J Rad One Biol Phys 1977;2:1109-1114. Durand RE. Repair during multifraction exposures: spheroids versus monolayers. Br J Cancer 1984;49(Suppl. 6):203-206. Durand RE. Use of a cell sorter for assays of cell clonogenicity. Cancer Res 1986;46:2775-2778. el-Deiry WS, Tokino T, Velculescu V E , Levy DB, Parsons R, Trent JM, Lin D, Mercer WE, Kinzler KW, Vogelstein B. WAF1, a potential mediator of p53 tumor suppression. Cell 1993;75(4):817-25. Erikson RL, Szybalski W. Molecular radiobiology of human cell lines. IV. Variation in ultraviolet light and X-ray sensitivity during the division cycle. Rad Res 1963;18:200-212. Fertil B, Deschavanne PJ, Lachet B, Malaise EP. In vitro radiosensitivity of six human cell lines. Rad Res 1980;82:297-309. Fertil B, Malaise EP. Inherent cellular radiosensitivity as a basic concept for human tumor radiotherapy. Int J Rad One Biol Phys 1981;7:621-629. Fertil B, Malaise EP. Intrinsic radiosensitivity of human cell lines is correlated with radioresponsiveness of human tumors: analysis of 101 published survival curves. Int J Rad One Biol Phys 1985; 11 (9): 1699-1707. Garkavtsev I, Grigorian IA, Ossovskaya VS, Chernov M V , Chumakov P M , Gudkov A V . The candidate tumour suppressor p33INGl cooperates with p53 in cell growth control. Nature 1998;391(6664):295-8. Garkavtsev I, Kazarov A, Gudkov A, Riabowol K. Suppression of the novel growth inhibitor p33INGl promotes neoplastic transformation. Nature Genetics 1996;14(4):415-20. Giaccia AJ, Schwartz J, Shieh J, Brown JM. The use of asymmetric-field inversion gel electrophoresis to predict tumor cell radiosensitivity. Radiother Oncol 1992;24(4):231-8. Gillespie CJ, Chapman JD, Reuvers AP, Dugle DL. The inactivation of Chinese hamster cells by X rays: synchronized and exponential cell populations. Rad Res 1975;64:353-364. Goodhead DT. Saturable repair models of radiation action in mammalian cells. Rad Res 1985;104:S58-S67. 209 References Gorczyca W, Gong J, Ardelt B, Traganos F, Darzynkiewicz Z. The cell cycle related differences in susceptibility of HL-60 cells to apoptosis induced by various antitumor agents. Cancer Res 1993;53:3186-3192. Graeber TG, Osmanian C, Jacks T, Housman DE, Koch CJ, Lowe SW, Giaccia AJ. Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours. Nature 1996;379:88-91. Grdina DJ. Variations in radiation response of tumor subpopulations. In: Meyn RE, Withers HR, editors. Radiation Biology in Cancer Research. New York: Raven Press; 1980. p. 353-363. Grdina DJ, Meistrich M L , Meyn RE, Johnson TS, White RA. Cell synchrony techniques. I. A comparison of methods. Cell Tissue Kinet 1984;17:223-236. Griffiths SD, Clarke AR, Healy L E , Ross G, Ford A M , Hooper M L , Wyllie A H , Greaves M . Absence of p53 permits propagation of mutant cells following genotoxic damage. Oncogene 1997; 14(5): 523-31. Gupta N , Vij R, Haas-Kogan DA, Israel M A , Deen DF, Morgan WF. Cytogenetic damage and the radiation-induced Gl-phase checkpoint. RadRes 1996;145:289-298. Haas-Kogan DA, Yount G, Haas M , Levi D, Kogan SS, Hu L, Vidair C, Deen DF, Dewey WC, Israel M A . p53-dependent G l arrest and p53-independent apoptosis influence the radiobiologic response of glioblastoma. Int J Rad One Biol Phys 1996;36:95-103. Hahn G M , Bagshaw M A . Serum concentration: effects on cycle and X-ray sensitivity of mammalian cells. Science 1966;151:459-461. Hall EJ. Radiobiology for the radiologist. Fourth ed. Philadelphia: J.B. Lippincott Company; 1994. Han A, Sinclair WK, Kimler BF. The effect of N-ethylmaleimide on the response to X -rays of synchronized HeLa cells. RadRes 1976;65:337-350. Harper JW, Adami GR, Wei N , Keyomarsi K, Elledge SJ. The p21 Cdk-interacting protein Cipl is a potent inhibitor of G l cyclin-dependent kinases. Cell 1993;75(4):805-16. Heichman K A , Roberts JM. Rules to replicate by. Cell 1994;79(4):557-62. Hopewell JW, van den Aardweg GJMJ. Studies of dose-fractionation on early and late responses in pig skin: a reappraisal of the importance of the overall treatment time and its effects on radiosensitization and incomplete repair. Int JRad One Biol Phys 1991;21:1441-1450. 210 References Horiot JC, Le Fur R, N'Guyen T, Chenal C, Schraub S, Alfonsi S, Gardani G, Van Den Bogaert W, Danczak S, Bolla M , Van Glabbeke M , De Pauw M . Hyperfractionation versus conventional fractionation in oropharyngeal carcinoma: final analysis of a randomized trial of the EORTC cooperative group of radiotherapy. Radiother Oncol 1992;25:231-241. Howard A, Pelc S. Synthesis of desoxyribonucleic acid in normal and irradiated cells and its relation to chromosome breakage. Heredity 1953;6(Suppl.):261-273. Hunter T, Pines J. Cyclins and cancer. II: Cyclin D and C D K inhibitors come of age. Cell 1994;79(4):573-82. Hwang A, Muschel RJ. Radiation and the G2 phase of the cell cycle. Rad Res 1998;150:S52-S59. Iliakis G, Cicilioni O, Metzger I. Measurement of D N A double-strand breaks in CHO cells at various stages of the cell cycle using pulsed field gel electrophoresis: calibration by means of 1-125 decay. IntJRadBiol 1991;59(2):343-357. Iliakis G, Nusse M . Evidence that repair and expression of potentially lethal damage cause the variations in cell survival after X irradiation observed through the cell cycle in Ehrlich ascites tumor cells. Rad Res 1983;95:87-107. Johns HE, Cunningham, J.R. The physics of radiology. Fourth ed. Springfield, Illinois: Charles C. Thomas; 1983. Joiner M C , Lambin P, Malaise EP, Robson T, Arrand JE, Skov K A , Marples B. Hypersensitivity to very-low single radiation doses: its relationship to the adaptive response and induced radioresistance. Mut Res 1996;358(2): 171-83. Karasuyama H, Melchers F. Establishment of mouse cell lines which constitutively secrete large quantities of interleukin 2, 3, 4 or 5, using modified cDNA expression vectors. Eur J Immunol 1988;18:97-104. Kastan M B , Zhan Q, El-Deiry WS, Carrier F, Jacks T, Walsh WV, Plunkett BS, Vogelstein B, Fornace AJ. A mammalian cell cycle checkpoint pathway utilizing p53 and GADD45 is defective in Ataxia-Telangiectasia. Cell 1992;71:587-597. Kellerer A M , Rossi H H . The theory of dual radiation action. In: Current topics in radiation research quarterly: North Holland; 1972. p. 85-158. Keng PC, Siemann DW, Wheeler KT. Comparison of tumour age response to radiation for cells derived from tissue culture or solid tumours. Br J Cancer 1984;50:519-526. 211 References Kerr JFR, Searle J. Apoptosis: its nature and kinetic role. In: Meyn E, Withers HR, editors. Radiation Biology in Cancer Research. New York: Raven Press; 1980. p. 367-384. Kerr JFR, Wyllie A H , Currie AR. Apoptosis: a basic biological phenomenon with wide ranging implications in tissue kinetics. Br J Cancer 1972;26:239-257. Kessis TD, Slebos RJ, Nelson WG, Kastan M B , Plunkett BS, Han SM, Lorincz AT, Hedrick L, Cho KR. Human papillomavirus 16 E6 expression disrupts the p53-mediated cellular response to D N A damage. Proc Natl Acad Sci USA 1993;90:3988-3992. King RW, Jackson PK, Kirschner MW. Mitosis in transition. Cell 1994;79(4):563-71. Kuerbitz SJ, Plunkett BS, Walsh WV, Kastan M B . Wild-type p53 is a cell cycle checkpoint determinant following irradiation. Proc Natl Acad Sci USA 1992;89:7491-7495. LaBaer J, Garrett M D , Stevenson LF, Slingerland JM, Sandhu C, Chou HS, Fattaey A, Harlow E. New functional activities for the p21 family of C D K inhibitors. Genes Dev 1997;ll(7):847-62. Laemmli U K . Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970;227:680-685. Lambin P, Malaise EP, Joiner MC. Might intrinsic radioresistance of human tumour cells be induced by radiation? Int J Rad Biol 1996;69(3):279-90. Langley RE, Quartuccio SG, Kennealey PT, Coleman CN, Bump EA. Effect of cell cycle stage, dose rate and repair of sublethal damage on radiation-induced apoptosis in F9 teratocarcinoma cells. RadRes 1995;144:90-96. Lawrence TS, Davis M A , Tang H Y , Maybaum J. Fluorodeoxyuridine-mediated cytotoxicity and radiosensitization require S phase progression. Int J Rad Biol 1996;70(3):273-80. Le X C , Xing JZ, Lee J, Leadon SA, Weinfeld M . Inducible repair of thymine glycol detected by an ultrasensitive assay for D N A damage. Science 1998;280:1066-1069. Lea DE. Actions of radiations on living cells: Cambridge University Press; 1946. Leadon SA. Repair of D N A damage produced by ionizing radiation: a minireview. Sem Radiat Oncol 1996;6(4):295-305. Lee JM, Bernstein A. p53 mutations increase resistance to ionizing radiation. Proc Natl AcadSciUSA 1993;90:5742-5746. 212 References Lee SE, Mitchell RA, Cheng A, Hendrickson EA. Evidence for DNA-PK-dependent and -independent D N A double-strand break repair pathways in mammalian cells as a function ofthe cell cycle. Mol Cell Biol 1997;17(3): 1425-33. L i C Y , Nagasawa H, Dahlberg WK, Little JB. Diminished capacity for p53 in mediating a radiation-induced G l arrest in established human tumor cell lines. Oncogene 1995;ll(9):1885-92. Ling CC, Chen CH, Fuks Z. An equation for the dose response of radiation-induced apoptosis: possible incorporation with the LQ model. Radiother Oncol 1994;33:17-22. Ling CC, Chen CH, L i W X . Apoptosis induced at different dose rates: implication for the shoulder region of cell survival curves. Radiother Oncol 1994;32:129-136. Little JB. Delayed initiation of D N A synthesis in irradiated human diploid cells. Nature 1968;218:1064-1065. Lowe SW, Schmitt E M , Smith SW, Osborne BA, Jacks T. p53 is required for radiation-induced apoptosis in mouse thymocytes. Nature 1993;362(6423):847-9. Lowy DR, Dvoretzky I, Shober R, Law MF, Engel L, Howley PM. In vitro tumorigenic transformation by a defined sub-genomic fragment of bovine papilloma virus DNA. Nature 1980;287:72-74. Maity A, McKenna WG, Muschel RJ. The molecular basis for cell cycle delays following ionizing radiation: a review. Radiother Oncol 1994;31:1-13. Marples B, Joiner M C . The response of Chinese hampster V79 cells to low radiation doses: evidence of enhanced sensitivity of the whole cell population. Rad Res 1993;133:41-51. Marples B, Lambin P, Skov K A , Joiner M C . Low dose hyper-radiosensitivity and increased radioresistance in mammalian cells. Int J Rad Biol 1997;71(6):721-35. McGinn CJ, Kinsella TJ. The clinical rationale for S-phase radiosensitization in human tumors. CurrProb Cane 1993;17(5):273-324. McGinn CJ, Miller E M , Lindstrom MJ, Kunugi K A , Johnston PG, Kinsella TJ. The role of cell cycle redistribution in radiosensitization: implications regarding the mechanism of fluorodeoxyuridine radiosensitization. Int J Rad One Biol Phys 1994;30:851-859. Mcllwrath AJ, Vasey PA, Ross G M , Brown R. Cell cycle arrests and radiosensitivity of human tumor cell lines: dependence on wild-type p53 for radiosensitivity. Cancer Res 1994;54:3718-3722. 213 References Meyn MS. Ataxia-telangiectasia and cellular responses to D N A damage. Cancer Res 1995;55:5991-6001. Meyn RE, Stephens L C , Ang K K , Hunter NR, Brock WA, Milas L, Peters LJ . Heterogeneity in the development of apoptosis in irradiated murine tumours of different histologies. Int J Rad Biol 1993;64(5):583-91. Milas L , Stephens L C , Meyn RE. Relation of apoptosis to cancer therapy. In Vivo 1994;8(5):665-73. Mirkovic N , Meyn RE, Hunter NR, Milas L. Radiation-induced apoptosis in a murine lymphoma in vivo. Radiother Oncol 1994;33(1): 11-6. Mozumder A. Early production of radicals from charged particle tracks in water. Rad Res 1985;104:S-33-S-39. Munro TR. The relative radiosensitivity of the nucleus and cytoplasm of Chinese hamster fibroblasts. Rad Res 1970;42:451-470. Murray AW, Kirschner MW. Dominoes and Clocks: the union of two views of the cell cycle. Science 1989;246(4930):614-621. Muschel RJ, Zhang HB, Iliakis G, McKenna WG. Effects of ionizing radiation on cyclin expression in HeLa cells. Rad Res 1992;132:153-157. Muschel RJ, Zhang HB, McKenna WG. Differential effect of ionizing radiation on the expression of cyclin A and cyclin B in Hela cells. Cancer Res 1993 ;53:1128-1135. Nagasawa H , L i C-Y, Maki CG, Imrich AC, Little JB. Relationship between radiation-induced GI phase arrest and p53 function in human tumor cells. Cancer Res 1995;55:1842-1846. Nagasawa H, Keng P, Maki C, Y u Y, Little JB. Absence of a radiation-induced first-cycle G l - S arrest in p53+ human tumor cells synchronized by mitotic selection. Cancer Res 1998;58:2036-2041. Neary GJ. Chromosome aberrations and the theory of R B E I. General Considerations. Int JRadBiol 1965;9(5):477-502. Noble JR, Willetts K E , Mercer WE, Reddel RR. Effects of exogenous wild-type p53 on a human lung carcinoma cell line with endogenous wild-type p53. Experimental Cell Research 1992;203(2):297-304. Nurse P. Ordering S phase and M phase in the cell cycle. Cell 1994;79(4):547-50. 214 References Olive PL, Banath JP, Durand RE. Heterogeneity in radiation-induced D N A damage and repair in tumor and normal cells measured using the "comet" assay. Rad Res 1990; 122:86-94. Olive PL, Banath JP, Durand RE. Development of apoptosis and polyploidy in human lymphoblast cells as a function of position in the cell cycle at the time of irradiation. Rad Res 1996;146(6):595-602. Olive PL, Durand RE. Apoptosis: an indicator of radiosensitivity in vitro? Int J Rad Biol 1997;71(6):695-707. Pellegata NS, Antoniono RJ, Redpath JL, Stanbridge EJ. D N A damage and p53-mediated cell cycle arrest: a reevaluation. Proc Natl Acad Sci USA 1996;93(26): 15209-14. Petterson EO, Christensen T, Bakke O, Oftebro R. A change in the oxygen effect throughout the cell cycle of human cells of the line NHIK 3025 cultivated in vitro. IntJ Rad Biol 1977;31(2): 171-184. Press W H , Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipies in Fortran. 2nd ed. Cambridge: Cambridge University Press; 1992. Puck TT, Marcus PI. Action of X-rays on mammalian cells. J Exp Med 1956;103:653-66. Quiet CA, Weichselbaum RR, Grdina DJ. Variation in radiation sensitivity during the cell cycle of two human squamous cell carcinomas. Int J Rad One Biol Phys 1991 ;20(4): 733-738. Rodrigues NR, Rowan A, Smith MEF, Kerr IB, Bodmer WF, Gannon JV, Lane DP. p53 mutations in colorectal cancer. Proc Natl Acad Sci USA 1990;87:7555-7559. Rowley R. Radiation-induced mitotic delay: a genetic characterization in the fission yeast. Rad Res 1992;132:144-152. Rudoltz MS, Kao G, Blank KR, Muschel RJ, McKenna WG. Molecular biology of the cell cycle: potential for therapeutic applications in radiation oncology. Sem Radiat Oncol 1996;6(4):284-294. Schultheiss TE, Zagars GK, Peters LJ. An explanatory hypothesis for early- and late-effect parameter values in the LQ model. Radiother Oncol 1987;9:241-248. Schwartz JL. The role of constitutive and inducible processes in the response of human squamous cell carcinoma cell lines to ionizing radiation. Rad Res 1994;138(1 Suppl): S37-9. 215 References Schwartz JL, Mustafi R, Beckett M A , Weichselbaum RR. D N A double-strand break rejoining rates, inherent radiation sensitivity and human tumour response to radiotherapy. Br J Cancer 1996;74(l):37-42. Sherr CJ. GI phase progression: cycling on cue. Cell 1994;79(4):551-5. Siles E, Villalobos M , Valenzuela MT, Nunez MI, Gordon A, McMillan TJ, Pedraza V , Ruiz de Almodovar JM. Relationship between p53 status and radiosensitivity in human tumour cell lines. Br J Cancer 1996;73:581-588. Sinclair WK. Cyclic X-ray responses in mammalian cells in vitro. Rad Res 1968;33:620-643. Sinclair WK, Morton RA. Variations in X-ray response during the division cycle of partially synchronized Chinese hamster cells in culture. Nature 1963;199:1158-1160. Sinclair WK, Morton RA. X-ray and ultraviolet sensitivity of synchronized Chinese hamster cells at various stages of the cell cycle. Biophys J 1965;5:1-25. Sinclair WK, Morton RA. X-ray sensitivity during the cell generation cycle of cultured Chinese hamster cells. Rad Res 1966;29:450-474. Skarsgard L D , Harrison I, Durand RE. The radiation response of asynchronous cells at low dose: evidence of substructure. Rad Res 1991;127:248-256. Skarsgard LD, Kihlman BA, Parker L, Pujara C M , Richardson S. Survival, chromosome abnormalities, and recovery in heavy-ion- and X-irradiated mammalian cells. Rad Res 1967;7(Suppl.):208-221. Skarsgard L D , Skwarchuk MW, Wouters BG, Durand RE. Substructure in the radiation survival response at low dose in cells of human tumor cell lines. Rad Res 1996; 146:3 88-398. Skarsgard L D , Wilson DJ, Durand RE. Survival at low dose in asynchronous and partially synchronized Chinese hamster V79-171 cells. Rad Res 1993;133:102-107. Skarsgard LD, Wouters BG. Substructure in the cell survival response at low radiation dose: effect of different subpopulations. Int J Rad Biol 1997;71(6):737-749. Skwarchuk M W , Wouters BG, Skarsgard LD. Substructure in the radiation survival response at low dose: asynchronous and partially synchronized V79-WNRE cells. Int J RadBiol 1993;64:601-612. 216 References Slichenmyer WJ, Nelson WG, Slebos RJ, Kastan MB. Loss of a p53-associated Gl checkpoint does not decrease cell survival following DNA damage. Cancer Res 1993;53:4164-4168. Steller H. Mechanisms and genes of cellular suicide. Science 1995;267:1445-1449. Stephens LC, Ang KK, Schultheiss TE, Milas L, Meyn RE. Apoptosis in irradiated murine tumors. RadRes 1991;127(3):308-16. Stephens LC, Hunter NR, Ang KK, Milas L, Meyn RE. Development of apoptosis in irradiated murine tumors as a function of time and dose. RadRes 1993;135(l):75-80. Story MD, Voehringer DW, Malone CG, Hobbs ML, Meyn RE. Radiation-induced apoptosis in sensitive and resistant cells isolated from a mouse lymphoma. Int J Rad Biol 1994;66(6):659-68. Tang H, Davis MA, Strickfaden SM, Maybaum J, Lawrence TS. Influence of cell cycle phase on radiation-induced cytotoxicity and DNA damage in human colon cancer (HT29) and Chinese hamster ovary cells. RadRes 1994;138 (Suppl.):S109-S112. Terasima T, Tolmach LJ. Changes in X-ray sensitivity of HeLa cells during the division cycle. Nature 1961;190:1210-1211. Terasima T, Tolmach LJ. Variations in several responses of HeLa cells to X-irradiation during the division cycle. Biophys J 1963;3:11-33. Terasima T, Tolmach LJ. X-ray sensitivity and DNA synthesis in synchronous populations of HeLa cells. Science 1963;140:490-492. Thames HD, White RA. State-vector models of the cell cycle. 1. Parametrization and fits to labelled mitoses data. J Theor Biol 1977;67:733-756. Thames HD, Withers HR, Peters LJ, Fletcher GH. Changes in early and late radiation responses with altered dose fractionation: implications for dose-survival relationships. Int J Rad One Biol Phys 1982;8:219-226. Thompson CB. Apoptosis in the pathogenesis and treatment of disease. Science 1995;267:1456-1462. Tobias CA. The repair-misrepair model in radiobiology: comparison to other models. Rad Res 1985;104:S-77-S-95. Tobias CA, Blakely EA, Ngo FQH, Yang TCH. The repair-misrepair model of cell survival. In: Meyn RA, Withers HR, editors. Radiation Biology and Cancer Research. New York: Raven Press; 1980. p. 195-230. 217 References Travis EL , Thames HD, Watkins TL, Kiss I. The kinetics of repair in mouse lung after fractionated irradiation. Int J Rad Biol 1987;52(6):903-919. Tsang N M , Nagasawa H, L i C, Little JB. Abrogation of p53 function by transfection of HPV16 E6 gene enhances the resistance of human diploid fibroblasts to ionizing radiation. Oncogene 1995;10(12):2403-8. Vegesna V, Withers HR, Thames HD, Mason K. Multifraction radiation response of mouse lung. Int J Rad Biol 1985;47(4):413-422. Vos O, Schenk H A E M , Bootsma D. Survival of excess thymidine synchronized cell populations in vitro after X-irradiation in various phases of the cell cycle. Int J Rad Biol 1966;ll(5):495-503. Wang Q, Fan S, Eastman A, Worland PJ, Sausville EA, O'Connor P M . UCN-01: a potent abrogator of G2 checkpoint function in cancer cells with disrupted p53. JNatl Cancer Inst 1996;88:956-965. Ward JF. Biochemistry of D N A lesions. RadRes 1985;8(Suppl.):S103-l 1. Weininger J, Guichard M , Joly A M , Malaise EP, Lachet B. Radiosensitivity and growth parameters in vitro of three human melanoma cell strains. Int J Rad Biol 1978;34:285-290. Wells J, Berry RJ, Laing A H . Reproductive survival of explanted human tumor cells after exposure to nitrogen mustard or X-irradiation; differences in response with subsequent subculture in vitro. RadRes 1977;69:90-98. West C M L , Keng PC, Sutherland R M . Growth phase related variation in the radiation sensitivity of human colon adenocarcinoma cells. Int J Rad One Biol Phys 1988;14:1213-1219. Whitaker SJ, Ung Y C , McMillan TJ. D N A double-strand break induction and rejoining as determinants of human tumour cell radiosensitivity. A pulsed-field gel electrophoresis study. Int J Rad Biol 1995;67(1):7-18. Whitmore GF, Gulyas S, Botond J. Radiation sensitivity throughout the cell cycle and its relationship to recovery. In: Cellular Radiation Biology. Baltimore: Williams and Wilkins; 1965. p. 423-441. Williams M V , Denekamp J, Fowler JF. A review of a/f3 ratios for experimental tumors: implications for clinical studies of altered fractionation. Int J Rad One Biol Phys 1985;11:87-96. 218 References Withers HR. Cell cycle redistribution as a factor in multifraction irradiation. Radiology 1975;114:199-202. Wlodek D, Hittelman W. The relationship of D N A and chromosome damage to survival of synchronized X-irradiated L5178Y cells. I. Initial damage. RadRes 1988;115:550-565. Wlodek D, Hittelman W. The relationship of D N A and chromosome damage to survival of synchronized X-irradiated L5178Y cells. II. Repair. RadRes 1988;115:566-575. Wouters B G , Giaccia AJ, Denko NC, Brown JM. Loss of p21 wafl/cipl sensitizes tumors to radiation by an apoptosis-independent mechanism. Cancer Res 1997;57:4703-4706. Wouters B G , Skarsgard L D . Low-dose radiation sensitivity and induced radioresistance to cell killing in HT-29 cells is distinct from the "adaptive response" and cannot be explained by a subpopulation of sensitive cells. RadRes 1997;148:435-442. Wouters B G , Sy A M , Skarsgard LD. Low-dose hypersensitivity and increased radioresistance in a panel of human tumor cell lines with different radiosensitivity. Rad Res 1996;146:399-413. Yount GL, Hass-Kogan DA, Vidair CA, Haas M , Dewey WC, Israel M A . Cell cycle synchrony unmasks the influence of p53 function on radiosensitivity of human glioblastoma cells. Cancer Res 1996;56:500-506. 219 Appendix 1: Effects of binucieate cells on classification of d , S, and G2/M phase cells To determine the fraction of cells in the G 2 peak which were initially binucieate Gi cells, mitotically selected synchronized HT-29 cells were fixed and fluorescently stained approximately 3 h after mitotic selection, and cells from the G 2 peak population were sorted onto slides and observed under the fluorescent microscope. A preliminary survey found that the large majority of cells from the G 2 peak were binucieate cells. In a more complete assessment, Giemsa-stained preparations of synchronized A549 cells at 3h after mitotic selection were counted by eye. This examination showed that the cells were 72%:28% singlefbinucleate by examination of Giemsa-stained slides compared to 78%: 22% Gi:G 2 D N A content by flow cytometric analysis. This 1:1 correspondence between G 2 D N A content and binucieate cells implied that binucieate cells accounted for approximately 100% of the cells in the G 2 peak at 3 h after mitotic selection in A549 cells. Based on this finding, a simple scheme for cell classification was devised, which was described in Section 3.2. For times from 0 h after mitotic selection to the approximate duration of (Gi+S) phase, all cells in the G 2 peak of D N A histograms were counted as Gi cells. After (Gi+S) phase, from the first time when there was a significant increase in the fraction of cells in the G 2 peak, all cells in the G 2 peak were counted as G 2 cells. The possible errors in cell classification which were introduced by this simple accounting of cells are as follows. During the first 3-4 h after mitotic selection, this accounting probably underestimates the fraction of G 2 / M cells (e.g. mitotic cells which had not completed 220 Appendix 1 division) in the synchronized population. From 3-4 h after mitotic selection until the end of (Gi+S) phase, this accounting would also underestimate the fraction of unsynchronized G2/M cells in the population, and overestimate the fraction of Gi cells, but this error would result in incorrect classification of certainly no more than 10% of Gi , S, or G2/M cells, based on the experimental counts of binucieate cells described above. After the end of (Gi+S) phase, the accounting would overestimate the number of G 2 phase cells, and underestimate the fraction of Gi phase cells, because some non-cycling binucieate Gi cells would have remained in the G2 peak. This error was estimated to cause the incorrect classification of no more than about 6% of the cells, since typically 80% of the approximately 30% of the total cell population that was in the G 2 peak during Gi phase left that peak by the end of (Gi+S) phase. Thus (100%-80%)x30% = 6% would be the approximate fraction of the total number of Gi cells incorrectly classified as G2 cells after the end of (Gi+S phase). A more complex accounting scheme was devised which attempted to account for G 2 / M phase cells in the first 0-3 h after mitotic selection, and also properly attribute non-cycling Gi cells after the end of (Gi+S) phase. This scheme was of course dependent on making diftlcult-to-confirm assumptions about the time course of mitosis and the fraction of non-cycling Gi cells; it had to be customized for each time course of D N A histograms; and application of the scheme did not alter the calculation of Gi , S, and G 2 / M fractions by more than about 6% at any time beyond 3h after mitotic selection (data not shown). Given these facts, the possible errors associated with the simple accounting of cells were deemed to be acceptable, and the original scheme was used as described above for all kinetic analyses. 221 Appendix 1 It should also be noted that later modelling of survival in synchronized cells used the calculated percentages of G i , S, and G 2 cells as they are described here (see section 3.6). If some or all of the cells in the binucleate population were not clonogenically viable (as might be suggested by the rapid disappearance of the binucleates from the synchronized populations), then the non-viable part of this population should be excluded from the cell cycle fractions for the purposes of survival modelling (which should only include viable cells). However, it was found that the effect of excluding the binucleates completely from the calculation of G i , S, and G2 fractions was negligibly small, because the binucleate population was a "mirror" of the singlet population, with close to the same relative fractions of G i , S, and G2 cells during the first cycle after mitotic selection. Specifically, excluding all cells with G2 complement or greater DNA content during (Gi+ S) phase in an alternative calculation of the fractions of G i , S, and G2 cells was found to alter the currently calculated fractions by no more than 7% at any time after mitotic selection. Since the effect of this change in calculation was small, and the rationale for such a change was not certain, this alternative calculation was not used. 222 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0085449/manifest

Comment

Related Items