UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The mechanism of malignancy associated changes Wilton, David William 1997

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

Item Metadata


831-ubc_1997-0493.pdf [ 5.1MB ]
JSON: 831-1.0087874.json
JSON-LD: 831-1.0087874-ld.json
RDF/XML (Pretty): 831-1.0087874-rdf.xml
RDF/JSON: 831-1.0087874-rdf.json
Turtle: 831-1.0087874-turtle.txt
N-Triples: 831-1.0087874-rdf-ntriples.txt
Original Record: 831-1.0087874-source.json
Full Text

Full Text

THE M E C H A N I S M OF M A L I G N A N C Y ASSOCIATED CHANGES by DAVID W I L L I A M WILTON B.Sc. (HON) Queen's University, 1994 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R OF SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES Department of Pathology We accept this thesis as conforming to f^he required standard THE UNIVERSITY O F ^ I T I S H C O L U M B I A August 1997 © David William Wilton, 1997 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for .extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be. allowed without my written permission. Department of 'Pq^ V*—O (o^ y The University of British Columbia Vancouver, Canada Date AuyxS>l 2/^.97 DE-6 (2/88) ABSTRACT Malignancy Associated Changes (MAC) can be defined as subtle morphologic and or physiologic changes which occur in ostensibly normal cells of patients harboring malignant or pre-malignant lesions. This phenomenon has been postulated to have great potential as a powerful tool in the fight against cancer. Recent evidence indicates that this phenomenon can be used in the detection of cancers in their earliest stages, for the objective diagnosis and prognosis of these cancers, as well as a means for monitoring of treatment outcome. Yet the mechanism of M A C remains poorly understood. There have been two main theories proposed to explain the mechanism of M A C . The first, commonly referred to as the "Field Cancerization Theory", postulates that M A C simply reflect mutations in the normal tissue surrounding the tumor resultant from concurrent exposure to carcinogens. If true, this theory implies that there is no clinical value of M A C . The second referred to as the "Humoral Theory of M A C " postulates a direct cause and effect relationship of malignancy to M A C . In this hypothesis, soluble chemical mediators are released from the malignant tissue which act on the normal tissue in the surrounding environment. In these studies an in vitro model was designed to mimic the interaction of malignant cells with normal bronchial epithelial cells. This model was employed to investigate the mechanism of M A C . It was shown that M A C can be induced in vitro demonstrating a cause and effect relationship of malignancy to M A C . Maximal induction of M A C was observed following 48 hours of direct co-culture of tumor cells with normal cells. The strength of M A C expression was found to be dose dependent indicating that this model can be utilized in future studies as a biological assay in the isolation of factors present in malignant conditioned medium capable of inducing M A C . These studies coupled with substantial evidence in vivo provide strong evidence supporting the Humoral mechanism of M A C . Wilton ii TABLE OF CONTENTS A B S T R A C T i i T A B L E OF CONTENTS i i i LIST OF FIGURES vi LIST OF T A B L E S vii LIST OF ABBREVIATIONS viii A C K N O W L E D G E M E N T S ix 1.0 INTRODUCTION 1.1 Malignancy Associated Changes: Historical Perspective 1 1.1.1 Qualitative Identification of Nuclear M A C 1 1.1.2 Quantitative Identification of Nuclear M A C 4 1.1.3 Non-Nuclear M A C 7 1.2 Clinical Significance of M A C 8 1.2.1 M A C as an Indicator for the Early Detection of Cancer 9 1.2.2 M A C as a Prognostic Indicator of Cancer 12 1.3 Proposed Mechanisms of M A C 14 1.3.1 Field Cancerization Theory of M A C 15 1.3.2 Humoral Mechanism of M A C 15 Humoral Mechanism of M A C : Rationale 18 1.4 Project Objectives 24 1.5 Proj ect Strategy 25 iii Wilton 2.0 MATERIALS AND METHODS 2.1 Primary Culture of Human Bronchial Epithelium 26 2.1.1 Growth Surface Preparation 26 2.1.2 Explant Culture 27 2.2 Models of Normal and Tumor Cell Interaction 28 2.2.1 Indirect Co-Culture Model 29 2.2.1 Metabolite Composition 30 2.2.3 Direct Co-Culture Model 31 Time Course of M A C Expression 32 Dose Dependency of M A C Expression 33 2.3 Quantitative D N A Staining 33 2.4 Cell Cycle Analysis 34 2.5 Image Cytometry 35 2.5.1 Nuclear Feature Calculations 38 Size and Shape Features 39 Photometric Features 39 Discrete Texture Features 39 Markovian Texture Features 40 Run Length Texture Features 40 Fractal Texture Features 40 Wilton 2.6 Statistical Analysis 41 2.6.1 Discriminant Function Analysis 41 2.6.2 Cell by Cell Analysis 43 2.6.3 Slide by Slide Analysis 44 2.6.4 t-test For Independent Samples 45 3.0 RESULTS 3.1 Primary Culture of Human Bronchial Epithelium 47 3.2 Tumor Cell Culture Conditions 48 3.3 Pilot Study of M A C Expression in vitro 49 3.4 Indirect Co-Culture Model of M A C 51 3.5 Procedural Controls 53 3.5.1 Metabolite Analysis 53 3.5.2 Discriminant Function Performance 55 3.5.3 Test for Statistical Overtraining 56 3.6 Direct Co-Culture Studies 57 3.6.1 Time Course of M A C Expression 58 3.6.2 Dose Dependency of M A C Expression 59 3.7 Cell Cycle Study 65 4.0 DISCUSSION 67 5.0 CONCLUSION 83 BIBLIOGRAPHY 84 Wilton LIST OF FIGURES Figure 1.1 M A C In Bronchial Biopsies 13 Figure 2.1 Indirect Co-Culture Model 29 Figure 2.2 Direct Co-Culture Model 31 Figure 2.3 What is Measured? 38 Figure 3.1 Growth of NCI-H460 in Several Culture Media 48 Figure 3.2 Pilot Study Sample Classification 50 Figure 3.3 M A C Induced in vitro : 52 Figure 3.4 Metabolite Analysis 54 Figure 3.5 Discriminant Function Performance 56 Figure 3.6 Time Course of M A C Expression 59 Figure 3.7 Strength of M A C Expression - Dose Dependency 1 60 Figure 3.8 ROC Curve for Dose Dependency 1 61 Figure 3.9 Frequency of M A C Expression - Dose Dependency 1 63 Figure 3.10 Dose Dependency 2 Trained to Middle Dose 64 Figure 3.11 Ploidy Distribution of Dose Dependency Experiments 65 Figure 4.1 Size Fractionation of M A C Conditioned Medium 81 vi Wilton LIST OF TABLES Table 1.1 Measurement of M A C by Qualitative Criteria 3 Table 1.2 Measurement of M A C by Quantitative Image Analysis 6 Table 1.3 Paraneoplastic Syndromes 22 Table 3.1 Classification Matrix for Statistical Overtraining Test 57 Table 3.2 Cell Cycle Analysis on Dose Dependency Experiments 66 Table 4.1 Molecular Weight of Candidate Growth Factors 82 Wilton vi i LIST OF ABBREVIATIONS M A C Malignancy Associated Changes NSCLC Non Small Cell Lung Cancer CIN Cervical Intraepithelial Neoplasia NHBE Normal Human Bronchial Epithelia B E G M Bronchial Epithelial Growth Medium BEBM Bronchial Epithelial Basal Medium TGF Transforming Growth Factor EGF Epidermal Growth Factor IGF-1 Insulin Like Growth Factor-1 PDGF Platelet Derived Growth Factor Sensitivity: ((true positive)/(true positive + false negative)) x 100% Specificity: ((true negative)/(true negative + false positive)) x 100% Wilton viii ACKNOWLEDGEMENTS The work in this thesis would not have been possible without the contribution and support of many people. I would like to thank my thesis supervisor Dr. Branko Palcic for his advice, optimism and encouragement throughout the course of this work. I would also like to extend a special thanks to Dr. Steven Lam and Dr. Calum MacAulay for their valuable contributions to this thesis. In addition, I would like to thank Dr. A . Autor for the time and energy she invested in training me in techniques of tissure culture and for the personal interest she showed in my work. I am also indebted to Dr. Alexi Dudkine for his assistance with the statistical methods of this project, Mr. Paul Lam for the many hours he spent staining the specimens in this work and Mr. David Abbott for the long hours spent managing volumes of data. I would also like to thank Dr. Sharon Sun for her assistance in completion of this work and wish her great success as she continues work on this project. I would also like to extend a special thank you to the staff of the Cancer Imaging department and Oncometrics Imaging Corp. for each of their individual contributions which made this work possible. The diversity of their interests and life experiences have provided me the opportunity to broaden my knowledge and my personal perspective. I enjoyed working with them in a friendly and cooperative work environment each day. Finally, I would like to thank my family and friends for their patience and support and for the imporant element of balance they bring to my life. ix Wilton 1.0 INTRODUCTION Malignancy Associated Changes (MAC) can be defined as subtle morphologic and or physiological changes which occur in ostensibly normal cells of patients harboring malignant or pre-malignant disease. These changes have been observed in histologically normal, diploid cells, both proximal and distal to malignant lesions in the majority of cancer patients. Although first reported by Gruner eight decades ago^, this phenomenon remains poorly understood and poorly exploited. 1.1 MALIGNANCY ASSOCIATED CHANGES : HISTORICAL PERSPECTIVE 1.1.1 QUALITATIVE IDENTIFICATION OF N U C L E A R M A L I G N A N C Y ASSOCIATED CHANGES The term Malignancy Associated Changes was first utilized by Nieburgs in the early 1960's to describe subtle changes in the nuclear organization of cells taken from the buccal mucosa in patients harboring distant malignancies . By viewing individual Feulgen stained cells with a light microscope at a total magnification of 20 000 times Nieburgs identified changes in the chromatin condensation pattern of normal cells of cancer patients and established the criteria for qualitative identification of M A C . These criteria include: 1. A n enlarged nucleus 2. Heterogeneous nuclear material with numerous (more than 10) small, round, clear areas of 0.5 to 1 micron in diameter. 1 Wilton 3. Clear areas of low density chromatin surrounded by semi-circular rings of denser, more darkly stained chromatin 4. Discontinuous Nuclear Membrane 5. A n absence of nucleoli and multinucleated cells. In subsequent studies Nieburgs applied this qualitative criteria to the identification of M A C in ostensibly normal cells from duodenal drainage fluid, uterine cervix, liver biopsies, skin of patients with internal tumors, polymorphonuclear leukocytes and lymphocytes from peripheral venous and capillary blood and in megakaryocytes in the 2 8 bone marrow. " Interestingly the morphological changes associated with malignancy were found to be consistent in all tissue types (with the exception of polymorphonuclear cells) and were independent of the site and histopathology of the tumors 5 . The majority of patients in these studies had advanced, often metastatic cancer. As summarized in Table 1.1, Nieburgs and a few of his contemporaries were relatively successful in the identification of M A C in patients with malignancy. In various trials, Nieburgs achieved sensitivities ranging from 81% to 90% and specificity's ranging from 82% to 96% ' ' . Although these results were significant, the majority of the studies were not conducted in a blind manner and did not set clear criteria for the number of M A C positive cells required for positive sample identification. Other investigators applied the criteria set out by Nieburgs and identified M A C with variable success. Romsdahi et al. reported nuclear clearing, and chromatin clumping which he likened to M A C in the megakaryocytes of patients with disseminated cancer9. Finch et al. later confirmed the 11 12 presence of M A C in the buccal mucosa of patients with various malignancies ' . 2 Wilton AUTHOR YEAR MAC SITE TUMOR SITE SPECIFICITY /SENSITIVITY Nieburgs 1959 Buccal Mucosa Various Nieburgs 1962 Duodenal Drainage Fluid Various Nieburgs 1962 Buccal Cell Mucosa Various 96.8%/81% Elias 1962 Liver Liver Romsdahi 1963 Megakaryocytes Various Nieburgs 1963 Cervix Cervix Howdon 1964 Cervix Cervix Nieburgs 1965 Liver Liver/Various 94.6% / 90% Kwittken 1966 Skin Various Nieburgs 1967 Venous Blood Various 82%/79% Capillary Blood Various 85%/79% Bone Marrow Various 88%/82% von Haam 1967 Monocytes Various Mattson 1967 Monocytes Various Nieburgs 1968 Cervix Cervix Clausen 1968 Leukocytes Various Martuzzi 1970 Sputum Lung Johnston 1970 Leukocytes Various 91.8%/88.8% Finch 1971 Buccal Smears Various 96.5%/50% Klawe 1972 Buccal Smears Various 85.8%/76.2% vanOppenToth 1975 Sputum Lung 76%/81.3% Table 1.1: Measurement of Malignancy Associated Changes by qualitative criteria. In this blinded study, he achieved a sensitivity of 50% and a specificity of 96.5%. Although Finch concluded that the subtle variation of M A C made it very difficult to use the qualitative test as a diagnostic indicator, he confirmed that the proportion of patients displaying M A C was significantly higher than that of the controls n . The poor SENSITIVITY achieved by Finch's group may be attributed in part to the difficulty in interpreting the subtle morphological changes based solely on a qualitative description communicated in the literature. Furthermore, in this study, the specimens were stained with the Papanicolaou process rather than the Feulgen process used by Nieburgs. The Papanicolaou staining procedure results in a poor resolution of the spatial distribution of chromatin within the nucleus, thereby limiting the ability to visualize M A C . Wilton Sputum smears were subsequently examined by van Oppen Toth et al. using the qualitative criteria outlined by Nieburgs. This study observed a sensitivity of 81.3% and a specificity of 76% 1 0. Klawe also used the qualitative criteria to achieve a sensitivity of 76.2% and a specificity of 85.8% in the buccal smears of patients with distant 13 malignancies . A n additional study by Johnston and Brady found a sensitivity of 88.8% and a specificity of 91.8% using M A C in peripheral leukocytes of patients1 4. The subtle changes observed in M A C cells were difficult to articulate and despite the seemingly meticulous detail outlined in Nieburgs' criteria for M A C , his results were difficult to reproduce by other researchers without substantial experience. Although some groups achieved moderate success, there remained much skepticism with respect to the validity, and much less the utility, of any diagnostic procedure based on the subjective identification of M A C . 1.1.2 QUANTITATIVE IDENTIFICATION OF N U C L E A R M A L I G N A N C Y ASSOCIATED CHANGES The existence of M A C was first proven in an objective way in 1974 by Klawe and Rowinski 1 5 . Using a simple quantitative image analysis system, this group measured the pixel area of hematoxylin stained nuclei at several light threshold levels. Using this data the researchers computed a parameter K which effectively measured the degree, and to a lesser extent, the distribution of D N A condensation in the nucleus. Using this quantitative parameter they demonstrated that there was a statistically significant Wilton difference between the D N A distribution in epithelial cells in buccal smears from healthy subjects or sick non-cancerous children versus those children who were afflicted with a cancerous disease. Furthermore, the parameter K changed in a manner consistent with Nieburgs' qualitative description of M A C . Thus, for the first time, M A C were identified using an objective means. Subsequent to Klawe and Rowinski, a number of investigators demonstrated the existence of M A C in cancer patients; their work is summarized in table 1.2. As summarized in this table, the majority of studies employing quantitative image analysis in the identification of M A C have been carried out using tissue samples from the cervix. Wied et al. reported finding significant changes in apparently normal intermediate cells in patients with malignant disease or even dysplasias of the cervix in 1980'°. This study and multiple subsequent studies confirmed that computed features describing nuclear shape and most prominently, nuclear texture (measuring chromatin condensation patterns) can be combined in a multifactoral analysis to effectively discriminate between normal cells found in healthy patients and normal cells found in patients with cervical dysplasia or • ™ 16-29 carcinoma In the last decade image cytometry has been utilized to demonstrate that M A C can be detected objectively in many other tissues including sputum cytology 3 3 ' 3 4 ' 4 0 , colon 35 36 37 38 39 mucosa ' , bronchial biopsies , breast , and laryngeal and pharyngeal smears . The work in these tissues coupled with that done in the cervix provides significant evidence 5 Wilton that objective differences exist in the nucleus of many apparently normal cells (i.e. those cells having normal 2N D N A amount, are normal in size and shape and histologically normal tissue ) found in tissues harboring malignancy. AUTHOR YEAR MAC SITE TUMOR SITE C E L L / C E L L SPECIFICITY / SENSITIVITY SAMPLE SPECIFICITY / SENSITIVITY Klawe 1974 Buccal Smear Various (+met) Wied 1980 Cervix Cervix 87.8%/77.8% /100% Dysplasia /80% CIS /100% Invasive Burger 1981 Cervix Cervix Bartels 1982 Cervix Cervix Vooijs 1982 Cervix Cervix Reinhardt 1982 Cervix Cervix King 1982 Breast Breast Rosenthal 1983 Cervix Cervix Sherman 1983 Bladder Bladder Wied 1984 Cervix Cervix 94 %/ 89% Boon 1986 Cervix Cervix Katzko 1987 Cervix Cervix 86% / 68% Haroske 1988 Cervix Cervix 77% / 52% Bibbo 1989 Colon Colon *76% / 75% 68.8%/65.8% Swank 1989 Sputum Lung *83 %/77% 83% / 56% Bibbo 1990 Breast Breast Haroske 1990 Cervix Cervix Zhanister 1991 Cervix Cervix 81.2%/87.8% Rieth 1991 Nasal Mucosa Nasal Mucosa Montroni 1991 Prostate Prostate Hutchinson 1992 Cervix Cervix 79% / 73% Palcic 1992 Cervix Cervix MacAulay 1994 Bronchial Biopsy Lung 89% / 86% Palcic 1993 Breast Breast Burger 1994 Buccal Mucosa Larynx / Pharynx Wilton 1995 Tissue Culture Tissue Culture 72% / 72.% 96% / 93% Bronchial Epith. Lung Payne 1996 Sputum Lung 78.8% / 77.5% Table 1.2: Measurement of MAC using quantitative image analysis •training set Wilton 6 1.1.3 N O N - N U C L E A R M A L I G N A N C Y ASSOCIATED CHANGES The term Malignancy Associated Changes has traditionally been used in the literature in reference to alterations in nuclear morphology. Various methods have also been used to demonstrate important staining, ultrastructural, histochemical and physiological differences in the tissue adjacent to and remote from cancerous lesions without evidence of cytological atypia. Nieburgs original definition of M A C included an increased nuclear to cytoplasmic ratio. This increase was attributed mainly to and increase in the nuclear area however a measurable decrease in the cytoplasmic area was also noted5. A recent study conducted by Ogden et al. found a significant reduction in the cytoplasmic area in the normal oral mucosa of patients with distant malignancy50. Other significant alterations observed in normal cells associated with malignancy include cytoplasmic "halo's" in mononuclear cells 1 4, subtle spectral shifts in cytoplasmic staining of Pap stained cervical intermediate cells 1 6 and an increased and more variable orangeophilic staining in exfoliated epithelial cells of the sputum. Saffos and Rhatigan reported consistent ultrastructural changes in the normal colonic mucosa adjacent to cancerous lesions. These changes included an increased proportion of goblet cells over absorptive cells in the surface epithelium, an increased depth in the crypts of the colon and an increased cell number and density in the crypts41. Filipe et. al. also noted immature and intermediate cells at higher levels of the crypt throughout the Wilton colon of cancer patients4^. Other studies have reported a shift in the mucous production of the normal epithelium of cancer patients from the normally predominant sulfomucins to sialomucins 4 1 " 4 3 . The altered mucin production of M A C mucosa resembles the mucous secretion pattern of the fetal colon 4 3. Caldero et al. have further demonstrated an altered expression of membrane glycoconjugates in the apparently normal mucosa in patients with colon carcinomas44. Several other studies indicate that there is an abnormal cell proliferation in the normal mucosa associated with malignancy. Terpstra et. al. found a significantly increased uptake of H-thymidine in colon crypts throughout the whole colon of patients with colon cancer45. Ngoi et. al. reported more than two times greater proliferative activity (S-phase fraction) in the normal mucosa up to 10 cm away from aneuploid tumors as measured by flow cytometry46. Finally, several studies report increased proliferating cell nuclear antigen (PCNA) expression in normal tissues adjacent to breast, liver and pancreas47"49. Thus, there is substantial evidence that significant utlrastructural and physiological changes occur in the ostensibly normal tissue associated with malignancy. 1.2 CLINICAL SIGNIFICANCE OF MAC It has long been recognized that detection of cancer in its early, noninvasive stage is one of the most effective ways to control the disease. This is especially true in cancers having epithelial origin, which have been demonstrated to have detectable pre-clinical phases Wilton that are locally confined. The majority of these lesions have prolonged precancerous periods of development and are well suited to successful clinical intervention. The potential of early detection programs is demonstrated by the success of mass screening programs for cervical cancer implemented in many developed countries over the last half century. In British Columbia a comprehensive cervical cancer screening program was implemented in 1955. As a result of early detection and clinical intervention , the incidence rate of invasive squamous carcinoma of the cervix has been reduced by 85% and the mortality rate has been reduced by 80% over the period between 1956 to 19885 1in B.C.. Strategies for the early detection of malignancies at other sites include regular palpation and mammography examinations for breast cancer, sputum cytology, chest x-rays and bronchoscopy for lung cancer, and non-invasive imaging techniques such as MRI and ultrasound, as well as use of needle aspirates for prostate and bladder cancer52. 1.2.1 M A C AS A N INDICATOR FOR THE E A R L Y DETECTION OF C A N C E R It has been proposed that the M A C phenomenon can be employed as a test for the detection of a wide variety of early cancers. In recent years several studies have demonstrated that M A C in ostensibly normal cells of cervical smears contribute information that when utilized in concert with diagnostic cell methodologies increases the effectiveness of detection of malignant and premalignant lesions as well as sample 30 32 classification over that of diagnostic cell classification methodologies alone " . This Wilton increased detection and diagnostic capability is derived from features extracted using computed cell images present in ostensibly normal cells which can not be readily perceived by human vision. Furthermore, diagnostic information is available in ostensibly normal cells which may indicate the presence of clinically relevant lesions in the instance where clinical sampling has failed to secure diagnostic cells. This second point is illustrated by the fact that approximately 10% of false negative cervical smears result from poor sampling technique by the primary physician resulting in a complete absence of diagnostic cells. It is possible that a significant number of these samples would be detected with M A C analysis despite the absence of diagnostic cells. The case for utilizing M A C detection in sputum testing for lung cancer is even stronger. Sputum cytology is the only non-invasive method that can detect pre-malignant lesions or carcinoma in situ in the tracheobronchial tree. Sputum cytology as performed by conventional methods has a very low yield of diagnostic cells. One hypothesis states that this inadequacy results from the fact that sputum cytology mainly detects centrally located squamous carcinoma. These tumors spread superficially and exfoliate cells readily into the bronchial lumen. However, large cell carcinoma and adenocarcinoma are usually located more peripherally in the smaller airways. They spread submucosally and tend to obstruct the bronchial lumen in the early stages. Thus, it becomes difficult to force air past these lesions and they do not exfoliate many cells until they become larger and invade the more central airways; at which point treatment outcome is poor. As a 1 0 Wilton result conventional sputum cytology has traditionally achieved a sensitivity of less than 30% for early stage lung cancer34. It has been postulated that the sensitivity of sputum cytology can be greatly enhanced with the utilization of M A C as these detection methods do not require traditional diagnostic cells for the detection of cancer. In a study conducted on sputum samples from the Mayo Clinic portion of the N.C.I. Cooperative Early Lung Cancer Study5 4 , M A C were detectable in sputum up to one year prior to any other clinical detection of cancer34. This study conducted by Payne et al. estimated that by setting the automated detection of M A C at a specificity of greater than 90% to minimize the cost of unnecessary bronchoscopies, M A C alone in sputum samples presenting no diagnostic cells would have a sensitivity of 20% over and above the conventional detection by sputum cytology. Although this figure may seems low, achievement of even such a conservative goal may result in the detection of approximately 30 000 cases of early lung cancer that would have otherwise gone undetected by conventional means each year in the United States alone34. Again, using conservative estimates; based on a 76% Stage I, 25% Stage II, and 7% stage Ilia and Illb five year survival rate these investigators propose that the utilization of M A C in a mass screening protocol could result in 10 000 lives saved annually34. Wilton 1 1 1.2.2 M A C AS A PROGNOSTIC INDICATOR OF C A N C E R It has been demonstrated that only 5-20% of mild dysplasia, 10-50% of moderate dysplasia, and 20-80% of severe dysplasias of the cervix will progress to malignant 52 disease i f left untreated . Therefore, detection of pre-malignant lesions presents clinicians with the difficult dilemma of how to interpret and respond to such poorly understood lesions. A significant proportion of premalignant lesions of the lung have 52 also been observed to regress spontaneously without treatment . Thus, it is clear that in order to avoid over treating patients not at risk of developing malignant disease it is very important to find a method to determine the malignant potential of their early lesions. Palcic et al. have postulated that nuclear texture features measured from both diagnostic cells and ostensibly normal cells of patients with cancer (MAC cells) can be used to 52 56 predict the progression/regression of precancerous lesions ' . In a study conducted by Anderson et al., 60 cases of cervical intraepithelial neoplastic (CIN) high grade lesions (CINII and CIN III) were studied in women who refused treatment, but were followed with cytology. Approximately one third of these women later developed invasive cancer, while two thirds regressed spontaneously. Using nuclear texture measurements taken from M A C cells, the progression/regression status of the lesion was successfully 52 predicted for 80%) of the cases. Similar results were achieved using diagnostic cells. In a subsequent study, this group observed a monotonic increase or decrease in cytometric features which correlated with the grade of cervical dysplasia55. It was therefore proposed that features measured in normal cells could be used to confirm diagnosis of the 12 Wilton grade of the lesion and could be employed in multiple consecutive samples to monitor the progression/regression of the lesion. (§) Resected O Non-Smoker • Ex-Smoker A Current Smoker • Invasive Cancer A i fiiRL • A CIS • f*Threshold • LA AD A ° • Dysplasia 1 ° n High Risk Son-Cancer • ~Q3 Low Risk Non-Cancer Q -2 -1 • 0 1 2 MEAN MAC SCORE Figure 1.1: MAC scores in bronchial biopsies taken from apparently normal regions of patients having dysplasia of cancer of the lung. Figure reproduced with permission from MacAulay C , Lam S. et al.: Malignancy associated changes in bronchial epithelial cells in biopsy specimens. Analyt. Quant. Cytol and Hist. 17(1):55-61, 1994. Additional indications of the prognostic value of the M A C phenomenon comes from studies on bronchial biopsies taken from apparently normal regions of patients having dysplasias or cancer of the lung37,93. Data from this study conducted by Payne and MacAulay et al. are shown in Figure 1.1. This figure illustrates the observation that as the severity of the lesion increases so does the strength of M A C expression in the normal tissue associated with it. Furthermore, within the dysplasia category the expression of M A C in normal tissue samples ranges from very weak to very strong. On follow up, six patients who originally had mild dysplasias of the lung but displayed strong M A C were later found to have more severe progressing lesions. From this data it was postulated Wilton 13 that those early lesions associated with normal cells displaying strong MAC may be more likely to progress than those with weak or non existing MAC. These lesions would then be closely monitored, entered into chemoprevention protocols or treated using low morbidity localized treatment methods such as photodynamic therapy, laser therapy or cryotherapy. Similarly, those early lesions displaying weak MAC may be considered 37 benign reactive changes and be marked for less frequent follow up . The ability to grade lesions and to determine their progression/regression status would have significant implications on the patient management strategy. In addition, this information would be very useful to gauge the effectiveness of chemoprevention trials. Furthermore, MAC may be employed to monitor the success/failure of treatment protocols as this phenomenon can provide an indication whether the lesion is responding to radiation or chemotherapy and also provide early warning signals of recurrence after apparent regression or cure of the cancer. 1.3 PROPOSED MECHANISMS OF MALIGNANCY ASSOCIATED CHANGES Despite the broad body of research employing the MAC phenomenon in the detection of cancer, the causal relationship of malignancy to MAC has not been clearly demonstrated. There are essentially two conflicting hypotheses which have gained favor in recent years which attempt to explain the mechanism of MAC. Wilton 1.3.1 FIELD CANCERIZATION THEORY OF M A C Field Cancerization is a theory originally proposed by Slaughter in 1953 to explain his observations of multiple epithelial tumors arising in the aerodigestive tract 86 . A natural extension of this theory has gained favor to explain the mechanism of M A C . This theory postulates that M A C arise from a concurrent exposure of normal tissue to the environmental carcinogen which caused the neoplastic lesion. In this scenario mutations 57 occur in the normal tissue as well as the tissue which eventually develops into cancer . Although the normal tissue has accumulated mutations, they are not sufficient for neoplastic transformation, thus this ostensibly normal appearing tissue, having acquired genetic mutations is detected as M A C . Nieburgs first proposed this mechanism; speculating that the localized atypical dysplasia observed in the cervix resulted from the same carcinogenic factor that induced the systemic cellular change identified as M A C . He believed that the local manifestation of malignancy was due to prolonged exposure to the same or other carcinogens as those which caused M A C 4 . It is unclear what Nieburgs is referring to as the systemic carcinogenic factor. He may have been hinting at viral infection when he postulates that the systemic M A C may constitute the disease entity of cancer with tumor growth as a local manifestation of the generalized cell disorder6. Haroske et al. also postulated that M A C may be pre-neoplastic changes in cells which appear normal^. Bibbo and Montag noted the fact that the feature values of M A C cells 15 Wilton changed toward the values typically found in nuclei from the malignant lesion 3 3. They also found that some cells located distally to the malignant lesion changed so drastically they were diagnosed as malignant cells based on their nuclear features. Although they did not commit to a hypothesis this group acknowledged that these observations may be interpreted as evidence supporting a field cancerization theory. The non nuclear changes associated with malignancy have also been attributed to a field cancerization effect. Shamsuddin et al. observed substantial morphological changes in the apparently normal mucosa (no evidence of cellular atypia) adjacent to and distant from malignancies of the colon^^. They also measured a shift in the predominance of sulphomucins to sialomucins in the secretions of colons harboring malignancy. They stated that these abnormalities (both morphological and histochemical) were identical to the presumed preneoplastic changes in rat colon epithelium after in vivo and in vitro carcinogen treatments. As a result, they concluded that the alterations observed in the transitional mucosa adjacent to neoplastic lesions of the colon, and those changes occurring to distal regions of the colon are due to carcinogenic stimuli rather than being nonspecific or secondary to the presence of tumor43. Further evidence that the morphological alterations measured as M A C may be subvisual physical manifestations of genetic damage resulting from exposure to carcinogen can be found in a recent study by MacAulay et al^^. In this study, the values of computed features from the images of 12 biopsies having one or more of 3p 21-22, 5q 21, and 9p Wilton 21-22 chromosomal deletions were compared to those from biopsies having no detected deletions. This study demonstrated that using only two features it was possible to correctly identify more than 80% of the samples as having deletions or not. This is an important observation as it indicates that a sophisticated quantitative image cytometry device is sensitive enough to measure genetic mutations which may not result in 58 morphological manifestations detectable by human eyes . Thus, it is possible that the subvisual alterations which have been referred to as M A C may in fact be early genetic damage caused by exposure of the apparently normal cells to the same carcinogen which was responsible for initiation of the neoplastic lesion. 1.3.2 H U M O R A L M E C H A N I S M OF M A C : WORKING HYPOTHESIS The working hypothesis of this thesis is referred to as "the Humoral Mechanism of M A C " . This theory postulates a direct cause and effect relationship of malignancy to M A C . We propose that the morphological manifestation in ostensibly normal cells called M A C results from signals released by malignant or pre-malignant lesions. These signals are in the form of soluble chemical mediators (e.g. growth factors , cytokines, eicosanoids et.) which act on the ostensibly normal tissue surrounding neoplastic lesions to alter the compliment of actively expressed genes. The altered gene expression is manifested as changes in the chromatin condensation pattern and therefore the morphology of ostensibly normal cells in the vicinity of cancer. Wilton 17 H U M O R A L THEORY OF M A C : RATIONALE The most compelling line of evidence to support the humoral theory of M A C comes from studies which observe that M A C expression is greatly reduced or disappears following curative surgery. The first important observations of this kind were made by Nieburgs et al.. They found that the expression of M A C was greatly reduced following surgical 2 7 removal of the tumor ' . In a study of the buccal mucosa of patients with various tumors Nieburgs found 81% of patients presented M A C pre-operatively and only 53% of post operative patients presented M A C . This phenomenon was also observed by Payne and MacAulay et al. in a recent study of M A C in the bronchial biopsies of cancer patients^ 93. Figure 1.1 summarizes the results of this study. From this figure it can be seen that the strength of M A C expression was greatly reduced following surgical resection of the tumor. Furthermore, only one patient in this group presented a M A C score above the set threshold. This patient was later found to have a residual cancer at the resection margin on a follow-up bronchial examination. An additional study conducted by Payne and Katsumi93 measured the expression of M A C in the sputum of patients pre and post operatively. In this study a statistically significant reduction for the expression of M A C was measured post operatively for patients having no cancer recurrence or metastasis for at least 30 months. Interestingly, those patients who were found to have metastasis and or recurrence within this time period were found to have no reduction in the expression in M A C in their sputum samples following surgical resection of the primary tumor. The disappearance of M A C following removal of the tumors of multiple patients in independent studies suggests a cause and effect relationship between cancer and M A C . 18 Wilton Additional support for the humoral mechanism can be found in the study of M A C in the 35 36 colon conducted by Montag et al. ' . This study demonstrated that the frequency of cells diagnosed as M A C was highest in apparently normal cells immediately adjacent to malignant lesions of patients colons and decreased as cells were measured at increasing 36 distances radially from the tumors . A similar trend was measured for many nuclear features important in diagnosing M A C cells. For these features (nuclear area, nuclear texture features) the mean value approached that of the tumor region immediately adjacent to the tumor and moved rapidly in the direction of control tissue values as a function of distance from 2mm to 1 Omm from the tumor margin. At 10mm away from the tumor a nearly constant plateau is reached which extends out to the farthest sampling site, at 50mm. For all the features the mean values observed at the most distal site still did not reach the values recorded for the controls. These results are supported by an earlier study conducted by Bibbo et al. in the cervical epithelium26. j n this study M A C were measured at lOOum, 200um, 300um, and 400um from the visual edge of the lesion into histologically normal tissue. As observed in the colon, the frequency of M A C cells correctly classified decreased with increasing distance from the lesion. These studies provide strong evidence that the local diffusion of one or many soluble chemical mediators released by the malignant lesion act in a paracrine manner on the local tissue inducing a percentage of cells to express the M A C phenotype. Wilton Further evidence supporting the humoral mechanism for M A C can be found in studies completed by Hall et al. on the expression of P C N A in normal tissue surrounding tumors^?. As previously mentioned, in several organs, including breast, liver and pancreas normal tissues adjacent to tumors show high levels of PCNA expression. Hall 48 suggests that in the majority of cases this expression is unrelated to proliferation . Using two model systems this group demonstrated that PCNA is induced in non-cycling cells by adjacent transplanted tumor cells and that this phenomenon can be mimicked by the in vitro administration of transforming growth factor alpha (TGFoc) and epidermal growth factor (EGF) 4 7 . This data suggests that tumors produce growth factors which alter the P C N A expression of the normal adjacent cells by paracrine stimulation. In 1980 Sporn et. al. introduced the concept that malignant cells produce growth factors which act on their own receptors59. This autonomous growth stimulatory mechanism was termed "autocrine growth stimulation"60 and has gained widespread acceptance as the knowledge of the pathogenesis of cancer has expanded. Since that time, many tumor cell lines from multiple tissue types, as well as primary cultures from patient tumors, have • 61*73 been demonstrated to produce supra-physiological levels of growth factors " . The altered growth factor production by tumor cells may play a pivotal role in the pathogenesis of malignancy by providing a selective growth advantage to the malignant cells via an autocrine loop. Alternatively, growth factor production may simply be a by product of the altered gene expression which occurs in neoplasia, conferring no growth advantage to the neoplastic cells themselves but having important local and systemic Wilton consequences. Inevitably, stromal cells located in the same micro environment having receptors for such growth factors will be affected in a paracrine manner. Of particular interest to the model system employed in this thesis, Non Small Cell Lung Cancers (NSCLC) have been shown to produce EGF (50% of primary adenocarcinomas), TGF-a (85-100% of NSCLC) , TGF-P, PDGF and Insulin-like GF 6 1 " 6 3 . As our knowledge of normal human physiology grows it is becoming clear that a delicate balance of negative and positive signals is at work regulating the expression of individual genes as well as the movement of quiescent or non-dividing cells into and out of the cell * 73 cycle in appropriate tissues at appropriate times . It is probable that disruption of this balance by elevated cytokine production by malignant lesions will result in an alteration of the morphologic and physiologic status of many normal cells exposed to the imbalance of signals. The morphologic alterations in the form of altered nuclear texture are those which are measured as M A C . In addition to the experimental data which demonstrates the local paracrine effects of growth factors released by tumors on normal tissue, there is a great deal of clinical experience which has documented profound systemic effects of tumor on the host. Systemic affects of cancer which are not attributable to hormones indigenous to the tissue of origin of the tumor are referred to as paraneoplastic syndromes*^. Table 1.3 lists several paraneoplastic syndromes and their causal mechanisms. As this table illustrates, many systemic manifestations of malignancy can be attributed to inappropriate Wilton growth factor release by the tumor directly or indirectly by induction of immune mechanisms resulting in cytokine release. Clinical Syndromes Major Forms of Underlying Cancer Causal Mechanism Fever Hodgkin's disease Renal cell carcinoma Osteogenic sarcoma others 11-1 TNF Immune Mediated Anorexia/Weight Loss many TNF-a? Endocrinopathies Cushing's Syndrome Hyponatremia Small cell lung cancer (SCLC) Pancreatic carcinoma Neural tumors SCLC Intracranial Neoplasms A C T H or ACTH like substance Antiduruetic hormone or Atrial Naturetic Factor Hypercalcemia SCLC Breast carcinoma Renal carcinoma PTH-like substance, TGF-a, Vitamin D Carcinoid syndrome Bronchial Carcinoid Pancreatic Carcinoma Gastric Carcinoma Sertonin, Bradykinin, Histamine? Polycythemia Renal Carcinoma Cerebellar Hemangioma Hepatocellular Carcinoma Erythropoietin Nerve and Muscle Syndromes Disorders of central and peripheral nervoes systems Myasthenia gravis SCLC Breast Carcinoma Thymoma Immunologic?, Toxic? Osseus, articular, and soft tissue changes Hypertrophic osteoarthropathy 'Polymyalgia rheumatica *Palmar fasciculitis •Most autoimmune CT disorders •Antiphospolipid Syndrome Clubbing of the Fingers Adenocarcinoma of the Lung hormones? immunoglobins? Vascular and hematologic changes Venous thrombosis (Trousseau's phenomenon) Pancreatic carcinoma Lung Carcinoma, Others Hypercoagulability Nonbacterial thrombotic endocarditis Advanced Cancers Hypercoagulabil ity Table 1.3: PARANEOPLASTIC SYNDROMES Adapted from: Kumar V, Cotran RS, Robbins SL: Basic Pathology, 5th ed. Philadelphia, W.B. Saunders Company, 1992. *Naschitz J.E., Yeshrum D.: Cancer-associated rheumatic disorders. The Cancer Journal 9(4), 1996. This phenomenon is classically revealed in a disorder appropriately called Malignancy Associated Hypercalcemia69. This disorder has both humoral and paracrine etiologies. Elevated serum levels of calcium have long been associated with a variety of Wilton 22 malignancies. Epidermoid (including NSCLC), renal cell and breast carcinomas have been found to release a novel peptide similar in sequence to parathyroid hormone (PTH) called parathyroid related peptide (PTHrP) into the systemic circulation69. This peptide is normally expressed during fetal development and is important in formation of the skeleton. When abnormally expressed in the adult it activates osteoclast activity thereby increasing bone resorption and elevating serum calcium. Furthermore, in the case of many lung and breast carcinomas, metastases to the bone release factors such as IL-1, TNF-oc and TNF-[3, prostaglandins, leukemia inhibitory factor, PTHrP, TGF-a which act in a paracrine manner activating osteoclasts with similar results69. Another common effect of cancer on the host is a symptom complex referred to as cachexia. This wasting syndrome is characterized by anorexia and weight loss. Although the patient often has decreased caloric intake this is not sufficient to explain the profound wasting. In addition, unlike starvation, which is associated with a lowered metabolic rate, cancer is often associated with an elevated metabolic rate. The most plausible explanation for these observations is a release of growth factors IL-1 and TNF-oc directly by the tumor or by tumor induced activated macrophages. TNF-a has been shown to reduce appetite, inhibit lipoprotein lipase, induce hormone sensitive lipase and down regulate transcription of key enzymes in fatty acid synthesis. As a result, fat storage is inhibited and fat stores are depleted. Furthermore, long term TNF-a administration has been shown to produce a wasting syndrome in experimental animals 84. This a n d other common paraneoplastic syndromes establish precedent that neoplasia commonly 23 Wilton produces growth factors which have profound local and systemic effects. Therefore, there is a great deal of current medical knowledge which provides affirmation of the humoral theory of M A C . 1.4 PROJECT OBJECTIVES Despite the growing body of evidence which supports the existence of Malignancy Associated Changes, the clinical significance of this phenomenon remains highly unexplored, primarily due to a lack of understanding of it's mechanism. If M A C are merely due to a field cancerization effect the relation of the observed morphological changes will be at best indirectly related to any neoplastic lesions which may be present. If this is true, the utility of this phenomenon in cancer screening will be reduced to a probability indicator as it will only indicate the extent to which the tissue as a whole has been exposed to carcinogen. Furthermore, under this scenario M A C will be of no utility in predicting the prognosis or monitoring treatment protocols of cancer patients. However, i f a cause and effect relationship between malignancy and M A C can be established one could postulate (as previously outlined) that M A C may be useful in the detection, diagnosis, prognosis and monitoring of treatment outcome of cancer patients. The multitude of in vivo studies published to this point provide indirect evidence which can be interpreted to support either of the two main theories of the mechanism of M A C . It is clear that in order to truly understand this phenomenon and utilize it in a meaningful way clinically, the mechanism of M A C must be understood. To accomplish this goal a Wilton 24 well designed, controlled in vitro model must be available. Thus the objectives of this thesis project are as follows: 1. Determine i f M A C can be induced in vitro 2. Develop a method which can be used to demonstrate the causal relationship of malignancy to M A C 1.5 PROJECT STRATEGY This thesis will describe the design and application of an in vitro model of M A C . This model will mimic the interaction between malignant cells and the normal tissue surrounding it in vivo. This model will be designed such that the tumor cell line and normal cells employed are both capable of growth in serum free medium and such that any communication between the cell types must occur through soluble chemical mediators. The model will be subsequently applied to demonstrate the induction of M A C in vitro by souluble factors released by the tumor cell lines into the culture medium. The causal relationship of malignancy to M A C in this model will be further investigated by determining the time course and dose dependency of this phenomenon. Appropriate control experiments to ensure that metabolite composition is adequate in the conditioned medium and that statistical analysis is sound will be completed. Wilton 25 2.0 MATERIALS AND METHODS 2.1 PRIMARY CULTURE OF HUMAN BRONCHIAL EPITHELIUM In a pilot study bronchial biopsy tissue was investigated as a source of primary cultured cells. Although the cells isolated with this protocol were not utilized in the MAC studies shown in this thesis, future work using the in vitro model may employ cells isolated from patient samples using the outlined protocol. Human bronchial epithelial cells were cultured from bronchial biopsies provided by Dr. S. Lam. The biopsies were obtained from consenting patients during white light and fluorescent bronchoscopic examination. Approval for this work was granted by a ethics review committee. Dr. W.A. Franklin of the Department of Pathology, The University of Colorado provided the protocol for the culture of human bronchial epithelium (personal communication). This protocol was 77 based on techniques developed by Lechner et. al. . 2.1.1 GROWTH SURFACE PREPARATION Stock solutions were prepared in advance as follows: Collagen stock- 0.1% (Sigma Chemicals), Fibronectin stock- 200ug/ml (Sigma Chemicals) prepared by asceptically adding 25ml warm sterile water to 5mg of sterile lyophilized fibronectin, store at -20 C. Bovine Serum Albumin (BSA) stock- lmg/ml (Sigma Chemicals) prepared by dissolving lOmg BSA into 10ml of Bronchial Epithelial Basal Media (BEBM)(Clonetics Corp, SanDiego). Wilton Coating medium was prepared fresh by combining 18.5ml of B E B M , 200ul of B S A stock, 1ml Fibronectin stock (thawed rapidly) and 1.5ml Collagen stock. Glassware/Plastic ware was coated for 12 hours at 37 C, 5% C O 2 and 98% humidity. 2.1.2 E X P L A N T C U L T U R E Using aseptic technique, fresh biopsy tissue of approximately 27mm3 was placed in 10ml cold B E B M supplemented with antibiotic/antimycotic solution (500u/ml). Biopsy tissue was transferred from B E B M to 1.5ml warm bronchial epithelial growth medium (BEGM) ( B E B M supplemented with 0.5ng/ml human recombinant Epidermal Growth Factor (EGF), 0.5 ug/ml Hydrocortisone, 5ug/ml Insulin, 10 ug/ml Transferrin, 0.5ug/ml Epinepherine, 6.5ng/ml Triidothyronine, 50ug/ml Gentamicin, 50ng/ml Amphotericin-B, Bovine Pituitary Extract and O.lng/ml Retinoic Acid)(Clonetics Corp, SanDiego) in 6 well corning tissue culture plates (Gibco B R L , Burlington). Biopsies were incubated at 37 C, 5% C O 2 and 98% humidity. The medium was changed on the fourth day after plating and every third day thereafter without disturbing the biopsy tissue. Successful primary culture from biopsy resulted in a radial growth of bronchial epithelium from the biopsy margin. At least 1 X 10 °" cells were required for successful passage in to traditional 75 cm^ tissue culture dishes. Wilton 27 2.2 MODELS OF NORMAL AND TUMOR C E L L INTERACTION The in vitro model systems employed in this thesis were designed to mimic the in vivo interaction between normal tissue and tumors. Specifically, these model systems were designed to mimic the interaction between non-small cell lung cancer (NSCLC) and the normal pseudostratified epithelium of the respiratory tract. For this model, primary cultured normal human bronchial epithelial cells (NHBE) were purchased from Clonetics Corp. of SanDiego California. These cells are readily expandable and do not require the 20 day lag time to acquire sufficient numbers of cells as was experienced with primary culture from bronchial biopsies. They also grow in a serum free medium (BEGM) supplemented with appropriate defined growth factors (see section 2.1.1). This factor was considered important as this model is intended to be used in subsequent studies involving the isolation and purification of unknown growth factors. The N H B E cells have a relatively well differentiated morphology in culture and are growth inhibited by cell to cell contact. Two N S C L C cell lines (NCI-H460, NCI-H358) employed in this model were isolated and donated to this project by Dr. A. Gazdar of Texas Southwestern Medical Center, Houston, Texas. These lines are available at the American Tissue Culture Collection. The NCI-H460 cell line was derived from the pleural fluid of a male with large cell carcinoma. It expresses wild type p53 and has a codon K61 ras mutation (CAT)78. NCI-H358 was isolated from a primary bronchioalveolar carcinoma of the lung. This cell line 28 Wilton has a codon K12 ras mutation (TGT) and a homozygous deletion for p53. Both tumor cell lines were isolated and have been cultured in RPMI 1640 medium supplemented with 5% fetal bovine serum (FBS) (Gibco, Burlington). 2.2.1 INDIRECT CO-CULTURE M O D E L The indirect co-culture model is shown graphically in Figure 2.1. This model was designed to investigate the changes that occur in normal cells exposed to soluble chemical mediators released by tumor cells in culture. Malignant Cells (NSCLC NCI-H460) Fresh Media Malignant Normal Conditioned Conditioned Media Media Exposure to Treatment Media D D D Confluent NHBE D D D Normal Treatment Media Cultured Independently ^ D i s i Fixation (Sedfix) Stain (Thionin) Image & Analyze Discriminant Function Analysis M A C ? Normal ? Figure 2.1: Indirect Co-Culture model In this model, primary normal human bronchial epithelial cells (NHBE passage 3, Clonetics Corp, SanDiego) were plated at a density of 5 X 10 4 cells/cm 2 and grown to 29 Wilton 100% confluency on fibronectin/collagen coated (see section 2.1.1 for protocol) slides . In the pilot study conventional 2.5 X 7.5 cm microscope slides were used while in subsequent studies Fisher 8 well tissue culture slides were used. In these experiments, the individual slides or wells on a slide were divided into two groups. One group was treated with the malignant cell conditioned medium. The control group was treated with normal cell conditioned medium. Malignant conditioned medium was prepared by combining B E B M conditioned on the appropriate N S C L C cell line for 48 hours adding an equal volume of fresh B E B M . Normal conditioned medium was prepared by combining B E B M conditioned by N H B E cells for 48 hours with an equal volume fresh B E B M . Treatment of the cells lasted 24 hours for the pilot study and 3 hours for the subsequent indirect model. After treatment, the slides were washed in Phosphate Buffered Saline (PBS) and the cells were fixed in Sedfix (Saccomano Fixative) and stained using Thionin-S02 Quantitative image cytometry and statistical analysis procedures are discussed later in this section. 2.2.2 M E T A B O L I T E COMPOSITION One ml aliquots of the treatment medium and the fresh medium were stored in air tight cryovials (Fisher Scientific) at 4°C for subsequent metabolic analysis. The metabolite composition of the malignant conditioned media and the normal conditioned medium was measured using a critical care analyzer (Nova Biomedical, Mississaga ON). Wilton 2.2.3 DIRECT CO-CULTURE M O D E L A direct co-culture model was developed which allowed for the culture of N H B E and N S C L C cell lines in the same culture vessel. This procedure is represented diagramatically in Figure 2.2. In this model passage 3 N H B E cells were plated on 0.64x0.64cm coverslips which had been autoclaved and pre-treated with B E G M for 12 hours. N H B E cells were cultured on coverslips at 5 X 104 cells/cm^ and allowed to grow to confluency (2 X 104 cells / coverslip). Each coverslip was placed in an individual well of a 24 well tissue culture plate (Fisher). N H B E cells were used for direct co-culture two days after reaching confluency. Malignant Cells (NSCLC NCI-H460) Log Phase NCI H460 BEGM medium DDD • • Normal Cells (NHBE) • • • • Confluent NHBE J3EGM medium 1 + 1 1 Direct Co-culture • • • • Control Fixation (Sedfix) Stain (Thionin) Image & Analyze Discriminant Function Analysis MAC ? Normal ? Figure 2.2: Direct Co-culture Model Wilton 31 NCI-H460 cells were plated on conventional glass slides (2.5 X 7.5 cm, Oncometrics Imaging Corp, Vancouver B.C.) which had been pre-treated with RPMI 1640 medium supplemented with 5% FBS for 12 hours. The cells were plated at a density of 2.5 X 104 cells/slide and allowed to reach log growth phase. At this stage the slides were washed with PBS and the medium was changed from RPMI medium to B E G M . The tumor cells were allowed 24 hours to adjust to the new conditions prior to co-culture. Direct co-culture was performed in 100 cm square tissue culture plates (Gibco, Burlington) in 8ml B E G M with the appropriate tumor cell/ N H B E cell ratio and for the appropriate duration according to the individual experimental design. The conditions were identical for the control group with the exception that no tumor cells were included in the square tissue culture plates with the coverslips of normal cells. TIME COURSE OF M A C EXPRESSION These experiments followed the direct co-culture protocol for cell preparation as outlined above. N H B E cells were directly cultured with NCI-H460 cells at a ratio of 10 tumor cells to one normal cell. Direct co-culture was performed for 24, 48, and 72 hours. Normal cells were washed with PBS and immediately fixed with Sedfix. The coverslips were air dried, glued to conventional slides with rubber cement and stained with Feulgen Thionin-S02 process. Wilton 32 DOSE D E P E N D E N C Y OF M A C EXPRESSION This experiment also followed the protocol for direct co-culture outlined above. N H B E were cultured directly wifhNCI-H460 cells at ratio's of 1:1, 5:1, and 10:1 tumor to normal cells respectively. Fixation and staining protocols were identical to those above. 2.3 QUANTITATIVE DNA STAINING 76 • • A l l cells were stained by the Feulgen method using Thionin-S0 2 stain. This staining procedure is very reproducible using a staining kit available from Oncometrics Imaging Corp., Vancouver B.C. The Feulgen process is a highly specific stoichiometric D N A stain employed routinely in quantitative image cytometry. In this method, hydrolysis with hydrochloric acid results in the cleavage of adenine and thymine from the deoxyribose backbone of DNA. The resulting aldehyde formed on the backbone sugar is then free to react with the Schiff reagent forming a covalent bond. In the Oncometrics procedure this Schiff reagent is Thionin-S02- This process provides an accurate quantitative assessment of the cells D N A content as it stains only the D N A molecule and not the nucleoproteins as do the haematoxylins. Further specificity in this procedure results from the fact that R N A is not stained. The reason for the different behavior of R N A is poorly understood. It has been suggested that most of the R N A is broken down to soluble substances during the hydrolysis procedure and furthermore, the ribosyl residues of any R N A that remains do not form aldehydes. Thus the intensity of staining with the Feulgen method is proportional to the concentration of DNA. These factors 33 Wilton make this procedure ideally suited to measuring the quantity and distribution of D N A in the nucleus of cells** 8. Although the Oncometrics process is highly reproducible all samples from within a particular experiment were stained within the same staining batch to minimize potential variance in staining intensity . Individual experiments were stained in different staining batches. 2.4 C E L L C Y C L E A N A L Y S I S The percentage of cells found in each compartment of the cell cycle (Gq/G\, S, G2/M) were calculated based on the histogram of integrated optical density available using the interactive image cytometry program "Classify" (Oncometrics Imaging Corp., Vancouver, B.C.). To accomplish this, the tails of the diploid and tetraploid peak were cut off at a width at which 10% of the peak height of the histogram was achieved. A l l cells found between the diploid and tetraploid peaks were grouped into the S-phase fraction. Cells registering integrated optical density values less than the lower margin of the diploid peak and greater than the upper margin of the tetraploid peak were excluded from the analysis. This strategy allowed for a consistent (repeatable) process by which the cells could be classified into the appropriate phase of the cell cycle. A minimum of 200 cells were collected for each sample with the average of approximately 600 cells per sample. The number of cells in each compartment was determined and normalized to a constant total sample cell number of 700. The total number of cells in each compartment Wilton was calculated for all M A C samples and all non-MAC samples for each treatment group for the two dose dependency experiments pooled. Statistical analysis was performed as described in the results section. The purpose of this experiment was to determine i f any significant differences existed in the cell cycle distribution of cells between the M A C and non-MAC groups. 2.5 I M A G E C Y T O M E T R Y A l l image cytometry was performed on the Cyto-Savant™ high resolution automated image cytometer developed by Oncometrics Imaging Corp. of Vancouver British Columbia using proprietary technology of the Cancer Imaging Department of the British Columbia Cancer Agency 7 4. This device was designed specifically to address some of the key requirements necessary to make accurate and reliable measurements of subtle differences in D N A condensation pattern in large populations of cells as is required for the detection and characterization of M A C . Optimal spatial resolution is one of the key factors on must achieve in order to make nuclear texture measurements sensitive enough to detect M A C . The Cyto-Savant employs a digital camera (Microlmager 1400) manufactured by Xi l l ix Technologies Corp, Richmond B.C.. The contribution of the camera individual sensor size and the objective achieves a spatial resolution approaching the theoretical limit that is attainable by visible light of 600nm. The camera employs a scientific CCD (charged coupled device) light transducer which has an advantage over conventional video cameras in that Wilton it has very small individual square sensors (6.8um per side) yielding a pixel picture 2 « +\ element size of O.lum utilizing a 20X objective. A 600 ~ 10 nm illumination broad band filter is employed which is optional for absorbance measurements using Feulgen-Thionin SO2 stain procedure. A typical video camera would require a magnifying lens of 60 to 80X in order to achieve the same resolution74. The ability to use a lower magnification lens has great advantages in terms of measuring nuclear texture. High magnification lenses typically have lower focal depths and therefore, most of the nuclear information will be out of focus at any one image plane. In contrast, low power lenses have a larger focal depth, the 20X objective (NA 0.75) captures most of the nucleus in a * 52 single focal plain . The use of a scientific CCD camera also offers significant advantages in terms of measuring nuclear texture features. Video cameras employing conventional CCD technology typically have only 30-40% fill factor, i.e. the percent of the area is used as active light sensor; leaving 60-70% of the surface "blind". The non-sensing area is reserved for continuous readout of the image74. In contrast, scientific CCDs utilize 100% fill factor i.e. the whole sensor surface area is used to measure the light. The disadvantage of such a system is that it must pause briefly between measurements to transfer the charge to read out. Such cameras need special design and are very expensive. The result is that the entire image is measured and a great deal more information 52 regarding nuclear texture is recorded . Wilton Reproducible focus and exact segmentation of the nuclear material from the background are two additional issues important for measuring nuclear texture. The Cyto-Savant employs novel focus and segmentation algorithms to achieve these goals in a fully automated way. The use of automated processes to achieve these goals has been shown to yield more accurate, and repeatable measurements of nuclear texture than compared to 52 such tasks performed by skilled technologists . In order to achieve population sizes sufficient to detect statistically significant changes in nuclear texture images of approximately 900 nuclei were captured from each sample and experiments were designed with a 20-40 samples per group. The pneumatic slide loader and motorized xyz stage allowed images of individual nuclei to be captured in a fully automated manner following a previously defined search pattern in the center of each individual sample. Images of cell nuclei from each sample were later sorted into the two categories, junk (overlapping cells, out of focus cells, cellular debris) or quality cell images using a binary decision tree composed of discriminant functions created with test data. Next the quality cell images were sorted interactively into three groups: G0\G1, S and G1\M phase cells using the raw histogram of total D N A amount (total integrated optical density). The nuclear feature values of all cells with in each sample were normalized to the mean optical density for the diploid peak of each sample to control for sample to sample variation in staining intensity. Wilton 37 2.5.1 N U C L E A R FEATURE CALCULATIONS One hundred and twenty six features were calculated for every nuclear image acquired by the Cyto-Savant. These features were derived from the literature and by the scientific staff of the Cancer Imaging Department at the B C C A . An in-depth discussion of the 75 • feature set used in this work was published by Doudkine et. al. This discussion will provide a brief overview of the basic principles upon which these features are based. What Is Measured Size Shape Total DNA Amount DNA Distributed Around the Edge of Nucleus or Clustered in the Center of Nucleus Large Intensity Contrast Between the Dark and f the Light Clumps or Little Intensity Change Large Irregular Clumps or Small Uniform Clumps Figuni-3, Six examples of the types of features measured by the high resolution quantitative microscopy system. Wilton 38 The feature set can be divided into eight categories which relate features based on their derivation and the nuclear texture phenomenon which they describe. These categories include: size features, shape features, photometric features, discrete texture features, Markovian texture features, run-length texture features and fractal texture features. A simple graphical representation of what these features measure in combination is shown in Figure 2.3. SIZE A N D SHAPE FEATURES The measurement of nuclear size and shape is the easiest set of features to visualize. Size features include nuclear area and radius etc., while shape features describe the circularity of the nucleus (e.g. sphericity) and the shape of the nuclear outline (e.g. eccentricity). PHOTOMETRIC FEATURES Photometric features give estimations of absolute intensity and optical density levels of the object as well as their distribution characteristics. For stoichiometric D N A stains (Feulgen-Thionin-S02) the total amount of D N A in a nucleus (DNA amount) is the raw measure of the integrated optical density of that object. Other photometric features can be used to measure the distribution of the various intensities of D N A within the nucleus. DISCRETE T E X T U R E FEATURES Discrete texture features are based on the segmentation of the nucleus into discrete regions of high, medium, and low chromatin condensation. These features describe the 39 Wilton size, shape, optical density and spatial distributions of pixels classified as one of the three states of chromatin condensation. M A R K O V I A N T E X T U R E FEATURES Markovian texture features attempt to characterize gray level variations between adjacent pixels in the image. For example the feature Contrast is based on the estimation of intensity differences between neighboring pixels. The value of this feature increases with the increasing magnitude of density variations between neighboring pixels and the relative number of pixels with such differences. The opposite of contrast is homogeneity which measures the "smoothness" of an image. Homogeneity increases in value when there are small and spatially distant variations in chromatin condensation within a nucleus. R U N L E N G T H TEXTURE FEATURES Run length texture features describe chromatin distribution in terms of the length of lines comprised of consecutive pixels with the same gray level value. F R A C T A L TEXTURE FEATURES The fractal texture features represent measurement of the areas of a three dimensional surface, created by the optical density of the chromatin throughout the nucleus. Fractal area is correlated to nuclear size however for nuclei of similar size large values of fractal area indicate images containing small chromatin clumps with high optical density 40 Wilton variations between them. This can be likened to a mountain range. The more the extreme the local variations in altitude are in a mountain range the more peaks and valleys will exist, therefore there will be a greater surface area of the land comprising this mountain range in comparison to a flat piece of land. 2.6 STATISTICAL ANALYSIS 2.6.1 DISCRIMINANT FUNCTION A N A L Y S I S The ability to discriminate between two morphologically similar populations of cells was fundamental to the achievement of the objectives of this thesis. The task of evaluating the hundreds of parameters measured by quantitative image analysis and achieving an objective criterion for the identification of M A C was accomplished via a procedure known as multivariate discriminant function analysis. Discriminant function analysis is used to determine which variables (nuclear features) discriminate between two or more naturally occurring (apriori defined) groups. In the cell by cell discrimination studies approximately 100 nuclear features were included in the analysis and a maximum of 12 were chosen which best discriminated between the two groups (those exposed to malignant conditioned medium, termed " M A C " , and those not exposed to malignant conditioned medium termed, "Control" .). In other words, for each experiment a "model" of nuclear feature expression was constructed which afforded the best prediction of which group a case belonged to. Wilton Discriminant function analysis was performed using program 7M of the Biomedical Data Processing (BMDP) statistical package (BMDP Statistical Software Inc., Berkley California). This program performs forward stepwise discriminant function analysis. With this process B M D P builds a "model" of discrimination step by step. Specifically, at each step B M D P reviews all variables and evaluates which variable has the greatest discriminating ability. That variable is then included in the model and the next step is initiated. The number of variables entered into the model was indirectly controlled by changing the "F to enter" and "F to remove" values in the program input file. The " F " value is computed as the ratio of the between-groups variance in the data over the pooled (averaged) within-group variance. Thus for features having between group variances significantly higher than within group variance, the "F" value is large. The features having large F values have greater ability to identify differences between populations and are utilized in the discriminant function. By increasing the F to enter/remove values more stringent criteria could be placed on which variables can be included in the model thereby reducing the number of features included. Using discriminant function analysis there is a risk of overtraining the model. This essentially means that the model is constructed so thoroughly as to recognize a particular training set with the consequence that it will fail to discriminate subsequent test sets. This occurs because the discriminant function has learned to recognize particular features of a training set which are not a generic phenomenon i.e. they do not occur in similarly treated test sets. A general convention used to avoid over training, states that there 42 Wilton should be at least twenty times more samples per group in a training set than the number 82 of features used in a discriminant function analysis . Procedural control experiments which deal with the issues of over training and discriminant function optimization are included in the results section 3.5. For the analysis in this thesis strict criteria were applied to the conditions for inclusion of features in the discriminant function analysis. Classifiers generated for cell by cell discriminant function analysis included a maximum of 12 features and sample by sample classification was accomplished using only a single feature to ensure that no over training took place in these analyses. 2.6.2 C E L L B Y C E L L A N A L Y S I S For each experiment the nuclear feature values of 200 diploid cells were taken from each sample and pooled into the M A C or CONTROL group according to the treatment protocol. Forward stepwise discriminant function analysis was performed to generate classification functions which discriminate optimally between cells of the two treatment groups. The discriminatory power of the discriminant function (how well are the two groups separated) can be interpreted from the classification matrix which reveals the specificity and sensitivity of the system in terms of cell by cell discrimination. The Jackknifed classification matrix can be utilized to determine the stability of the classification i.e. would this discriminant function perform as well on a test data set. Jackknifing refers to a statistical procedure in which one sample is removed from the data set and a 43 Wilton discriminant function is generated on the remainder of the data. This discriminant function is then applied to the data which has been excluded and the classification results are recorded. This procedure is repeated until each data point has been excluded from the generation of the discriminant function and has acted as a test set. This procedure allows one to use a data set as both a training set and a test set. If the classification matrix achieved using Jackknife analysis is similar to the training classification matrix one can be relatively confident that the discriminant function is robust (i.e. the system was not over trained). 2.6.3 SLIDE B Y SLIDE ANALYSIS The discriminant function generated in the cell by cell analysis of each experiment was then utilized to generate population statistics on each sample allowing for sample classification. The M A C score of an individual cell was computed by entering the nuclear feature values into the discriminant function for that particular analysis. This score represents the weighted combined nuclear feature expression for a particular cell. The sample by sample data is represented in terms of " M E A N M A C SCORE". This term refers to the average value of individual M A C SCORES obtained for 200 diploid cells per sample. In some figures a decision boundary has been chosen to obtain a representation of the discriminating power of the discriminant function on a sample by sample basis. This decision boundary was chosen to maximize the chance of correctly classifying a normal sample (specificity) and to maximize the chance of correctly classifying a M A C sample (sensitivity). Wilton A Receiver Operating Characteristic curve (ROC curve) is a more common way of representing sample classifications. A n ROC curve is a plot of false negative versus false positive values for a particular system (MAC/control discrimination experiment) along a continuum of decision boundaries. For this type of a curve two populations that can be perfectly discriminated (i.e. an orthogonal test) will have ROC curve points along the x and y axes i.e. for all operating points the samples are correctly classified. The better the discrimination is between the samples the closer the ROC curve will follow the x and y axes. A system having absolutely random discrimination will fall along a straight line which extends from 100% false negative to 100% false positive. The farther the ROC curve for a system is from this line the better the discrimination power is for the system. This analysis was applied to the dose dependency experiment and can be seen in figure 3.8 . 2.6.4 t-TEST FOR INDEPENDENT SAMPLES There were several instances where the statistical significance of the means of two groups were analyzed in these studies. The Statistica (Statsoft Inc., Tulsa OK) statistical software packaged was used for this purpose. The p-level for statistically different means used in these analysis was 0.05. Thus, tested populations were considered significantly different i f the p-level using a t-test for independent samples was less than 0.05. The p-value reported represents the probability of error involved in accepting the hypothesis of the existence of a difference in the sample groups. Technically, this represents the Wilton probability of error associated with rejecting the hypothesis of no difference between the two categories of observations in the population i f , in fact, the hypothesis is true. Wilton 3.0 RESULTS 3.1 PRIMARY CULTURE OF HUMAN BRONCHIAL EPITHELIUM The use of bronchial biopsies as a source of primary cultured N H B E and tumor cells was investigated. Bronchial biopsies obtained during bronchoscopic examination were successfully cultured from three of four patients using the protocol outlined in section 2.1. Immediately following plating of the bronchial biopsy, numerous ciliated epithelial cells were observed in the culture as part of the large tissue mass and in smaller pieces of tissue fragments. The cilia could be seen beating in the culture medium. Immediately after plating, a significant number of red blood cells were seen surrounding the tissue. Approximately ten days following plating, numerous well differentiated epithelial cells were visualized in a radial pattern surrounding the explanted tissue. At this point the red blood cells were dead. Approximately 20 days subsequent to plating, the primary culture bronchial epithelial cells were sufficiently numerous (1 X 106 cells) to be passaged to conventional 75 cm growth area tissue culture plates. For the first two passages of the primary cultured cells a small population of macrophages persisted. These cells were identifiable by their dendrite like extensions and were eventually eliminated as N H B E cells were selected for by the high calcium growth medium. Primary culture of bronchial epithelial cells was successful from a lesion which appeared dysplastic according to LIFE™ (Xillix Technologies Corp., Richmond, B.C.) bronchoscopic examination from a 64 year old male. Primary cells were also cultured from a squamous cell carcinoma of a 47 Wilton 71 year old male and the normal tissue of a 75 year old male. Although primary culture from patients' tissue was successful it was decided to purchase normal cells from a commercial source and to utilize a characterized cell line in these models. This decision was based on reasons of quality control and convenience. 3.2 TUMOR C E L L CULTURE CONDITIONS As described in section 2.2 the tumor cell line NCI-H460 was isolated and grown in RPMI medium supplemented with 5% FBS. The normal bronchial epithelial cells utilized in this model were grown in a serum free defined growth media (BEGM). To accommodate future objectives for this model and to ensure reproducible growth conditions between different media batches the defined growth medium was preferred for the co-culture experiments. Thus, it was necessary to ensure that the defined growth medium could support log phase growth of the tumor cell line. GROWTH OF NCI-H460 IN SEVERAL CULTURE MEDIA G R O W T H MEDIA H O U R S F O L L O W I N G C O N V E R S I O N Figure 3.1: The tumor cell line NCI-H460 was passaged in RPMI supplemented with 5% FBS. Twenty four hours later treatment groups were grown in one of four media for a five day period. The confluency was estimated at 24, 72 and 120 hours to indicate the growth rate for each media. Wilton 48 Attempts to convert the tumor cell line NCI-H460 to B E G M outright achieved poor results. This cell line was found to require serum supplementation to initiate the cells to begin cycling following passage. Following initiation, the cells grew very well in serum free media. Figure 3.1 shows the confluency of the tumor cells at 24, 72 and 120 hours after conversion from RPMI supplemented with 5% FBS. As the figure illustrates, the tumor cell line NCI-H 460 grows best in RPMI 5% FBS as it reached a confluency of 95% after 120 hours in comparison to the 60% confluency achieved in the other growth media. This figure also demonstrates that this tumor cell line grows equally well in the serum free defined media B E G M , ACL-3 (RPMI with no serum) and B E G M supplemented with 2% FBS. Under all three growth conditions the cells reached 40% confluency in 72 hours and 60% confluency within 120 hours. Furthermore, the cells appeared morphologically identical to those grown in RPMI medium supplemented with 5% FBS. Although these measures of growth rate were crude they provided sufficient evidence that the cell line NCI-H460 is healthy and remains in growth phase when converted to the B E G M following passage in RPMI supplemented with 5% FBS. Thus, the co-culture experiments could be performed using defined media (BEGM) with no serum supplementation following passage of the tumor cell line. 3.3 PILOT STUDY OF MAC EXPRESSION IN VITRO A pilot study was conducted to determine whether malignancy associated changes can be measured in vitro using the proposed methodology. This study was performed according to the indirect co-culture protocol outlined in section 2.2.1. N H B E were cultured to 49 Wilton confluency on conventional glass slides in square tissue culture dishes. The malignant conditioned media was composed of 25% B E G M media conditioned on NCI-H460 cells and 25% B E G M medium conditioned on NCI-H358 cells for four days plus 50% fresh B E G M . The control conditioned medium (BEGM) was conditioned on N H B E for four days and supplemented with an equal volume of fresh B E G M . The test sets were treated with their respective treatment media for 24 hours. The results of this protocol are summarized in figure 3.2. PILOT STUDY SAMPLE CLASSIFICATION 1 * A A A / A 100% correc MAC classifi t cation • Control 1 A MAC , i i r < — decision boi mdary 100% correct control classific ation ' 1 i -I -2 ' -1 0 1 2 3 4 MEAN MAC SCORE Figure 3.2: Pilot study demonstrating the validity of the methodology proposed for the MAC in vitro studies. M A C samples were treated with medium conditioned on NCI-H460 and NCIH-358 for 4 days combined then supplemented with equivolume fresh medium as described above. Control samples treated with medium conditioned on NHBE for 4 days supplemented with equivolume fresh media. M A C samples and Control samples were classified using the four feature discriminant function generated on a cell by cell analysis of the data. The discriminant function generated on a cell by cell analysis achieved a 70% correct cell by cell classification on the data from this experiment. As figure 3.2 illustrates, the mean Wilton 5 0 M A C score per sample (mean value of the discriminant function applied to the nuclear features of 200 cells) resulted in a clear separation of the two treatment groups. A decision boundary chosen at -0.58 resulted in 100% discrimination between the two treatment groups. Although the sample number was very low in this experiment (n=l 6) these results indicated that the methodology for this study was worthy to be tested on a larger scale. 3.4 INDIRECT CO-CULTURE MODEL OF MAC Malignancy Associated Changes in N H B E cells grown in vitro were induced by the tumor cell line NCI-H460 using the indirect co-culture model. In this experiment N H B E were cultured on Fisher 8 well chamber slides in B E B M two days past confluency. A volume of 100 ml B E B M conditioned with NCI-H460 tumor cells for 48 hours was concentrated to a total of 20ml using Kwikspin^M Macro Ultrafiltration Units 10 000 MWCO(Pierce, Rockford, IL.). The control treatment group was treated with medium conditioned with N H B E cells also for 48 hours. This medium was not concentrated. Equivolume of fresh B E B M was added to all treatment wells to ensure adequate metabolites. The treatment period with conditioned medium was 3 hours in this experiment. Wilton 51 MAC INDUCED in vitro [ A A A A A A A A A 1 / A& \l * A A A A A A o Control A MAC 71.4% correct MAC classification | 1 80.6% correct control classification r< decision b oundary o o o 1 -2.5 -1.5 -0.5 0.5 1.5 2.5 3.5 MEAN MAC S C O R E Figure 3.3: Indirect co-culture of NHBE cells with medium conditioned for 48 hours by NCI-H460 cells concentrated by a factor of 5 and diluted by 50% with fresh medium (MAC). Control group consists of NHBE cells treated with medium conditioned by NHBE cells and diluted with equivolume fresh BEBM (control). Median scores are marked by vertical rectangles. For the experiment shown in Figure 3.3, a seven feature discriminant function generated with 13200 cells resulted in a cell by cell classification efficiency of 71.4% according to Jackknife analysis. The mean M A C score per sample was used to discriminate samples from the two treatment groups. This score represents the average weighted nuclear feature expression for 200 cells per sample. Based on the decision boundary shown, this classifier achieved 81% correct classification of control samples and 71% correct classification of M A C samples. The mean values of M A C score for the two treatments groups were found to be significantly different using an independent sample t-test (p<0.05). Wilton 52 3.5 PROCEDURAL CONTROL EXPERIMENTS Due to the complexity of the culture protocol and statistical analyses, several experiments were performed in order to ensure that the conditions outlined in this protocol were sound. 3.5.1 M E T A B O L I T E ANALYSIS The metabolite composition of the two treatment media from the pilot study as well as fresh B E B M were analyzed using a Critical Care Analyzer (Nova Biomedical, Mississaga ON). The conditioning protocol for the two treatment medium was outlined in section 3.3. The results of these analyses are summarized in Figure 3.4. This figure shows that both conditioning protocols resulted in a significant lactate production as well as glucose consumption. Also shown in Figure 3.4, the pH dropped slightly in both conditioned media, however it remained well within the range of normal cell culture media (7.4-7.7)87 7 ^ K + ,C1" and N a + ion concentrations all increased slightly but similarly in the two conditioned media and the PO2 of the media dropped insignificantly. Thus, we concluded that there were no significant differences in the metabolite concentrations for the two conditioning protocols. Wilton 53 5 1 0 30-| 25 | 20 S 15J Q_ 10-5 1.5. 0.0 i 24 ncrrralcordfoned rafignantcordioned 140-120-100-p 80-jlcn 60-2. 40-20-0-ncrrrelcorcHoned raligrantconciloned 5. CD ncmdccndtoned mSgiartccrrffciied III I f i l l I I . I n i l •••] II Nal normal cordicned rraSgnartccrdicned 140 120 100 3 80 g 60 | . 40 20 0 rarralconcHcned rrtfgpartccncHoned Figure 3.4 Metabolite analysis of fresh BEBM (fresh), equivolume fresh media and BEGM conditioned on NHBE for 4 days (normal conditioned media) and 50% media combined with 25% BEGM conditioned on NCI-H358 cells and 25% BEGM conditioned on NCI-H460 tumor cell lines for 4 days (malignant conditioned media). Wilton 54 3.5.2 DISCRIMINANT FUNCTION.PERFORMANCE As outlined in section 2.6.1, discriminant function analysis is considered statistically valid when the ratio of samples per group to features is greater than 20:1. In the methodology outlined for these experiments a maximum of 12 features were included in the cell by cell analysis although a very large number of features could be used for a population of cells as large as 10000. In order to determine if the stringent criteria placed on the discriminant function analysis limited the ability of this process to separate the two treatment populations, repeated discriminant function analysis was performed on a single data set incorporating a range of 2 to 28 features in the analyses. Figure 3.6 shows the cell by cell classification efficiency based on the number of variables included in discriminant function analyses. This figure clearly shows that the classification efficiency of the discriminant function reached a plateau at approximately 12 features. From this point the inclusion of more features in the analysis did not improve the discriminatory power of the analysis. Wilton 55 Discriminant Function Performance 0 4 8 12 16 20 24 28 Number Features Used in Discriminant Function Analys is Figure 3.5:Cell by cell classification efficiency of discriminant function analysis as the function of features included in the analysis. Both training set classification efficiency and Jackknife classification efficiency are shown for each function. The experimental data were generated using NCI-H358 tumor cells according to the indirect co-culture protocol. 3.5.3 TEST FOR STATISTICAL OVERTRAINING A major concern when using discriminant function analysis to generate a sample classifier based on a priori defined groups is overtraining. In overtraining, when too many features are used, discriminant function analysis characterizes the training set so well that the classifier effectively becomes useless when applied to a test set of data. To test i f the discriminant functions were overtrained using the outlined methodology, individual cells from the same set of identically treated N H B E cells were randomly assigned to two test sets. Discriminant function analysis on these two groups was performed as before. In this experiment the 3046 cells were classified with an accuracy Wilton 56 of 51.2%. This is effectively a random classification, reflecting the random nature of the group assignment. The classification matrix for this analysis is shown in Table 3.1. Based on this result one can conclude that there were no differences detected in the nuclear features of the two test sets as expected. Therefore, using the outlined methodology for statistical analysis does not over train the classifier and the reported results are based on true morphological differences between populations. Group % Correct # Classified A # Classified B Group A 44.0 1324 1682 Group B 58.4 1247 1747 Total 51.2 2571 3429 Table 3.1: Classification matrix for discriminant function generated on two groups (A and B) of randomly assigned control cells. Overall 51.2% correct classification efficiency indicates that the current statistical protocol does not over train the classifier. 3.6 DIRECT CO-CULTURE STUDIES The indirect co-culture protocol was modified to investigate the time course and dose dependency of M A C expression. To achieve this, N H B E cells were cultured to confluency on glass coverslips. M A C were induced by directly co-culturing the normal cells on coverslips with tumor cells grown on glass slides in common square tissue culture dishes. A detailed description of this protocol is available in section 2.2.3. Although the cells were immersed in the same culture medium it is probable that there was no physical contact between the N H B E and the tumor cells as both groups adhere to Wilton their respective growth surfaces. Thus, this protocol is well suited to measure the effect on normal cells of soluble chemical mediators released by tumor cells. 3.6.1 TIME COURSE OF M A C EXPRESSION The duration of direct co-culture required for optimal M A C expression was investigated for time points of 24, 48 and 72 hours. The expression of M A C was found to be highly significant at the 24 hour time point. A discriminant function generated from cells in this time period versus the controls resulted in an overall cell by cell classification efficiency of 82.7 percent according to Jackknife analysis. The peak strength of M A C expression was found at 48 hours. A discriminant function generated on cells from this time point versus the control resulted in an overall cell by cell classification efficiency of 85.1 percent according to Jackknife analysis. The strength of M A C expression declined by the 72 hour time point while remaining highly significant. The discriminant function generated for this time point versus the control resulted in a classification efficiency of 75.8 percent using the same analysis. Figure 3.6 shows the average of the weighted nuclear feature expression (Mean M A C Score) for each slide according to the discriminant function generated on the cell by cell analysis of the 48 hour time point. From this figure it is clear that the strength of M A C expression is significant by the 24 hour time point, it reaches a maximum at 48 hours and declines, yet remains significant at 72 hours. A t-test for independent samples determined that the difference in the Mean Mac Score for all three time points was highly statistically significant (p<0.001). Wilton TIME C O U R S E O F M A C E X P R E S S I O N M A C 72 hours * A A A A » AA A . A - —-—— ft M A C 48 hours ° 6 >o °<> o°c>, o o . — 0 M A C 24 hours Control o 0 1 2 3 M E A N M A C S C O R E Figure 3.6: Time course of MAC expression. NHBE on coverslips were cultured directly with NC1-H460 tumor cells grown on glass slides. Co-culturing periods lasted 24,48 and 72 hours. The discriminant function was generated on the 48 hour time point. 3.6.2 DOSE DEPENDENCY OF M A C EXPRESSION The strength of M A C expression in relation to tumor burden or dose of tumor released stimulus was investigated using the direct co-culture model. A detailed account of this protocol is found in section In brief, N H B E grown to confluency were cultured directly in tissue culture plates concurrently with NCI-H460 tumor cells proliferating on glass slides. The dosage of M A C stimulus was controlled by setting the ratio of tumor cells to N H B E at 1:1, 5:1 and 10:1 respectively. Direct co-culture lasted 24 hours. The average of the weighted nuclear feature expression (MAC score) for each sample in this experiment is displayed in Figure 3.7. The 12 feature discriminant function used to Wilton 59 generate the M A C score on a sample by sample basis for all doses was generated using cells from the 10:1 dose versus those from the control group. As the figure illustrates, a clear trend from 57% in the low dose, to 70% in the medium dose, and finally 95% correct sample classification of the high dose M A C treatment group was observed. It is important to note that cells from the intermediate doses were not included in the training set for the discriminant function. This point makes the gradual increase in strength of M A C expression very significant as the intermediate doses represent true test sets. S T R E N G T H O F M A C E X P R E S S I O N - D O S E D E P E N D E N C Y 1 JL.,.. 1 A L -A_ A A A A _ — jQ.... t\ **» LI iX 9 5 % correct M A C 10:1 classification A A A . A - A • - AJI A A . - A A - A - - A A .g. A A u A a a a, aaa. a a a. a 7 0 % correct M A C 5:1 classification der is ion boundary A A A M A £ a J A A A a a — A — AA..H.*.—Hlfl——y^.-Ai nfl Or ;Qj 5 7 % correct M A C 1:1 classification A t A A A A 7 3 % correct control classification 1 -1 -0.6 -0.2 0.2 0.6 1 1.4 M E A N M A C S C O R E Figure 3.7:The M A C score for NHBE samples cultured directly with NCI-H460 tumor cells at ratio's of 0:1, 1:1, 5:1 and 10:1 tumor to normal cells respectively. The discriminant function was generated to discriminate between the control group and the 10:1 dose tumor cell group Wilton 60 A t-test for independent samples determined that the average weighted nuclear feature expression was significantly different (p<.001) for samples from the 10:1 and 5:1 dose treatment groups compared to the control group. There was no statistically significant difference for the expression of these features in the 1:1 dose treatment group. F^LSEPCSTTlvE Figure 3.8: A Receiver Operating Characteristic curve for the direct co-culture dose dependency experiment. The solid line represents zero discrimination and the x and y axis represent perfect discrimination. The data are presented in the form of percent false negative sample classification and percent false positive classification for a continuum of operating points. Figure 3.7 shows the classification results for the dose dependency experiment utilizing a single decision boundary. Figure 3.8 is the Receiver Operating Characteristic curve (ROC curve) for the same data set. Representation of the data in the form of ROC curves allows one to visualize the classification efficiency of the experiment across the full range Wilton 61 of potential decision boundaries. The shape of the curve reveals a great deal about the characteristics of the experimental system. The figure clearly shows that for all operating points the false negative and false positive classification are smaller as the dose of tumor cells is increased, i.e. the curve for the 10:1 dose is closest to the x and y axis (perfect discrimination line) followed by the curve for the 5:1 dose and finally the 1:1 dose. The figure also illustrates that the classification efficiency is greatest for the 10:1 dose followed by the 5:1 and finally 1:1 doses across all potential decision boundaries. Further analysis of the data revealed that the frequency of cells expressing M A C in each sample also increased with the dose of malignant cells present in the direct co-culture system. As seen in Figure 3.9, the sample classification efficiency based on the frequency of M A C cells expressed in each sample shows a clear trend from 42% for 1:1 dose to 60% for the 5:1 dose and finally to 83% for the 10:1 dose based on a decision boundary at 42%. Statistical analysis also revealed that the difference between the frequency of M A C cells expressed is significant for the 5:1 and 10:1 doses but not for the 1:1 dose (p<0.05). Wilton 62 F R E Q U E N C Y OF M A C E X P R E S S I O N - D O S E D E P E N D E N C Y 1 1 A - | -t . . . A . A A i . . . J! A ....AtA A 83% correc classificatio t M A C 10:1 n A i _ A A \ A .™.~~ A 1 d A — A - — - A A A .1 I fifl% nnrrpr t MAP. fV1 decision boundary I —> classification A A A A A A ^ A a flf* ~ 42% correc classificatic t M A C 1:1 >n A A A A A A A AA A A A * * A * A A —™- " •••A 7b% correct control classification [ , . , . I . . I , — i — . — . — . — . — I — . — . — . — . — i — . — . — • — • — I 0.1 0.3 0.5 0.7 0.9 1.1 F R E Q U E N C Y OF M A C C E L L S Figure 3.9: The frequency of cells classified as MAC based on their nuclear feature expression for the various doses of malignant cells in the direct co-culture protocol. The median value for each dose is shown by a vertical rectangle. The dose dependency protocol was repeated using tumor to normal doses of 0:1 (control), 5:1 and 10:1. In this experiment the cell by cell discriminant function was generated to discriminate cells from the control treatment group and the 5:1 M A C treatment group. Thus, the training set for this experiment was the middle dose treatment group and the test set was the high dose treatment group. Wilton 63 D O S E D E P E N D E N C Y 2 TRAINED TO MIDDLE D O S E „ O. o Control o M A C 5:1 o M A C 10:1 o 1 89% correct M A C 10:1 classification <o 1 $ o o o o o . O o o 0 i 76% correc classificatic :t M A C 5:1 )n O D 9.. ... Q • ° • 1_J < decision boundary 8 1 % correct control classification 1 o 0 i o ° 0 -1.2 -0.6 0 ' 0.6 1.2 1.8 2.4 M E A N M A C S C O R E Figure 3.10: NHBE cells directly co-cultured with NCI-H460 tumor cells at ratio's of 0:1 (control), 5:1 and 10:1 tumor to normal cell number respectively. The cell by cell discriminant function was generated to discriminate between cells from the 5:1 ratio and the o: 1 ratio (control) cells. It is clear from Figure 3.10 that the strength of the weighted nuclear feature expression is greater for the samples cultured with a greater dose of tumor cells. The evidence to support this is considerable as the nuclear feature changes identified as M A C in this experiment were determined by comparing the nuclear feature expression of the 5:1 training set to that of the control group. The fact that the nuclear features are not only the J same but are even greater in magnitude in the 10:1 test set provides convincing evidence that there is a dose response relationship between the strength of M A C expression and the amount of an as of yet uncharacterized soluble chemical mediator released by tumor cells. Wilton 64 3.5 C E L L C Y C L E STUDY The data from the two time course experiments were combined and the effect of M A C on the distribution of cells throughout the cell cycle was investigated. The cells in each sample were classified into G1/G0, S and G2/M phases of the cell cycle based on the amount of D N A each contained. The number of cells classified in each group was multiplied by a normalization factor which brought the total number of cells per sample to equal 700. This step was done to ensure that each sample had equal weight in the analysis. The percentage of cells classified in each phase of the cell cycle for the control and the two doses of M A C treatment can be seen in figure 3.11. Control Ploidy Distribution G2 14% 73% 5:1 Mac Ploidy Distribution G2 18% 10:1 Mac Ploidy Distribution G2 18% 68% ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 68% Figure 3.11: Distribution of normal cells in the cell cycle for two dose dependency experiments pooled. Control distribution obtained from 49 samples, 5:1 distribution is pooled from 50 M A C samples and 10:1 distribution is pooled from 48 M A C samples. The data from this analysis is also summarized in table 3.2. Based on this data one can conclude that there was a slight drop in the number of cells classified as G l phase in both the 5:1 and 10:1 dose M A C treatment groups with respect to the control. Although the Wilton 65 drop in the number of cells classified as Gl i s small, this difference was found to be statistically significant (p<.05) based on a t-test for the difference of means. Furthermore, there was a statistically significant increase in the number of cells classified as G2/M phase cells in the treatment groups versus the control group. The difference between the number of cells classified as S-phase for the control and both treatment groups was not statistically significant. Control (n=49) 5:1 (n=50) p-value (5:1 vs Control) 10:1 (n=48) p-value (10:1 vs Control) G l phase 507 474 .002492 470 .00978 S-phase 92 99 .486692 101 .348849 G2/M phase 101 128 .000092 128 .000015 Table 3.2: Pooled results from two dose dependency studies. The integer number in the columns represent the mean value of the number of cells classified as G l , S or G2/M phase per sample based on the histogram of the DNA amount. P values were calculated using a t-test for the difference of means between the control group and the respective treatment group. Wilton 66 4.0 DISCUSSION Malignancy Associated Changes (MAC) can be defined as subtle morphologic and or physiologic changes which occur in ostensibly normal cells of patients harboring malignant or pre-malignant lesions. Recent evidence indicates that this phenomenon can be used in a practical and objective way in the detection of cancers even in their earliest stages-^?, 93 Authors have further postulated that M A C may be utilized as a prognostic indicator and be used to monitor the success of treatment or effect of chemoprevention protocols^. If true, this phenomenon has great potential to become one of the most powerful tools in the fight against cancer. Yet, the mechanism of M A C remains poorly understood. There have been two main theories proposed to explain the mechanism of M A C . The first, so called "Field Cancerization Theory", postulates that M A C simply reflect mutation(s) in the normal tissue surrounding the tumor resultant from concurrent exposure of the tissue to carcinogen. In this theory, M A C could at best be used as a probability indicator of susceptibility to cancer and could not be directly related to the presence of malignancy. The second main theory has been the working hypothesis of this thesis, and is commonly referred to as the "Humoral Mechanism of M A C " . This theory postulates a direct cause and effect relationship between malignancy and M A C . In this hypothesis soluble chemical mediators are released from the malignant tissue which act on the normal tissue in the surrounding environment. Under these circumstances the presence and strength of M A C expression may be directly related to the presence, type 67 Wilton and grade of malignancy. Clearly, these two divergent theories for the mechanism of M A C have great implications on the value of this phenomenon as a tool for cancer control. In these studies, an in vitro model designed to mimic the interaction of N S C L C tumors with the adjacent normal bronchial epithelial tissue was employed to demonstrate the humoral mechanism of M A C . The success of this model can be attributed to several key elements in its design. The model employed primary cultured N H B E rather than a transformed cell line for the normal tissue model. There were several reasons for this decision. First, primary cells display a greater degree of differentiation in culture in comparison to the transformed cell lines. Thus, the normal primary cultured cells in this model were more likely to express receptors and react physiologically in a similar way as normal tissue in vivo than would be expected from transformed cells. Secondly, the primary cultured N H B E cells expressed a consistent nuclear morphology in culture. This included consistent nuclear size and texture features when cultured under controlled conditions. In contrast, most transformed (normal) cells have a highly variable nuclear morphology with wide ranging values of nuclear size and texture features. In fact, one could argue, that the process of immortalization itself changes the state of the D N A , and thus, inevitably the nuclear morphology of transformed (normal) cells. Thus, the subtle nuclear morphological changes that comprise M A C are better measured in the primary cultured cells. 68 Wilton Another key element of this model was our ability to culture cells in a defined, serum free medium (BEGM). This was important as the growth factor composition was known, easily controlled and constant from one medium batch to the next. Furthermore, it is envisioned that a future application of this model will be in the isolation of growth factors capable of inducing M A C in media conditioned by tumor cells. Clearly, a defined medium is essential in such studies. The tumor cell line (NCI-H460) primarily used in these studies required serum supplementation (RPMI 5% FBS) for initiation of cell division following passage. Based on morphological appearance and cell counts shown in Figure 3.1 this tumor cell line was healthy following conversion to a serum free (BEGM) growth medium after passage in serum supplemented medium. Thus, for all experiments in this study the tumor cell line was passaged and cultured in RPMI 5% FBS for 24 hours prior to converting it to B E G M . The growth rate experiment also revealed an interesting characteristic in the NCI-H460 cell line. Following passage in serum supplemented medium these cells grew equally well in defined medium supplemented with growth factors (BEGM), the latter medium supplemented with only 2.5% serum and RPMI with no growth factor or serum supplementation. This suggests that this cell line has an autocrine growth stimulation mechanism which is active after the cells are initiated in to the cell cycle allowing growth independent of growth factor supplementation in the medium. Therefore, this cell line is likely to produce and elaborate growth factors in to the medium. 69 Wilton Using this model system, a pilot study was conducted to determine i f M A C could be measured in vitro using an indirect co-culture methodology. In this study, medium conditioned by both the tumor cell lines NCI-H 460 and NCI-H358 was used to induce M A C in N H B E cultured on conventional glass slides. The 70% correct cell by cell classification and 100% correct sample by sample discrimination achieved in this study provided strong evidence that the methodology was worth pursuing. However, the low number of samples in the pilot study was a limitation for adequate statistical significance of the results. In subsequent trials of the indirect co-culture model a greater number of independent samples was achieved while conserving medium and cell resources by employing Fisher eight well tissue culture slides. These slides allow for growth of cells directly on the slides in eight independent wells separated by a water tight plastic barrier. These conditions provided sample sizes of up to 30 independent samples per group while limiting costs due to experimental materials. At this point a strong cautionary note is required. Although convenient for achieving high sample numbers, significant staining irregularities were observed on these slides. This problem was solved for the results shown in figure 3.3 by agitating the slides vigorously with each solution change during the staining procedure. However, it is recommended that future application of the indirect co-culture model should be performed using coverslips (as outlined for the direct co-culture model) to culture normal cells. Wilton In the experiment shown in figure 3.3, medium conditioned by NCI-H460 cells for 48 hours was used to induce M A C . The discriminant function generated for this experiment resulted in a cell by cell classification efficiency of 71.4 %. Based on the average of the weighted nuclear feature expression of cells in each sample (mean M A C score), 81% of the normal samples and 71.4% of M A C samples were correctly classified. This study represented the first statistically significant result demonstrating that M A C can be induced in vitro. Furthermore, in this experiment M A C were induced by soluble factors present in the medium conditioned by the N S C L C tumor cells. This fact was assured as the conditioned media was centrifuged and passed thru a 0.22 urn filter to remove all tumor cells and debris prior to co-culture. Several precautions were taken to ensure that there was no bias in the design of the indirect co-culture model or in the statistical analysis of the data. In the indirect model, two steps were taken to certify that the control cells were treated with medium at the same metabolite concentration as the test cells treated with medium conditioned by malignant cells. First, the control cells were treated with a medium conditioned on normal cells for an equal time period as the malignant conditioned medium was conditioned on tumor cells. Thus, approximately equal amounts of metabolites were consumed or produced in both treatment media. Secondly, an equal volume of B E B M ( B E G M without growth factor supplementation) was added to both conditioned media to supplement any essential metabolites consumed by the conditioning of the respective cell lines. As illustrated in figure 3.4 these steps succeeded in creating equal metabolite Wilton composition the control and M A C treatment media. Similarly, a difference in the metabolite concentration was not considered to be a factor in the direct co-culture model. This assertion is based on the fact that co-culture occurred for a shorter duration than the medium was conditioned for in the pilot indirect model. Analysis of the metabolite concentration of medium conditioned on NCI-H460 cells for 48 hours resulted in negligible changes in the metabolite concentration. Therefore, any difference in the cellular morphology is likely not have been attributed to metabolite depletion in the treatment medium for either model. As mentioned in the main body of the thesis, a major concern when using multivariate discriminant function is over training the classifier. If careful measures are not taken to avoid over training, it is possible to obtain results that are artificially improved by the statistical process and which are not entirely the result of a biological phenomenon. For this reason very strict criteria were placed on the number of features included in the discriminant function for each classifier. On a cell by cell basis the ratio of cells to features was generally 1000 to 1 or greater. On a slide by slide basis there was never more than one feature used to classify samples. Two experiments were performed in order to closely scrutinize the statistical methodology utilized in these studies. The first was designed to determine if the steps taken to avoid over training limited the ability of the discriminant function to correctly classify samples. In this experiment the discriminant function performance on a cell by cell analysis was investigated as a function of the number of features included. The results shown in figure 3.5 72 Wilton demonstrated that the ability of the function to discriminate between two treatment groups reached a plateau at approximately 12 features. Thus, the strict criteria implemented to prevent over training did not sacrifice the ability of the discriminant function to correctly classify samples. The second experiment was designed to determine if there was any detectable over training occurring using the above methodology. This test randomly assigned normal cells into two test sets. Discriminant function analysis was performed in attempt to classify these two artificially generated groups. This exercise resulted in a 51.2% correct classification efficiency, which is essentially random. This test also indicated that the statistical methodology employed in these studies does not over train the classifier. Therefore, the discrimination achieved in each experiment is a true measure of morphological difference between the two populations. The direct co-culture model was designed to obtain further evidence supporting the cause and effect relationship of malignancy to M A C . The first investigation undertaken with this model was the determination of the time course of M A C induction. This study demonstrated that a significant induction of M A C was achieved within 24 hours of direct culture of normal cells with tumor cells. The peak strength of M A C induction occurred at the 48 hour time point as measured by the cell by cell discrimination efficiency. The strength of M A C expression declined slightly by 72 hours, however this expression still remained highly significant. The latter observation was very interesting as it indicated that M A C represent at least a partially reversible dynamic phenomenon. It is unclear why the M A C expression declined in this model. It is possible that the metabolic activity and 73 Wilton growth factor production of the tumor cells declined as the metabolites in the medium were consumed and/or as the density of tumor cells increased on the slide. Alternatively the normal cells may, with time, down regulate their receptors for the growth factor produced by the tumor cells. If M A C is in fact a reversible event the mechanism of this phenomenon can not be due to irreversible D N A damage as the Field Cancerization theory postulates. The most compelling evidence which demonstrated that soluble chemical mediators released by malignant cells are directly responsible for inducing M A C could be found in the dose dependency studies. In the study shown in figure 3.7 and 3.8, a clear correlation between the strength of M A C expression and the dose of tumor cell burden was seen. In this experiment slides were identified as M A C based on the average of the weighted nuclear feature expression for the discriminant function generated on the 10:1 dose versus the control. This classifier produced an accuracy of 57% for the dose having 1:1 tumor to normal cell numbers . A classification accuracy of 70% was achieved for the 5:1 dose and 95% for the 10.T tumor to normal cell dose. This result is very convincing as the intermediate doses are true test sets. The dependency of the strength of induction of M A C in vitro on the dose of tumor burden could be interpreted to be due to the concentration of a soluble chemical mediator(s) released by these malignant cells. A n equally convincing result was obtained in a subsequent dose dependency experiment. In this experiment M A C were induced in a dose dependent manner, however the 74 Wilton classifier was trained on a cell by cell analysis of the 5:1 tumor to normal cell dose. This protocol resulted in the 10:1 dose being a true test set. Interestingly, the classifier generated on the 5:1 dose performed better on the 10:1 dose test set. Thus, the nuclear feature expression changed in a similar direction but to a larger degree in the higher tumor burden dose in comparison to the medium tumor dose. This provided further evidence that M A C induction in vitro was reproducible and that the strength of expression of this phenomenon is dependent on the tumor burden or dose of soluble factor released by the tumor. This result provides support to theories which postulate that the strength of M A C expression is dependent on the type and size of tumor and therefore may have significant implications on clinical prognosis. The expression of M A C in these studies was quantified in terms of Mean Mac Score. This measure indicates the average of the discriminant function score for 200 individual cells per sample. Therefore, this represents an average of the nuclear morphology of the entire population in a sample. The first dose dependency study is also represented in terms of frequency of individual cells classified as M A C per sample in Figure 3.9. This representation of the data reveals an interesting phenomenon suggesting that within the population of each sample there were cells which have normal morphology and those which display M A C morphology. Furthermore as this figure illustrates, the induction of M A C in vitro results in an increase in the frequency of cells classified as M A C in a dose dependent manner. This observation could be interpreted as meaning the M A C phenotype is a normal physiological state which is present at a low frequency in normal Wilton cell populations. It is plausible that the induction of cells in to the M A C state requires a threshold of stimulus (e.g. growth factors) which when reached causes each individual cell to alter its physiology and therefore its morphology. The physiological changes which translate to the physical morphological changes interpreted as M A C are unknown. Current models of gene regulation postulate that D N A topology and high order chromatin structure is an important contributor to gene regulation89-92 Based on these models, it can be postulated that the local regions of chromatin distribution changes observed in M A C nuclei are alterations in D N A supercoiling which expose the naked D N A to the transcriptional machinery. This transition from a condensed state to a more exposed decondensed state removes the steric hindrance of higher order chromatin structure and allows for transcription of genes within the newly decondensed domain. Thus, the morphological manifestation observed as M A C could be the result of alterations in the compliment of actively expressed genes in a subset of the ostensibly normal cells surrounding the lesion. A key to understanding the physiologic manifestations of M A C and the substances responsible for their induction may be found in measurements of the cell cycle distribution of M A C tissues. There are conflicting reports in the literature with respect to the physiologic effect of malignancy on the normal tissue adjacent to it. Several studies indicate that there is an increased P C N A expression in tissues adjacent to cancer4?, however the physiologic significance of this is debatable. Some claim that there is an 76 Wilton increased cycling fraction in this tissue4 5 while others believe that the altered P C N A expression is induced by growth factors produced by the tumor and is not accompanied by increased cycling fraction4 8. The cell cycle analysis performed on the data from the two dose dependency experiments described in this work indicated that M A C induced in vitro resulted in a small but statistically significant decrease in the number of cells in G1/G0 phase and a statistically significant increase in the number of cells in the G2/M phase of the cell cycle. There was no statistically significant change in the S-phase fraction. This data could be interpreted in two ways. The first explanation suggests that M A C induces a growth inhibition associated with an accumulation of cells in G2/M phase of the cell cycle. The alternate, explanation for this observation would be that M A C is associated with a growth stimulus associated with a short S-phase relative to the G2/M phase. The G2/M phase could be prolonged as the cells in this compartment of the cell cycle are poised to divide but are delayed due to a lack of space. The simplest way to resolve these conflicting interpretations would be to perform a ^H-Thymidine incorporation assay using the indirect model to determine if the induction of M A C is accompanied by an increased D N A synthesis in the normal cells. In this model the normal cells were utilized two days after reaching confluency. As these cells were growth inhibited by cell to cell contact any growth enhancing or inhibition stimulus would not have had a profound effect on the distribution of cells in the cell cycle. This is also true for the in vivo situation as the majority of cells used to measure M A C are differentiated. Although M A C is likely to be the manifestation of altered gene 77 Wilton expression, the subtle cell cycle effect measured in vitro may provide clues as to which growth factors are likely candidates based on their physiological effect on normal bronchial epithelium. A ^H-thymidine incorporation assay coupled with this analysis would provide more sensitive clues as to the physiological stimulus associated with M A C . The gross physiological stimulus associated with several of the candidate growth factors for the induction of M A C in this model are relatively well understood. Using neutralizing TGF-a antiserum and tryphostin TGF-oc/EGF receptor tyrosine kinase inhibitor, Nettesheim et al. showed that TGF-a is utilized by normal and transformed cells as an autocrine mitogenic factor94. Interestingly, in this model normal cells were found to down regulate TGF-a expression when they reached confluency, whereas transformed cells did not. This would explain an autocrine growth advantage for tumor cells which produce this growth factor. In another study the response of cultured bronchial tumor cells to TGF-a was found depend on the dose of growth factor 95, i n this model a growth stimulatory response was observed at levels of 0. lng/ml whereas at levels of 10 ng/ml of TGF-a the cells were growth inhibited. Therefore the physiologic manifestation of this growth factor may also depend on its concentration in our model. A similar observation for EGF95 suggests that these growth factors can induce differentiation as well as growth stimulation under proper conditions. TGF-P has been shown to inhibit the growth of normal human bronchial epithelial cells however this growth restraining mechanism has been demonstrated to be inoperative in many 78 Wilton transformed airway epithelial tumor cell lines93. This would also confer a growth advantage to tumor cells producing this growth factor. Finally Insulin-like growth factor has been shown to have a mitogenic effect on airway epithelial ce l l s^ . Each of these growth factors has been shown to be produced in supraphysiological amounts by N S C L C and therefore could be considered as prime candidates for the induction of M A C in normal tissue via a paracrine mechanism^. The indirect co-culture model employed in these studies can be utilized in future studies to determine the growth factor responsible for inducing M A C in vitro. This model allows for treatment of the medium and subsequent testing of its biological activity on the normal cultured cells. It has also been demonstrated that the strength of M A C expression is dose dependent, therefore this model can be refined into a useful biological assay for M A C activity. In order to utilize the model for this purpose, the expression of M A C will need to be reliably quantifiable. To achieve this a very large quantity of conditioned medium will need to be generated. A n adequate concentration of conditioned medium required to illicit a significant M A C response in the normal cells must be determined. Following this, several tightly controlled experiments will have to be run in which an identical amount of conditioned medium should be used to stimulate M A C in the N H B E cells. A discriminant function should then be generated using cells pooled from several identical experiments. Such a derived discriminant function should then be used in subsequent trials to quantitate the strength of M A C expression. 79 Wilton The isolation of the factor(s) responsible for inducing M A C can begin by determining i f the conditioned medium retains its activity following heat treatment and, independently, acid treatment. The tertiary structure of many growth factors can be destroyed under appropriate conditions of heating or extremes of pH. This information will assist in determining the characteristics of the unknown factor (e.g. i f no activity is lost following heat denaturation the factor could be lipid based or a short polypeptide or amine). The next step in the identification of the factor (assuming it is a protein) would be to fractionate the proteins in the conditioned medium into known size fragments and test these fractions for their ability to induce M A C . In a preliminary study, medium conditioned on NCI-H460 cells for 48 hours was fractionated using 10 000 M W C O KwikspinTM Macro Ultrafiltration Units (Pierce, Rockford, IL.). Figure 4.1 shows the result of classification of samples treated with the two fractions of conditioned medium. This sample classification is based on a discriminant function generated on a cell by cell basis for the greater than 10 000 M W fraction versus the control. Wilton 80 S i ze Fractionation of M A C Conditioned Medium o N O R M A L o M A O 10000 kDA o M A C < 10000 kDA 81 % correct classification o 80% correct classification ' o o -Co-P-O OO! • • • ra 87.5% correct classification i -2.5 -1.5 -0.5 0.5 M E A N M A C S C O R E 1.5 2.5 Figure 4.1: Indirect co-culture of NHBE treated with size fractionated medium conditioned on NCI-H460 cells for 48 hours. Medium was fractionated with a 10 000 MWCO filter This preliminary experiment indicated that there was MAC-inducing activity in both fractions of the conditioned medium. This could be explained by the fact that the growth factor responsible may be smaller than 10 000 M W and therefore could potentially be present in both fractions. Another explanation for this result could be that there are more than one size of the active growth factor secreted by this cell line. This theory is supported by evidence that secreted products of different molecular weights which can bind to the EGF receptor can be released by a single cell line96. For example TGFs of M W 6000-7000 and 20 000 have been purified from the same melanoma cell line. Alternatively, there may be more than one growth factor present in the cultured medium having the ability to induce M A C . 81 Wilton This experiment must be repeated, however its results already suggest that a greater resolution of the cultured medium into multiple size fractionations will be required for isolation of the growth factor(s). This can be accomplished using many commercially available versions of size exclusion column chromatography. Each fraction should then be tested and its ability to induce M A C quantified. The results of this procedure can be compared to known molecular weights of growth factors to obtain an idea of the growth factor(s) responsible. The molecular weights of some candidate growth factors are shown in table 4.1. Growth Factor Molecular Weight TGF-a 5.5 kD TGF-pl 25 kD EGF 6 k D PDGF 30-33 kD IGF-1 7.5 kD Table 4.1: Molecular weight by SDS-page of growth factors likely to induce MAC in lung tumors. Data from Damstrup et al."' The next step in the identification process would be to perform the indirect co-culture experiment with conditioned medium treated with neutralizing antibodies directed at hypothesized growth factor(s). If the hypothesis is correct, this procedure should eliminate the ability of the medium to induce M A C . Following this, the growth factor Wilton could be identified directly in the growth medium using a radio immunoassay and mRNA for the growth factor could be detected in the tumor cell by Northern blot analysis. Should the M A C activity not be attributable to one of the suspected growth factors using this method, one could attempt to identify the protein by treating the active fraction of the medium with trypsin under defined conditions. The size of fragments resulting from this degradation process could be determined using mass spectroscopy and compared to a database of known growth factors thereby identifying the unknown growth factor. 5.0 CONCLUSION The studies presented in this M.Sc. thesis represent the first time M A C have been induced successfully in vitro. Using this model, it was demonstrated that M A C can be induced in Normal Human Bronchial Epithelial cells by soluble chemical mediators released by a malignant N S C L C cell line in culture. Furthermore, the strength of M A C induction was found to be dose dependent. These results coupled with the body of evidence in vivo provide convincing evidence supporting the Humoral mechanism of M A C . They suggest a direct causal relationship of malignancy to Malignancy Associated Changes thereby lending strength to hypotheses which propose an important role for M A C in the detection, prognosis and treatment of malignancy. In future work, the in vitro model outlined in these studies can be further characterized into a sensitive biological assay for the isolation and identification of those factors Wilton 83 responsible for inducing M A C . It is probable that there are multiple chemical mediators responsible for inducing M A C in this model and, likewise, in vivo. Identification of these factors may lead to a better understanding of the pathophysiology of carcinogenesis and greater knowledge of how tumors interact with the host. This information will lead to a better understanding of the M A C phenomenon. Further investigation using the in vitro model coupled with on going studies of M A C in vivo may eventually lead to the implementation of this phenomenon as a powerful tool in the fight against cancer. Wilton 84 BIBLIOGRAPHY 1. Gruner O.C.:Study of the changes met with the leukocytes in certain cases of malignant disease. British Journal of Surgery. 3:505-522, 1916. 2. Nieburgs H.E., Herman B.E., Reisman H.: Buccal cell changes in patients with malignant tumors. Laboratory Investigation 11(1): 80-88, 1962. 3. Nieburgs H.E., Dreiling D.A., Rubio C , Reisman H.: The morphology of cells in duodenal-drainage smears: histologic origin and pathologic significance. American Journal of Digestive Diseases 7(6):489-505, 1962. 4. Nieburgs H.E., Reisman H. , Pacheco B.: Interpretation of cellular changes preceding invasive uterine cervix carcinoma. Cancer April, 1963. 5. Nieburgs H.E., Parets A . D., Perez V. , Boudreau C : Cellular changes in liver tissue adjacent and distant to malignant tumors. Archives of Pathology 80, 1965. 6. Nieburgs H.E., Goldberg A.F. , Bertini B. , Silagi J., Pacheco B., Reisman H. : Malignancy associated changes (MAC) in blood and bone marrow cells of patients with malignant tumors. Acta Cytologica ll(5):415-423, 1967. 7. Nieburgs H.E.: Recent progress in the interpretation of malignancy associated changes (MAC). Acta Cytologica 12(6):445-453, 1968. 8. Nieburgs H.E., Allen D., Reisman H. : Malignancy associated changes (MAC) in human skin sections distal to internal tumors. 11th Annual Meeting. Am. Soc. of Cytol., Columbus Ohio, 1963. 9. Romsdahi M . , Valaitis J., McGrath R.G., McGrew E.A.: Cytological changes of megakaryocytes in patients with carcinoma. 10. van Oppen Toth B., Meisels A. , St-Pierre S.: Malignancy associated changes in sputum: a correlated study of 315 patients. Acta Cytologica 19(6):573-576, 1975. 11. Finch R.R., von Haam E.: Malignancy associated changes in buccal smears. Acta Cytologica 15(l):46-49, 1971. 12. Finch R.R.: A classification of nuclear aberration in relation to malignancy associated changes (MAC). Acta Cytologica 18(6),1971. 13. Klawe, H. : Malignancy associated changes (MAC) in buccal mucosa distant from malignant tumors. Congressus Oncologicus, Kralovy Vary, Czechoslovakia, 1972. Wilton 85 14. Johnston B., Brady J.M.,: Malignancy related changes in peripheral blood smears. Acta Cytologica 14:399-403, 1970. 15. Klawe H. , Rowinski J.: Malignancy associated changes (MAC) in cells of buccal smears detected by means of image analysis. Acta Cytologica 18(l):30-33, 1974. 16. Wied G.L., Bartels P.H., Bibbo M . et. al.: Cyomorphometric markers for uterine cancer in intermediate cells. Analyt. Quant. Cytol. 2:257-263, 1980. 17. Bartels P.H., Bibbo M . , Dytch H.E. et al.: Diagnostic marker displays for intermediate cells from the uterine cervix. Acta Cytologica 26:29-34, 1982. 18. Burger G., Jetting U . , Rodenacker K. , : Changes in benign cell populations in cases of cervical cancer and its precursors. Analyt. Quant. Cytol. 3:261-271, 1981. 19. Voojis P.G., Oud P.S., Zahniser D.J. et al.: Chromatin measurements in feulgen-stained intermediate cells from normal and abnormal cervical smears. Analyt. Quant. Cytol. 4:154, 1982. 20. Reinhardt E.R., Lenz R., Greiner W. et al.: Diagnostic relevance of visually non-suspect cells for cancer diagnosis: I. results of feature evaluation and their consequences for automated specimen analysis. Analyt. Quant. Cytol. 4:157, 1982. 21. Rosenthal D.L., Missirlian N . , McLatchie C. et al.: Predicting cervical epithelial disease states using subcategorizes on intermediate cells. Analyt. Quant. Cytol. 5:217, 1983. 22. Boon M.E. , Jetting U . , Rodenacker K et al.: Differences in chromatin pattern of benign cells in smears with condylomatous CIN I and III. In: Proc. 5th Int.Conf. on Automation of Diagnostic Cytol. and Histol, Brussels, Belgium, p. 23, 1986. 23. Kwikkel H.J., Timmers T., Boon M.E. et al.: Relation of quantitative features of visually normal intermediate cells in cervical intraepithelial neoplasia I and II smears to progression or non-progression of the lesion. Analyt Quant. Cytol. Histol. 9:405-410, 1987. 24. Haroske G., Kunze K.D. : Comparative image analysis studies of cervical paraneoplasias on smears and tissue sections. In: Proc. 5thInt.Conf. on Automation of Diagnostic Cytol. and Histol., Brussels, Belgium, p. 31-32, 1986. 25. Montag A . G . , Bartels P.H., Lerma-Puertas E., Dytch H.E., Leelakusolvong S., Bibbo M . : Karyometric marker features in tissue adjacent to in situ cervical carcinomas. Analyt. Quant. Cytol. ll(4):275-285,1989. Wilton 8 6 26. Bibbo M., Montag A.G., Lerma-Puertas E., Dytch K.E., Leelakusolvong S., Bartels P.H.: Karyometric marker features in tissue adjacent to invasive cervical carcinomas. Analyt. Quant. Cytol. 11:281-285, 1989. 27. Minna J.D., Pass H., Glatstein E., Ihde D.C.: Lung cancer. In: Principles and Practice of Oncology. V.T. DeVita, S. Rosenberg , and S. Hellman (eds), W.B. Saunders, Philadelphia, pp. 591-705,1989. 28. Wied G.L., Bibbo M. Pishotta F.T., Bartels P.H.: Intermediate cell markers for malignancy, consistency of expression. Analyt. Quant. Cytol. 6(4):243-246, 1984. 29. Haroske G., Berander S. Konig R.: Frequency and diagnostic reliability of subvisual morphological markers for malignancy in the cervical epithelium. Arch Gesechwulstforsch 58(3):159-168, 1988. 30. Haroske G., Berander S., Konig R., Meyer W.: Application of malignancy-associated changes of the cervical epithelium in a hierarchic classification concept. Analytical Cellular Pathology 2:189-198, 1990. 31. Zhanister D.J., Wong K.L., Brenner J.F., Ball H.G., Garcia G.L., Hutchinson M.L.: Contextual analysis and intermediate cell markers enhance high-resolution cell image analysis for automated cervical smear diagnosis. Cytometry 12:10-14, 1991. 32. Isenstein L.M., Zhanister D.J., Hutchinson M.L.: Combined malignancy associated change and contextual analysis for computerized classification of cervical cell monolayers. Analytical Cellular Pathology 9:83-93, 1995 33. Swank P.R., Greenberg S.D., Hunter N.R., Spjut H.J., Estrada R. et al.: Identification of features of metaplastic cells in sputum for the detection of squamous-cell carcinoma of the lung. Diagnostic Cytopathology 5(3):98-103. 34 Payne P.W., Sebo T.J., Doudkine A., Garner D., MacAulay C , Lam S., LeRiche J.C., Palcic B.: Sputum screening by quantitative microscopy: a reexamination of a portion of the national cancer institute cooperative early lung cancer study. Mayo Clinic Proceedings 72(8):In Press. 35. Bibbo M., Michelassi F., Bartels P.H., Bania C , Lerma E., Montag A.G.: Karyometric marker features in normal-appearing glands adjacent to human colonic adenocarcinoma. Cancer Research 50:147-151, 1990. 36. Montag A.G., Bartels P.H., Dytch H. E., Lerma-Puertas E., Michelassi F., Bibbo M.: Karyometric features in nuclei near colonic adenocarcinoma, statistical analysis. Analyt. Quant. Cytol. and Hist. 13(3): 159-167, 1991. Wilton 87 37. MacAulay C , Lam S., Payne P., LeRiche J.C., Palcic B.: Malignancy associated changes in bronchial epithelial cells in biopsy specimens. Analyt. Quant. Cytol. and Hist. 17(1):55-61, 1994. 38. Susnik B., Worth A. , Palcic B., Poulin N . , LeRiche J.: Differences in quantitative nuclear features between ductal carcinoma in situ (DCIS) with and without accompanying invasive carcinoma in the surrounding breast. Analytical Cellular Pathology 8:39-52, 1995. 39. Burger G., Aubelle M . Clasen B., Juttig U . , Gais P., Malignancy associated changes in squamous epithelium of the head and neck region. Analytical Cellular Pathology 7(3):181-93, 1994. 40. Greenberg S.D., Spjut H.J., Estrada R.G., Hunter N.R., Grenia C : Morphometric markers for the evaluation of preneoplastic lesions in the lung, diagnostic evaluation by high-resolution image analysis of atypical cells in sputum specimens. Analyt. and Quant. Cytol. and Hist. 9(l):49-54,1987. 41. Soffos R.O., Rhatigan R .M. : Benign (nonpolypoid) mucosal changes adjacent to carcinomas of the colon. Human Pathology 8(4):441-449, 1977. 42. Dawson P. A . , Filipe M.I.: A n ultrastructural and histochemical study of the mucus membrane adjacent to and remote from carcinoma of the colon. Cancer 37:2388, 1976. 43.Shamsuddin A . K . M . , Weiss L. , Phelps P.C., Trump B.F.: Colon Epithelium. IV human colon carcinogenesis, changes in human colon mucosa adjacent to and remote from carcinomas of the colon. Journal of the National Cancer Institute 66(2):413-419, 1981. 44. Caldero J. Campo E. Ascaso C , Ramos J., Panades M.J. , Rene J.M.: Regional distribution of glycoconjugates in normal, transitional and neoplastic human colonic mucosa, a histochemical study using lectins. Virchows Archiv A Pathol Anat. 415:347-356, 1989. 45. Terpstra O.T., Blankenstein M . , Dees J., Eilers G . A . M . : Abnormal pattern of cell proliferation in the entire colonic mucosa of patients with colon adenoma or cancer. Gastroenterology 92:704-8, 1987. 46. Ngoi S.S., Staiano-Coico L. , Goodwin T.A., Wong R.J., DeCosse J.J.: Abnormal D N A ploidy and proliferative patterns in superficial colon epithelium adjacent to colorectal cancer. Cancer 66:953-59,1990. 47. Hall P.A., Coates P.J., Goodlad R.A., Hart I.R., Lane D.P.: Proliferating cell nuclear antigen expression in non-cycling cells may be induced by growth factors in vivo. Br. J. Cancer 70:244-47, 1994. 88 Wilton 48. Hall P.A., Levison D.A., Woods A.L., Kellock D.B., Watkins J.A. et al.: Proliferating cell nuclear antigen (PCNA) immunolocalization in paraffin sections: an index of cell proliferation with evidence of deregulated expression in some neoplasms. Journal of Pathology 162:285-294, 1990. 49. Harrison R.F., Reynolds G.M., Rowlands D.C.: Immunohistochemical evidence for the expression of proliferating cell nuclear antigen (PCNA) by non-proliferating hepatocytes adjacent to metastatic tumors and in inflammatory conditions. Journal of Pathology 171:115-122 (1993). 50. Ogden G.R., Cowpe J.G., Green M.W.: The effect of distant malignancy upon quantitative cytological assessment of normal oral mucosa. Cancer 65:477-480, 51 Benedet J.C., Anderson G.H., Matisic J.P.: A comprehensive program for cervical cancer detection and management. Am. J. Obstet. Gynecol. 166(4), 1992. 52. Palcic B.: Nuclear texture: can it be used as a surrogate endpoint biomarker? Journal of Cellular Biochemistry 19:40-46, 1994. 53. Frost J.K., Ball W.C., Levin M.I.: Sputum cytopathology: use and potential in monitoring the workplace environment by screening for biological effects of exposure. J. Occup. Med. 28:692-703, 1986. 54. Fontana R.S., Sanderson D.R., Woolner C.B., Taylor W.F., Miller W.E., Muhm J.R.: Lung cancer screening : the mayo program. J. Occup. Med. 28:746-750, 1986. 55. Guillaud M., Doudkine A., Garner D., MacAulay C , Palcic B.: Malignancy associated changes in cervical smears: systemic changes in cytometric features with the grade of dysplasia. Analytical Cellular Pathology 9:191-204, 1994. 56. Palcic B., MacAulay C : Malignancy associated changes, can they be employed clinically? In: Wied G.L., Rosenthal D.L., eds. Compendium on computerized cytology and histology laboratory. Tutorial of Cytology, Chicago, IL, USA 1994; 157-165. 57. Strong M.S., Incze J., Vaughan C.W.: Field cancerization in the aerodigestive tract-its' etiology, manifestation and significance. J. Otolaryngology 13:1-6, 1980. 58. MacAulay C , Payne P., Wilton D., Garner D. Doudkine A., Lam S., Palcic B.: MACs: field cancerization or signal response? Presented at ISAC Congress XVIII, Rimini, Italy. April 13-18, 1996. 59. Sporn M.B., Todaro G.J.: Autocrine secretion and malignant transformation of cells N.Engl. J. Med. 303:878-880,1980. 89 Wilton 60. Sporn M.B . , Roberts A .B . : Autocrine growth factors and cancer. Nature 313(28):745-747, 1985. 61. Salomon D.S., Brandt R., Ciardiello F.: Epidermal growth factor-related peptides and their receptors in human malignancies. Critical Reviews in Oncology/Hematology 91:183-232,1991. 62. Kelly K. , Kane M.A. , Bunn P.A.: Growth factors in lung cancer: possible etiologic role and clinical target. Medical and Pediatric Oncology 19:449-458, 1991. 63. Damstrup L. , Rorth M . , Skovgaard P.: Growth factors an growth factor receptors in human malignancies with special reference to human lung cancer. Lung Cancer 5:49-68,1989.. 64. Carney D.N.: Lung cancer biology. European Journal of Cancer 27(3):369-372, 1992. 65. Shuller H . M . : Receptor mediated mitogenic signals and lung cancer. Cancer Cells 3(12):496-503, 1991. 66. Nettesheim P., Ferriola P., Robertson A.T.: Abnormalities of growth factor systems in transformed airway epithelial cells. In: Multistage Carcinogenisis, C.C. Harris et al. eds., Japan SCI Press, Boca Raton, 171-181, 1992. 67. Gorgoulis V . , Aninos D., Knavaros P.: Expression of EGF, TGF-a and EGFR in squamous lung carcinomas. Anticancer Research 12:1183-1188, 1992. 68. Tateishi M . , Ishida T., Mitsuodomi T.: Immunohistochemical evidence of autocrine growth factors in adenocarcinoma of the human lung. Cancer Research 50:7077-7080, 1990. 69. Seymour J.F.: Malignancy associated hypercalcemia. Scientific American Science & Medicine. September\October:48-57, 1995. 70. Sporn M.B . , Roberts A .B . : Autocrine secretion 10 years later. Annals of Internal Medicine 117(5):408-413, 1992. 71. Schuller H . M . : Receptor-mediated mitogenic signals and lung cancer. Cancer Cells 3(12):496-503, 1991. 72. Cross M . , DexterM.T.: Growth factors in development transformation and tumorigenisis. Cell 64:271-280, 1991. 73. Brattain M.G. , Howell G., Sun C , Wilson J.K.: Growth factor balance and tumor progression. Current Opinion in Oncology 6:77-81, 1994. 90 Wilton 74. Gamer D.M., Harrison A., MacAulay C , Palcic B.: Cyto-Savant and its use in automated screening of cervical smears. In Wied G.L., Bartels P.H. eds. Compendium on computerized cytology and histology laboratory. Tutorial of Cytology, Chicago, II, USA, 1994, 75. Doudkine A., MacAulay C , Poulin B., Palcic B.: Nuclear texture measurements in image cytometry. Pathologica 87:286-299, 1995. 76. Feulgen R., Rosenbeck H.: Mikroskopischchemischer nachweiseiner nukleinsaure nom typus der thymonuklunsqure und die darauf beruhende elektive farbung von zellkernenin mikroskopischen paraparatin. Z Physiol Chem. 125:203-248, 1924. 77. Lechner, J.F et al.: In vitro human bronchial epithelial model systems for carcinogenisis studies. In, In Vitro Models For Cancer Research, Volume VI, Webber M.W. eds. CRC Press, 1988. 78 Gazdar A.F., Texas Southwestern Medical Center, Houston Texas. Personal Communication. 79 Bravo R., Rainer F., Blundell P.A., Macdonald-Bravo H.: Cyclin/PCNA is the auxilliary protein of DNA polymerase-5. Nature 326:515-517, 1987. 80. Sntisteban M.S., Brugal G.: Image analysis of in situ cell cycle related changes of PCNA and Ki-67 proliferating antigen expression. Cell Prolif. 27:435-453, 1994. 81. Bartels P.H.: Numerical evaluation of cytological data VII. multivariate significance tests. Analytical and Quantitative Cytology 3(1): 1-8, 1981. 82 Baricikowski R., Stevens J.P.: A monte carlo study of the stability of canonical correlations, canonical weights, and canonical variate-variable correlations. Multivariate Behavioral Research, 10:353-364, 1975. 83. Bartels P.H.: Numerical evaluation of cytologic data IV. discrimination and classification. Analytical and Quantitative Cytology 2(l):19-24, 1980. 84. Rubin E., Farber J.: Pathology, second ed. Philadelphia, J.B. Lippincott Company, 1994. 85. Kumar V, Cotran RS, Robbins SL: Basic Pathology, fifth ed. Philadelphia, W.B. Saunders Company, 1992. 86. Slaughter D.P., Southwick H.W., Smejkal W.: "Field Cancerization" in oral stratified squamous epithelium: clinical implications of multicentric origin. Cancer, 6, 1954. Wilton 91 87 Freshney I.R.: Culture of Animal Cells a Manual of Basic Techniques, 2nd ed. New York, Alan R. Liss Inc., 1987. 88. Kiernan J.A.: Histological and Histochemical Methods Theory and Practice, Second ed. New York, Pergamon Press, 1990. 89 Paranjape S.M., Kamakaka R.T., Kadonaga J.T.: Role of chromatin structure in the regulation of transcription by RNA polymerase II, Annual Revue of Biochemistry, 65:265-97, 1994. 90 Grunsteiri M.:Histones as regulators of genes, Scientific American, October:68-74, 1992. 91 Felsenfeld G.: Chromatin structure and gene expression, Current Opinion in Cell Biology, 4:436-43, 1992. 92 Dillon N., Grosveld F.: Chromatin domains as potential units of eukaryotic gene function, Current Opinion in Genetics and Development, 4:260-64, 1994. 93 Payne P. Cytometric Detection of Nuclear Features Associated With Pre-Malignancy or Malignancy in Human Bronchial Specimens. PhD thesis, The University of British Columbia, May 1997. 94 Nettesheim P., Ferriola P.C., Robertson A.T. et al.: Abnormalities of growth factor systems in transformed airway epithelial cells. In Multistage Carcinogenisis. C.C. Harris et al (EDS), Japan Sci. Soc. Press, Boca Raton, pp. 171-181, 1992. 95. Seigfried J.M.: Owens S.E.: Response of primary human lung carcinomas to autocrine growth factors produced by a lung carcinoma cell line, Cancer Research, 48:4976-4971,1988. 96. Siegfried J.M.: Detection of human lung epithelial cell growth factors produced by a lung carcinoma cell line: use in culture of primary solid lung tumors. Cancer Research, 47:2903-2910, 1987. Wilton 92 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items