Despite an initial response to chemotherapy, most patients with ovarian cancer eventually progress and succumb to their disease. Understanding why effector T cells that are known to infiltrate the tumor do not eradicate the disease after cytoreduction is critically important to the development of novel therapeutic strategies to augment tumor immunity and improve patient outcomes. Such studies have been hampered by the lack of a suitable in vivo model. We report here a simple and reliable model system in which ovarian tumor cell aggregates implanted intraperitoneally into severely immunodeficient NSG mice establish tumor microenvironments within the omentum. The rapid establishment of tumor xenografts within this small anatomically well-defined site enables the recovery, characterization, and quantification of tumor and tumor-associated T cells. We validate here the ability of the omental tumor xenograft (OTX) model to quantify changes in tumor cell number in response to therapy, to quantify changes in the tumor vasculature, and to demonstrate and study the immunosuppressive effects of the tumor microenvironment. Using the OTX model, we show that the tumor-associated T cells originally present within the tumor tissues are anergic and that fully functional autologous T cells injected into tumor-bearing mice localize within the tumor xenograft. The transferred T cells remain functional for up to 3 days within the tumor microenvironment but become unresponsive to activation after 7 days. The OTX model provides for the first time the opportunity to study in vivo the cellular and molecular events contributing to the arrest in T cell function in human ovarian tumors.
This article was published in Cancer Immunity, a Cancer Research Institute journal that ceased publication in 2013 and is now provided online in association with Cancer Immunology Research.
It is becoming increasingly apparent that in order to design therapies that successfully eradicate primary tumors and prevent their recurrence and dissemination, a more complete understanding is required of the complex interactions between human tumors and the tumor-associated leukocytes (TALs), vasculature, and fibroblasts that comprise the tumor microenvironment (1–4). Increased understanding of the molecular and cellular events that facilitate human ovarian tumor cell survival and dissemination would enable the development of therapeutic strategies to simultaneously target multiple biological processes that contribute to tumor growth. However, attempts to address these issues have been hampered by the lack of suitable models in which fresh and frozen human ovarian tumors, along with all the nonmalignant tumor-associated cells that comprise the human ovarian cancer microenvironment, can be rapidly established as xenografts that can be serially monitored for changes in the number and function of these cells after treatment with the therapeutic approaches of interest.
Previous attempts have been made to study the dynamic interactions between human tumor cells and the tumor-associated stroma by implanting small pieces of human tumor tissues subcutaneously (s.c.) into immunodeficient mice. Using this approach, xenografts have been established in which the tumor microenvironment, including tumor cells, tumor-associated inflammatory cells, and tumor-associated fibroblasts are maintained to varying degrees (5–8). These models have established that lymphocytes within the tumor can be activated by cytokines to orchestrate an anti-tumor response. However, such models have several significant limitations. The xenografts established s.c. often fail to progress and metastasize and therefore do not reflect the patterns of tumor progression seen in patients. While the histology and immunohistochemistry of the s.c. xenografts often reflect properties of the original tumor, it is not possible to accurately quantify changes in the number of tumor cells or in the number and function of tumor-associated cells. Other xenograft models have been reported in which human tumors are established orthotopically and some are shown to metastasize (9–12). However, these models do not include tumor-associated stromal cells and/or fail to provide a method to monitor and quantify changes in the xenografts at the single cell level. Many of the previous xenograft models are also limited by the requirement of fresh tumor tissues, a prolonged period of time to establish the tumor engraftment and by a strong host vs. graft response in which the recipient’s natural killer (NK) cells and granulocytes ultimately invade and destroy the xenografts (13).
Newer SCID mouse mutant strains lack both NK cells and adaptive immune function (14). These newer generations of SCID mice with targeted mutations in the Il2rg locus have significantly improved the survival of human tissues including peripheral blood monocytes, hematopoietic cells, and a number of diverse tumor cell types (7, 15–17). Using one of the newer immunodeficient mouse strains (NOD-scid IL2Rγnull or NSG mice), we developed the omental tumor xenograft (OTX) model in which it was possible to rapidly establish ovarian tumor xenografts and to monitor and quantify changes in the number of tumor and tumor-associated stromal cells. The design of the OTX model is based in part upon several observations made previously by others. For example, human intra-abdominal tumors, such as ovarian cancer, metastasize most often to the omentum (18, 19), an anatomically well-defined organ that is well vascularized and composed primarily of adipocytes that provide fatty acids for rapid tumor growth (20). In addition, murine tumor cell lines injected intraperitoneally (i.p.) into immunocompetent mice preferentially localize within the omentum and exhibit aggressive growth (21, 22). In view of these findings, we evaluated whether tumor cell aggregates derived from fresh or frozen human ovarian tumor biopsy tissues when injected i.p. into NSG mice would establish in the omentum of the recipient mice. We determined that human ovarian tumor cell aggregates localize rapidly in the omentum and these xenografts establish and progress within the omentum. Immunofluorescent staining of whole mounts of unfixed omental tissues and immunohistochemical staining of fixed tissues revealed the presence of dividing tumor cells, TALs, fibroblasts, and hyperplasia of omental microvessels. Importantly, because it was possible to obtain single-cell suspensions from the omenta, the phenotype and quantity of the different cell types present within the xenograft were easily determined by flow cytometry. We report here that this OTX model allows the recognition and quantification of changes in the number and function of tumor-associated T cells, changes in the tumor-associated microvessels, evaluation of the cytoreduction of tumor in the omentum, and the subsequent prevention of the metastatic dissemination of the tumor following chemotherapy and chemoimmunotherapy.
Human ovarian tumors and tumor-associated stroma initially engraft within the omentum following i.p. injection of tumor cell aggregates into NSG mice
Tumor cell aggregates were derived from a mild disruption of fresh primary serous epithelial ovarian tumor tissues. Tumor cell aggregates (that include cytokeratin-positive tumor cells, CD45+ leukocytes, and human fibroblasts characterized by their positive staining with D7-FIB, an antibody that recognizes human fibroblasts) were injected i.p. into NSG mice. Using this approach, we previously reported that ovarian tumor xenografts established in multiple organ sites including the ovary, pancreas, uterus, spleen, liver, and lung (17). However in this initial NSG xenograft model, no gross or histological evidence of tumors was observed in these major organ sites until 10–25 weeks post-tumor injection. Another limitation of this model was that it was not possible to recover, quantify, and assess the function of tumor-associated T cells and, after this prolonged period, there was the risk of a xenograft vs. host reaction. Based upon the findings of others that human intra-abdominal tumors such as ovarian cancer metastasize most often to the omentum (18, 19), the possibility that the human tumor cell aggregates localize very early in the mouse omentum was investigated.
In mice, the omentum is a very small strip of well-vascularized fatty tissue that is located between the stomach, pancreas, and spleen. By focusing on this tiny membranous but anatomically well-defined site, we were able to obser ve consistently microscopic evidence of tumor engraftment in the omenta of the mice at one week following the injection of the tumor cell aggregates that ultimately progressed and spread into other organ sites. This observation led to the design of the OTX model that is reported here that has made possible both the early and late monitoring of changes in tumor cell numbers and vascular changes in the tumor microenvironment, as well as the tracking and monitoring of changes in tumor-associated T cell function.
Within 2 to 6 weeks post-injection, tumor xenografts were well established in the omenta of the NSG mice. Hematoxylin and eosin (H&E) staining of OTX showed evidence of tumor cells and TALs in juxtaposition with blood vessels (Figure 1A), and trichrome staining revealed the presence of fibroblasts (Figure 1B). Immunohistochemical staining of the omenta revealed the presence of human CD3+ T cells, cytokeratin-positive tumor cells, and CD68+ macrophages (Figure 1C, E, and F, respectively). Tumor cell and TAL proliferation was indicated by Ki67 staining (Figure 1D).
The early engraftment and establishment of OTX in the omentum of NSG mice has been repeated with 11 of 11 different patient-derived primary serous epithelial tumor cell aggregates derived from fresh ovarian tumor biopsy tissues obtained 4 to 8 hours after surgical resection. Tumor xenografts were also established with cryopreserved tumor cell aggregates. This has been repeated four times using tumor cell aggregates from a single ovarian tumor that was stored for over six months and recovered from liquid nitrogen (data not shown). The xenografts established with the frozen and thawed tissues, like those generated with fresh tissues, included tumor cells, TALs (including CD3+, CD4+, CD8+ T cells, CD68+ macrophages, and FOXP3+ Tregs), and tumor-associated fibroblasts (data not shown). Xenografts established with the cryopreserved cell aggregates expanded in size in the omentum from 7 to 28 days and ultimately spread to other organs as similarly seen with the fresh tissue.
We conclude that tumor cell aggregates derived from primary fresh and cryopreserved tumors attach rapidly and preferentially in the omenta of NSG mice, establishing viable xenografts in which tumor cells and tumor stroma implant and proliferate. Because the omentum is an anatomically well-defined tissue, we sought to determine whether by removing the omentum and generating single-cell suspensions it would be possible to recover, identify, and quantify the absolute number of human EpCAM+ tumor cells and CD45+ human leukocytes present within this initial site of engraftment by using flow cytometric analysis.
EpCAM+ tumor cells and CD45+ leukocytes present in the omental xenografts evaluated by flow cytometry of single-cell suspensions
Four to six weeks following the i.p. injection of 300 mg of tumor cell aggregates into NSG mice, the mice were euthanized and the omenta removed. As shown in Figure 2, gross evidence of tumor is observed in the omentum of 5 of 5 mice six weeks after the injection of the tumor cell aggregates (Figure 2A). Single-cell suspensions were generated by collagenase disruption of the omenta and cells stained with an anti-EpCAM antibody. The percentage of EpCAM+ tumor cells was determined by flow cytometry (Figure 2B) and, based upon the total number of cells derived from each omentum, the absolute number of tumor cells was calculated and reported for each mouse (Figure 2C). EpCAM+ tumor cells were found in the omentum of all 5 of the tumor-inoculated mice.
Single-cell suspensions derived by collagenase digestion of a pool of omenta were next assayed for the presence of human TALs in the tumor xenografts. The recovery and phenotype of the TALs are presented in Figure 3. In this experiment, 13% of the cells present in a cell suspension derived from a pool of tumor-bearing omenta were human CD45+ leukocytes (Figure 3). The presence and percentage of human leukocytes derived from omental xenografts established with other ovarian tumors varied from 0.5% to > 20%. The great majority (85–90%) of the CD45+ tumor-associated human leukocytes in all experiments were CD3+ T cells (data not shown). Using a monoclonal antibody (D7-FIB) that binds human fibroblasts, we have also recovered and quantified tumor-associated fibroblasts from the xenografts (data not shown).
Having established the ability to recover, phenotype, and quantify cells from omenta bearing human tumor xenografts, we next sought to determine whether this approach could be exploited to preclinically evaluate the efficacy of cancer therapies by quantifying the degrees of tumor cell cytoreduction in xenografts following treatment with chemotherapy and a combination of chemotherapy with human IL-12-loaded liposomes.
Quantification of EpCAM+ tumor cells in the omenta of mice following chemotherapy and a combination of chemotherapy and IL-12 liposomes
To test the ability of the OTX model to recognize and quantify early cytoreductive changes in tumor cell number following two different therapeutic protocols, 30 NSG mice were injected with 300 mg of tumor cell aggregates i.p. derived from a high-grade serous ovarian carcinoma. Mice were treated with two chemotherapeutic drugs commonly used to treat ovarian cancer patients. Twenty tumor-bearing mice were treated with i.p. injections of paclitaxel (25 mg/kg) and carboplatin (50 mg/kg) in normal saline in five split doses (day 18–22 post-aggregate injection) and 10 tumor-bearing mice were injected i.p. with normal saline as control on day 18. On day 25, 10 of the chemotherapy-treated mice were injected i.p. with a single dose of IL-12-loaded liposomes (40 μg of IL-12/mouse).
On day 41, 5 of the chemotherapy-treated mice, 5 of the chemotherapy + IL-12 liposome-treated mice, and 5 of the saline-treated control mice were euthanized. The omenta were removed, photographed, and single-cell suspensions made, and then stained with an EpCAM-specific antibody and analyzed by flow cytometry. As shown in Figure 4A, gross evidence of tumor was observed in all 5 of the control saline-treated group, and 4 of 5 of the chemotherapy-treated group had gross evidence of tumor. Three of 5 of the chemotherapy + IL-12 liposome-treated mice showed no gross evidence of tumor in the omentum and 2 showed minimal gross evidence of tumor.
While these results from gross inspection suggest a therapy-associated tumor cytoreduction, it is not possible to draw a definitive conclusion regarding the experimental outcome. However, based upon the flow cytometric data on the single-cell suspensions, there was a decrease in the total number of EpCAM+ tumor cells isolated from the pooled omenta in both the chemotherapy group (sixfold reduction) and a statistically significant 634-fold reduction in the chemotherapy + IL-12 liposome-treated group compared to the control saline-treated group (p < 0.01) (Figure 4B). Also noted was a 99-fold greater decrease in the number of EpCAM+ tumor cells present in the chemotherapy + IL-12 liposome-treated group compared to the chemotherapy-treated group (p < 0.05). We conclude that the flow cytometric quantification of tumor cells recovered from the omentum provides a viable approach for the generation of data to assess and confirm a predicted cytoreduction in response to two different therapeutic protocols.
To determine whether the short-term (at day 41) initial cytoreduction in tumor cell number resulting from the two therapeutic protocols had an effect upon the long-term survival of the mice, the remaining mice (5 in chemotherapy group, 5 in chemotherapy + IL-12 liposome group, and 5 in the control group) were monitored weekly for 27 weeks for evidence of the development of tumor ascites, at which time the mice were euthanized. One of the control (saline-treated) mice was found dead at day 60 and was not further evaluated. All 4 of the remaining control mice developed tumor ascites and were euthanized between 90 to 112 days post-tumor injection (Figure 5A). Gross and microscopic evidence of tumor was observed in these 4 control mice in the omentum, and in one or more of the following organs (ovary, liver, and spleen), and multiple peritoneal cavity sites had large tumor masses, as well as tumors in the diaphragm (Figure 5B). Two of the 4 control mice showed micrometastatic evidence of tumor in the lung (Figure 5B). None of the chemotherapy-treated or chemotherapy + IL-12 liposome-treated mice had developed tumor ascites by 27 weeks post-tumor inoculation. These mice were then euthanized and examined for evidence of tumor. No gross evidence of tumor was observed in any of the organs examined (ovary, uterus, spleen, liver, pancreas, kidney, omentum, diaphragm, or lung) in the 5 chemotherapy-treated or the 5 chemotherapy + IL-12 liposome-treated mice. No histological evidence of tumor was observed in any of the organs of the chemotherapy + IL-12 liposome-treated mice. However, histological evidence of tumor was detected in the diaphragm in 2 of 5 mice and in the lung of 1 of 5 mice treated with chemotherapy only (Figure 5C).
We conclude that the initial cytoreduction of the tumor in the omentum resulting from a combination of chemotherapy + IL-12 liposome therapy resulted in the longer term prevention of tumor progression, development of ascites, and dissemination of the tumor into other organ sites. While 3 of the 5 chemotherapy only-treated group showed no evidence of tumor, 2 of these mice had histological evidence of tumor, suggesting the possibility that the addition of IL-12 enhanced the therapeutic efficacy. In contrast, in all of the mice left untreated, the tumor progressed and spread into multiple different organs.
These results establish the OTX model as a valuable tool with which to monitor and quantify both the short-term and long-term effects of single and combination therapeutic protocols.
Multicolor whole mount histology of unfixed tissues reveal the presence of EpCAM+ tumor cells, CD45+ human leukocytes, and a hyperplasia of CD31+ murine microvessels in the omentum induced by the tumor cells and tumor-associated fibroblasts
To monitor the omentum for possible changes in the vasculature following the i.p. injection of tumor cell aggregates, and to characterize and confirm the attachment of tumor and CD45+ human leukocytes, a multicolor whole mount histological examination was performed. Seven and 117 days after the i.p. injection of the tumor cell aggregates, omenta were removed and the entire omentum (unfixed) was placed in a tube and stained with fluorescent antibodies specific for EpCAM+ tumor cells, CD45+ human leukocytes, and CD31+ mouse blood vessels. The stained omentum was mounted on a slide and fluorescent digital images acquired by confocal microscopy (see Materials and methods). Tumor cells and TALs were present in the omentum in juxtaposition to the mouse vessels (Figure 6A). A striking and consistent observation was the marked increase in the number and tortuosity of the murine CD31+ microvessels 7 days following the attachment of the tumor cell aggregates that persisted for up to 16 weeks post-cell injection (Figure 6A). The disorderly and chaotic pattern of the microvessels observed in the omentum-bearing tumor xenografts was in striking contrast to the more well-organized vasculature present in the naïve (without tumor) omentum. The hyperplasia and tortuosity of the capillary networks observed in the microenvironment of the omental tumor xenografts closely resemble vascular changes observed in human tumor microenvironments (23).
The marked vascular changes that are observed in patient tumors and those found in the tumor xenograft may well contribute to the survival and spreading of the tumor. Therefore, it was of interest to determine in the OTX model which cells may provoke the tortuosity and increase in the microvessel densities observed in the tumor xenograft-bearing omenta. The three most dominant human cell types in the xenograft are the leukocytes, fibroblasts, and tumor cells. To determine which of these cells contributed to the vascular changes in the omentum, primary cultures of ovarian tumor cells, and tumor-derived fibroblasts, and suspensions of peripheral blood leukocytes (PBLs) derived from normal donors were injected i.p. into naïve (tumor-free) NSG mice. The omenta were removed 7 days later and stained for each cell type and for CD31+ mouse blood vessels. All three of the human cells (tumor cells, tumor-derived fibroblasts, and leukocytes) were found to localize within the omentum. The microvessel densities were quantified using the software program CellProfiler to analyze the CD31+ fluorescent digital images. It was established that either the tumor cells alone or the tumor-derived fibroblasts alone induced a significant enhancement in the number of microvessels in the omentum (Figure 6B). Human PBLs failed to induce the microvessel hyperplasia that was observed with tumor cells or tumor-associated fibroblasts. The recognition that the tumor cells and tumor-associated fibroblasts induce a marked hyperplasia of blood vessels in the omentum is novel and suggests that both of these cell populations may contribute to the survival and later dissemination of the tumors. An understanding of the mechanism by which these cells induce the vessel hyperplasia may lead to new targets to enhance the efficacy of chemo- and immuno-therapeutic strategies for ovarian cancer.
Use of the OTX model to monitor the modulatory effects of the tumor microenvironment on endogenous T cells present in the xenograft and on exogenous T cells following their entry into the xenograft
T cells derived from cancer patients’ peripheral blood have been shown to be fully responsive to activation via the T cell receptor (TCR), while a significant proportion of the CD4+ and CD8+ T cells derived from patient solid tumors are hyporesponsive (anergic) to the same TCR stimulus (24, 25). These observations have led to the hypothesis that functional T cells eventually become anergic or unresponsive to activation following their entry into the tumor microenvironment in the patient. This hypothesis has never been tested in humans for obvious ethical restrictions to such studies with patients.
Using the OTX model, we have established that the majority of the T cells derived from TALs recovered from the omental tumor xenografts (defined as endogenous TALs present in the original tumor tissue used to establish the xenograft) are unresponsive to TCR activation generated by exposure to immobilized antibodies to human CD3 and CD28. The translocation of NF-κB from the cytosol into the nucleus is an early and reliable indicator of T cell activation. Only about 10% of these endogenous TALs translocated NF-κB into the nucleus in response to the TCR activation signal (Figure 7). This is consistent with the findings of others (26) and our own data (24, 25) that a majority of the T cells present within the microenvironment of both mouse and human tumors are anergic and remain in this quiescent state for up to six weeks in the tumor xenograft.
To determine whether functional T cells become anergic following their entry into a solid tumor microenvironment, lymphocytes derived from the ascites fluid of an ovarian cancer patient (defined as exogenous TALs) were isolated, stained with CellTrace Violet (CTV), and injected i.p. into NSG mice in which tumor xenografts were previously established in the omentum with ovarian tumor tissue derived from the same cancer patient. We determined that these exogenous TALs present within the CTV-stained population were fully responsive to TCR activation prior to their entry into the tumor xenograft (60–70% of the T cells were activated one hour after incubation with immobilized antibodies to CD3 and CD28). Mice were euthanized one hour (day 0), 3 days (day 3), or 7 days (day 7) after injection, the omenta removed, single-cell suspensions prepared, and the CTV-labeled exogenous TALs were recovered and monitored for their ability to respond to a TCR stimulus. The exogenous TALs recovered from the xenografts one hour (day 0) and three days after entry into the tumor microenvironment remained responsive to TCR activation (60–70% of the T cells showed translocation of NF-κB into the nucleus one hour after exposure to immobilized anti-CD3 and -CD28) (Figure 7). However, the exogenous TALs derived from the xenograft 7 days after their entry into the tumor microenvironment were anergic (only about 10% of these cells were activated after an identical TCR stimulus).
These results establish that fully functional T cells that enter the microenvironment of the OTX become anergic after 7 days within the xenograft. The possibility that the induction of the anergy in the exogenous T cells is a result of the mouse xenoenvironment of the host and not the tumor xenograft seems unlikely since we determined that human T cells in normal donor PBLs remain functional for at least 7 days following their entry into the omentum of naïve (no tumor xenograft) NSG mice (Figure 7). We anticipate that the OTX model will greatly facilitate the ability to study many complex interactions between human tumor cells and tumor-associated stroma. Moreover, the possibility of using this model to test the efficacy of novel therapeutic protocols and to preclinically screen patients for individual differences in response to therapeutic protocols present viable issues for future studies.
The need for developing improved tumor xenograft models for evaluating anti-cancer therapies is essential as has previously been documented (27). The development of new genetically modified immunodeficient mouse models has greatly facilitated the successful engraftment and prolonged survival of normal and neoplastic human tissues (14). These newer generations of “humanized” mice bearing human tissue xenografts have begun to provide valuable insights into complex human biological systems including hematopoiesis, innate and adaptive immunity, autoimmunity, infectious diseases, regenerative medicine, and cancer (14). Our goal has been to utilize severely immunodeficient mice to establish xenografts of human tumors that include both the tumor cells and the associated tumor stroma, i.e., fibroblasts and inflammatory leukocytes. The implantation of human lung (7) and ovarian tumors (17) into NSG mice resulted in the long-term engraftment of the tumor and the tumor-associated nonmalignant cells that make up the tumor microenvironment. While these models encompass many significant improvements over previous xenograft models, the relatively slow development (10–25 weeks) of tumors and the inability to recover and precisely quantify tumor cells, or to recover, phenotype, and assess the function of tumor-associated inflammatory cells, limited their use in studying the dynamic and complex interactions that occur between the tumor and nonmalignant associated cells and limited the testing of newly developed therapeutic strategies.
We report here a novel mouse model of ovarian cancer in which tumor xenografts establish rapidly and both tumor and tumor-associated stromal cells can be easily recovered from the xenografts, quantified (at the single cell level), and the tumor-associated inflammatory leukocytes phenotyped and studied for their functional responsiveness to activation. This model has several unique and significant advantages (that are discussed in detail below) over previous xenograft models that are made possible by the rapid and preferential localization of the tumor tissues (both fresh and frozen) into a very small and anatomically well-defined site (the omentum) following the i.p. injection of tumor cell aggregates derived from primary tumor tissues.
The omentum is a well-vascularized membranous fatty tissue localized between the spleen, stomach, and pancreas (28) and is the primary site of metastasis in ovarian cancer patients (29, 30). The combination of a relatively dense network of microvessels and adipocytes has been suggested to contribute to the preferential omental localization of intraperitoneal tumors and to their survival and subsequent growth (21). Multiple different murine tumor cell lines in three different strains of normal immunocompetent mice were reported to initially attach and grow within the omentum (21, 22). It has now been established in ovarian cancer patients that adipocytes promote the initial homing of tumor cells to the omentum through the secretion of adipokine, and the adipocytes provide fatty acids to ovarian cancer cells that utilize them as a rich energy source (20). The adipocytes, along with the dense network of microvessels, are now thought to be responsible for the survival and rapid growth and metastasis of tumors in the omentum (20). This likely contributes to the rapid and preferential localization of human tumor cell aggregates in the omentum of the NSG mice in this omental xenograft model and may underlie the spread of tumor cells to other organ sites in this humanized model of ovarian cancer.
One of several advantages of the localization of tumor xenografts in the omentum has been the ability to recover and quantify EpCAM+ tumor cells. We have exploited this ability to assess and compare the therapeutic efficacy of chemotherapy and the combination of IL-12-loaded liposomes with chemotherapy. We report here that both therapeutic approaches resulted in a cytoreduction of the tumor in the omentum shortly after treatment and coincided with a suppression of the systemic spreading of the tumor to other organ sites and the development of ascites that was observed much later in the control (untreated) mice. Moreover, the combination therapy appeared be more effective in preventing further dissemination compared to chemotherapy alone. Patient to patient variability in the response to therapy is commonly observed in the clinic. Importantly, we have recently found in the OTX model that some tumors exhibit variability in response to the therapies employed in the current studies that may reflect the variability found in patients (data not shown). While recognizing that much larger studies that correlate the response to chemotherapy or other therapies in the OTX model with clinical observations of responsiveness to the same therapies are needed, this model offers the possibility of addressing this issue directly. Moreover, the ability to rapidly establish viable tumor xenografts with frozen and thawed tumor cell aggregates offers a way to establish a panel of tumors from patients with known clinical responses which can then be tested with the same therapeutic regimes in the OTX model. Importantly, if validated in this fashion, the OTX approach might have practical utility in prospectively directing therapy since it can be performed much more quickly than previous strategies, which would be of considerable value in tailoring therapies in personalized medicine.
Another particularly significant advantage of the preferential localization of the tumors in the omentum has been the ability to recover from the xenograft, phenotype, and assay the functional capacity of endogenous and exogenous T cells present within the tumor microenvironment at different time intervals. Most of the endogenous T cells (i.e., those present in the original tumor tissues used to establish the xenograft) failed to respond to activation via the TCR. This finding was anticipated as tumor-associated endogenous T cells in both mice (26) and humans (24, 25) have been reported to be anergic. A reasonable, but previously untested, hypothesis has been that fully functional T cells entering the tumor microenvironment become anergic at some point following their entry into the immunosuppressive environment of the tumor. Consistent with this hypothesis, our results showed that functional T cells derived from patients remained functional for at least 3 days following their entry into an OTX (established with the same patient’s tumor). However, a significant proportion of these T cells (exogenous TALs) when recovered from the xenograft 7 days after their entry were found to be anergic. We conclude that the microenvironment of the OTX, like that of the tumor in the patient, is able to modulate T cell function. It is therefore likely that following the migration and localization within the tumor, T cells survive but may eventually become unresponsive to activation via their TCR. It will be important to determine the phenotype of the anergic T cells and the phenotype of the T cells that remain functional within the microenvironment of the tumor xenograft. It should be possible to identify the cellular and molecular factors contributing to the TCR signaling arrest and to determine whether it is possible to target and block these immunosuppressive mechanisms using the OTX model.
Phase III clinical trials of anti-angiogenic drugs have shown only limited success or have failed altogether (31). Many possible reasons for this have been suggested but no consensus has been reached that adequately explain the failures. Several possible strategies have been discussed for improving drug efficacy including combinations with chemotherapeutic regimens such as long-term metronomic chemotherapies addressing individual patient response differences, and combinations with other anti-cancer agents that either directly attack the tumor or target tumor stromal cells (31). The ability to recognize and quantify changes in the vasculature network in the OTX model reported here provides a potential preclinical model with which to test many of these strategies and to possibly determine some of the reasons for the failures of the current clinical trials with anti-angiogenic drugs.
We anticipate that the OTX model can be exploited to (a) evaluate preclinically the efficacy of novel therapeutic approaches to cancer, (b) determine individual differences in patient responses to currently approved treatment protocols, (c) determine the cellular and molecular mechanisms by which T cell functions are arrested upon entry into the tumor microenvironment and, (d) design and test ways in which the tumor-augmented microvessel densities and tortuosity can be decreased and the immunosuppressive effects of the tumor microenvironment overcome.
While we have discussed many of the strengths of the OTX model, we recognize there are limitations and potential pitfalls with the current model. While we have not observed any xenograft vs. host reactions with ovarian tumors, this is a real potential problem with other tumor types where there are more numerous leukocytes present within the tumor tissue as with hematological tumors. Another limitation of the OTX model as currently configured is that it lacks the influx of other cell types that occur in patients as the tumor progresses. Finally the ability of the OTX model to be used for other human tumor types has not yet been fully established. We have only determined that, in addition to ovarian tumors, it also works well for primary non-small cell lung tumors. And since lymphocytes migrate and survive within the omentum, we predict that the OTX model has a potential for studying different hematological malignancies.
Many other xenograft models have been reported using NSG mice to study and characterize cancer stem cells and other facets of tumor biology (32–34). Recent models have been successfully designed to evaluate antibody-mediated cancer therapies (35), and the therapeutic efficacy of the chimeric antigen receptor strategy has been demonstrated for EpCAM+ tumors (36) and for a mesothelioma (37). The in vivo analysis of these and other human tumor-specific therapies is possible because the human tumors grow more efficiently in NSG mice and these mice can be engrafted with functional immune cells. Use of the OTX model and these other NSG-based models provides a real potential to preclinically evaluate novel single and combination therapeutic approaches for cancer patients (32).
Materials and methods
Human primary and metastatic serous high-grade ovarian solid tumor tissue was provided through the Roswell Park Cancer Institute Tissue Procurement Facility. All tissue samples were obtained under sterile conditions using Institute Review Board-approved protocols.
Preparation of tumor cell aggregates and implantation into mice
NOD.Cg-PrkdcscidIl2rgtm1Wjl, abbreviated NSG mice (8 to 12 weeks old), raised in a research colony at The Jackson Laboratory were implanted with human ovarian tumors as approved by the Institutional Animal Care and Use Committee. Briefly, human ovarian solid tumor was minced with scissors, placed into DMEM medium, and then passed through a 50 gauge stainless steel wire mesh (Buffalo Wire Works, Buffalo, NY) with a Teflon policeman. The resulting tumor aggregate suspension was washed with medium, pelleted, and weighed. NSG mice were injected i.p. with 300 mg of tumor aggregates resuspended in 1 ml of medium. NSG mice were housed in specific pathogen-free conditions in the animal care facility at SUNY Buffalo.
Omental whole mount fluorescent staining
At various times post-engraftment, the NSG mice were euthanized and the greater omentum was surgically removed and transferred to polypropylene tubes containing PAB buffer (PBS with 1% BSA and 0.1% Sodium Azide). The tissue was placed directly into cold PBS, pH 7.2. To prevent nonspecific Fc binding, tissue was blocked using Fc Block (BD Biosciences, San Diego, CA). PE-labeled anti-mCD31, Cy5-labeled anti-huCD45 (BD Biosciences, San Diego, CA), and FITC-labeled anti-human EpCAM (Dako) were added and incubated for 2 hours with rocking. Samples were washed after incubation and mounted on a slide. Fluorescent digital images were acquired using a Zeiss Axio Imager Fluorescence Microscope. The images were visualized using AxioVision Rel. 4.6. To quantify microvascular vessel density, CellProfiler r10997 software was used to analyze the photomicrographs.
Chemotherapy and IL-12 liposome treatment
Intraperitoneal injection of paclitaxel (25 mg/kg) and carboplatin (50 mg/kg) in 0.9% saline, in 0.2 ml volume, total concentration administered in split daily doses for 5 days into NSG mice bearing omental tumor xenografts. On day 25, IL-12 liposomes were prepared as previously reported (17) and were administered i.p. (40 μg IL-12/mouse). Mice were euthanized at different times and the omenta were excised and single-cell suspensions generated by collagenase digestion (long-term observation).
Collagenase treatment of omenta
Omenta were excised, minced with scissors, and collagenase digested at 37°C for 1 hour with a cocktail containing 2 mg/ml Collagenase P (Roche), 1 mg/ml Hyaluronidase type 1, and 0.1 mg/ml DNase 1 in DMEM media with gentle swirling every 15 minutes. Debris was processed through a 70 μm mesh and cells were washed and centrifuged. Omental cells were counted and viability determined by trypan blue exclusion.
Single-cell suspensions of omental tissue were prepared by placing the tissue in DMEM/F12 with 10% FCS and then passing through a # 50 mesh as described above. The cell suspension was then passed through a 70 μm cell strainer (BD Falcon, Bedford, MA). The cell suspension was counted and set up at 106 cells/tube in PBS for staining. The cells were blocked with normal mouse IgG for 10 minutes, and directly labeled antibodies were added and incubated on ice for 15 minutes.
Antibodies used were FITC-conjugated anti-human EpCAM (clone Ber-EP4, Dako) and APC-conjugated anti-human CD45 (clone H130, BD Biosciences). Cells were washed twice with PBS, then fixed with 2% formaldehyde, filtered with cell strainer capped polystyrene tubes (BD Falcon, # 352235), and analyzed on a FACSCalibur flow cytometer.
Histology and immunohistochemical staining
SUNY Buffalo Histology Service Laboratory performed the H&E staining of the tumor tissue and omental tissue. Fresh tissue was fixed in 10% neutral-buffered formalin and processed for paraffin embedding. Tissues sampled were stained with antibodies specific for human CD3, CD19, CD68, cytokeratin, and EpCAM and stained for Ki67 and trichrome as previously reported (17). Portions of the immunohistochemistry were performed at Roswell Park Cancer Institute by the Pathology Resource Network.
NF-κB nuclear translocation assay by confocal microscopy
TALs or normal donor PBLs were labeled with 5 μM CellTrace Violet (Invitrogen, # C34557) for 20 minutes at 37°C. TALs or PBLs (1×107) were injected i.p. into OTX-bearing mice or naïve (no tumor xenograft) NSG mice. Mice were then euthanized 1 hour, 3 days, and 7 days after the injection of the cells and the omentum removed and mechanically disrupted into a single-cell suspension. Omental cells were counted and viability determined. To determine T cell activation, antibodies to human CD3 (BioXCell, # BE0001-2) at 0.1 μg and human CD28 (Invitrogen, # CD28004) at 5 μg were immobilized onto MaxiSorp Star tubes, and 5×105 cells per tube were added and incubated at 37°C for 2 hours. NF-κB nuclear translocation of T cells was determined by confocal microscopy as described previously by us (6, 25, 26). After the 2 hour incubation, the cells derived from the omentum were attached to cationized coverslips, fixed, and permeabilized. The cells were stained for CD3 using mouse anti-human CD3 antibody (BD, # 347340) followed by goat anti-mouse AlexaFluor 568 (Invitrogen, # A11004). After washing, the cells were incubated with rabbit anti-NF-κB antibody (Santa Cruz Biotechnology, # SC-372) followed by goat anti-rabbit AlexaFluor 488 (Invitrogen, # A11008). NF-κB nuclear localization was visualized using a Zeiss LSM 510 Meta Confocal Microscope and photomicrographs analyzed using AxioVision software. T cell activation was assessed by enumerating cells with an NF-κB nuclear staining pattern that were positive for CD3. The percentage of CD3+ T cells with nuclear NF-κB staining (activated cells) vs. the total NF-κB cytoplasmic stained cells (unactivated cells) and NF-κB nuclear stained cells was determined by visual analysis of at least 100 cells.
The significance of the differences in tumor cell number between therapeutic treatment groups was assessed using the non-parametric Rank Sum test. The significance of the differences between the means of other groups was analyzed using Student’s t-test. A p value ≤ 0.05 was considered statistically significant.
This work was supported by National Institutes of Health grants CA108970 and CA131407 (to R.B.B) and CA034196 (to L.D.S.).We thank Anthony Miliotto, Amy Beck, Sayeema Daudi, and the Tissue Procurement Facility of Roswell Park Cancer Institute (RPCI) for their assistance in providing tumor tissues. We thank Dr. Maurice Barcos, a pathologist at RPCI, for his help and guidance in reviewing the histology and immunohistochemistry of the ovarian tumor xenografts. We thank Dr. Steven Bernstein for his critical reading and suggested edits of the manuscript.
- Copyright © 2013 by Richard B. Bankert