Trop souvent diagnostiqués à des stades tardifs du fait de leur quasi asymptomatie, les adénocarcinomes séreux ovariens posent un véritable problème de santé publique. Malgré les progrès récents de prise en charge chirurgicale, l’émergence d’une maladie résiduelle microscopique chimiorésistante impacte grandement le pronostic des patientes.Le microenvironnement tumoral est un acteur clé de la progression tumorale et de l’émergence de résistances aux traitements anticancéreux. Durant ces travaux de thèse, nous nous sommes intéressés à deux composants majeurs du stroma tumoral, d’une part les cellules souches mésenchymateuses, d’autre part les cellules endothéliales.Nous avons pu démontrer que les cellules souches mésenchymateuses participent à la progression tumorale et l’émergence de résistances. Enfin nous avons démontré que les cellules endothéliales, via la production de facteurs angiocrines, participent à la chimiorésistance des cellules tumorales ovariennes.Dans ce travail, nous avons pu définir de nouvelles cibles thérapeutiques mettant en jeu la relation entre les cellules tumorales ovariennes et l’hôte. / Ovarian cancers constitute a poor prognosis disease. Due to their absence of symptoms, ovarian cancers are generally diagnosed at late stages. Despite major breakthrough regarding ovarian cancer surgery, minimal residual disease-induced relapse is still a hurdle for clinicians.Tumor microenvironment is a key actor on disease progression and resistance to therapy. In this study, we have focused on two major components of the tumor stroma, on one hand, the mesenchymal stem cells, and the endothelial cells on the other hand.We were able to demonstrate that mesenchymal stem cells are critically involved in ovarian cancer progression and resistance to therapy, while the endothelium, through production of angiocrine factors, is deeply involved in resistance of ovarian cancer cells to platinum and taxane based therapy.Here, we set forth the idea that disrupting the relationship between ovarian cancer cells and their host stroma constitute a new therapeutic window.

The American Cancer Society predicts there will be 110,070 new cases and 32,120 deaths due to gynecological malignancies in 2018. A major contributing factor to the high mortality associated with gynecological cancers is the recurrence of treatment-resistant tumors. Ovarian cancer (OC) remains the most lethal gynecological malignancy, yet the mechanisms responsible for regulating tumor resistance and vulnerability are largely unknown or undruggable. Therefore, the goal of this research is to identify mechanisms responsible for therapeutic resistance in gynecological cancers and discover innovative approaches to circumvent these molecular alterations. Our efforts began in OC where secondary analysis of gene expression data from OC studies revealed that Sab, an outer mitochondrial membrane (OMM) scaffold protein, was down-regulated in OC tumors compared to normal tissue controls. Our previous studies demonstrate that Sab-mediated OMM signaling induces cell death in cervical cancer. In the current study, we found that Sab concentrations corresponded to chemoresponsiveness in a panel of OC cells; wherein, OC cells with low Sab levels were chemo resistant. Dynamic BH3 profiling revealed that cells with high Sab expression were primed for apoptosis. Furthermore, over-expression of Sab in chemo resistant cells enhanced apoptotic priming and restored cellular vulnerability to cisplatin/paclitaxel treatment. Additionally, an examination of treatment-resistant metastatic uterine cancer (UC) cells were found to have low Sab concentrations compared to vulnerable primary site-derived UC cells. Ectopic expression of Sab in chemo resistant UC cells enhanced the susceptibility towards megestrol acetate and BH3-mimetic ABT-737. To exploit the relationship between Sab concentrations and chemo-responsiveness in gynecological cancer cells, we developed a high-throughput screening assay to detect Sab levels in chemo-resistant OC cells. In collaboration with the Torrey Pines Institute for molecular studies, we have identified compounds that can increase Sab levels in resistant OC cells. The identified compounds improved the effectiveness of cisplatin/paclitaxel therapy. We propose that Sab may be a prognostic marker to discern personalized treatments for gynecological cancer patients. Furthermore, pharmacologically enhancing Sab-mediated signaling may increase the efficacy of chemotherapeutic agents, which would mean lower doses that would limit toxic side-effects.

<p>OVARIAN cancer research focuses on answering important questions related to the disease, determining whether new approaches are feasible to contribute towards improving current treatments or discovering new ones. This study focused on the transcriptional regulation of genes that have been implicated in ovarian cancer, based on the occurrences of single nucleotide polymorphisms (SNPs) within transcription factor binding sites (TFBSs). Through the application of several in silico tools, databases and custom programs, this research aimed to contribute toward the identification of potentially bio-medically important genes or SNPs for pre-diagnosis and subsequent treatment planning of ovarian cancer. A total of 379 candidate genes that have been experimentally associated with ovarian cancer were analyzed. This led to the identification of 121 SNPs that were found to coincide with putative TFBSs potentially influencing a total of 57 transcription factors that would normally bind to these TFBSs. These SNPs with potential phenotypic effect were then evaluated among several population groups, defined by the International HapMap consortium resulting in the identification of three SNPs present in five or more of the eleven population groups that have been sampled.</p>

Ovarian cancer is one of the most common female cancer diseases in the world today and in Sweden more than 800 new cases are diagnosed every year. The standard treatment consists of chemotherapy with paclitaxel in combination with carboplatin after initial cytoreductive surgery. The response to treatment and the severity of adverse drug reactions after chemotherapy varies greatly among individuals, and one of the most important factors responsible for these differences is now recognized to be the genetic variability. One of the major obstacles to successful treatment is drug resistance. Several potential mechanisms have been suggested for the resistance to paclitaxel, such as mutations in the target protein β-tubulin, single nucleotide polymorphisms (SNPs) in the gene ABCB1, which encodes the transport protein P-glycoprotein. P-glycoprotein can mediate efflux of various drugs from cancer cells as well as from the circulation into the intestinal lumen, and overexpression and/or high activity leads to drug resistance and/or increased elimination. Another reason might be the high interindividual variability of paclitaxel plasma concentrations, which has been suggested to be influenced by variability in metabolic enzymes, such as CYP2C8 and CYP3A4, and transport proteins e.g. P-glycoprotein. In the studies constituting this thesis we have investigated the possibilities of predicting the pharmacokinetics of paclitaxel as well as the tumor response and adverse drug reactions after chemotherapy in the preparation of personalized chemotherapy. We studied the correlation between the response and the presence of mutations in the dominant β-tubulin gene and SNPs in ABCB1. DNA from 40 ovarian tumors was screened for sequence variations in the β-tubulin gene without finding any, showing that β-tubulin mutations are rare and unlikely to be a clinically relevant resistance mechanism for paclitaxel. The SNPs G2677T/A and C3435T in the ABCB1 gene were determined in 53 ovarian cancer tumors from patients with poor (progressive disease or relapse within one year) or good (disease-free survival of more than one year) response to paclitaxel-carboplatin chemotherapy. Patients homozygously mutated for G2677T/A had a higher probability of responding to chemotherapy. There was also a dose-dependent influence of the number of mutated alleles on the response to paclitaxel treatment. No correlation was found for the C3435T variant. By using a newly developed quantitative LC/MS method for the simultaneous determination of paclitaxel and its hydroxymetabolites in human plasma we assessed the individual elimination of paclitaxel in 33 ovarian cancer patients. The patients were genotyped for SNPs in the ABCB1, CYP2C8 and CYP3A4 genes and their in vivo CYP3A4 enzyme activity, tumor response and toxicity, especially the neurotoxicity, were determined. Patients heterozygous for G/A in position 2677 in ABCB1 had a significantly higher clearance of paclitaxel than patients with the wild type or homozygously mutated, but not compared to patients carrying the G/T alleles. A lower clearance of paclitaxel was also found for patients heterozygous for CYP2C8*3 when stratified according to the ABCB1 G2677T/A genotype. The CYP3A4 enzyme activity in vivo affected the relative influence of CYP2C8 and CYP3A4 on the metabolism, but not the total clearance of paclitaxel. The exposure to paclitaxel was correlated to the neurotoxicity, but not to the treatment response. In conclusion, our findings suggest that the SNP G2677T/A in the ABCB1 gene, but not β-tubulin mutations, might be a predictor for paclitaxel response and that the interindividual variability in paclitaxel pharmacokinetics might be predicted by ABCB1 and CYP2C8 genotypes and provide useful information for individualized chemotherapy. / Ovarialcancer (äggstockscancer) är en av de vanligaste cancerformerna hos kvinnor i Sverige idag. Behandlingen består vanligen av tumörreducerande kirurgi följd av kemoterapi med paklitaxel och karboplatin. Målsättningen med detta avhandlingsarbete har varit att förbättra cytostatikabehandlingen (cellgiftsbehandlingen) med framförallt paklitaxel vid ovarialcancer genom att lägga grunden för individualisering av doser och förutsäga tumörsvaret vid behandlingen. Ett problem med dagens cancerbehandling är att många cancerceller så småningom blir resistenta mot olika cytostatika. För att angripa den mest resistenta cellen innan den induceras att öka uttrycket av, eller utveckla, fler resistensmekanismer vore det en fördel om vi före behandlingen kunde prediktera vilken dos av cytostatika som är bäst lämpad för individen samt om tumören kommer att reagera på behandlingen eller ej. En av de viktigaste faktorerna för skillnader i behandlingseffekt tros vara genetiska variationer mellan olika individer. I våra studier har vi använt genetiska metoder för att studera om vi kan prediktera tumörsvaret vid behandlingen genom att bestämma mutationer i genen för paklitaxels målprotein, β-tubulin, samt bestämma genetiska variationer i ABCB1-genen, kodande för transportproteinet P-glykoprotein. Tanken är att ett förändrat målprotein eller en förändrad förmåga hos cancercellerna eller kroppen att transportera ut paklitaxel skulle leda till en skillnad i påverkan på tumören. DNA från 40 ovarialtumörer analyserades utan att en enda sekvensvariation hittades i genen för β-tubulin, vilket tyder på att genetiska förändringar i genen för β-tubulin sannolikt inte är en klinisk relevant resistensmekanism. De normalt förekommande genetiska variationerna G2677T/A och C3435T i ABCB1-genen bestämdes i DNA från 53 ovarialtumörer där behandlingen endera givit en bra (tumörfri minst ett år) eller dålig (progression av tumören eller tumörfri mindre än ett år) anti-tumöreffekt. Patienter som var dubbelmuterade i position 2677 dvs hade endera T/T eller T/A (A/A hittades inte i materialet) i denna position hade en högre sannolikhet att få ett bra anti-tumörsvar vid behandlingen. Även antalet muterade baser påverkade utfallet, ju fler muterade baser i position 2677, desto högre sannolikhet att få ett bra svar på behandlingen. Andelen T eller A var också högre i den grupp av patienter som fått en lyckad behandling. För att kunna prediktera patientens individuella förmåga att bryta ner paklitaxel studerade vi inverkan av sekvensvariationer i generna för de nedbrytande enzymerna, CYP2C8 och CYP3A4, och transportproteinet P-glykoprotein (genen ABCB1) på eliminationen av läkemedlet i kroppen. Vi utvecklade en metod för att mäta paklitaxelkoncentrationerna i blodet och använde den för att studera hur snabbt 33 ovarialcancer patienter eliminerade cytostatikat från blodbanan. Hos dessa patienter bestämde vi förekomsten av kända genetiska variationer i generna ABCB1, CYP2C8 och CYP3A4 samt deras CYP3A4 enzymaktivitet i kroppen. Biverkningarna och tumörsvaret vid behandlingen utvärderades också. Eliminationen av paklitaxel hos dessa patienter var beroende av vilken bas som fanns i position 2677 i ABCB1-genen och förekomsten av den genetiska varianten CYP2C8*3. Enzymaktiviteten hos CYP3A4 kunde inte påvisas påverka eliminationen av paklitaxel utan snarare vilket enzym, CYP2C8 eller CYP3A4, som var relativt dominant i respektive patient. Exponeringen av paklitaxel korrelerade till den neurologiska påverkan som patienten orsakades av cytostatikat, men kunde inte korreleras till tumörsvaret vid slutet av cytostatikabehandlingen. Sammanfattningsvis ger patientens genetiska variationer i ABCB1, men inte β-tubulin, information om behandlingsutfallet. Genetiska variationer i CYP2C8 och ABCB1 påverkar patientens förmåga att eliminera paklitaxel och kan förhoppningsvis användas för att individualisera doserna. Vår förhoppning är att resultaten i denna avhandling skall kunna användas för att individualisera och ytterligare förbättra cytostatikabehandlingen vid ovarialcancer.

Ovarian cancer is thought to be derived from the ovarian surface epithelium (OSE), but it is often diagnosed during the late stages and therefore the events that contribute to the initiation and progression of ovarian cancer are poorly defined. Epidemiological studies have indicated an association between the female reproductive hormones and ovarian cancer etiology, but the direct effects of 17β-estradiol (E2), progesterone (P4), luteinizing hormone (LH) and follicle stimulating hormone (FSH) on disease pathophysiology are not well understood. A novel transgenic mouse model of ovarian cancer was generated that utilized the Cre/loxP system to inducibly express the oncogene SV40 large and small T-Antigen in the OSE. The tgCAG-LS-TAg mice developed poorly differentiated ovarian tumours with metastasis and ascites throughout the peritoneal space. Although P4 had no effect; E2 significantly accelerated disease progression in tgCAG-LS-TAg mice. The early onset of ovarian cancer was likely mediated by E2’s ability to increase the areas of putative preneoplastic lesions in the OSE. E2 also significantly decreased survival time in ovarian cancer cell xenografts. Microarray analysis of the tumours revealed that E2 mainly affects genes involved in angiogenesis and cellular differentiation, proliferation, and migration. These results suggest that E2 acts on the tumour microenvironment in addition to its direct effects on OSE and ovarian cancer cells. In order to examine the role of the gonadotropins in ovarian cancer progression, the tgCAG-LS-TAg mice were treated with 4-vinylcyclohexene-diepoxide (VCD) to induce menopause. Menopause slowed the progression of ovarian cancer due to a change in the histological subtype from poorly differentiated tumours to Sertoli tumours. Using a transgenic mouse model, it was shown that E2 accelerated ovarian cancer progression, while P4 had little effect on the disease. Menopause (elevated levels of LH and FSH) altered the histological subtype of the ovarian tumours in the tgCAG-LS-TAg mouse model. These results emphasize the importance of generating animal models to accurately recapitulate human disease and utilizing these models to develop novel prevention and treatment strategies for women with ovarian cancer.

Resistance to cisplatin (CDDP)-based therapy is a major hurdle to the successful treatment of human ovarian cancer (OVCA) and the chemoresistant phenotype in OVCA cells is associated with Akt-attenuated, p53-mediated apoptosis. Pro-apoptotic functions of p53 involve both transcription-dependent and -independent signaling pathways and dysfunctional localization and/or inactivation of p53 contribute to the development of chemoresistance. PARC is a cytoplasmic protein regulating p53 subcellular localization and subsequent function. Little is known about the molecular mechanisms regulating PARC. Although PARC contains putative caspase-3 cleavage sites, and CDDP is known to induce the activation of caspases and calpains and induce proteasomal degradation of anti-apoptotic proteins, if and how PARC is regulated by CDDP in OVCA is unknown. Here we present evidence that CDDP promotes calpain-mediated PARC down-regulation, mitochondrial and nuclear p53 accumulation and apoptosis in chemosensitive but not resistant OVCA cells. Inhibition of Akt is required to sensitize chemoresistant cells to CDDP in a p53-dependent manner, an effect enhanced by PARC down-regulation. CDDP-induced PARC down-regulation is reversible by inhibitor of calpain but not of caspase-3 or the 26S proteasome. Furthermore, in vitro experiments confirm the ability of calpain in mediating Ca2+-dependent PARC down-regulation. The role of Ca2+ in PARC down-regulation was further confirmed as ionomycin induced PARC down-regulation in both chemosensitive and chemoresistant ovarian cancer cells. The data presented here implicates the regulation of p53 subcellular localization and apoptosis by PARC as a contributing factor in CDDP resistance in OVCA cells and Ca2+/calpain in PARC post-translational processing and chemosensitivity.

Studies of prophylactic salpingectomy specimens from BRCA1/2 mutation carriers, at risk for tubal and ovarian high-grade serous carcinoma (SerCa), have consistently revealed occult carcinomas and putative histological cancer precursors in the distal fallopian tube epithelium (FTE), supporting the FTE as the source of SerCa. In this thesis I molecularly characterized and compared non-malignant FTE from mutation carriers (FTEb) and control patients (FTEn) to identify alterations that may predispose to malignant transformation. Gene expression profiling of laser capture microdissected FTEn, FTEb and SerCa indicated that SerCa have similar molecular profiles whether of presumed ovarian or tubal origin, supporting the notion they share a common cell of origin within the FTE. Furthermore, FTEb samples obtained during the post-ovulatory luteal phase showed gene expression profiles closely resembling SerCa samples, suggesting that the luteal phase milieu may contribute to serous carcinogenesis. An initial hypothesis was that FTEb may respond differently to luteal progesterone compared to FTEn, via differential expression of progesterone receptor (PR) isoforms. However, similar relative isoform expression in FTEn and FTEb samples suggested that a luteal phase-associated factor other than progesterone directs gene expression changes in FTEb. The possibility that FTEb respond differently to ovulation-associated inflammatory cytokines that are locally elevated during the luteal phase was next investigated. Importantly, FTEb specimens previously found to cluster with SerCa based on their global gene expression profiles showed evidence of increased nuclear factor-κB (NFκB)-dependent (pro-inflammatory) signalling and diminished glucocorticoid receptor (GR)-dependent (anti-inflammatory) signalling. Furthermore, I demonstrate that disabled homolog 2 (DAB2), an adaptor molecule decreased in SerCa and FTE luteal samples, enhances both GR-mediated transactivation and suppression of NFκB signalling, implicating DAB2 as a crucial determinant of inflammatory signalling and ovarian cancer risk. Altogether, this thesis identifies gene expression changes in FTE from BRCA mutation carriers during the post-ovulatory luteal phase that parallel those detected in SerCa. The data support a proposed novel testable model for predisposing events contributing to SerCa that centres on an altered ability to quickly resolve the pro-inflammatory environment created by the ovulatory event.

Studies of prophylactic salpingectomy specimens from BRCA1/2 mutation carriers, at risk for tubal and ovarian high-grade serous carcinoma (SerCa), have consistently revealed occult carcinomas and putative histological cancer precursors in the distal fallopian tube epithelium (FTE), supporting the FTE as the source of SerCa. In this thesis I molecularly characterized and compared non-malignant FTE from mutation carriers (FTEb) and control patients (FTEn) to identify alterations that may predispose to malignant transformation. Gene expression profiling of laser capture microdissected FTEn, FTEb and SerCa indicated that SerCa have similar molecular profiles whether of presumed ovarian or tubal origin, supporting the notion they share a common cell of origin within the FTE. Furthermore, FTEb samples obtained during the post-ovulatory luteal phase showed gene expression profiles closely resembling SerCa samples, suggesting that the luteal phase milieu may contribute to serous carcinogenesis. An initial hypothesis was that FTEb may respond differently to luteal progesterone compared to FTEn, via differential expression of progesterone receptor (PR) isoforms. However, similar relative isoform expression in FTEn and FTEb samples suggested that a luteal phase-associated factor other than progesterone directs gene expression changes in FTEb. The possibility that FTEb respond differently to ovulation-associated inflammatory cytokines that are locally elevated during the luteal phase was next investigated. Importantly, FTEb specimens previously found to cluster with SerCa based on their global gene expression profiles showed evidence of increased nuclear factor-κB (NFκB)-dependent (pro-inflammatory) signalling and diminished glucocorticoid receptor (GR)-dependent (anti-inflammatory) signalling. Furthermore, I demonstrate that disabled homolog 2 (DAB2), an adaptor molecule decreased in SerCa and FTE luteal samples, enhances both GR-mediated transactivation and suppression of NFκB signalling, implicating DAB2 as a crucial determinant of inflammatory signalling and ovarian cancer risk. Altogether, this thesis identifies gene expression changes in FTE from BRCA mutation carriers during the post-ovulatory luteal phase that parallel those detected in SerCa. The data support a proposed novel testable model for predisposing events contributing to SerCa that centres on an altered ability to quickly resolve the pro-inflammatory environment created by the ovulatory event.

Resistance to cisplatin (CDDP)-based therapy is a major hurdle to the successful treatment of human ovarian cancer (OVCA) and the chemoresistant phenotype in OVCA cells is associated with Akt-attenuated, p53-mediated apoptosis. Pro-apoptotic functions of p53 involve both transcription-dependent and -independent signaling pathways and dysfunctional localization and/or inactivation of p53 contribute to the development of chemoresistance. PARC is a cytoplasmic protein regulating p53 subcellular localization and subsequent function. Little is known about the molecular mechanisms regulating PARC. Although PARC contains putative caspase-3 cleavage sites, and CDDP is known to induce the activation of caspases and calpains and induce proteasomal degradation of anti-apoptotic proteins, if and how PARC is regulated by CDDP in OVCA is unknown. Here we present evidence that CDDP promotes calpain-mediated PARC down-regulation, mitochondrial and nuclear p53 accumulation and apoptosis in chemosensitive but not resistant OVCA cells. Inhibition of Akt is required to sensitize chemoresistant cells to CDDP in a p53-dependent manner, an effect enhanced by PARC down-regulation. CDDP-induced PARC down-regulation is reversible by inhibitor of calpain but not of caspase-3 or the 26S proteasome. Furthermore, in vitro experiments confirm the ability of calpain in mediating Ca2+-dependent PARC down-regulation. The role of Ca2+ in PARC down-regulation was further confirmed as ionomycin induced PARC down-regulation in both chemosensitive and chemoresistant ovarian cancer cells. The data presented here implicates the regulation of p53 subcellular localization and apoptosis by PARC as a contributing factor in CDDP resistance in OVCA cells and Ca2+/calpain in PARC post-translational processing and chemosensitivity.

The purpose of this Thesis is to present a Bayesian analysis of oncological data sets with particular focus on cervical carcinomas and ovarian cancers. Bayesian methods of data analysis have a very long history, and have been used with great success in many disciplines, from Physics to Econometrics. Nonetheless, they remain very controversial among statisticians who belong to the orthodox - i.e, frequentist school, and are not well known by the medical community. To help in that direction, we reviewed in the introductory chapter the basic philosophical and practical differences between the two schools, and in the second chapter, we briefly reviewed the history of Bayesian methodology, from the early efforts of Thomas Bayes and of Pierre Simon de Laplace to the modern contributions of Harold Jeffreys, Richard Cox, and Edwin Jaynes. In many aspects of medical research, we deal with experimental data from which a certain proposition or hypothesis is validated. Unlike in physics, where we have strong and solid foundations such as Newton's law of motion, Snell's optical laws, Kirchoff's laws, Einstein's relativity theory, and many more, we do not have such privileges in medical research. Hence, many hypotheses are constantly tested as new evidence becomes available. One of the actively-researched medical areas is cancer research about which our understanding is still in its infancy. Numerous experiments (both in vivo and in vitro) and clinical trials have been conducted to further our knowledge; thus, Bayesian methodology finds its place to aid us in obtaining scientific inferences about certain propositions or hypotheses from available data and resources. In this work, we use data given to us by our medical collaborators at the Princess Margaret Hospital (PMH) in Toronto to carry out two main projects: Firstly, to make an inference about the oxygenation status (oxygen partial pressure, pO$_2$) within human cervical carcinomas and secondly, an inference about the effectiveness of various molecularly-targeted agents (MTAs) in phase II clinical trials of relapsed ovarian cancer patients. In the first problem, we address the challenges of tumor hypoxia - a state of oxygen deprivation in tumors. Currently, there are two methods to obtain tumor oxygen status, namely the direct Eppendorf needle electrode and the indirect immunohistochemical assay of a protein marker, Carbonic Anhydrase IX (CAIX). In this project, we introduce Bayesian probability theory to obtain inferences about tumor oxygenation from each technique and the concordance between the two techniques. From this study, we conclude that under certain conditions, two biopsies are sufficient to infer the tumor oxygenation level based on the immunohistochemical assays of CAIX. Additionally, there is a fair concordance between the direct and the indirect measurements of tumor oxygenation. In the latter problem, ovarian cancer is the topic of study. Ovarian cancer has the highest mortality rate among gynecological cancers and one that is known to relapse. CA-125 is still the most inexpensive biomarker for monitoring ovarian cancers. From the phase II clinical trial data, we demonstrate the survival advantage of CA-125 responsive group of patients by means of a non-parametric Kaplan-Meier statistic.