Search

Development of targeted therapeutics required translationally relevant preclinical models with well-characterized cancer genome alterations. Here, by studying 52 colorectal patient-derived tumor xenografts (PDX), we examined key molecular alterations of the IGF2–PI3K and ERBB–RAS pathways and response to cetuximab. PDX molecular data were compared with that published for patient colorectal tumors in The Cancer Genome Atlas. We demonstrated a significant pattern of mutual exclusivity of genomic abnormalities in the IGF2–PI3K and ERBB–RAS pathways. The genomic anomaly frequencies observed in microsatellite stable PDX reproduce those detected in nonhypermutated patient tumors. We found frequent IGF2 upregulation (16%), which was mutually exclusive with IRS2, PIK3CA, PTEN, and INPP4B alterations, supporting IGF2 as a potential drug target. In addition to maintaining the genomic and histologic diversity, correct preclinical models need to reproduce drug response observed in patients. Responses of PDXs to cetuximab recapitulate also clinical data in patients, with partial or complete response in 15% (8 of 52) of PDXs and response strictly restricted to KRAS wild-type models. The response rate reaches 53% (8 of 15) when KRAS, BRAF, and NRAS mutations are concomitantly excluded, proving a functional cross-validation of predictive biomarkers obtained retrospectively in patients. Collectively, these results show that, because of their clinical relevance, colorectal PDXs are appropriate tools to identify both new targets, like IGF2, and predictive biomarkers of response/resistance to targeted therapies. Cancer Res; 75(8); 1560–6. ©2015 AACR.

Identifying common patterns of regulation in cancer and placental cells might shed new light on cellular programs allowing aggressive tumor development. Indeed, as it has been described for the first time more than a century ago, cancer and placental cells (trophoblasts) share astonishingly similar phenotypes : migration and invasion, immune escape, and angiogenesis induction. These common phenotypes, relying on common genomic sequences and transcriptomic profiles, may result from a shared epigenomic regulation program. This might be especially true for tumor cells leading eventually to metastasis and cytotrophoblasts (a trophoblasts subset) during placental implantation at the beginning of pregnancy. In order to investigate such common epigenomic patterns, we carried out comparisons of DNA methylation in cancer and placental cell genomes. This study involved: cancer samples (primary tumor vs. normal tissue) across various tissues (including breast, colon, liver, lung, prostate, thyroid, uterus), on one hand; and placenta samples (early vs. late term) either heterogeneous (chorionic villi) or homogenous (ex vivo cytotrophoblasts), on the other hand. Cancer data were data downloaded from The Cancer Genome Atlas web portal. Placenta data were downloaded from the Gene Expression Omnibus web portal. In addition, our group generated original data from ex vivo cytotrophoblasts samples. All data were generated on the Illumina Infinium 450K array. Data analysis was carried out thanks to computational methods for epigenomics recently described. This first direct comparison of cancer and placental cells epigenomes, leveraging both published data and original data, led to the identification of large hypomethylated blocks common to cancer and placental cells. Such common patterns have recently been described as a universal defining epigenetic alteration in human solid tumors. These megabase-scale DNA methylation marks differentiate primary tumors from normal tissues. Likewise, they differentiate early term placentas from late term placentas. Moreover, genes belonging to genomic regions displaying common large hypomethylated blocks overlapping in cancers and placentas are enriched for pathways involved in migration and invasion, immune escape, and angiogenesis induction. Common DNA methylation patterns in cancer and placental cells identified in this pilot study might contribute to the epigenomic regulation of cellular programs allowing aggressive tumor development. Further analyses of these common patterns, as well as analyses of differences in cancer and placental cell epigenomes, could eventually lead to the identification of critical epigenomic switches that prevent healthy placentas to degenerate into tumors, while they allow aggressive tumors to develop. Ultimately, such epigenomic switches could also represent innovative targets in oncology. Citation Format: Akpeli V. Nordor, Sophie Richon, Thierry Fournier, Dominique Bellet, Virginie Dangles-Marie, Martin J. Aryee. Common DNA methylation patterns in cancer and placental cells involved in migration and invasion, immune escape, and angiogenesis induction. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 2762.

Obtaining representative human colon cancer cell lines from fresh tumors is technically difficult. Using 32 tumor fragments from patients with colon cancer, the present study shows that prior xenograft leads to more efficient cell line establishment compared with direct establishment from fresh tumors (P < 0.05). From 26 tumor specimens, we successfully established 20 tumor xenografts in nude mice (77%); among 19 of these xenografts, 9 (47%) led to cell lines, including four from liver metastases. Only 3 of 31 tumor specimens (9.7%) grew immediately in vitro, and all were derived from primary tumors. To compare major phenotypic and genotypic characteristics of human colon cancer cell lines derived from the same tumor fragment using two protocols, the two pairs of cell lines obtained from 2 of 32 tumor fragments were extensively studied. They displayed similar morphology and were able to form compact spheroids. Chemosensitivity to 5-fluorouracil, CPT11, and L-OHP differed between cell lines obtained from patient tumors and those derived from xenografts. Matched cell lines shared a common core of karyotype alterations and distinctive additional chromosomal aberrations. Expression levels of genes selected for their role in oncogenesis evaluated by real-time quantitative PCR were found to be statistically correlated whatever the in vitro culture model used. In conclusion, xenotransplantation in mice of tumor fragments before establishment of cell lines enables generation of more novel human cancer cell lines for investigation of colon cancer cell biology, opening up the opportunity of reproducing the diversity of this disease. [Cancer Res 2007;67(1):398–407]

We described previously a CTL clone able to lyse the autologous carcinoma cell line IGR-Heu after specific recognition of an HLA-A2/mutated alpha-actinin-4 peptide complex. Here, we used IGR-Heu, cultured either as standard two-dimensional monolayers or as three-dimensional spheroids, to further analyze the influence of target architecture on CTL reactivity. Interestingly, we found that changes in the tumor structure from two- to three-dimensional induced a dramatic decrease in its capacity to activate autologous CTL, as measured by IFN-gamma and tumor necrosis factor-alpha secretion. These functional alterations were attributable neither to MHC class I expression nor to tumor antigen (Ag) down-regulation, because IGR-Heu, cultured as two- or three-dimensional, expressed similar levels of HLA-A2 and alpha-actinin-4. More importantly, incubation of three-dimensional cells with synthetic epitope completely restored cytokine release by CTL. This defective Ag presentation correlated with a decrease in heat shock protein (hsp)70 expression by three-dimensional tumors compared with two-dimensional cells. Furthermore, transfection of the tumor cells with hsp70 cDNA completely restored the Ag-presenting potential of spheroids and, therefore, cytokine production by T cells. These data strongly suggest that hsp70 down-regulation in three-dimensional cells may result in tumor resistance to the immune response.

There is a recent increase in the use of patient-derived tumor xenografts (PDX) engrafted into immunodeficient mice as improved preclinical models of patient tumors. An important component in the validation of disease-specific PDXs for clinical relevance is comparing the genomic alterations and the response to standard agents. The CReMEC collection of 54 colorectal PDX has been recently reported to mimic the clinical situation for histopathological and molecular diversity of colorectal cancer. We further analyze here this colorectal PDX collection in regard to robust human patient molecular features: 1) Alterations in both PI3K and RTK-Ras pathways; 2) Role of oncogenic activation of EGFR-Ras downstream signaling on response to cetuximab. The Cancer Genome Atlas Network (TCGA) newly reported a large genome-scale analysis of 276 colorectal cancer tissue samples, showing common genetic alterations in the PI3K and RTK-RAS pathways, with mutual exclusions in the PI3K pathway. The analysis of the PDX panel by CGH array, RNA expression (microarray, real-time qRT-PCR) and exome sequencing confirmed activation with mutual exclusion in PI3K pathway (IGF2 focal amplification/overexpression; IRS2 overexpression, mutation of PIK3CA, PIK3R1 and PTEN homozygous deletion). In RTK-Ras pathway, frequencies of genomic abnormalities (ERBB2 mutation/amplification; mutation of ERBB3, NRAS, KRAS and BRAF) in PDXs are fully in line with the TCGA patient collection. The response to cetuximab of 52 subcutaneous engrafted PDX (including 24 KRAS mutated PDX) has been analyzed according to translated clinical criteria: xenograft regression as been defined as a partial response (decrease of at least 70% of the tumor volume measured at the beginning of the treatment) or as a complete response. In this unselected population, tumor regression occurred in 8 out of 52 cases (15%); all were KRAS wild type tumors. The percentage of responders was enriched up to 30% (7/23) when PTEN deletion and mutations of KRAS, BRAF, and PIK3CA are concomitantly excluded. These results completely fit with recent publication of data in patients treated with anti-EGFR antibodies. Taken together, these results demonstrate that colorectal PDXs are representative clinical colorectal tumor models. They also underline their interest as appropriate tools to identify and test new targeted therapeutics. Citation Format: Manoel Nunes, Louis-Bastien Weiswald, Patricia Vrignaud, Sophie Vacher, Edouard Turlotte, Sophie Richon, Dominique Bellet, Sergio Roman-Roman, Ivan Bieche, Virginie Dangles-Marie. Similar PI3K and RTK-Ras status in patient derived colorectal cancer-xenografts and patients. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr LB-34. doi:10.1158/1538-7445.AM2013-LB-34

Objectives. Establishment of pre-clinical models which reflect molecular alterations of primary tumour is a major goal in cancer research. In order to identify novel therapeutic targets and new treatment, xenografts models are powerful tools to analyse pathways and genes involved in carcinogenesis. The aim of this study is to compare gene expression profilings of 4 cancer types (breast, colon, ovary cancer and uveal melanoma) for which primary tumors (PT) and corresponding xenografts (PTDX) are available. We will highlight differentially expressed gene-sets and pathways involved in stromal environment, cancer cells environment, engraftment process and cancer specific development. In addition, this characterisation was associated to survival analysis in breast cancer. Materials and Methods. Gene expression profiling have been determined using Affymetrix technology. Breast cancer dataset was composed of 8 patient/xenograft paired models corresponding to 16 samples, colon, ovary and uveal melanoma datasets was composed of 24, 11 and 12 paired models respectively. Differential analysis between patients and corresponding xenografts was performed on difference matrix (PTDX-PT). Differentially expressed genes were those with a difference unequal to 0. In order to better identified our gene lists, we analysed large public tumors and cell lines datasets. Results. Comparison between PT and corresponding PTDX identified 150 genes always differentially expressed in the different cancer datasets. A large part are down regulated in PTDX. These genes could be divided into two different environments, immunity on one side and angiogenesis, cellular adhesion and development on the other side. Twenty percent are up-regulated and were mainly associated to cell cycle. A large patient and cell lines collections were used for each cancer type in order to differentiate stromal dependant genes and tumor specific genes. A large part of genes were stromal dependant. However some of them presented same expression between patients and cell lines and are lightly deregulated in mice. Survival and response treatment analysis were performed in breast data. In this cancer, genes related to angiogenesis, cellular adhesion and development were associated to survival. Discussion. This descriptive work based on microarray gene-expression profiling allowed to present common genetic background in four different cancer types and increase knowledge in gene micro-environment in carcinogenesis. This study will then focus on breast cancer and we will characterize patient tumors with high probability to grow into mice in order to concentrate on the response to treatment and improve patient management strategies. Citation Format: Cécile Laurent, Elisabetta Marangoni, Philippe Hupé, Fariba Némati, Nathalie Cassoux, David Gentien, Oumou Goundiam, Virginie Dangles-Marie, Sophie Piperno-Neumann, Pascale Mariani, Didier Decaudin, Laurence Desjardins, Xavier Sastre, Sergio Roman-Roman, Fabien Reyal. Identification of gene-expression modifications related to the tumor engraftment in patient-derived xenograft models of solid tumors. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr LB-263. doi:10.1158/1538-7445.AM2013-LB-263

There is increasing evidence that three-dimensional (3D) tissue culture technologies have many uses within the preclinical assays in cancer. The 3D tumor model accurately reproduces the in vivo tumor phenotype and represents an additional tool for studying tumor biology and allowing better preclinical evaluation of anticancer drugs. One of the used models involves small tumor aggregates, termed spheres, that are obtained from mechanical tumor dissociation and that have shown the superiority of 3D culture over standard two-dimensional cell culture for mimicking the tumor biology and drug response observed in vivo. The aim of this study was to generate new ex vivo 3D models from breast (BC) cancer xenografts established from patients’ tumor fragments. These ex vivo mastospheres are easily obtained from mechanically dissociated fresh human BC tissue xenografted in Nude mice, in a similar way as colospheres from human colon cancer (Weiswald et al, Br J Cancer 2009, 101:473). In contrast to mammospheres described in the literature, 3D mastospheres are obtained from tumor fragments, without enzymatic tumor dissociation, without matrix substratum and in SVF supplemented medium. From a large panel of patient-derived BC xenografts already well characterized (Marangoni et al, Clin Cancer Res 2007, 13:3989; Cottu et al, Breast Cancer Res Treat 2011; Reyal et al, Breast Cancer Res 2012, 14:R11), we get mastospheres from 26 out of 36 (72%) xenografts. Mastosphere formation is scored on day 1 after culture according to the number of spheres/mg of dissociated xenograft tissue. Within mastospheres, we clearly distinguish 3 distinct morphologies: round, grape-like and aggregates. Histological analyses show also that mastospheres were formed only with proliferating cancer cells. In a reproducible way, dissociation of a given xenograft leads to a similar score and to a similar morphology. All BC subtypes (luminal, triple negative and HER2+ tumors) give rise to mastospheres. We noted that the 5 out of 5 xenografts (100%) able to metastasise in Nude mice (lung metastasis) form mastospheres, suggesting that the capacity to give mastospheres could be related to tumour aggressiveness, as already reported with colospheres and colon cancer. These different features prompted us to test the potential of these mastospheres in chemosensitivity assays. We first demonstrated that mastospheres can be kept viable a couple of days, consistent with ex vivo assay aim. For this purpose, we tested the sensitivity to cisplatin in two xenograft models with different in vivo response. We found that the paired mastospheres tested in viability assays mimicked these different response profiles (mean of IC50: 1.2 µM versus 15.5 µM). In conclusion, according to these preliminary data, mastospheres deserve additional investigation because of their interest as new ex vivo microtumour model. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr LB-111. doi:1538-7445.AM2012-LB-111

Well characterized models representing the heterogeneity of human colorectal cancers (CRC) are needed to develop effective therapeutic agents. Establishment of such tools will allow a better prediction of the clinical outcome, taking into account the diversity of each patient tumor phenotype and genotype. For this purpose and with the financial support of the French Ministry of Industry, we have associated efforts from hospitals, academic groups, biotech and pharmaceutical companies. From May 2007 to January 2009, 86 surgical specimens [59 primary (P) tumors, 19 metastasis (M), and 8 peritoneal carcinomatosis (C)] were collected from CRC patients (with informed consents and negative HBV, HCV, and HIVs serologies). Tumor samples were subcutaneously xenografted in nude mice. The take rate was 57% (P), 55% (C), and 82% (M) with a difference in the take rate between tumor stages (p= 0.0184). We further engrafted in nude mice, SCID mice and nude rats. No significant difference in the take rate and growth was observed between the 2 mouse strains. Most of the mouse-growing tumors (95%) were successfully engrafted in nude rats, however their growth was significantly slower showing a more pronounced stromal component when compared with mice. Characteristics of our models are in accordance with the CRC clinical heterogeneity. Sequence analysis of 54 models has been performed and mutations were observed for KRAS (46%), APC (44%), TP53 (59%), BRAF (11%), PI3KCA (11%), FBXW7 (6%), CTNNB1 (2%), EGFR (2%), and 5/21 present a MSI status. Besides molecular markers, we compared the gene copy number using CGH technology before and after engraftment. The results exhibit high similarity between early passages of xenografts and the original clinical tumor samples. The histological structure and molecular profile were preserved, indicating the relevance of these models. Nevertheless the quality control is required to follow potential molecular deviation after multiple in vivo passages. The established models are being evaluated for ex vivo and in vivo sensitivities to colon anticancer drugs (5-FU, oxaliplatin, irinotecan and cetuximab). In vivo studies performed in our panel show 14/19 models sensitive to 5-FU, 1/19 to oxaliplatin, 7/8 to irinotecan, and 6/19 to cetuximab. We will present the correlation between pharmacological studies, molecular profile and patient clinical history. Preclinical studies on patient-derived tumor models will bring benefits to evaluate novel targeted therapeutic strategies and potentially help the stratification strategy for cancer patients. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 4169.

The ex vivo colospheres are a newly characterised three-dimensional colon cancer multicellular model, derived from mechanically dissociated human colon cancer tissue (Br J Cancer 2009, 101:473-82). We have previously demonstrated that this short term culture model is exclusively formed by cancer cells and associated with tumor aggressiveness. To further investigate the potential interest of colospheres as micrometastasis or cancer stem cell model, we used 3 human colon cancer xenografts directly established from 3 patient colon adenocarcinoma in Nude mice (Cancer Res 2007, 67:398-407). These xenograft tissues are able to give rise to numerous colospheres in only 3 days after mechanical dissociation. This approach allows working with reproducible biological tumor material, in which all human part is known to be cancer cells. Xenograft tissues and derived colospheres were collected for gene expression study. Expression of 78 genes, involved in stemness, proliferation, epithelial-to-mesenchymal transition and metastasis, has been studied using real-time RT-PCR. All the probes used are human specific and do not cross with mouse genome. Consequently, the comparison of gene expression profiles corresponds to the comparison of gene expression in colosphere-forming cells versus that in the cancer cell counterpart from xenograft tissue. Gene expression clustering analysis clearly showed that for the 3 colon adenocarcinoma, colospheres matched with their parent xenografts. This first point is important because it demonstrated 1) the lack of ex vivo culture artefact (which would have led to classification into 2 groups: all xenografts opposed to all colospheres); 2) the relevance of colospheres for mimicking in vivo tumor cells. Nevertheless, in all 3 pairs xenograft/colospheres, 3 genes were found overexpressed: ALDH1, KLF4 and PLAUR. This overexpression has to be put in line with high tumorigenicity of these colospheres when injected into mice (Br J Cancer 2009, 101:473-82). Gene expression increase will be confirmed at protein level. To gain also information about location of the cells expressing ALDH1, KLF4 and PLAUR, we developed an immunostaining protocol for confocal microscopy and in situ protein detection (BMC Cancer, in revision). In conclusion, these preliminary data underline the interest of colospheres as ex vivo microtumor model with cancer initiating-cell functions. Note: This abstract was not presented at the AACR 101st Annual Meeting 2010 because the presenter was unable to attend. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr LB-161.