Successful eradication of established peritoneal ovarian tumors in SCID-Beige mice following adoptive transfer of T cells genetically targeted to the MUC16 antigen

Abstract

Purpose

Most patients diagnosed with ovarian cancer will ultimately die from their disease. For this reason, novel approaches to the treatment of this malignancy are needed. Adoptive transfer of a patients own T cells, genetically modified ex vivo through the introduction of a gene encoding an chimeric antigen receptor (CAR) targeted to a tumor associated antigen, is a novel approach to the treatment of ovarian cancer.

Experimental design

We have generated several CARs targeted to the retained extracellular domain of MUC16, termed MUC-CD, an antigen expressed on a majority of ovarian carcinomas. We investigate the in vitro biology of human T cells retrovirally transduced to express these CARs by co-culture assays on artificial antigen presenting cells (AAPCs) as well as by cytotoxicity and cytokine release assays utilizing the human MUC-CD⁺ ovarian tumor cell lines and primary patient tumor cells. Further, we assess the in vivo anti-tumor efficacy of MUC-CD targeted T cells in SCID-Beige mice bearing peritoneal human MUC-CD⁺ tumor cell lines.

Results

CAR modified MUC-CD targeted T cells exhibited efficient MUC-CD specific cytolytic activity against both human ovarian cell lines as well as primary ovarian carcinoma cells in vitro. Furthermore, expanded MUC-CD targeted T cells infused either through intraperitoneal injection or intravenous infusion into SCID-Beige mice bearing orthotopic human MUC-CD⁺ ovarian carcinoma tumors either delayed progression or fully eradicated disease.

Conclusion

These promising pre-clinical studies justify further investigation of MUC-CD targeted T cells as a potential therapeutic approach for patients with high risk MUC-16⁺ ovarian carcinomas.

INTRODUCTION

Ovarian cancer is the sixth most common cancer worldwide and the seventh leading cause of cancer-related deaths in women (1, 2). Despite multimodality therapy with surgery and chemotherapy, most patients with ovarian carcinomas have a poor prognosis. For this reason, alternative approaches to treating this disease are urgently needed.

Infusion of a patient’s own T cells genetically targeted ex vivo to antigens expressed on the surface of tumor cells is a promising novel approach to the adoptive immunotherapy of cancer, and one which has only recently been explored in earnest in the clinical setting. T cells may be genetically modified to target tumor associated antigens through the retroviral introduction of genes encoding artificial T cell receptors termed chimeric antigen receptors (CARs). CARs are most commonly composed of a single chain fragment length antibody (scFv), derived from a murine monoclonal antibody targeting a given tumor associated antigen, fused to a transmembrane with or without an additional cytoplasmic signaling domain (typically derived from CD8, CD28, OX-40, or 4-1BB), fused to the TCR ζ chain cytoplasmic signaling domain (3–13). When expressed by the T cells, the resulting construct, upon engagement with the targeted antigen, induces T cell activation, proliferation, and lysis of targeted cells. To date, preclinical studies utilizing CAR-modified T cells have demonstrated promising results in a wide variety of malignancies (3, 4, 11, 14–18). More recently this approach been investigated in phase I clinical trials (6, 19–21).

Ovarian carcinomas appear to be relatively immunogenic tumors capable of inducing an endogenous immune response based on the fact that long-term prognosis of patients is markedly influenced by the degree and quality of the endogenous immune response to the tumor. Specifically, it has been well documented that the presence of endogenous effector T cells within the ovarian cancer tumor microenvironment directly correlates to prolonged patient survival (22–25). In contrast, increasing numbers of immune suppressive CD4⁺ CD25^hi regulatory T cells (Tregs) within the tumor, which in turn presumably abrogate the anti-tumor activity of infiltrating effector T cells, correlates with shorter patient survival (26–29). In fact, it appears that it is the ratio of Tregs to effector T cells within the tumor microenvironment which ultimately dictates whether the endogenous immune response to the cancer is of benefit or detriment to the patient (24, 28). In this setting, the ability to generate and subsequently expand a population of tumor targeted effector T cells ex vivo which are subsequently infused back into the patient, may in turn skew the Treg to effector T cell ratio to one more favorable to eradicating the disease.

Mucins are important biomolecules for cellular homeostasis and protection of epithelial surfaces (1). MUC16 is one such mucin which is over expressed on most ovarian carcinomas and is an established surrogate serum marker (CA-125) for the detection and progression of ovarian cancers (30–33). MUC16 is a high-glycosylated mucin composed of a large cleaved and released domain, termed CA-125, consisting of multiple repeat sequences, and a retained domain (MUC-CD) which includes a residual non-repeating extracellular fragment, a transmembrane domain, and a cytoplasmic tail (34). Since the antigen is otherwise only expressed at low levels in the uterus, endometrium, fallopian tubes, ovaries, and serosa of the abdominal and thoracic cavities, MUC16 is a potentially attractive target for immune-based therapies.

The fact that most of the extracellular domain of MUC16 is cleaved and secreted limits the utility of MUC16 as a target antigen on ovarian carcinomas. To date, all reported MAbs to MUC16 bind to epitopes present on the large secreted CA-125 fraction of the glycoprotein, with none known to bind to the retained extra-cellular fraction (MUC-CD) of the antigen (35–37). Since the MUC-CD fraction of the antigen is retained on cell surface, generating T cells specific to this portion of MUC16 may largely overcome the limitation of MUC16 as a target for adoptive cellular immunotherapy. To this end, we have previously generated a series of murine MAbs specific to the retained MUC-CD extracellular domain (38). Utilizing a hybridoma which expresses one such MAb, 4H11, we have successfully constructed several CARs specific to the MUC-CD antigen.

In this report, we demonstrate highly efficient retroviral transduction of these MUC-CD targeted CARs into human T cells with resulting T cells able to specifically target and lyse MUC-CD⁺ tumor cell lines in vitro. In addition, we demonstrate significant cytotoxicity mediated by 4H11-28z⁺ patient T cells and healthy donor T cells targeting primary ascites-derived ovarian carcinoma cells. In vivo studies in immune compromised SCID-Beige mice bearing established peritoneal orthotopic MUC-CD⁺ human ovarian tumors demonstrate marked anti-tumor efficacy of MUC-CD targeted T cells, following either intraperitoneal (i.p.) or intravenous (i.v.) injection. These data serve as a rationale for future clinical trials utilizing this approach in patients with high risk ovarian carcinomas.

MATERIALS AND METHODS

Cell lines and T cells

The OV-CAR3 and SK-OV3 tumor cell lines were cultured in RPMI 1640 (Invitrogen, Grand Island, NY) supplemented with 10% heat-inactivated FBS, nonessential amino acids, HEPES buffer, pyruvate, and BME (Invitrogen). T80 cells originated from normal ovarian surface epithelium (NOSE) which were immortalized by transfection with SV40 large T antigen were kindly provided by Dr. Robert C. Bast (39). T80 cells were cultured in 1:1 ratio of MCDB 105 medium (Sigma-Aldrich Co., St. Louis, MO) and Medium 199 (Invitrogen) supplemented with 10% heat-inactivated FBS. The PG13, HeLa and gpg29 retroviral producer cell lines were cultured in DMEM (Invitrogen) supplemented with 10% FCS. NIH-3T3 artificial antigen-presenting cells (AAPCs), described previously (3), were cultured in DMEM supplemented with 10% heat-inactivated donor calf serum. Human T cells were isolated from peripheral blood of healthy donors under IRB approved protocol #95-054, using BD Vacutainer CPT tubes (Becton Dickinson, Franklin Lakes, NJ) as per the manufacturers instructions. T cells were cultured RPMI 1640 media as above supplemented with 20 IU/ml IL-2 (Novartis Pharmaceuticals, East Hanover, NJ) and where indicated, medium was supplemented with 10ng/mL interleukin 15 (IL-15) (R&D Systems, Minneapolis, MN). Primary ovarian cancer cells derived from ascites specimens were cultured in RPMI 1640 media supplemented with 10% FBS. All media were supplemented with 2 mmol/L L-glutamine (Invitrogen), 100 units/mL penicillin, and 100 μg/mL streptomycin (Invitrogen).

Isolation of patient ascites-derived ovarian cancer cells

Primary human ascites-derived cancer cells were obtained from ovarian cancer patients undergoing surgery for newly diagnosed ovarian carcinomas under IRB approved protocol #97-134. The tumor cells were isolated from ascities fluid of patients by centrifugation at 600g for 10 min at room temperature. Cells were washed once with 1x PBS and cultured in FBS supplemented RPMI 1640 media in tissue culture flasks. After 5 days in culture, non-adherent cells were removed retaining the tumor cell enriched adherent fraction for further study.

Generation of the MUC-CD targeted 4H11z and 4H11-28z CARs

The heavy and light chain variable regions of the 4H11 monoclonal antibody were derived from the hybridoma cell line 4H11. Utilizing cDNA generated from 4H11 RNA we isolated the V_H coding region by RACE PCR utilizing modified primers as described elsewhere (40, 41). The V_L chain variable region was cloned by standard PCR utilizing modified primers as described by Orlandi et al (42, 43). The resulting V_H and V_L fragments were subcloned into the TopoTA PCR 2.1 cloning vector (Invitrogen) and sequenced. The V_H and V_L fragments were subsequently ligated to a (Gly₄Ser)₃ spacer domain, generating the 4H11 scFv and fused to the human CD8 leader peptide (CD8L) by overlapping PCR (9, 42). In order to construct the MUC-CD targeted 4H11 CARs, the coding region of the CD8L-4H11 scFv was fused to the human CD8 hinge and transmembrane domains (to generate the 4H11z CAR), or alternatively to the CD28 transmembrane and cytoplasmic signaling domains (to generate the 4H11-28z CAR), fused to the T cell receptor CD3ζ-signaling domain (3, 9, 44). The resulting CAR constructs were subsequently sub-cloned into the modified MMLV retroviral vector SFG (45). VSV-G preudotyped retroviral supernatants derived from transduced gpg29 fibroblasts were used to construct stable PG13 gibbon ape leukemia virus (GaLV) envelope-pseudotyped retroviral producing cell lines (42).

Retroviral gene transfer

Isolated healthy donor peripheral blood mononuclear cells (PBMCs) were activated with phytohemagglutinin (PHA) at 2μg/ml (Sigma. St. Louis, MO) while patient T cells derived from fresh ascites samples were isolated, activated, and expanded with Dynabeads® ClinExVivo™ CD3/CD28 beads (Invitrogen, Carlsbad, CA) following the manufacturer’s recommendations. Activated T cells were retrovirally transduced on retronectin coated non-tissue culture plates as described previously(46). Gene transfer was assessed on day 7 by FACS.

In order to generate the relevant NIH-3T3 murine fibroblast artificial antigen presenting cells, a MUC-CD construct encoding the retained extracellular, transmembrane and cytoplasmic domains of the MUC-16 antigen was initially subcloned into SFG retroviral vector, SFG(MUC-CD). 3T3(MUC-CD) AAPCs were generated by retroviral transduction of SFG(MUC-CD) into wild-type NIH-3T3 fibroblasts, while 3T3(MUC-CD/B7.1) AAPCs were generated by retroviral transduction of previously established 3T3(B7.1) fibroblasts (42, 47). Enriched cell lines were isolated by FACS.

OV-CAR3(MUC-CD), SK-OV3(MUC-CD), HeLa(MUC-CD), OV-CAR3(MUC-CD/GFP-FFLuc) and SK-OV3(MUC-CD/GFP-FFLuc) cell lines were generated by retroviral transduction with SFG(GFP-FFLuc)(48) and/or SFG(MUC-CD) VSV-G pseudotyped retroviral supernatants derived from gpg29 fibroblasts as described elsewhere (45). T80(MUC-CD) cells were generated by transfection of the MUC-CD gene using the Vitality phrGFP II C vector expression system (Stratagene, LaJolla, CA). Resulting tumor cells were sorted by FACS for MUC-CD expression.

In vitro analyses of CAR⁺ human T cells

To assess in vitro expansion and cytokine release upon stimulation, 3 × 10⁶ transduced T cells were co-cultured for 7 days after retroviral transduction in 6-well tissue culture plates (BD Biosciences) on confluent NIH 3T3 AAPCs in the absence of supplemented cytokines. For in vivo studies, transduced T cells were expanded by co-culture on 3T3(MUC-CD/B7.1) AAPCs in RPMI medium supplemented with 20 IU IL-2/mL and 10 ng/mL IL-15 as described previously (3, 44).

Western Blot analysis of CAR expression

Western blot analysis of T-cell lysates under reducing conditions with 0.1 mol/L DTT (Sigma) was performed as previously described (47).

Cytotoxicity assays

In vitro modified T cell cytotoxicity was assessed using the DELFIA® EuTDA assay (PerkinElmer LAS, Inc, Boston, MA) following manufacturer’s recommendations. Cytotoxocity was assessed at 2 hours at serially diluted effector T cell to target cell (E:T) ratios. Effector T cells in these assays represent the CD8⁺ CAR⁺ T cell fraction.

Cytokine detection assays

Cytokine assays were performed as per manufacturer’s specifications using a multiplex Human Cytokine Detection assay to detect IL-2 and IFNγ (Millipore Corporation, Billerica, MA) utilizing the Luminex IS100 system. Cytokine concentrations were assessed using IS 2.3 software (Luminex Corp., Austin, TX).

In vivo SCID-Beige mouse tumor models

In all in vivo studies, 8–12 week-old FOX CHASE C.B.-17 (SCID-Beige mice) (Taconic, Hudson, NY) were injected i.p. with 3 × 10⁶ tumor cells. Subsequently, CAR⁺ T cells, at indicated doses, were injected either i.p. or i.v. on either day 1 or 7 following tumor cell injection. Mice were monitored for distress as assessed by increasing abdominal girth, ruffled fur, and decreased response to stimuli. Distressed mice were euthanized. All murine studies were done in the context of an Institutional Animal Care and Use Committee-approved protocol (#00-05-065).

Bioluminescent imaging (BLI) of GFP-FFLuc⁺ human ovarian tumor cells in SCID-Beige mice

BLI was performed using Xenogen IVIS imaging system with Living Image software (Xenogen; Alameda, CA). Briefly, GFP-FFLuc⁺ tumor bearing mice were injected by ip with D-luciferin (150 mg/kg; Xenogen) suspended in 200 μl PBS and imaged under 2% isoflurane anesthesia after 10 min. Image acquisition was done on a 25-cm field of view at medium binning level for 0.5-min exposure time (3, 44).

Flow cytometry

All flow cytometric analyses of T cells and tumor cells was performed using a FACScan cytometer with Cellquest software (BD Biosciences). T cells were analyzed using CAR-specific polyclonal goat Alexa Fluor 647 antibody (Molecular probes, Eugene, OR), phycoerythrin-labeled anti-human CD4, CD8, B7.1, CD45RO (Caltag Laboratories, Burlingame, CA), B7.2 (Invitrogen, Camarillo, CA), phycoerythrin-conjugated anti-human CCR7, and FITC-conjugated anti-human CD62L (eBioscience, San-Diego, CA). MUC-CD expression was measured by FACS with Alexa Fluor 647 labeled 4H11 antibody (generated and labeled in the MSKCC monoclonal antibody core facility).

CFSE labeling of CAR⁺ T cells

CAR⁺ T cells were stained with CFSE using the CellTrace^TM CFSE cell proliferation kit following the manufacturer’s recommendations (Molecular Probes, Eugene, OR).

Assessment of in vivo T cell persistence

SCID-Beige mice were infused i.p. with OVCAR3(MUC-CD) tumor cells and 7 days later were treated by i.p. infusion with CAR⁺ T cells. Subsequently, mice were sacrificed at days 2, 14, 21 and 28 following T cell infusion. Peritoneal washes were collected using 10 ml of PBS, red blood cells were lysed with ACK lysing buffer (Lonza, Walkersville, MD), remaining cells were washed with 2% FBS/PBS and analyzed by FACS for persistence of human CD3⁺ T cells.

Statistics

Survival data assessed by log-rank analysis and cytokine secretion data assessed by Student’s two-tailed t-test were analyzed using GraphPad Prism software (GraphPad Prism software, San Diego, CA). In the comparison between survival of i.p. versus i.v. modified T cells infusion, two experiments were performed with different follow-up times. For combined analysis, the stratified log-rank statistic (stratified by experiment) was used to test whether the survival rates differed between the i.p. and i.v. treatment groups.

RESULTS

We have constructed SFG retroviral vectors encoding first (4H11z) and second generation, co-stimulatory (4H11-28z) CARs targeted to the MUC-CD antigen using the 4H11 hybridoma which generates a MAb specific to the MUC-CD antigen (figure 1A). We confirmed expression of appropriately sized CAR proteins by Western blot analysis of resulting PG-13 retroviral producer cells (SFG-4H11z and SFG-4H11-28z) probed with a ζ-chain specific antibody (data not shown).

Figure 1.

Design and in vitro analysis of MUC-CD targeted CARs. (A) Schematic diagram of the first generation 4H11z and second generation 4H11-28z retroviral vectors. 4H11scFv: MUC16 specific scFv derived from the heavy (V_H) and light (V_L) chain variable regions of the monoclonal antibody 4H11; CD8: CD8 hinge and transmembrane domains; CD28: CD28 transmembrane and cytoplasmic signaling domains; ζ chain: T cell receptor ζ chain cytoplasmic signaling domain; LTR: long terminal repeat; black box: CD8 leader sequence; grey box: (Gly₄Ser)₃ linker; arrows indicate start of transcription. (B) FACS analysis of human T cells retrovirally transduced to express either the 4H11z or 19z1 CAR. (C) 4H11z⁺ but not 19z1⁺ T cells expand on 3T3(MUC-CD/B7.1) AAPC. CAR⁺ were co-cultured on 3T3(MUC-CD/B7.1) AAPC monolayers at 3 × 10⁶ CAR⁺ T cells/well of a 6 well plate. Proliferation of CAR⁺ T cells was assessed by FACS in combination with viable T cell counts as assessed by trypan blue exclusion assays.

In order to assess the function of the first generation 4H11z CAR, healthy donor T cells isolated from peripheral blood were retrovirally transduced to express the 4H11z and control T cells modified to express the irrelevant CD19-targeted 19z1 CARs (figure 1B). Function of the 4H11z CAR was assessed by proliferation of 4H11z transduced T cells following co-culture on 3T3(MUC-CD/B7.1) AAPCs. Results demonstrate specific proliferation of 4H11z transduced T cells, when compared to 19z1 modified T cells (figure 1C). These data are consistent 4H11z CAR mediated specific binding to the MUC-CD antigen and subsequent T cell activation.

Since most malignancies fail to express co-stimulatory ligands, we further modified the 4H11z CAR to express the CD28 transmembrane and cytoplasmic co-stimulatory signaling domains, generating the 4H11-28z CAR (figure 1A). To assess whether the 4H11-28z CAR, when expressed by human T cells, was capable of generating both a primary activating signal (termed “signal 1”) through the ζ chain, as well as a co-stimulatory signal (termed “signal 2”) through the CD28 cytoplasmic domain, which in turn allows for efficient T cell proliferation in the absence of exogenous co-stimulatory ligands, we compared T cell proliferation following co-culture on either 3T3(MUC-CD) or 3T3(MUC-CD/B7.1) AAPCs in the absence of exogenous cytokines. As expected, the second generation 4H11-28z⁺ T cells markedly expanded when compared to 4H11z⁺ T cells upon co-culture with 3T3(MUC-CD) AAPCs (p=0.0047). In contrast, both 4H11z⁺ and 4H11-28z⁺ T cells expanded equally well on 3T3(MUC-CD/B7.1) AAPCs (p=0.18) (figure 2A). Co-stimulation mediated by the 4H11-28z CAR was further verified by analysis of day 2 tissue culture supernatants from co-culture experiments on 3T3(MUC-CD) AAPCs demonstrating significantly enhanced IL-2, a cytokine typically secreted in the context of T cell co-stimulation, and IFNγ secretion when compared to control 19-28z⁺ T cells and first generation 4H11z⁺ T cells (figure 2B).

Figure 2.

In vitro comparison of T cells modified to express the first generation 4H11z CAR to T cells modified to express the second generation co-stimulatory 4H11-28z CAR. (A) CAR⁺ T cells were co-cultured on MUC-CD monolayers with (right panel) or without B7.1 (left panel). 1 × 10⁶ CAR⁺ T cells per well were co-cultured on AAPC monolayers in 12 well tissue culture plates in cytokine-free medium. Total viable T cell counts were assessed on days 2, 4 and 7, by trypan blue exclusion assays. (B) Antigen specific cytokine’s secretion by MUC-CD⁺ targeted T cells. (C) Expansion of CAR⁺ T cells following 3 cycles of stimulation on 3T3(MUC-CD/B7.1). Arrows indicate 1st and 2nd cycles of restimulation on AAPCs. (D) FACS analysis of the CAR⁺ T cell fraction of 4H11-28z⁺ T cells increased following each weekly cycle of stimulation. (I) FACS following initial transduction, (II) FACS at 7 days following first stimulation on AAPCs, (III) FACS at 7 days following second stimulation on AAPCs.

We next assessed the ability of 4H11-28z⁺ T cells to expand following weekly re-stimulations on 3T3(MUC-CD/B7.1) AAPC monolayers in the context of exogenous IL-2 and IL-15 (3). In this setting, 4H11-28z⁺ T cells expanded greater than 2 logs over 3 weeks compared to no expansion of control 19-28z⁺ T cells (figure 2C). T cells transduced with the 4H11-28z were further analyzed by FACS for CAR expression 7 days after initial transduction and following two subsequent co-stimulations on AAPCs demonstrating an expected enrichment of the CAR⁺ T cell fraction (figure 2D).

In vitro MUC-CD specific activation of 4H11-28z⁺ T cells

In order to assess MUC-CD specificity of 4H11-28z⁺ T cells, we initially conducted a series of cytotoxicity assays using the MUC-CD⁻ OVCAR3 and SK-OV3 cell lines with or without further genetic modification with MUC-CD. As expected, 4H11-28z⁺ T cells efficiently lysed MUC-CD⁺ but not unmodified tumor cell lines consistent with MUC-CD specificity mediated through the 4H11-28z CAR (figure 3A). To further verify these findings, we conducted similar studies on non-ovarian carcinoma cell lines T80 and HeLa demonstrating similar results (supplemental figure 1). We next assessed the activation of CAR modified T cells, as measured by proliferation and cytokine secretion, by co-culture with these same cell lines. Consistent with 4H11-28z CAR specificity for MUC-CD, T cell proliferation as well as IL-2 and IFNγ secretion only occurred in the setting of MUC-CD⁺ cell lines co-cultured with 4H11-28z⁺ T cells (figure 3B-C, supplemental figure 1).

Figure 3.

Upon co-culture, 4H11-28z⁺ T cells specifically expand and lyse MUC-CD⁺ tumor cells. (A) Cytotoxicity assays of 4H11-28z⁺ T cells targeting OV-CAR3(MUC-CD) and SK-OV3(MUC-CD) tumor cells demonstrate efficient cytotoxicity mediated by 4H11-28z⁺ T cells T cells from healthy donors compared to control 19-28z⁺ T cells. (B) Antigen specific proliferation of MUC-CD targeted T cells after co-culture with OV-CAR3(MUC-CD) or SK-OV3(MUC-CD) tumor cells. CAR⁺ T cells were co-cultured with wild type or MUC-CD expressing tumor cells at 1:1 ratio for 7 days. Absolute T cell numbers assessed on days 2, 4 and 7 following co-culture with ovarian tumor cells. The T cell seeding density for T cell proliferation assay was 1 × 10⁶/mL. (C) Antigen-specific IL-2 and IFN-γ secretion by 4H11-28z transduced T cells after 2 days of co-culture on ovarian cancer cells. (D) Healthy donor T cells or autologous T cells isolated from ascites, modified to express the 4H11-28z CAR lyse primary patient ascites-derived MUC-CD⁺ tumor cells when compared to T cells modified to express the control 19-28z CAR.

Utilizing more clinically relevant MUC-CD⁺ ovarian carcinoma cells enriched from fresh acsites samples, healthy donor T cells modified to express the 4H11-28z CAR similarly exhibited lysis of primary ovarian carcinoma cells when compared to 19-28z transduced T cells (figure 3D). Moreover, in 3 of 3 cases, we found that patient peripheral blood T cells modified to express the 4H11-28z CAR similarly lysed matched autologous primary ovarian carcinoma cells (figure 3D).

In vivo anti-tumor activity of MUC-CD targeted T cells in SCID-Beige mice

To assess the in vivo anti-tumor activity of 4H11z⁺ and 4H11-28z⁺ T cells, we next generated an orthotopic xenotransplant ovarian cancer tumor model by i.p. injection of OV-CAR3(MUC-CD) tumor cells into SCID-Beige mice. If left untreated, these mice developed marked ascites and multiple nodular peritoneal tumors by 5 weeks following tumor cell injection (figure 4A). All untreated tumor bearing mice had to be euthanized by 7 weeks following tumor cell injection due to abdominal distention and evidence of distress.

Figure 4.

Eradication of OV-CAR3(MUC-CD) tumors after i.p. treatment with first and second generation of MUC-CD targeted T cells. (A) Intraperitoneal injection of OV-CAR3(MUC-CD) tumors in untreated SCID-Beige mice results in abdominal distension and nodular peritoneal tumors. At 5 weeks post intraperitoneal injection of OV-CAR3(MUC-CD) tumor cells mice developed ascites as evidenced by a distended abdomen (center panel) when compared to a tumor free mouse (left panel). Post mortem visualization of the peritoneum demonstrates nodular tumor masses (arrows) within the abdominal cavity (right panel). (B) Intraperitoneal injection of 4H11z⁺ and 4H11-28z⁺ T cells either delay tumor progression or fully eradicate disease. Kaplan-Meier survival curve of SCID-Beige mice treated with first or second generation of MUC-CD targeted T cells.

For in vivo studies, cohorts of SCID-Beige mice, injected i.p. with 3 × 10⁶ OV-CAR3(MUC-CD/GFP-FFLuc) tumor cells on day 1, were initially treated with dose escalating levels of CAR-modified T cells by i.p. injection on day 2. FACS analysis of modified in vitro expanded T cells utilized in our in vivo studies demonstrated a majority of infused T cells with a retained a central memory phenotype as assessed by CD62L, CCR7, CD28, and CD45RO expression (49) (supplemental figure 2). These studies demonstrated a dose dependent in vivo anti-tumor response to MUC-CD targeted T cell therapy at a treatment dose of 1 × 10⁷ 4H11-28z⁺ T cells found to be the minimum dose required achieve a meaningful long-term survival at day 70, with all mice treated at lower dose levels demonstrating 100% mortality (supplemental figure 3). Based on these findings we chose the highest tested dose level of 3×10⁷ CAR⁺ T cells for further studies. We subsequently repeated these studies with larger cohorts of mice treated at this dose level of 4H11z⁺ or 4H11-28z⁺ T cells. For negative controls, tumor bearing mice were either untreated or treated with T cells modified to express the irrelevant CD19-targeted 19z1 CAR. Both MUC-CD-targeted T cell treated cohorts demonstrated statistically significant enhanced survival when compared to untreated or 19z1⁺ T cell treated control cohorts with no statistically significant difference in survival when comparing the 4H11z⁺ and 4H11-28z⁺ T cell treated cohorts (figure 4B and supplemental figure 4A).

As a further in vivo control for 4H11-28z MUC-CD specificity, we repeated these studies utilizing SCID-Beige mice bearing i.p. MUC-CD⁻ SK-OV3(GFP-FFLuc) or SK-OV3(MUC-CD/GFP-FFLuc) human ovarian carcinoma tumors similarly treated with 3 × 10⁷ 4H11-28z⁺ T cells one day after tumor cell injection. As expected, treatment with 4H11-28z⁺ T cells failed to result in any long-term survival of SK-OV3(GFP-FFLuc) tumor bearing mice, while 40% of mice bearing SK-OV3(MUC-CD/GFP-FFLuc) tumors remain alive at day 70 (supplemental figure 4B) with no evidence of disease as assessed by BLI (data not shown).

To determine whether systemically infused MUC-CD-targeted T cells successfully traffic to peritoneal tumors, we next compared i.p. to i.v. infusion of 4H11-28z⁺ T cells in SCID-Beige 1 day after i.p. infusion of OV-CAR3(MUC-CD/GFP-FFLuc) tumors. Both i.p. and i.v. 4H11-28z⁺ T cell treated mice exhibited statistically enhanced survival when compared to untreated or 19-28z⁺ T cell treated control cohorts as assessed by BLI imaging (figure 5A) as well as overall survival (figure 5B, supplementary figure 5). Stratified log-rank analysis of merged data from 2 independent experiments demonstrated statistically equivalent anti-tumor efficacy between i.p. and i.v. MUC-CD targeted T cell treated cohorts (p=0.092).

Figure 5.

MUC-CD targeted 4H11-28z⁺ T cells traffic to peritoneal OV-CAR3(MUC-CD/GFP-FFLuc) tumors following systemic intravenous infusion resulting in efficient anti-tumor efficacy. (A) BLI of tumor progression of representative i.p. and i.v. 4H11-28z⁺ T cell treated mice with ultimately progressive disease following treatment compared to BLI of tumor progression in a representative control 19-28z⁺ T cell treated mouse. (B) Kaplan-Meier survival curve of SCID-Beige mice treated i.p. or i.v. with 4H11-28z⁺ T cells. Tumor eradication is enhanced after either ip or iv infusion of 4H11-28z⁺ T cells when compared to control treated mice. Both ip and iv 4H11-28z⁺ T cell treated mice exhibited statistically enhanced survival when compared to 19-28z⁺ T cell treated control cohorts while survival between the i.p. and i.v. treated 4H11-28z⁺ T cell cohorts was not statistically significant (p=0.22). (C) Systemically injected CFSE stained 4H11-28z⁺ T cells traffic to advanced i.p. OV-CAR(MUC-CD) tumors. Presence of i.v. injected CFSE labeled 19-28z⁺ control T cells (left panel) and 4H11-28z⁺ T cells (right panel) 1 day following infusion into SCID-Beige mice with OV-CAR(MUC-CD) tumors injected 7 days earlier as assessed by FACS analysis of single cell OV-CAR3(MUC-CD) tumor suspensions.

We further confirmed trafficking of i.v. infused CFSE labeled 4H11-28z⁺ T cells to the peritoneum by FACS analysis of single cell suspensions of macerated OV-CAR3(MUC-CD) peritoneal tumor deposits (figure 5C). Presence of i.v. injected CFSE labeled 19-28z⁺ control T cells and 4H11-28z⁺ T cells 1 day following infusion into SCID-Beige mice with advanced OV-CAR(MUC-CD) tumors (injected 7 days earlier), reveals a marked population of human CD3/CFSE⁺ T cells within peritoneal OV-CAR3(MUC-CD) tumor deposits of 4H11-28z⁺ but not control 19-28z⁺ T cell treated mice.

We next treated SCID-Beige mice bearing peritoneal OV-CAR3(MUC-CD/GFP-FFLuc) tumor injected 7 days prior to adoptive T cell therapy, a time point at which the mice had evidence of overt disease as assessed by BLI. A subset of 4H11-28z⁺ T cell treated mice assessed to be tumor free by BLI on day 60 after tumor cell infusion were sacrificed and peritoneal washes were analyzed for presence of MUC-CD/GFP-FFLuc⁺ tumor cells by FACS. At this time point, in contrast to control treated mice sacrificed and analyzed at day 30, peritoneal washings from treated mice demonstrated <0.01% of analyzed cells to be MUC-CD/GFP⁺ (Supplemental Fig. 6A). Once more, we found that therapy with MUC-CD targeted T cells initially eradicated most BLI evident disease in all treated mice (figure 6A) with 75% of mice ultimately developing relapsed disease at later time points while 25% of treated mice survive at 120 days post-tumor cell infusion with no evidence of disease as assessed by BLI (figure 6B). Significantly, FACS analyses of tumor cell suspensions obtained from all 4H11-28z⁺ T cell treated mice with relapsed disease demonstrated persistent expression of the MUC-CD antigen (data not shown).

Figure 6.

Eradication of BLI evident OV-CAR3(MUC-CD) tumors in SCID-Beige mice by ip infusion of 4H11-28z⁺ T cells. (A) BLI of 4H11-28z⁺ T cell treated mice with either relapsed disease (middle row) or eradicated disease (bottom row) compared to a representative 19-28z⁺ T cell treated control mouse. (B) Kaplan-Meier survival curve of SCID-Beige mice with advanced OV-CAR3(MUC-CD/GFP-FFLuc) tumors treated i.p. with 4H11-28z⁺ T cells.

To test for the persistence of modified T cells over time in these studies, additional mice were infused i.p. with OV-CAR3(MUC-CD) tumors and treated on day 7 with modified T cells, and sacrificed serially following therapy. We analyzed for the presence of CAR⁺ T cells in peritoneal washes of 4H11-28z and 19-28z T cell treated tumor-bearing mice on days 2, 14, 21 and 28 after T cell infusion by FACS, demonstrating a decreasing but persistent population of modified T cells out to 28 days post T cell infusion (the latest time point studied) (supplemental figure 6B).

DISCUSSION

Based on analyses of patient tumor samples, ovarian carcinomas appear to be relatively immunogenic tumors. Specifically, researchers have found a direct correlation between prognosis following surgery and chemotherapy and the quantity of TILs in pretreatment tumor samples (25, 50, 51). Furthermore, others have described an inverse correlation between prognosis following therapy and pre-treatment levels of Tregs within the tumor, which in turn presumably inhibit the anti-tumor function of tumor specific effector TILs (26, 28, 52). Both of these findings imply a role for an endogenous effector T cell response to tumor in controlling disease progression both prior to and following initial therapy and strongly support the contention that ovarian carcinomas may be susceptible to killing by adoptive infusion of autologous T cells targeted to ovarian tumor cell antigens.

While endogenous effector TILs are one source for presumably tumor specific T cells, an alternative approach to adoptive T cell therapy is to isolate autologous peripheral blood T cells which in turn may be genetically modified ex vivo to target tumor cell antigens. One such genetic approach is to retrovirally transduce patient T cells with CARs targeted to surface exposed antigens either unique to or over-expressed by the tumor. To this end, promising preclinical studies utilizing this approach in other malignancies have recently been translated to the clinical setting (6, 16, 19, 53).

Application of this approach to ovarian carcinomas requires the identification of suitable target antigens expressed on the tumor cell surface. To this end, other investigators have studied this approach in vivo in the pre-clinical setting utilizing CAR⁺ T cells targeted to disparate antigens over-expressed on ovarian carcinomas including the α-folate receptor, the Lewis Y antigen, Her2/neu, NKG2D ligands (MICA, MICB, and UL-16 binding proteins), mesothelin, and MUC-1 (4, 11, 54–61). Specifically, in an elegant series of studies utilizing a syngeneic orthotopic 1D8 tumor model of ovarian carcinoma in C57BL6 mice, Barber et al demonstrate efficient eradication of well established i.p. tumors when treated with i.p. injections of syngeneic murine T cells modified to express the chNKG2D CAR (54). Furthermore, Hwu et al demonstrated significant delays in tumor progression in immune compromised nude mice bearing orthotopic human IGROV tumors following a single infusion of murine T cells modified to express an α-folate receptor-targeted CAR (4). In the xenogeneic setting, several groups have demonstrated delayed tumor progression or complete anti-tumor responses of subcutaneous human ovarian carcinoma cell lines in immune compromised mice following intratumoral and/or intravenous infusion of human T cells expressing CARs specific to the Her2/neu and Lewis-Y antigens (59, 61, 62). Similarly, Wilkie et al and Carpentino et al have recently published reports demonstrating efficient anti-tumor efficacy in xenotransplant models of human breast and mesothelioma tumors treated with human T cells modified respectively with CARs targeted to the MUC-1 and mesothelin antigens, antigens also over-expressed on ovarian carcinomas(57, 60).

In the clinical setting, Kershaw et al recently published the results of a phase I dose escalation trial treating patients with relapsed ovarian carcinomas with autologous T cells modified to express a first generation CAR specific to the α-folate receptor (6). The authors of this study found that therapy with targeted T cells was well tolerated, but noted a lack of anti-tumor response in these studies related to poor persistence of modified T cells over time as well as a yet undefined T cell inhibitory factor in the serum of several treated patients.

In our studies, we have chosen to target the MUC-16 glycoprotein which is over-expressed on a majority of ovarian carcinomas (1, 30, 32, 33). The utility of MUC-16 as a target antigen for adoptive T cell therapy is compromised by the fact that most of the extracellular portion of this molecule is cleaved by the tumor cell, secreted, and may be detected in the serum as the CA-125 tumor marker. However, following cleavage of this secreted fraction of MUC-16, there remains a residual extracellular fraction of the glycoprotein, termed MUC-CD, which is retained on the tumor cell surface and is therefore an attractive target for immune-based therapies. To this end, we utilized a murine hybridoma, 4H11, generated to the MUC-CD antigen (38) to construct a first generation (4H11z) as well as a second generation co-stimulatory CAR (4H11-28z) specific to MUC-CD. Significantly, the antigen to the 4H11 antibody is highly expressed on a majority of pre-treatment ovarian carcinoma tumor samples obtained from patients treated at our institution as assessed by immunohistochemistry (38, 63).

Consistent with previous studies, we found that T cells transduced to express the second generation 4H11-28z CAR, but not the first generation 4H11z CAR, efficiently expanded upon co-culture with 3T3(MUC-CD) fibroblasts in the absence of exogenous co-stimulation. This conclusion is further supported by the finding that 4H11-28z⁺ T cells secreted significantly higher levels of IL-2, a cytokine indicative of T cell co-stimulation, and IFNγ upon co-culture on 3T3(MUC-CD) fibroblasts when compared to T cells transduced to express the first generation 4H11z CAR.

Specificity of the 4H11-28z CAR to the MUC-CD antigen was subsequently verified in vitro by comparing 4H11-28z⁺ T cell cytotoxicity and proliferation on a series of MUC-CD⁻ ovarian carcinoma cell lines, as well as HeLA cells and the T80 immortalized normal ovarian surface epitherlial cell line, to the same cell lines further genetically modified to express the MUC-CD antigen. In order to additionally validate the clinical relevance of our findings, we subsequently demonstrated specific in vitro lysis of primary ascites-derived tumor cells isolated from untreated ovarian carcinoma patients by both healthy donor allogeneic 4H11-28z⁺ T cells and, more significantly, by autologous 4H11-28z⁺ patient peripheral blood T cells. These data strongly support the contention that treatment with autologous 4H11-based CAR⁺ T cells have promise in future clinical applications.

In order to assess the in vivo relevance of our in vitro findings, we next generated several orthotopic human ovarian cancer tumor models in SCID-Beige mice. These studies demonstrated eradication of early as well as more established BLI evident peritoneal tumors following i.p. injection of healthy donor 4H11-28z⁺ T cells. In the setting of delayed therapy, tumor imaging by BLI could demonstrate initial marked eradication of disease. However, loss of bioluminescent signal in these studies did not preclude future relapse of disease since a majority of apparently tumor free mice, as assessed by BLI at earlier time points of follow-up, developed relapsed disease within the peritoneum over time, consistent with the notion that BLI lacks the sensitivity to measure in vivo minimal residual disease in demonstrated persistent expression of the targeted MUC-CD antigen consistent with the notion that loss of target antigen expression by the tumor, or immune selection of MUC-CD⁻ tumors was not the cause of tumor relapse. While the source of relapsed disease in these mice remains speculative, studies of i.p. infused 4H11-28z⁺ T cell persistence demonstrate rapidly declining numbers of T cells over time. Although modified T cells were still detectable in peritoneal washes out to 28 days post T cell infusion, these numbers were declining, suggesting a loss of modified T cell persistence as a potential source of disease relapse which typically occurred at later time points.

We further demonstrate trafficking of i.v. injected MUC-CD targeted T cells to peritoneal tumors by FACS. Significantly, tumor bearing mice treated with i.v. infused 4H11-28z⁺ T cells exhibited similar anti-tumor efficacy when compared to i.p. treated mice as assessed by combined survival data from 2 separate experiments using stratified log-rank analysis.

While insightful, these xenotransplant murine tumor models have significant limitations. Specifically, the biology of human T cells in immune compromised mice may significantly differ from autologous modified patient T cells in the clinical setting wherein these T cells encounter an intact immune system which may elicit an immune response to the CAR, and enter into a hostile tumor microenvironment containing immune suppressive regulatory T cells, immune inhibitory cytokines, including IL-10 and TGFβ, and myeloid derived suppressor cells (MDSCs) (64–66). To address these limitations, we are currently generating a more clinically relevant syngeneic immune competent murine tumor model of ovarian carcinoma to further study the in vivo biology, immunogenicity, and anti-tumor efficacy of MUC-CD targeted T cells.

A further limitation of xenotransplant models is the inability of these studies to address potential unforeseen off-target toxicities which may occur in the clinical setting wherein infused CAR-modified T cells recognize antigen not only on tumor cells but also on normal tissues. Based on previously published adverse events in recent clinical trials utilizing CAR-modified T cells, this is a very real concern with respect to the clinical feasibility of this adoptive T cell approach to cancer therapy. Specifically, three patients with metastatic renal carcinoma treated with autologous T cells transduced to express the G250 CAR specific to the TAA carboxyanhydrase IX (CAIX) developed significant liver toxicity (CTC grade 4 in patient 1, grade 2 in patient 2 and grade 3 in patient 3) due to an off target modified T cell response to CAIX expressed by bile duct epithelial cells (20, 53). More recently, Morgan et al reported the death of a patient with metastatic colon cancer treated with ERBB2 targeted autologous 4D5-CD8-28BBz⁺ T cells. Following an extensive post-mortem analysis, the investigators of this study postulate that the patient’s death resulted from off target recognition of the targeted ERBB2 antigen expressed by normal lung tissue resulting in marked modified T cell release of inflammatory cytokines, including TNF-α and IFN-γ, leading to pulmonary toxicity, edema, and a subsequent cascading cytokine storm resulting in multi-organ failure and death (67).

While similar concerns may be raised in the setting of treating patients with autologous T cells targeted to the MUC-CD antigen, we have conducted extensive immunohistochemistry studies using the 4H11 antibody to assess for off target binding to normal human tissues. Significantly, these studies demonstrated no binding of the 4H11 MAb in normal adult colon, rectum, small intestine, ectocervix, ovary, liver, pancreatic ducts, spleen, kidney, brain and skin tissues. However, 4H11 did weakly bind cytoplasm of endocervical gland cells and thymic corpuscles, the luminal surface of esophageal glands, as well as intracytoplasmic granules of bronchial epithelium and gastric glands (38). Significantly, these studies failed to demonstrate membrane bound antigen on the vascular surfaces of any normal human tissues. Nevertheless, the potential for off target toxicity in the clinical setting are not precluded by these studies. To this end, we acknowledge the likely requirement that further safety measures be added in the form of additional T cell modification with a suicide gene vector in order to enhance safety in the clinical setting.

In conclusion, herein we present the first published data demonstrating the feasibility of targeting MUC-16, an antigen over-expressed on a majority of ovarian carcinomas, through adoptive therapy with genetically modified T cells targeted to the retained MUC-CD portion of the MUC-16 antigen. Further, this report is the first to demonstrate efficient targeting of T cells in an orthotopic murine model of ovarian cancer, demonstrating efficacy of a single T cell infusion of modified T cells in the absence of exogenous IL-2 cytokine support. Collectively, these data support the further planned translation of this approach to the clinical setting in the form of a phase I clinical trial in patients with persistent or relapsed ovarian carcinomas following initial therapy with surgery and chemotherapy.

Statement of Translational Relevance.

Ovarian carcinomas appear to be immunogenic tumors based on the fact that increased numbers of tumor infiltrating T cells (TILs) present in the pre-treatment tumor is associated with an enhanced survival following surgery and chemotherapy. Patient T cells may be genetically modified to recognize tumor associated antigens (TAAs) through the introduction of chimeric antigen receptors (CARs) specifically targeted to these antigens. We have generated a CAR, 4H11-28z, specific to the retained extracellular portion (MUC-CD) of MUC-16, a glycoprotein over expressed on most ovarian carcinomas. We demonstrate that T cells, when modified to express the 4H11-28z CAR, specifically lyse human ovarian cancer cells in vitro and further demonstrate efficient anti-tumor efficacy in orthotopic xenotransplant tumor models. Collectively, these data support the translation of these preclinical findings to the clinical setting in a planned phase 1 trial treating patients with relapsed or refractory ovarian cancer.