Abstract
Purpose: In previous studies, we have used human embryonic stem cells (hESC) to generate a tissue microenvironment in immunocompromised mice as an experimental approach for studying human tumorigenesis. We now examine the attributes of such a cellular microenvironment in supporting the growth of human cancer cells freshly harvested from malignant ovarian ascites and to determine whether there are differences among subsets of ascites-derived cancer cells in terms of tumorigenic capacity in the conventional murine xenograft model and in the hESC-derived microenvironment.
Experimental Design: Freshly harvested malignant ovarian ascites-derived cancer cells and six derivative ovarian cancer cell subpopulations (CCSP) were characterized for ovarian cancer–associated biomarker expression both in vitro and in vivo and for their capacity to generate tumors in the two models.
Results: Ovarian cancer–associated biomarkers were detected in the ascites-derived cancer cells and in the six newly established CCSPs. Nevertheless, certain CCSPs that did not develop into tumors in a conventional murine xenograft model did generate tumors in the hESC-derived cellular microenvironment, indicating variable niche dependency for the tumorigenic capacity of the different CCSPs. The hESC-derived microenvironment provided an improved niche for supporting growth of certain tumor cell subpopulations.
Conclusions: The results highlight the experimental utility of the hESC-derived cellular microenvironment to enable functional distinction of CCSPs, including the identification of cells that do not grow into a tumor in the conventional direct tumor xenograft platform, thereby rendering such cells accessible to characterization and testing of anticancer therapies.
In previous studies, we have shown that a cellular microenvironment comprising human embryonic stem cells (hESC) in immunocompromised mice provides an improved experimental approach for studying human tumorigenesis using established human cancer cell lines. The current study extrapolates this approach to human cancer cells freshly harvested from malignant ovarian ascites. Using the mixed population of malignant ascites cells, as well as isolated cancer cell subpopulations (CCSP), we were able to show growth of tumors, which preserve the full spectrum of ovarian cancer–associated biomarkers and the capacity for differentiation into structures characteristic of clinical ovarian cancer. However, certain CCSPs that did not develop into tumors in a conventional murine xenograft model did generate tumors in the hESC-derived cellular microenvironment, suggesting variable niche dependency for the tumorigenic capacity of the different CCSPs derived from the same parent tumor. These results underscore the potential experimental utility of the hESC-derived cellular microenvironment to enable functional distinction of CCSPs, including the identification of cells that do not grow into a tumor in the conventional direct tumor xenograft platform and therefore may escape characterization and testing of anticancer therapies. The current model renders such potentially important tumor cells amenable to biological analysis as well as anticancer therapy testing.
The process of tumorigenesis is associated with extensive remodeling of adjacent tissues to provide a supportive microenvironment for cancer cell proliferation, migration, and invasion. Promotion and maintenance of solid tumor cancer growth invasion and metastasis clearly depend on the surrounding stroma, which provides supportive cells and functions such as tumor-induced neoangiogenesis (1, 2), inflammatory cells (3, 4), and cancer-associated fibroblasts (5–8) that together with the extracellular matrix (ECM) create the complexity of the tumor mass (9). Genetic and epigenetic changes provide cancer cells with the advantages of proliferation and the survival capacity to evade censoring mechanisms, such as apoptosis, necrosis, and senescence (10).
The heterogeneous population of tumor cell types and abnormal tissue structure comprising solid tumors suggests that tumors may be regarded as aberrant organs in which genetically altered malignant cells coexist with normal cells and tissues with which they interact to execute the process of tumor progression (10–12). This recognition has led to hypothesizing the existence of distinct subpopulations of tumor cells that may contribute to the process of tumor repopulation, asymmetric division, and generation of heterogeneity, metastasis, and drug resistance (13–18).
Xenograft injections of human cancer cells into mouse tissues have long been an accepted in vivo model for investigating tumor initiation and progression and for the development and preclinical testing of novel anticancer therapeutic agents targeted to frustrate stromal response factors, which support progressive tumor growth (19). However, this model suffers the drawback of not being able to fully recapitulate the relationship between the tumor cells and the complex surrounding human microenvironment (20, 21). Accordingly, we have established a novel in vivo tumor microenvironment model that is composed of a wide variety of disorganized but normal differentiated human tissue and cellular structures using teratomas generated following injection of human embryonic stem cells (hESC) into immunocompromised mice (22–24). Previous studies using this model in conjunction with a variety of different established cancer cell lines suggested that the human cellular microenvironment seems to provide a preferential experimental niche for supporting tumor development: (a) tumor cell viability was shown to be greater in the setting of growth within the hESC-derived microenvironment compared with the corresponding direct tumor xenograft; (b) prominent tumor cell invasion into the heterogeneous surrounding human cellular tissue was evident; (c) tumor vasculogenesis comprising blood vessels of both human and murine origin was evident; and (d) anticancer immunotherapy directed against tumor cell–related epitopes induced the complete regression of direct tumor xenografts as opposed to corresponding tumors growing within the hESC-derived teratoma microenvironment, wherein remnant foci of viable cells were detected and resulted in full tumor recurrence. We postulated that such an optimized niche permits intercellular interactions between subsets of tumor cells and corresponding subsets of differentiated human stromal cells within the teratoma, conducive to supporting tumor growth and resistance to anticancer therapy. Because such a heterogeneous human cellular microenvironment is not available for cancer cells in direct tumor xenografts, we sought to examine whether this novel hESC-derived platform might be used as a tool that could distinguish cancer cell subpopulations (CCSP) based on niche dependency and thereby enable study of subsets of cancer cells that cannot execute their tumorigenic potential in conventional murine xenograft systems. For this purpose, we turned to clinical samples of cancer cells harvested from malignant ovarian ascites fluid.
Using this approach, we derived CCSPs, which were functionally distinguished by a differential capacity for tumorigenesis in a conventional xenograft model, but all of which were robust in generating tumors in the hESC-derived cellular microenvironment. These results suggest that the hESC-derived cellular microenvironment has the additional experimental advantage of providing a tumor-promoting niche, permissive to the growth of CCSPs that do not grow into a tumor as direct human tumor murine xenografts and hence are otherwise not readily accessible to anticancer therapy testing in currently used murine xenograft models.
Materials and Methods
Harvesting of cancer cells from malignant ovarian cancer ascites fluid. Ovarian cancer cells were harvested at the time of palliative paracentesis of 6 L of malignant ascites fluid of a 64-y-old woman with stage IV ovarian cancer, diagnosed as clear cell adenocarcinoma in the Gynecologic Oncology Unit at Rambam Medical Center (Haifa, Israel). Her primary treatment included total abdominal hysterectomy with bilateral salpingo-oophorectomy followed by a standard chemotherapy regimen, including Taxol, carboplatin, topotecan, doxorubicin, and gemcitabine. The ascites fluid was centrifuged at 4°C for 10 min at 2,000 rpm, and hemolysis of erythrocytes was done using lysis buffer [10 nmol/L potassium bicarbonate, 155 mmol/L ammonium chloride, 0.1 mmol/L EDTA (pH 7.4)] for 5 min and centrifugation for 10 min at 2,000 rpm at 4°C. The pellet was resuspended in 30 mL RPMI 1640 (Biological Industries) per 1 cm of pellet. Cell suspensions (30 mL) were layered over 10 mL Ficoll and centrifuged for 30 min at 2,000 rpm at 25°C. The cell layer was removed from the Ficoll and washed thrice with RPMI 1640 followed by centrifugation for 10 min at 1,200 rpm at 25°C. Cells (1-2 × 107) were seeded on 20 μg/mL fibronectin-coated culture dishes in RPMI 1640 supplemented with 20% FCS and 1% penicillin/streptomycin solution (Biological Industries) at 37°C in 5% CO2. Collection of ascites fluid was done with informed consent and the protocol was approved by the institutional Ethics Review Committee of the Rambam Medical Center.
Animals. Severe combined immunodeficient (SCID)/beige mice were purchased from Harlan Laboratories Ltd. The mice were housed and maintained under specific pathogen-free conditions. The facilities and experimental protocols were approved by the Committee for Oversight of Animal Experimentation at the Technion-Israel Institute of Technology (Haifa, Israel).
Teratomas and tumor formation. For teratoma formation, undifferentiated hESC clone H9.1 (kindly provided by J. Itskovitz-Eldor, Rambam Medical Center, Haifa, Israel) was injected into the hind limb of SCID/beige mice (∼5 × 106 cells per injection). At 7 to 8 wk following initial injection of hESC, 4 × 106 cancer cells were injected into the teratoma and allowed to grow for an additional 20 to 60 d. Tumors derived from direct injection of 4 × 106 cells into the hind limb musculature were harvested at 20 to 60 d following injection.
Histologic analysis. Teratomas and hind limb musculature injected with cancer cells were harvested, fixed for 24 h in 10% neutral buffered formalin, transferred into 70% ethanol, and processed using a routine wax embedding procedure for histologic examination. Six-micrometer paraffin sections were mounted on SuperFrost Plus microscope slides (Menzel-Glaser) and stained with H&E.
Immunohistochemistry. Slides were deparaffinized using xylene and rehydrated through a series of gradients of alcohol to water. Antigens were retrieved using microwave exposure at 90°C for 20 min in a pH 6.1 citrate buffered solution. Endogenous peroxidase enzyme activity was blocked using 3% hydrogen peroxidase in methanol for 30 min at room temperature. Slides were washed in distilled water and then in PBS (pH 7.4) and blocked using 10% nonimmune goat serum for 24 h at 4°C. Slides were incubated for 24 h at 4°C with the primary antibodies monoclonal mouse anti-CA-125 (1:80), monoclonal mouse anti-MUC1 (1:75), monoclonal mouse anti-CA-19.9 (1:100; Zymed Lab, Inc.), and mouse monoclonal anti-human epithelial antigen (Ber-EP4, 1:250; DakoCytomation), followed by incubation with goat anti-rabbit or anti-mouse biotinylated secondary antibody. Preimmune rabbit or mouse sera were used as negative controls. Detection was accomplished using the Histostain-SP (AEC) kit (Zymed Lab).
Immunofluorescence. Cells were seeded on fibronectin-coated cover glasses in six-well plates, fixed with 4% paraformaldehyde for 10 min at room temperature, and perforated using 0.1% Triton in 0.1% citric acid for 10 min at 4°C. Slides were washed in PBS, blocked with 10% nonimmune goat serum, and incubated with primary antibodies (at 1:50 dilution) against CA-125, CA-19.9, MUC1, and Ber-EP4 and mouse monoclonal anti-human E-cadherin and N-cadherin (1:50; Zymed Lab). Cells were incubated for 1 h at room temperature with secondary antibodies as follows: goat anti-mouse CY3 (1:150), donkey anti-rabbit CY3 (1:150), and donkey anti-mouse FITC (1:100; Jackson ImmunoResearch Laboratories, Inc.).
Photomicrographs. An Olympus BX50 microscope equipped with an Olympus DP-70 digital camera was used for documentation and evaluation of the different expression intensities. The expression level was determined by the exposure time in milliseconds when the camera was in autoexposure mode (see Supplementary Data 1). Pictures were processed using the analySIS 5 software (Soft Imaging Systems).
Semiquantitative one-step reverse transcription-PCR. Absolutely RNA kit (Stratagene) was used to extract total RNA from the different ovarian CCSPs. Similar amounts of total RNA were separated on agarose gels stained with ethidium bromide. The rRNA of all the samples was read by a densitometer for calibration. Samples of total RNA extracted from each cell subpopulation were subjected to semiquantitative one-step real-time reverse transcription-PCR (RT-PCR) using the QuantiTect SYBR Green RT-PCR kit (Qiagen, Inc.) and the Rotor-Gene 2000 (Corbett Research) as previously described (24). For each sample, RT-PCR analysis was done for a reference gene (β-actin gene) as an internal control. Primers used for gene amplification of the genes described in this study are provided in Supplementary Data 2.
Results
Examination of the tumorigenic properties of freshly harvested cancer cells from ascites fluid in direct tumor xenografts and in the hESC-derived microenvironment. Using commercially available established cell lines, we have recently shown that the composition of heterogeneous human cellular tissue in a mouse, derived from in vivo differentiation of hESCs, seems to confer advantageous properties as a microenvironment for supporting solid tumor development (23, 24). To determine whether this experimental platform could also be extrapolated to freshly harvested human cancer cells, we examined the tumorigenic capacity of a bulk sample of malignant ovarian ascites cells in a direct tumor xenograft model and in the hESC-derived cellular microenvironment. For this purpose, cells collected directly after harvesting were injected into the hind limb musculature of SCID/beige mice (i.m. injection) or into a hESC-derived teratomas [intrateratoma (i.t.) injection], which developed within SCID/beige mice. Injected mouse muscle and teratoma tissues were harvested and subjected to histologic examination of paraffin sections using H&E staining and immunohistochemistry detection. Two of three i.m. injections and five of seven i.t. injections generated tumors with the characteristic appearance of adenocarcinoma in both models. However, a longer period was needed for the tumors to develop within the murine tissue (45-65 days) compared with tumors growing within the teratoma tissue (21-24 days). Furthermore, the tumors growing within the hESC-derived cellular tissue showed more obvious differentiated glandular structures typical of ovarian origin (Fig. 1A and B).
Tumorigenic properties of cancer cells freshly collected from the malignant ovarian ascites fluid. Paraffin sections stained with H&E show the tumorigenic capacity of cancer cells derived from the malignant ascites fluid following i.m. injection (A) and i.t. injection (B). CA-125–positive, CA-19.9–positive, and MUC1-positive cancer cells were detected in tumors generated in the murine (C–E) and in the teratoma tissue (F–H) following immunohistochemistry staining of paraffin sections with specific antibodies. In both models, the tumor cancer cells were stained positively with Ber-EP4 antibody (I and J). Bars, 200 μm (A and B) and 100 μm (C–J).
Tumorigenic properties of cancer cells freshly collected from the malignant ovarian ascites fluid. Paraffin sections stained with H&E show the tumorigenic capacity of cancer cells derived from the malignant ascites fluid following i.m. injection (A) and i.t. injection (B). CA-125–positive, CA-19.9–positive, and MUC1-positive cancer cells were detected in tumors generated in the murine (C–E) and in the teratoma tissue (F–H) following immunohistochemistry staining of paraffin sections with specific antibodies. In both models, the tumor cancer cells were stained positively with Ber-EP4 antibody (I and J). Bars, 200 μm (A and B) and 100 μm (C–J).
The tumor cells in both experimental platforms expressed the ovarian cancer–associated markers CA-125, CA-19.9, and MUC1 that are routinely used in the clinic to screen for ovarian cancer progression (Fig. 1C-H). In both in vivo models, the cells in the tumor stained positively with the Ber-EP4 antibody, which recognizes the epithelial-specific antigen on the cell surface and indicates an epithelial origin of the cancer cells (Fig. 1I and J; refs. 25, 26).
Isolation and characterization of CCSPs from the mixed population of ascites-derived cancer cells. We have previously shown that following anticancer immunotherapy of A431 epidermoid carcinoma cells growing within the teratoma model, there remained a subset of surviving malignant cells, and proposed that the i.t. model might be permissive to the growth of those CCSPs of possibly greater clinical and therapeutic relevance (24). Therefore, from the bulk of heterogeneous malignant ovarian cells, we sought to propagate CCSPs, which might escape detection using currently available experimental direct murine xenograft approaches and whose growth might require the hESC-derived niche, with the idea that just such CCSPs might be of particular interest in understanding cancer growth and effects of anticancer therapy. For this purpose, the bulk of ascites-derived cancer cells (108) was subjected to a clonal expansion process that yielded six different CCSPs that were further characterized and are now maintained as stable cell lines. These six CCSPs displayed the morphologic appearance of epithelial cells (Fig. 2A), and indeed, all of them stained positively in vitro with Ber-EP4 antibody (Fig. 2B).
In vitro characterization of the different CCSPs and detection of ovarian cancer–associated biomarkers. A, phenotypic appearance of the six isolated CCSPs. B, immunofluorescence staining with the primary monoclonal antibody Ber-EP4 and CY3 as a secondary antibody indicates the epithelial origin of the CCSPs. Staining of the cell nuclei was done using 4′,6-diamidino-2-phenylindole (DAPI). Bars, 100 μm. C, detection of expression of CA-125, CA-19.9, and MUC1 biomarkers by representative CCSPs. D, expression of N-cadherin and E-cadherin in representative CCSPs and nonmalignant control IOSE80 cells in vitro. Detection was done by using immunofluorescence with corresponding primary specific antibodies. Positively stained cells were detected using CY3 or FITC as a secondary antibody. Staining of the cell nuclei was done using 4′,6-diamidino-2-phenylindole. Bars, 100 μm.
In vitro characterization of the different CCSPs and detection of ovarian cancer–associated biomarkers. A, phenotypic appearance of the six isolated CCSPs. B, immunofluorescence staining with the primary monoclonal antibody Ber-EP4 and CY3 as a secondary antibody indicates the epithelial origin of the CCSPs. Staining of the cell nuclei was done using 4′,6-diamidino-2-phenylindole (DAPI). Bars, 100 μm. C, detection of expression of CA-125, CA-19.9, and MUC1 biomarkers by representative CCSPs. D, expression of N-cadherin and E-cadherin in representative CCSPs and nonmalignant control IOSE80 cells in vitro. Detection was done by using immunofluorescence with corresponding primary specific antibodies. Positively stained cells were detected using CY3 or FITC as a secondary antibody. Staining of the cell nuclei was done using 4′,6-diamidino-2-phenylindole. Bars, 100 μm.
In vitro detection of ovarian cancer–associated biomarkers. The level of expression of three ovarian cancer–associated biomarkers, CA-125, CA-19.9, and MUC1, was examined in the different CCSPs. Although exceedingly useful, the biomarker CA-125, which is widely used to screen for ovarian cancer progression, is not perfectly sensitive for ovarian cancer as 20% of ovarian cancer tissues lack expression of this antigen, nor is it highly specific because its level may be elevated in conditions not associated with ovarian cancer (27). To increase the sensitivity and specificity of diagnosis in screening and monitoring, two additional ovarian cancer–associated biomarkers are also used in the clinic: CA-19.9, which is also strongly associated with ovarian cancer of mucinous histology and which is thought to have an important role in metastasis (28), and MUC1, which is expressed on the cell surface and interferes with cell-cell and cell-ECM adhesion (29). In vitro immunofluorescence staining showed expression of these biomarkers by all the different CCSPs (Fig. 2C); however, differential expression intensity profile of the biomarkers was observed in cell membranes and cytoplasm (see Supplementary Data 1). As shown, CA-125 displayed an expression profile spanning the range from high expression in CCSPs C2 and C13 to low levels of expression in cell subpopulations C1, C5, C12, and C16. These diverse levels of CA-125 expression among the different CCSPs are consistent with the clinically observed heterogeneity of CA-125 levels, both interindividual and within tumors (30). The expression pattern of CA-19.9 and MUC1 is relatively uniform among the different CCSPs, showing relatively weak expression of CA-19.9 and relatively strong expression of MUC1 that might reflect a high metastatic potential of the cells extracted from the ascites fluid (30). Of note, cells that stained positively for CA-125 and CA-19.9 antigens showed intercellular heterogeneity of expression levels of these biomarkers within the same culture, consistent with the observation that surface expression of these antigens is cell cycle related (31).
Determination in vitro of the status of the different CCSPs in relation to the “cadherin switch.” An important transition that drives the transformation of normal ovarian surface epithelium (OSE) into cancer-forming epithelium is a switch in the expression of the predominant cadherin subtype, which involves the down-regulation of N-cadherin in normal cells and the up-regulation of E-cadherin in cancerous cells that is characteristic for ovarian carcinoma effusions (32, 33). Therefore, the level of expression of these cadherins was examined in the different CCSPs. SV40 large T antigen–infected immortalized OSE nonmalignant ovarian epithelial cells were used as ovarian control cells (34). Immunofluorescence analysis revealed positive staining of IOSE80 cells for N-cadherin and negative staining for E-cadherin. In contrast, all the malignant ascites-derived CCSPs showed a high level of expression of E-cadherin with no detectable expression of N-cadherin (Fig. 2D; Supplementary Data 1). These results indicate that all the malignant ascites-derived CCSPs have traversed this step in ovarian malignant transformation and express E-cadherin as has been shown in ovarian cancer secondary metastatic sites and effusions.
Expression of matrix metallopeptidases MMP9 and MMP2. The lower survival rate of ovarian cancer patients is attributed to the diagnosis of the disease at advanced stages when the peritoneal lining is studded with metastatic implants and contains ascites fluid with disseminated cancer cells (35). The matrix metallopeptidases (MMP), and in particular MMP9 and MMP2, which are implicated in the promotion of proliferation, tumor-associated angiogenesis, tumor metastasis and survival of neoplastic cells, and in release of mitogenic growth factors from the mesothelial cell surface, have been suggested to be key regulators for the invasive and metastatic pattern of ovarian carcinoma cells during progression of the disease (36, 37). Because the different CCSPs were isolated from detached cancer cells contained in ascites fluid collected from a patient with stage IV ovarian carcinoma, we examined the level of expression of MMP2 and MMP9 in the six different CCSPs. Samples containing equivalent total RNA extracted from the CCSPs were subjected to semiquantitative RT-PCR analysis using MMP2- and MMP9-specific primers. For comparison, RNA was extracted from hESC H9.1. As shown in Fig. 3A and B, MMP9 transcript was detected in all six CCSPs, albeit expression was highly variable with lowest levels observed in CCSPs C1, C2, and C12 and relatively higher levels of expression in C13. However, even the highest level of expression in C13 was ∼0.3 of the level of MMP9 detected in hESC H9.1. Notably, no expression of MMP2 was observed in any of the CCSPs. The latter finding is further consistent with the results of several studies, suggesting that in metastatic ovarian cancer the expression of the MMP2 gene is induced by the host mesothelial cells that cover the peritoneum and the omentum, whereas the MMP9 gene is expressed by the ovarian cancer cells themselves (37).
Expression of MMPs. RT-PCR analysis for detection of MMP2 and MMP9 genes (A and B) in the different CCSPs. Total RNA was extracted from the different CCSPs cultivated in vitro and from hESC H9.1 control cells. RNA samples were subjected to one-step RT-PCR analysis for detection of gene expression with (+) and without (−) reverse transcriptase. β-Actin served as an internal control. The relative expression levels of each gene were quantified by densitometry analysis and normalized for the internal control β-actin.
Expression of MMPs. RT-PCR analysis for detection of MMP2 and MMP9 genes (A and B) in the different CCSPs. Total RNA was extracted from the different CCSPs cultivated in vitro and from hESC H9.1 control cells. RNA samples were subjected to one-step RT-PCR analysis for detection of gene expression with (+) and without (−) reverse transcriptase. β-Actin served as an internal control. The relative expression levels of each gene were quantified by densitometry analysis and normalized for the internal control β-actin.
Tumorigenic capacity of the CCSPs using in vivo models. Motivated by the previous report that the hESC-derived cellular microenvironment might provide a more supportive microenvironment for relevant subsets of tumor cells (24), we searched for distinct tumor cell subpopulations that might preferentially grow in such a microenvironment in comparison with the conventional xenograft microenvironment. To this end, we took advantage of the six different newly established CCSP cell lines derived from the bulk malignant ovarian ascites fluid and compared their tumorigenic capacities in the direct tumor xenograft (i.m. injection) platform and in the hESC-derived cellular teratoma microenvironment (i.t. injection).
Repeated i.m. and i.t. injections were done for each of the different CCSPs as described in Table 1. Despite the absence of observable differences in the expression of the gene products and markers noted above, reproducible variation was observed in the tumorigenic capacity among the different CCSPs in the murine compared with the hESC-derived microenvironment. For example, CCSP C12 generated tumors very readily, within 33 and 26 days in both the murine and the human cellular microenvironment, reaching a large tumor mass in the murine musculature tissue. Other CCSPs such as the C5, C13, and C16 also developed into tumors in both in vivo models; however, a longer period was needed for the tumor to develop in the murine tissue (58, 58, and 43 days in the murine tissue and 30, 22, and 19 days in the hESC-derived tissue, respectively), and the tumor sizes were variable. Strikingly, two CCSPs, C1 and C2, developed into barely palpable masses of only 0.7 to 0.9 mm2 (respectively) at 56 days following murine i.m. tissue injection in only one of five injections, with no evident tumor growth whatever in the remaining injections. In contrast, robust tumors developed consistently within a short period of 25 and 23 days (respectively) following injection into the hESC-derived cellular tissue. The above pattern of variation in tumorigenic capacity of the different CCSPs using the two models was observed consistently along 3 years of research with these CCSPs. The measurements of tumor dimension presented in Table 1 were done only for i.m. tumors. Such measurements are uninformative for i.t. tumors, as the tumor cells are intertwined within the teratoma as an integrated three-dimensional structure.
Tumorigenic capacities of the six CCSPs isolated from the bulk malignant ovarian ascites in the hESC-derived microenvironment and in direct tumor xenograft
CCSP . | No. instances of tumor formation i.m. . | No. instances of tumor formation i.t. . | Estimated average tumor size (mm2) i.m. . |
---|---|---|---|
C1 | 1/5 (56 d) | 4/4 (25 d) | 0.73 |
C2 | 1/5 (56 d) | 4/5 (23 d) | 0.96 |
C5 | 3/4 (58 d) | 2/2 (30 d) | 19.43 |
C12 | 4/4 (33 d) | 2/2 (26 d) | 14.83 |
C13 | 5/5 (58 d) | 2/3 (22 d) | 9.22 |
C16 | 4/4 (43 d) | 2/2 (19 d) | 4.75 |
CCSP . | No. instances of tumor formation i.m. . | No. instances of tumor formation i.t. . | Estimated average tumor size (mm2) i.m. . |
---|---|---|---|
C1 | 1/5 (56 d) | 4/4 (25 d) | 0.73 |
C2 | 1/5 (56 d) | 4/5 (23 d) | 0.96 |
C5 | 3/4 (58 d) | 2/2 (30 d) | 19.43 |
C12 | 4/4 (33 d) | 2/2 (26 d) | 14.83 |
C13 | 5/5 (58 d) | 2/3 (22 d) | 9.22 |
C16 | 4/4 (43 d) | 2/2 (19 d) | 4.75 |
Each of the tumor and teratoma-bearing tumor tissues harvested at the indicated times (see Table 1) were subjected to histologic and immunohistochemical analyses. Paraffin sections stained with H&E of the tumors generated by the various CCSPs in both in vivo models showed the presence of a hyaluronic acid–containing ECM characteristic of clear cell ovarian adenocarcinoma. This phenotype matches the histologic characterization of the tumor type determined for the patient from whom the cells were collected. Nevertheless, various degrees of differentiation and hyaluronic acid content could be observed among the tumors generated by the different CCSPs and among tumors generated by any given CCSP in the murine compared with hESC-derived tissues (Fig. 4A and B). Previous studies have shown heterogeneity of ovarian cancers classified as clear cell, with variability in the histologic type of clear cell and in the relative proportion of clear cells to nonclear cells within the tumor (38). The current study highlights the fact that such heterogeneity can constitute part of the tumor evolution, generated by the clonal expansion of a single-source transformed cell. For example, the CCSP C1 displayed features of a highly differentiated tumor type following both i.m. and i.t. injections as manifested by the formation of glandular structures with fibroblast-like cells lining the spaces among microscopic substructures of the tumor. The i.t. tumor could be readily harvested 25 days following injection compared with the i.m. tumor, which only grew to a very small size (see Table 1) but nevertheless exhibited a highly differentiated pattern at 59 days following injection. The CCSP C2 displayed a poorly differentiated, hyaluronic acid–containing ECM tumor in both models, although the tiny i.m. tumor was harvested at 59 days following injection. The CCSP C5 generated robust tumors in both models that were very poorly differentiated but contained high level of hyaluronic acid. This tumor was prominently characterized by many fibroblast-like cells, which lined the borders of the i.m. as well as the i.t. tumor. CCSP C12 exhibited highly differentiated tumors with both the i.m. and i.t. protocols. In these tumors, the clear cells were arranged in sheets or in tubules with many fibroblast-like cells lining the spaces. In contrast, CCSP C13 generated tumors that displayed a hyaluronic acid–containing ECM, but the cells within the tumor masses that generated both i.m. and i.t. were poorly differentiated and fibroblast-like cells were concentrated at the outer border of the tumors. The CCSP C16 also generated tumors that displayed a hyaluronic acid–containing ECM, but the tumor-generated i.t. presented a higher degree of differentiation compared with the tumor-generated i.m, as indicated by the presence of hobnail-shaped clear cells in the tumor (38). In this case, the i.m. tumor was also surrounded by a layer of fibroblast-like cells.
Ovarian clear cell carcinomas generated by the different CCSPs in in vivo models. Paraffin sections stained with H&E of tumors generated following i.m. or i.t. injections of the different CCSPs (4 × 106 cells per injection) were analyzed. The histologic appearances of the tumors in both in vivo models are presented in A and B. Higher power magnification of the areas indicated by the filled arrows in each tumor is shown in the small inset. Open arrows indicate fibroblast-like cells lining the spaces between the glandular structures in tumors generated by CCSP C1 and C12 and fibroblast-like cells lining the i.m. tumors generated by CCSP C13 and C16. The brown open arrow in the small inset of i.t. tumor generated by CCSP C16 points at the highly differentiated hobnail-shaped cells. Bars, 200 μm (large tumor picture) and 50 μm (small inset). The cancer cells in the tumors generated in both models by the six different CCSPs express the ovarian cancer–associated biomarkers CA-125, CA-19.9, and MUC1 as observed by immunohistochemistry analysis with specific antibodies and maintain their epithelial features as indicated by positive staining for Ber-EP4 antibody. Representative positively stained CCSPs are presented in C. Bar, 100 μm.
Ovarian clear cell carcinomas generated by the different CCSPs in in vivo models. Paraffin sections stained with H&E of tumors generated following i.m. or i.t. injections of the different CCSPs (4 × 106 cells per injection) were analyzed. The histologic appearances of the tumors in both in vivo models are presented in A and B. Higher power magnification of the areas indicated by the filled arrows in each tumor is shown in the small inset. Open arrows indicate fibroblast-like cells lining the spaces between the glandular structures in tumors generated by CCSP C1 and C12 and fibroblast-like cells lining the i.m. tumors generated by CCSP C13 and C16. The brown open arrow in the small inset of i.t. tumor generated by CCSP C16 points at the highly differentiated hobnail-shaped cells. Bars, 200 μm (large tumor picture) and 50 μm (small inset). The cancer cells in the tumors generated in both models by the six different CCSPs express the ovarian cancer–associated biomarkers CA-125, CA-19.9, and MUC1 as observed by immunohistochemistry analysis with specific antibodies and maintain their epithelial features as indicated by positive staining for Ber-EP4 antibody. Representative positively stained CCSPs are presented in C. Bar, 100 μm.
The cancer cells within the solid tumors generated by each CCSP in both in vivo models expressed the ovarian cancer–associated biomarker panel of CA-125, CA-19.9, and MUC1 (Fig. 4C). In addition, the tumor cells also maintained their epithelial features as indicated by their morphology and by positive staining for Ber-EP4 (Fig. 4C).
Overall, these results agree with previous clinicopathologic observations indicating that clear cell carcinomas of the ovary do not constitute a homogeneous entity but rather vary by the histologic type of clear cells of which they are composed and by the proportion of clear cells to nonclear cell elements within the neoplasms (38). Furthermore, these results indicate that the CCSPs faithfully recapitulate the full spectrum of heterogeneity known for human ovarian cancer.
Discussion
The concept that a regulatory niche provides an essential and supportive microenvironment for tumor development has been developed in conjunction with the increasing evidence of cancer stem cells sharing properties with other stem cell types whose growth is niche dependent. Accumulating evidence in support of this formulation led us to take advantage of a novel experimental system of growing tumors within a hESC-derived cellular microenvironment for examination of the influence of such a supportive sanctuary on tumorigenesis properties of cancer cells. We have recently shown that provision of such a microenvironment allows tumor cells to evade complete eradication following anticancer treatment as opposed to complete regression of corresponding tumors developed within a solely murine muscle tissue microenvironment (24). The differences between the features of these tissue microenvironments motivated their use in the current study as experimental platforms to examine the tumorigenic heterogeneity among tumor cell subpopulations derived from freshly harvested ovarian cancer cells.
Various versions of the “cancer stem cell” hypothesis postulate the existence of rare subsets of cells of as yet unclear origin, which are responsible possibly for tumor initiation, repopulation, dissemination, and relative protection from antitumor mechanisms and therapies (13–16, 39). Such a paradigm could account for the heterogeneity observed within solid tumors because tumors can be composed of such putative cancer stem cells as well as their derivative cells generated through amplification and aberrant differentiation. This paradigm has important implications both for the understanding of tumorigenesis processes and for the development of novel anticancer therapies (10–12).
Accordingly, we took advantage of the availability of cancer cells in freshly collected malignant ovarian ascites concentrates to investigate whether two distinct cellular niches can distinguish the tumorigenic capacities of different CCSPs. The six different CCSPs derived from the bulk malignant ovarian ascites were maintained in culture for more than 3 years. Despite this duration of continuous propagation in culture following harvesting and separation, these six CCSPs remarkably still maintained the same “bona fide” ovarian cancer characteristics that were evident on harvesting. These include expression of ovarian cancer–associated markers CA-125, CA-19.9, and MUC1; expression of genes involved in self-dissemination such as metallopeptidases; and genes involved in self-protection such as the ATP-binding cassette transporters (data not shown). In contrast, in the different in vivo models, we observed not only evidence for an overall more supportive microenvironment using the hESC-derived teratomas but also the ability to distinguish CCSPs that seem to display greater niche dependency irrespective of the expression of the set of markers described above. Consistent heterogeneity in tumorigenic properties among the six CCSPs was most strikingly reflected in their niche-dependent in vivo tumorigenic capacities and tumor cellular phenotypes. Greater in vivo tumor cell viability and growth was observed in the context of the hESC-derived cellular microenvironment. This may be attributed to the heterogeneous composition of human cellular tissues, which we propose confers a supportive niche with greater fidelity to the actual clinical tumor microenvironment due to mutual interactions of the tumor cells with one or more normal human cell or tissue types available in the teratoma tissue. The hESC-derived cellular microenvironment thereby “jump starts” tumor cell growth by circumventing hurdles that certain tumor cell subpopulations face in the direct murine xenograft model (23, 24). This formulation was greatly strengthened by the observation of different tumorigenesis properties of the different CCSPs isolated from the bulk malignant ovarian ascites fluid using the two in vivo models. The consistent and reproducible differences observed in the average estimated tumor size measured within the mouse muscle tissue show that the murine muscle microenvironment/niche confers different and inferior developmental supporting conditions for tumor development for the different CCSPs. On the other hand, each of the different CCSPs injected into the hESC-derived cellular microenvironment/niche consistently resulted in generation of robust tumor growth. These observations point to the different tumorigenic characteristics among the CCSPs and also suggest that realization of CCSPs C1 and C2 tumorigenic capacities requires niche-dependent factors that they obtain from the complex multicellular teratoma tissue (40). In contrast, the CCSPs C5, C12, C13, and C16 display lesser niche dependency in terms of tumor growth that might be attributed to vigorous recruitment of fibroblast-like cells into the tumor tissue that facilitate and further support their tumorigenesis features of interest, as has been described before (7, 8). Such a self-producing supportive niche has been recently shown as an inherent feature of cultivated undifferentiated hESCs (41), and niche dependency has been considered as a hallmark for stem cell activity in the hematopoietic and other systems (42). Of note, in the very small tumors that were generated by CCSPs C1 and C2 in the mouse musculature, fibroblast-like cells could also be observed. However, this recruitment was not sufficient to overcome their niche dependency features, which were provided in the supportive heterogeneous hESC-derived cellular microenvironment. Interestingly, although CCSP C16 showed niche-independent tumorigenic capacity, the differentiation potential of the cancer cells seemed to be niche dependent as it could be executed only in the context of the human teratoma tissue. The evaluation of differences in gene expression between tumors generated in murine tissue versus tumors generated in a hESC-derived cellular tissue might serve to elucidate niche-dependent factors. In parallel, quantitative assessment of tumorigenic capacities and treatment responsiveness within the hESC-derived teratoma will require ectopic expression of luciferase in the CCSPs, which would enable the in vivo measurement of tumor size within the hESC-derived cellular microenvironment using vital imaging systems such as IVIS.
Taken together, the results of the current study suggest that the hESC-derived cellular microenvironment has the additional experimental advantage of the capacity to unveil certain CCSPs that do not grow into a tumor in direct tumor xenograft and therefore are most probably not accessible to testing of anticancer therapies used in the conventional xenograft model. The current model renders such potentially important tumor cells amenable to biological analysis as well as anticancer therapy testing.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Daniel M. Soref Charitable Trust, Skirball Foundation, Ed Satell Stem-Cell Research Program, American Technion Society Stem Cell Fund, and Israel Science Foundation grant 453/06.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
Acknowledgments
We thank Daniel Beck and Amnon Amit (Gynecology Department, Rambam Medical Center) for providing the malignant ascites fluid; Raymond Coleman, Ofer Ben-Itzhak, and Bernard Czernobilsky for helpful input; and Irena Reiter, Galit Paor, and Shoshana Ben-Eliezer for excellent technical assistance.