Human Regulatory T Cells Kill Tumor Cells through Granzyme-Dependent Cytotoxicity upon Retargeting with a Bispecific Antibody

Despite maximal surgical resection, dose-intensive radiotherapy and multimodal chemotherapy, glioblastoma remains uniformly lethal with a median survival of less than 15 months (1). Novel therapies are desperately needed to improve the prognosis of this disease and immunologic targeting of tumor-specific mutations has emerged as a promising approach (2).

A major barrier that has impeded translation of efficacious immunotherapy is the inability to overcome profound immunosuppression associated with malignant disease (3). Regulatory T cells (T_regs) in particular are thought to play a central role in tumor escape from immune-mediated rejection. One mechanism by which T_regs are known to suppress and even kill autologous immune cells is through the granzyme-perforin pathway (4, 5). Despite this well-characterized cytotoxic capacity, whether T_regs can co-opt cytotoxic mechanisms to kill tumor cells has yet to be evaluated.

T-cell–activating bispecific antibodies (bscAb), particularly those of the bispecific T-cell engager (BiTE) subclass, represent a new therapeutic strategy that has the potential to treat even bulky, invasive disease (6, 7). BiTEs are tandem single-chain molecules that possess dual specificity for tumor-associated surface antigens and the CD3 complex on T cells, which allows them to divalently bind and afford potent, specific target cell lysis (8). Because CD3 is universally expressed among T cells, BiTEs have the theoretical capacity to redirect and activate even T_regs that are elevated and present in tumors of patients with cancer.

Among the few known tumor-specific antigens, perhaps the most widely characterized is the truncated mutant EGF receptor variant type III (EGFRvIII). EGFRvIII is a constitutively activated tyrosine kinase that is frequently expressed on the surface of glioblastoma and other common neoplasms but is completely absent from healthy tissues (9).

In this study, we show that an EGFRvIII-specific BiTE, bscEGFRvIIIxCD3 (7), successfully redirects highly purified T_regs and activates them in the presence of tumors expressing EGFRvIII. Despite their known suppressive properties, purified CD4⁺CD25⁺CD127^dim/− T cells efficiently lysed EGFRvIII-expressing glioblastoma in vitro upon redirection and activation with bscEGFRvIIIxCD3. This activity was found to be dependent on the granzyme-perforin pathway. Immunohistochemical (IHC) analysis from human primary GBMs also displayed diffuse infiltration of activated, granzyme-producing FoxP3⁺ cells, showing that T_regs with potent effector functions may already be present in tumors even under natural conditions.

Previous efforts to enhance antitumor immunity via T_reg depletion have been limited, in part due to an inability to efficiently eliminate suppressive cells that infiltrate tumor tissue (10). Alternatively, our data suggest that BiTEs can activate even suppressive T_regs to lyse tumor cells by redirecting their natural granzyme-mediated cytotoxic potential. These findings not only highlight a new mechanism by which BiTEs may circumvent certain aspects of T_reg-mediated suppression, but also have broader implications with regard to the natural functional role of activated, tumor-infiltrating T_regs that express granzyme and perforin in the tumor microenvironment.

The human glioma cell line U87MG and its subline U87MG.ΔEGFR, which expresses EGFRvIII, are described elsewhere and were kind gifts from Dr. Webster Cavenee at University of California, San Diego (11). EGFRvIII expression was verified by flow cytometry using the L8A4 antibody (12). These cell lines were regularly confirmed to be Mycoplasma-free by nucleic acid hybridization assay (GEN-PROBE MTC-NI). Characterization, production, and purification of bscEGFRvIIIxCD3 as well as control antibody constructs were conducted as previously described by us (7). Antibodies to CD4 (RPA-T4), CD25 (M-A251), CD69 (L78), CD152 (BNI3), Granzyme A (CB9), Granzyme B (GB11), and Perforin (δG9) were purchased from BD Biosciences. Anti-FoxP3 antibody (PCH101) and the FoxP3 Staining Buffer Set were purchased from eBioscience and intracellular staining was conducted according to manufacturer's instructions. Antibodies against human FoxP3 (259D, BioLegend) and granzyme B (Cat No. ab4059) were used for IHC staining.

All human samples were obtained at Duke University Medical Center (Durham, NC) from individuals who had given written, informed consent. Human PBMCs were prepared by density gradient centrifugation from buffy coats of healthy donor leukapheresis. Highly purified regulatory T cells were isolated from PBMCs by magnetic separation using the CD4⁺CD25⁺CD127^dim/− Regulatory T Cell Isolation Kit II and AutoMACS Separator (Miltenyi Biotec) according to the manufacturer's instructions. Following isolation, purity was confirmed by flow cytometric analysis to be more than 95% (CD25⁺FoxP3⁺). Where noted, purified T_regs were prepared with the T_reg Expansion Kit (Miltenyi Biotec) at a MACSiBead Particle-to-T_reg ratio of 4:1 according to manufacturer instructions, with recombinant interleukin 2 (rIL-2) at 500 U/mL, without rapamycin. Cells were verified for their suppressive ability in vitro before all cytotoxicity assays.

Activation, proliferation, cytokine secretion, and measures of specific lysis were conducted as previously described by us (7). In assays assessing activation and proliferation, freshly thawed and sorted lymphocytes were incubated in 96-well round-bottom plates with 1 × 10⁴ target tumor cells and EGFRvIII BiTE [E:T ratio, 20:1; incubation time, 48 hours at 37°C; (bscEGFRvIIIxCD3), 10 μg/mL] in a total volume of 200 μL. Supernatant was removed after incubation and analyzed by cytometric bead array analysis according to manufacturer instructions (BD Biosciences). Cells were also harvested and assessed for surface and activation markers as above. In experiments assessing proliferation 1 μCi, ³H-thymidine was added to each well of a 96-well round-bottom plate for an additional 24 hours and cells were collected by a cell harvester. Counts were conducted using a Wallac 1450 Microbeta Trilux Liquid Scintillation/Luminescence Counter (Perkin-Elmer). Similarly, cytotoxicity assays were conducted by a standard chromium release assay as previously described by us (7). In brief, target cells were labeled with ⁵¹Cr and incubated with EGFRvIII BiTE and effector cells [E:T ratio, 20:1; incubation time, 18 hours at 37°C; (bscEGFRvIIIxCD3), 10 μg/mL] in a total volume of 200 μL. Following incubation, supernatants were removed and measured by gamma counter. Where noted, for inhibition of the granzyme-perforin axis, concanamycin A (CMA, 100 nmol/L; Sigma), Granzyme B Inhibitor I (Z-AAD-CMK, 50 μmol/L; Calbiochem), or ethylene glycol tetracetic acid (EGTA, 4 mmol/L; Calbiochem) was added to each well. Inhibitor concentrations were chosen based on previously published studies to establish dependence of cell-mediated cytotoxicity on perforin-granzyme activity (4, 13, 14). Before use, each inhibitor was found to have insignificant effects on the viability of T_regs following 18 hours incubation at 37°C as assessed by LIVE/DEAD Fixable Violet Dead Cell Stain Kit (Invitrogen). Blockade of FasL- and TRAIL-mediated apoptosis was carried out with antibody clones NOK-1 (25 μg/mL; BD Biosciences) and RIK-2 (25 μg/mL; BD Biosciences), respectively.

Tissues for IHC analysis were derived from human brain tumor biopsy material. Paraffin-embedded tissues were fixed, cut into 5 μm sections, mounted on glass slides and subjected to primary and secondary staining using the MACH 2 Double Stain 1 polymer Detection Kit (Biocare Medical) according to the manufacturer's instructions. Diaminobenzidine (DAB) and AminoEthyl Carbazole (AEC) were used to detect granzyme B and FoxP3, respectively.

Groups were compared using a two-sample two-tailed t test and statistical significance was determined at a value of P < 0.05.

The EGFRvIII BiTE, bscEGFRvIIIxCD3, has previously been shown to redirect polyclonal T-cell populations against EGFRvIII-expressing gliomas (7); however, whether this BiTE also signals activation of FoxP3⁺ T_regs, which are not only elevated among tumor-infiltrating lymphocytes (15) but also known to possess potent suppressive properties in the setting of glioblastoma (16), is currently unknown. To evaluate the impact of the EGFRvIII BiTE on human T_regs, we first isolated naturally occurring human T_regs according to phenotypic CD4⁺CD25⁺CD127^dim/− expression (17) and cultured these cells in the presence of EGFRvIII BiTE and glioma cells expressing EGFRvIII (U87MG.ΔEGFR). A control BiTE directed against a nonspecific antigen was also tested to exclude the possibility that the CD3-binding portion of bscEGFRvIIIxCD3 alone was sufficient for activation. After 24 hours, T cells were harvested and examined for surface and intracellular activation markers.

In line with previous literature, isolation of CD4⁺CD25⁺CD127^dim/− cells yielded a highly pure population of T cells expressing elevated levels of transcription factor FoxP3, consistent with the phenotypic definition of the T_reg subset (18) (Fig. 1A). Following coculture with target cells expressing EGFRvIII, flow cytometric analysis of sorted T_regs showed that, in the presence of control BiTE, surface activation markers CD69 and CD25 remained unchanged. However, in wells cultured with EGFRvIII BiTE, T_regs exhibited significant upregulation of both early activation associated glycoprotein, CD69, as well as IL-2 receptor alpha chain, CD25 (Fig. 1B). These results were also consistent on repeat using sorted T_regs isolated from three separate healthy donors (P < 0.05; Fig. 1C). On the basis of this evidence of activation, we also sought to determine functional characteristics of T_regs stimulated by the EGFRvIII BiTE in the presence of tumor. Importantly, we found that following BiTE-mediated activation, T_regs remained refractory both in terms of their ability to produce inflammatory cytokines (Fig. 1D) and proliferate when compared with CD4⁺CD25⁻ helper T cells (T_H; Fig. 1E). Together, these results indicate that although the EGFRvIII BiTE clearly has the capacity to activate T_regs, the functional outcome of this activation may be similar to what would otherwise be expected through standard TCR-restricted engagement of T_regsin vitro.

Despite their lack of classical proinflammatory mechanisms, it is well-known that even naturally activated T_regs possess significant cytotoxic capacity, largely through the production of high levels of granzyme B expression (19). In fact, the perforin-granzyme axis is a major mechanism by which T_regs are known to suppress immune responses, through the direct killing of effector T cells (4, 13). Given this well-characterized cytotoxic function, we hypothesized that the EGFRvIII BiTE might actually have the ability to co-opt the perforin-granzyme axis in T_regs to kill cells other than lymphocytes, resulting in the redirected lysis of even tumor cells expressing a target antigen of interest, in this case EGFRvIII.

To test this hypothesis, we first sought to evaluate the impact of bscEGFRvIIIxCD3-mediated activation on the expression of perforin and granzymes among T_regs in the presence of target tumor cells expressing EGFRvIII. Using flow cytometric analysis of CD4⁺CD25⁺FoxP3⁺ cells, we found that T_regs cultured with control BiTE and target tumor cells expressed only low levels of perforin, granzyme A (GrA), and granzyme B (GrB). However, T_regs activated in the presence of bscEGFRvIIIxCD3 exhibited significant upregulation of these cytotoxic molecules (Fig. 2A). These results were repeated with consistent and significant upregulation of GrA, GrB, and perforin in samples isolated from three separate healthy donors (P < 0.05; Fig. 2B). Moreover, when analyzed for surface markers, back-gating analysis of GrB-positive cells (dots) displayed a distinct population with dual positivity for T_reg phenotypic markers, FoxP3 and CD25 (Fig. 2C). The level of FoxP3 expression did not appear to change significantly when gating for T_regs with either high or low levels of GrB (data not shown). Although the impact of BiTEs on T_reg expression of GrB has not been previously described, these results are consistent with previous studies which have shown substantial GrB upregulation in T_regs on stimulation with anti-CD3 antibody (5), which in theory provides the same signal delivered through the T-cell engaging arm of bscEGFRvIIIxCD3.

Because of the significant upregulation of the perforin-granzyme pathway in BiTE-activated T_regs, we next decided to test whether the EGFRvIII BiTE could redirect the cytotoxic payload expressed in T_regs to lyse target tumor cells expressing EGFRvIII. CD4⁺CD25⁺CD127^dim/− T cells were expanded between 4 and 7 days to provide sufficient numbers for in vitro cytotoxicity, a period of time known to preserve both suppressive capacity and FoxP3 expression (20). Under these conditions, sorted T_regs efficiently lysed target cells expressing EGFRvIII when redirected by the EGFRvIII-specific BiTE. Importantly, this effect was not detected in the presence of a nonspecific control bscAb or T_regs alone (Fig. 3A). Moreover, redirected T_reg-mediated lysis was shown to be actually dependent on the perforin-granzyme pathway, as cytotoxicity was not impacted by TRAIL or FasL blockade, but significantly abrogated in the presence of previously characterized inhibitors of perforin- and granzyme-mediated cytotoxicity (Fig. 3B). Adding clinical relevance to these findings, infiltrating FoxP3⁺ T_regs in human glioma samples were actually found to coexpress detectable levels of GrB by IHC analysis (Fig. 3C). Consistent with what has been reported elsewhere, GrB expression appeared to be largely cytoplasmic while subcellular localization of FoxP3 was exhibited prominently in the nucleus (21). Overall, these data suggest that activated T_regs with potent effector functions may be present within glioblastomas, and could potentially serve as effector cells upon infusion with T-cell–engaging bscAb therapies in vivo.

In light of recent findings showing that T_regs may actually be required for the priming of high-avidity CD8⁺ T-cell responses (22), innovative methods to reappropriate the T_reg compartment without ablating these cells completely could provide an attractive alternative to currently available depletive strategies. As such, we have provided evidence and a mechanism by which T_regs might be redirected to kill tumor cells through engagement with a bscAb. One previous study has explored the impact of bscAbs on T_regs and shown that bscAb-mediated activation of T_regs suppresses effector cell proliferation and abrogates antitumor efficacy (23). However, the direct effects of bscAb-redirected T_regs on target tumor cells were not discussed. Furthermore, because their work employed the use of rapamycin to expand T_regsex vivo, an additive which is known to completely suppress GrB expression in T_regs (19), it may ultimately not be feasible to directly compare the results of our studies. One limitation of our work is the use of a glioblastoma cell line that may elicit low-level allogeneic responses and thus a permissive effect on the mechanisms of T_reg activation explored here. Further studies will be necessary to explore the degree to which this is the case, likely through the use of syngeneic mouse models or autologous human tumors.

In the current study, redirected cytotoxicity was dependent on the presence of bscEGFRvIIIxCD3; although, it is reasonable to suspect that similar results might be observed upon even endogenous TCR-specific engagement of granzyme-producing T_regs present in the tumor microenvironment. Supporting this, previous studies have shown that, in general, BiTE-mediated antitumor activity is indeed contact-mediated, and has the ability to induce cytolytic synapses that are indistinguishable from those naturally occurring between effector TCR and MHC (24). Thus, while our findings here were shown using a single test molecule against EGFRvIII, we expect that BiTEs targeting other tumor antigens may likewise mediate similar effects. Overall, this work advances our understanding of potential functions for suppressive immune cells in the tumor microenvironment; however, the true functional relevance of T_reg-mediated tumor killing remains unknown. The implications of these findings will need to be explored in future studies across broader areas of autoimmunity and immunotherapy where T_reg biology is relevant.

No potential conflicts of interest were disclosed.

Conception and design: B.D. Choi, P.C. Gedeon, E.A. Reap, L. Sanchez-Perez, J.H. Sampson

Development of methodology: B.D. Choi, P.C. Gedeon, G.A. Archer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.D. Choi, P.C. Gedeon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.D. Choi, P.C. Gedeon, J.E. Herndon, G.A. Archer, E.A. Reap, L. Sanchez-Perez, D.A. Mitchell

Writing, review, and/or revision of the manuscript: B.D. Choi, P.C. Gedeon, J.E. Herndon, D.A. Mitchell, D.D. Bigner, J.H. Sampson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.D. Choi, G.A. Archer, D.D. Bigner, J.H. Sampson

Study supervision: B.D. Choi, J.H. Sampson

This work was supported by grants from the NIH 5R01-CA135272-04 (to J.H. Sampson), 5P50-NS020023- 29 (to D.D. Bigner and J.H. Sampson), 3R25-NS065731-03S1 (to J.H. Sampson), 1F30CA177152-01 (to B.D. Choi) as well as grants from the Pediatric Brain Tumor Foundation (to D.D. Bigner and J.H. Sampson), Ben and Catherine Ivy Foundation (to J.H. Sampson), Duke Cancer Institute (to J.H. Sampson and B.D. Choi), Cancer Research Institute (to B.D. Choi), and Miami Brain Tumor Coalition (to B.D. Choi).