Activating and propagating polyclonal gamma delta T cells with broad specificity for malignancies

Abstract

Purpose

To activate and propagate populations of γδT cells expressing polyclonal repertoire of γ and δ TCR chains for adoptive immunotherapy for cancer, which has yet to be achieved.

Experimental Design

Clinical-grade artificial antigen presenting cells (aAPC) derived from K562 tumor cells were used as irradiated feeders to activate and expand human γδT cells to clinical scale. These cells were tested for proliferation, TCR expression, memory phenotype, cytokine secretion, and tumor killing.

Results

γδT cell proliferation was dependent upon CD137L expression on aAPC and addition of exogenous IL-2 and IL-21. Propagated γδT cells were polyclonal as they expressed Vδ1, Vδ2, Vδ3, Vδ5, Vδ7, and Vδ8 with Vγ2, Vγ3, Vγ7, Vγ8, Vγ9, Vγ10, and Vγ11 TCR chains. Interferon-γ production by Vδ1, Vδ2, and Vδ1^negVδ2^neg subsets was inhibited by pan-TCRγδantibody when added to co-cultures of polyclonal γδT cells and tumor cell lines. Polyclonal γδT cells killed acute and chronic leukemia, colon, pancreatic, and ovarian cancer cell lines, but not healthy autologous or allogeneic normal B cells. Blocking antibodies demonstrated that polyclonal γδT cells mediated tumor cell lysis through combination of DNAM1, NKG2D, and TCRγδ. The adoptive transfer of activated and propagated γδT cells expressing polyclonal versus defined Vδ TCR chains imparted a hierarchy (polyclonal>Vδ1>Vδ1^negVδ2^neg>Vδ2) of survival of mice with ovarian cancer xenografts.

Conclusions

Polyclonal γδT cells can be activated and propagated with clinical-grade aAPC and demonstrate broad anti-tumor activities, which will facilitate the implementation of γδT cell cancer immunotherapies in humans.

INTRODUCTION

Human γδT cells exhibit an endogenous ability to specifically kill tumors and hold promise for adoptive immunotherapy. They have innate and adaptive qualities exhibiting a range of effector functions, including cytolysis upon cell contact (1, 2). Recognition and subsequent killing of tumor is achieved upon ligation of antigens to heterodimers of γ and δ T-cell receptor (TCR) chains. The human TCR variable (V) region defines 14 unique Vγ alleles, 3 unique Vδ alleles (Vδ1, Vδ2, and Vδ3), and 5 Vδ alleles that share a common nomenclature with Vα alleles (Vδ4/Vα14, Vδ5/Vα29, Vδ6/Vα23, Vδ7/Vα36, and Vδ8/Vα38-2) (3). T cells expressing TCRα/TCRβ heterodimers compose approximately 95% of peripheral blood (PB) T cells and recognize peptides in the context of major histocompatibility complex (MHC) (4). In contrast, TCRγδligands are recognized independent of MHC and these cells are infrequent (1-5% of T cells) in PB (1, 5, 6). Many conserved ligands for TCRγδare present on cancer cells, thus an approach to propagating these T cells from small starting numbers while maintaining a polyclonal repertoire of γδTCRs has appeal for human application.

Clinical trials highlight the therapeutic potential of γδT cells, but numeric expansion is needed for adoptive immunotherapy because they circulate at low frequencies in PB. Methods to propagate αβ T cells, e.g., using interleukin-2 (IL-2) and/or antibody cross-linking CD3, cannot sustain proliferation of γδT cells (7, 8). Aminobisphosphonates, e.g., Zoledronic acid (Zol), have been used to initiate a proliferative signal in γδT cells (5, 9), but only one lineage of γδT cells, expressing Vγ9Vδ2 TCR, can be reliably expanded by Zol. The adoptive transfer of Vγ9Vδ2 T cells has yielded clinical responses for investigational treatment of solid and hematological cancers (10-14). Furthermore, long-term remission of leukemia among recipients of haploidentical αβ T cell-depleted hematopoietic stem-cell transplantation (HSCT) correlated with increased engraftment frequency of donor-derived Vδ1 cells (8, 15-17). However, direct administration of Vδ1 cells or other non-Vγ9Vδ2 cell lineages has yet to be performed. In addition, no reports to date have described the therapeutic impact of Vδ1^negVδ2^neg cells in cancer immunotherapy and this subset has not been directly compared to T cells expressing Vδ1 and Vδ2 TCRs. Thus, there are significant gaps in the knowledge and human application of non-Vγ9Vδ2 lineages.

Given that γδT cells have endogenous anti-cancer activity, such as against K562 cells (8, 18), we hypothesized that malignant cells would serve as a cellular substrate to propagate polyclonal γδT cells. K562 cells have been genetically modified to function as artificial antigen presenting cells (aAPC) to ex vivo activate and numerically expand αβ T cells and NK cells (19-23). We determined that interleukin-2 (IL-2), IL-21, and γ-irradiated K562-derived aAPC (designated clone #4, genetically modified to co-express CD19, CD64, CD86, CD137L, and a membrane-bound mutein of IL-15 (mIL15); used in selected clinical trials at MD Anderson Cancer Center) can sustain the proliferation of γδT cells with polyclonal TCR repertoire. Polyclonal γδT cells exhibited broad tumor reactivity and displayed a multivalent response to tumors as evidenced by the ability of separated Vδ sub-populations to kill and secrete cytokine against the same tumor target. Further, killing by polyclonal populations was multifactorial being mediated through DNAM1, NKG2D, and TCRγδ. Tumor xenografts were eliminated by both polyclonal and distinct γδT-cell subsets, and mice treated with polyclonal γδT cells had superior survival. Given the availability of aAPC as a clinical reagent, trials can for the first time, evaluate polyclonal populations of γδT cells as a cancer immunotherapy.

MATERIALS AND METHODS

Cell lines

HCT-116, Kasumi-3, and K562 were acquired from American Type Culture Collection (ATCC; Manassas, VA). Jurkat was purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ; Germany). cALL-2 and RCH-ACV were gifts from Dr. Jeff Tyner (Oregon Health & Science University). BxPC-3, MiaPaCa-2, and Su8686 (pancreatic cancer) were donated by Dr. Vijaya Ramachandran (MD Anderson Cancer Center). A2780, CAOV3, EFO21, EFO27, Hey, IGROV1, OAW42, OC314, OVCAR3, and UPN251 (ovarian cancer) were provided by Dr. Robert C. Bast, Jr. (MD Anderson Cancer Center). Identities of all cell lines were confirmed by STR DNA Fingerprinting at MD Anderson Cancer Center’s “Characterized Cell Line Core” and cells were used within 6 months of authentication.

Propagation of γδT cells

Peripheral blood mononuclear cells (PBMC) and umbilical cord blood (UCB) were isolated from healthy volunteers by Ficoll-Hypaque (GE Healthcare) after informed consent (24). Thawed PBMC (10⁸) were initially treated with CD56 microbeads (cat# 130-050-401, Miltenyi Biotec, Auburn, CA) and separated on LS columns (cat# 130-042-401, Miltenyi Biotec) to deplete NK cells from cultures because they proliferate on aAPC (23) and would contaminate the purity of the γδT-cell product. Unlabeled cells from CD56 depletion sorting were then labeled with TCRγ/δ+ T-cell isolation kit (cat# 130-092-892, Miltenyi Biotec) and placed on LS columns to separate γδT cells in the unlabeled fraction from other cells attached to magnet. γδT cells were co-cultured at a ratio of one T cell to two γ-irradiated (100 Gy) aAPC (clone #4) in presence of exogenous IL-2 (Aldesleukin; Novartis, Switzerland; 50 IU/mL), and IL-21 (cat# AF20021; Peprotech, Rocky Hill, NJ; 30 ng/mL) in complete media (CM; RPMI, 10% FBS, 1% Glutamax). Cells were serially re-stimulated with addition of γ-irradiated aAPC every 7 days for 2 to 5 weeks in presence of soluble cytokines, which were added three times per week beginning the day of aAPC addition. K562 were genetically modified to function as aAPC (clone #4) as previously described (25, 26). Validation of co-expression of CD19, CD64, CD86, CD137L, and eGFP (IL-15 peptide fused in frame to IgG4 Fc stalk and co-expressed with eGFP) on aAPC clone #4 was performed before addition to T-cell cultures (25). Fluorescence activated cell sorting (FACS) was used to isolate Vδ1 (TCRδ1⁺TCRδ2^neg), Vδ2 (TCRδ1^negTCRδ2⁺), and Vδ1^negVδ2^neg (TCRδ1^negTCRδ2^neg) populations, which were stimulated twice as above with aAPC clone #4, phenotyped, and used for functional assays. γδT cells from UCB were isolated by FACS from thawed mononuclear cells using anti-TCRγδand anti-CD3 monoclonal antibodies (mAbs) and were stimulated for five weeks on aAPC/cytokines as per PBMC.

Abundance and identity of mRNA molecules by DTEA

At designated times after co-culture on aAPC, T cells were lysed at a ratio of 160 μL RLT Buffer (Qiagen) per 10⁶ cells and frozen at −80 °C. RNA lysates were thawed and immediately analyzed using nCounter Analysis System (NanoString Technologies, Seattle, WA) with “designer TCR expression array” (DTEA), as previously described (27, 28). DTEA data was normalized to both spike positive control RNA and housekeeping genes (ACTB, G6PD, OAZ1, POLR1B, POLR2A, RPL27, Rps13, and TBP). Spiked positive control normalization factor was calculated from the average of sums for all samples divided by the sum of counts for an individual sample. Spiked positive control normalization factor was calculated from the average of geometric means for all samples divided by the geometric mean for an individual sample. Normalized counts were reported.

Flow cytometry

Cells were phenotyped with antibodies detailed in Supplemental Table 1. Gating strategy is displayed in Supplemental Figure 1. Samples were acquired on FACS Calibur (BD Biosciences, San Jose, CA) and analyzed with FlowJo software (version 7.6.3).

Cytokine production and cytolysis assays

Expression of cytokines was assessed by intracellular staining and secretion of cytokines into tissue culture supernatants was evaluated by Luminex multiplex analysis. In vitro specific lysis was assessed using a standard 4-hour CRA, as previously described (25). Additional information can be found in the supplemental materials and methods.

Mouse experiments

In vivo anti-tumor efficacy was assessed in NSG mice (NOD.Cg-Prkdc^scidIl2rγ^tm1Wjl/SzJ; Jackson Laboratories). CAOV3 ovarian cancer cell line was transduced with recombinant lentivirus (Supplemental Figure 2) encoding mKate red fluorescence protein (29) to identify transduced cells and enhanced firefly luciferase (effLuc) for non-invasive bioluminescence imaging (30). CAOV3-effLuc-mKate (clone 1C2; 3×10⁶ cells/mouse) tumors were established by intraperitoneal (i.p.) injection and mice were randomly distributed into treatment groups. Eight days later (designated Day 0), a dose escalation regimen was initiated with γδT cells administered i.p. and PBS administered i.p. as a negative control. T-cell doses were 3×10⁶, 6×10⁶, 10⁷, and 1.5×10⁷ on days 0, 7, 14, and 21, respectively. Non-invasive BLI was performed during the course of the experiments to serially measure tumor burden of CAOV3-effLuc-mKate following subcutaneous administration of d-Luciferin (cat#122796, Caliper, Hopkinton, MA) as detected with IVIS-100 Imager (Caliper). BLI was analyzed using Living Image software (version 2.50, Xenogen, Caliper).

RESULTS

Ex vivo numeric expansion of γδT cells on aAPC depends on co-stimulation and cytokines

The adoptive transfer of γδT cells requires ex vivo propagation as starting numbers from PBMC are limiting (gating on lymphocyte pool: 3.2% ± 1.2%; mean ± standard deviation (SD); n=4; Figure 1A). γδT cells from PBMC were isolated by “negative” paramagnetic bead selection and co-cultured for 22 days with weekly addition of γ-irradiated K562-derived aAPC (clone #4) in the presence of soluble recombinant IL-2 and IL-21 in alignment with protocols at MD Anderson for propagation of clinical-grade αβ T cells. This resulted in the outgrowth of a population of T cells homogeneously co-expressing CD3 and TCRγδ(97.9% ± 0.6%). NK cells (CD3^negCD56⁺) and αβ T cells (TCRαβ⁺) were absent from these cultures supporting the purity of the γδT-cell product. Populations of TCRδ1⁺TCRδ2^neg, TCRδ1^negTCRδ2⁺, and TCRδ1^negTCRδ2^neg were detected indicating that aAPC, IL-2, and IL-21 supported polyclonal γδT cell proliferation (Figure 1A far right). Cells were activated as marked by expression of CD38 (93.5% ± 3.5%) and CD95 (99.7% ± 0.1%) (Supplemental Figure 3). This approach to propagation yielded >10⁹ γδT cells from <10⁶ total initiating cells (Figure 1B), which represented a 4.9×10³ ± 1.7×10³ (mean ± SD; n=4) fold increase. Thus, aAPC with recombinant human cytokines supported the robust numeric expansion of polyclonal γδT cells from small starting numbers of γδT cells derived from PBMC.

Figure 1. Sustained proliferation of polyclonal PBMC-derived γδT cells on γ-irradiated aAPC in presence of soluble IL-2 and IL-21.

(A) Frequency of γδT cells before (Day 0) and after (Day 22) co-culture on γ-irradiated aAPC, IL-2, and IL-21 where expression of CD3, CD56, TCRαβ, TCRγδ, TCRδ1, and TCRδ2 is shown at Day 22 of co-culture. One of 7 representative donors is shown. Quadrant frequencies (percentage) within flow plots are displayed in upper right corners. (B) Inferred cell counts of polyclonal γδT cells are displayed calculated based on weekly yields and relative fold changes, where three arrows represent addition of aAPC. Black line is mean ± SD (n=4) pooled from 2 independent experiments and each gray line is an individual donor. (C) Fold increase over 9 days of γδT cells co-cultured with IL-2 and IL-21 along with aAPC expressing membrane-bound IL-15 (mIL15), CD86, and/or CD137L. Data are mean ± SD (n=3) pooled from 2 independent experiments and each shape represents an individual donor. Two-way ANOVA with Bonferroni’s post-tests was used for statistical analysis. *p<0.05 and **p<0.01 (D) Fold increase over 9 days of γδT cells co-cultured with aAPC (clone #4) in the presence of either soluble recombinant IL-2 and/or IL-21. Data are mean ± SD (n=3) pooled from 2 independent experiments where each shape represents an individual donor. Two-way ANOVA with Bonferroni’s post-tests was used for statistical analysis. *p<0.05

The addition of exogenous cytokines and presence of mIL15, CD86, and CD137L on clinical-grade aAPC were assessed for their ability to support the outgrowth of γδT cells. Parental K562 cells were stably transfected with Sleeping Beauty (SB) transposons to introduce individual stimulatory molecules, cloned to achieve homogeneous expression (Supplemental Figure 4), and then used to assess their impact on γδT-cell proliferation. Co-cultures with exogenous IL-2 and IL-21 were initiated with paramagnetic bead-purified γδT cells and five sets of γ-irradiated K562: (i) parental, (ii) mIL15⁺, (iii) mIL15⁺CD86⁺, (iv) mIL15⁺CD137L⁺, and (v) mIL15⁺CD86⁺CD137L⁺ (clone #4). γδT cells cultured in parallel without APC demonstrated that soluble IL-2 and IL-21 sustained only limited numeric expansion of γδT cells (Figure 1C). Propagation improved upon addition of parental K562 cells, indicating that endogenous molecules on these cells can activate γδT cells for proliferation. The expression of mIL15 with or without CD86 did not further improve the ability of γδT cells to propagate compared with parental K562. In contrast, improved rates of propagation of γδ T cells were observed upon co-culture with mIL15⁺CD137L⁺ and mIL15⁺CD86⁺CD137L⁺ aAPC. Thus, it appears that CD137L on aAPC clone#4 provides a dominant co-stimulatory proliferative signal for γδT cells. In the absence of IL-2 and IL-21 the proliferation of γδT cells ceased on aAPC clone#4, and together these cytokines exhibited an additive benefit to the rate of γδT-cell propagation (Figure 1D). This validated our approach to combining aAPC clone #4 with cytokines to sustain the proliferation of polyclonal γδT cells ex vivo, and demonstrated that CD137L on aAPC, IL-2, and IL-21 were driving factors for proliferation of polyclonal γδT cells to clinical scale.

Ex vivo numeric expansion of neonatal γδT cells on aAPC in presence of IL-2 and IL-21

Allogeneic UCB is an important source of γδT cells for adoptive transfer, because it contains younger cells and a more diverse TCRγδrepertoire relative to PBMC, which could increase the number of ligands targeted by the engrafted cells and result in long-term engraftment in the recipient (31). However, the limited number of mononuclear cells within a banked UCB unit curtails the number of neonatal γδT cells directly available for adoptive transfer. Thus, we evaluated whether aAPC could sustain proliferation from small starting numbers of neonatal γδT cells. Fluorescence activated cell sorting (FACS) was used to isolate 10⁴ UCB-derived γδT cells (~0.01% of a typical UCB unit) which were co-cultured on aAPC clone #4 with IL-2 and IL-21. After 35 days, there was a 10⁷-fold increase in cell number, as an average of 10¹¹ UCB-derived γδT cells (Range: 6×10⁹ – 3×10¹¹; n=5) were propagated from the 10⁴ initiating γδT cells (Supplemental Figure 5A). Two additional stimulations were performed for γδT cells derived from UCB compared to PBMC highlighting their potential for proliferating to clinically-appealing numbers. The propagated γδT-cell populations exhibited uniform co-expression of CD3 and TCRγδand lacked TCRαβ⁺ T cells or presence of CD3^negCD56⁺ NK cells (Supplemental Figure 5B-D). Collectively, these data demonstrate that aAPC clone #4 with IL-2 and IL-21 could sustain the ex vivo proliferation of γδT cells from a small starting population of neonatal UCB.

Ex vivo activated and propagated γδT cells express polyclonal and defined TCRγδrepertoire

Upon establishing that γδT cells could numerically expand on aAPC and selected cytokines, we sought to determine the TCR repertoire of the propagated cells. Prior to numeric expansion, resting γδT cell repertoire followed TCRδ2>TCRδ1^negTCRδ2^neg>TCRδ1 by flow cytometry (Supplemental Figure 6). However, the γδT-cell repertoire followed TCRδ1>TCRδ1^negTCRδ2^neg>TCRδ2 following expansion, suggesting that there was a proliferative advantage for Vδ1 cells within polyclonal γδT-cell cultures. To look more in-depth at TCRγδdiversity in aAPC-expanded γδT cells, we adapted a non-enzymatic digital multiplex assay used to quantify the TCR diversity in γδT cells expressing a CD19-specific chimeric antigen receptor (CAR) (27) termed “direct TCR expression array” (DTEA). Following expansion (Day 22), four of eight Vδ alleles (Vδ1, Vδ2, Vδ3, and Vδ8) were detected in PBMC-derived γδT cells (Figure 2A) and were co-expressed with Vγ2, Vγ7, Vγ8 (two alleles), Vγ9, Vγ10, and Vγ11 (Figure 2B). Similarly, a polyclonal assembly of Vδ and Vγ chains was observed in γδT cells from UCB following expansion (Day 34-35), albeit with reduced abundance of Vδ2 cells, more Vγ2, and presence of Vγ3, Vδ5, and Vδ7 cells not seen from PBMC (Figure 2C-D). Similar patterns of Vδ and Vγ mRNA usage were detected in PBMC and UCB before and after expansion (Supplemental Figure 7) although overall mRNA counts were fewer in the resting cells (Day 0) relative to the activated γδT cells. Thus, aAPC-expanded γδT cells maintain a polyclonal TCR repertoire from both PBMC and UCB.

Figure 2. Abundance of Vδ and Vγ mRNA species in γδT cells propagated and activated ex vivo.

Quantification of mRNA species coding for (A) Vδ and (B) Vγ alleles in PBMC-derived γδT cells by DTEA at day 22 of co-culture on aAPC/IL-2/IL-21. Quantification of mRNA species coding for (C) Vδ and (D) Vγ alleles in UCB-derived γδT cells by DTEA at day 34-35 of co-culture on aAPC/IL-2/IL-21. Box-and-whiskers plots display 25% and 75% percentiles where lines represent maximum, mean, and minimum from top to bottom (n=4). Solid lines at bottom of graphs represent limit-of-detection (LOD) calculated from mean ± 2xSD of DTEA negative controls. Student’s paired one-tailed t-tests were performed for each allele relative to the sample LOD. *p<0.05 and **p<0.01

We sought to validate these mRNA data by sorting polyclonal populations with TCRδ-specific antibodies and repeating DTEA on isolated cultures. There are only two TCRδ-specific mAbs commercially available and they identified 3 discrete Vδ populations (Vδ1: TCRδ1⁺TCRδ2^neg, Vδ2: TCRδ1^negTCRδ2⁺, and Vδ1^negVδ^neg: TCRδ1^negTCRδ2^neg) within aAPC-expanded γδT cells from PBMC (Figure 1A) and UCB (Supplemental Figure 8) with abundance following Vδ1>Vδ1^negVδ^neg>Vδ2. FACS isolated subsets from PBMC-derived γδT-cell pools were propagated with clone #4 as discrete populations and maintained their identity as assessed by expression of TCRδ isotypes (Figure 3A). Each of the separated subsets could be identified by a pan-specific TCRγδantibody confirming that these cells were indeed γδT cells (Figure 3B). Furthermore, each population could be differentiated based on pan-TCRγδantibody mean fluorescence intensity (MFI) where Vδ2, Vδ1^negVδ^neg, and Vδ1 T cells corresponded to the TCRγδ^low (43 ± 9; mean ± SD; n=4), TCRγδ^intermediate (168 ± 40), and TCRγδ^hi (236 ± 56) groupings, respectively. No differences in proliferation kinetics on aAPC were observed between isolated Vδ-sorted subsets (Figure 3C) indicating that the observed inversion of Vδ1 and Vδ2 frequencies in polyclonal cultures before versus after expansion was not due to a proliferative defect in one of the subsets. DTEA demonstrated that isolated and propagated Vδ1, Vδ2, and Vδ1^negVδ^neg sub-populations were homogeneous populations as they predominantly expressed Vδ1*01, Vδ2*02, and Vδ3*01 mRNA species at 261 ± 35, 3910 ± 611, and 5559 ± 1119 absolute counts, respectively (Figure 3D). Therefore, there were fewer Vδ1*01 mRNA species expressed by Vδ1 cells relative to the Vδ2*02 expressed by Vδ2 cells and Vδ3*01 expressed by Vδ1^negVδ2^neg cells. Moreover, these data indicated that the relatively low counts observed for Vδ1*01 in polyclonal populations with a preponderance of TCRδ1⁺ cells was not a defect in DTEA detection but rather a product of fewer total mRNA transcripts relative to other Vδ species. Given the wide range of mRNA transcript quantities for each allele, DTEA was not useful for calculation of relative frequencies of Vδ subsets in polyclonal populations but rather was indicative of presence or absence of a particular γδT cell subset. Expression of other Vδ2 alleles (Vδ1*01_07 and Vδ1*01_75) was absent from polyclonal γδT cells (Figure 2A) and each of the sorted subsets (data not shown). Small amounts of Vδ4, Vδ5, Vδ6, and Vδ7 mRNA species were detected in the three subsets of T cells sorted for Vδ expression (Supplemental Figure 9). Vδ8 mRNA was exclusively present in sorted Vδ1^negVδ^neg cells and these T cells are likely the main contributors of Vδ8 in bulk γδT cells. The same Vγ mRNA present in polyclonal cultures was detected in Vδ-sorted cultures (Supplemental Figure 10). Furthermore, Vδ1 and Vδ1^negVδ^neg were not different (p=0.419; Two-way ANOVA) but Vδ2 was different to both Vδ1 (p<0.0001) and Vδ1^negVδ^neg (p<0.0001) in Vγ usage. Collectively, these results confirmed DTEA from unsorted cultures and strongly supported the polyclonal TCRγδexpression on γδT cells activated to proliferate by aAPC and cytokines.

Figure 3. Sustained proliferation of PBMC-derived Vδ T-cell subsets expanded on γ-irradiated aAPC/IL-2/IL-21.

After two 7-day stimulations with aAPC (clone #4) and IL-2/IL-21 the bulk population of γδT cells were separated into Vδ1, Vδ2, and Vδ1^negVδ2^neg subsets by FACS based on staining of T cells defined as TCRδ1⁺TCRδ2^neg, TCRδ1^negTCRδ2⁺, and TCRδ1^negTCRδ2^neg, respectively. (A) Expression of TCRδ1 and TCRδ2 chains on Vδ1, Vδ2, and Vδ1^negVδ2^neg subsets of γδT cells (from left to right) after 15 days of numeric expansion on aAPC and cytokines as isolated groups. One of 4 representative donors is shown pooled from 2 independent experiments. Quadrant frequencies (percentage) within flow plots are displayed in upper right corners. Frequency of TCRδ1⁺TCRδ2^neg (open bars), TCRδ1^negTCRδ2⁺ (black bars), and TCRδ1^negTCRδ2^neg (gray bars) cell surface protein expression in subsets of γδT cells after 15 days numeric expansion on aAPC and cytokines as isolated groups. Data are mean ± SD (n=4) pooled from 2 independent experiments. (B) Flow cytometry plots of CD3 and TCRγδexpression in Vδ1, Vδ2, and Vδ1^negVδ^neg subsets (from left to right). Mean fluorescence intensity (MFI) of TCRγδstaining in Vδ1, Vδ2, and Vδ1^negVδ^neg T-cell subsets where each shape represents a different donor and data are mean ± SD (n=4) pooled from 2 independent experiments. (C) Proliferation of each isolated Vδ subset stimulated twice with aAPC clone #4 (arrows) in presence of cytokines and total cell counts are displayed. Data are mean ± SD (n=4) pooled from 2 independent experiments. (D) DTEA was used to identify and measure abundance of mRNA species coding for Vδ1*01, Vδ2*02, and Vδ3*01 (from left to right) in γδT-cell sub-populations after 15 days of proliferation on aAPC and cytokines as separated subsets. Box-and-whiskers plots display 25% and 75% percentiles where lines represent maximum, mean, and minimum from top to bottom (n=4). Student’s paired, two-tailed t-tests were undertaken for statistical analyses between groups. **p<0.01 and ***p<0.001

Interferon-γ produced in response to tumors is dependent on TCRγδ

A multiplex analysis of cytokines and chemokines was performed to determine whether aAPC-propagated γδT cells might foster a pro-inflammatory response in a tumor micro-environment (Figure 4A). The T_H1-associated cytokines interferon-γ (IFNγ) and tumor necrosis factor-α (TNFα) were secreted in abundance by γδT cells upon exposure to leukocyte activated cocktail (LAC; PMA and Ionomycin for non-specific mitogenic stimulation), in addition to small amounts of IL-2 and IL-12 p70. In contrast, no significant production of the T_H2-associated cytokines IL-4, IL-5, and IL-13 was observed from LAC-treated γδT cells, but there was a small increase in IL-10 production over baseline. Similarly, Th17-associated cytokines IL-1RA, IL-6, and IL-17 were secreted at low levels by LAC-treated γδT cells. The chemokines CCL3, CCL4, and CCL5 were detected in abundance. Minor contributions of non-γδT cells in the culture that could have been activated by LAC to secrete cytokines could not be ruled out, but given that the cells tested were 97.9% ± 0.6% CD3⁺TCRγδ⁺ these data indicate that it was activation of γδT cells that led to a largely pro-inflammatory response. IFNγ was the most responsive of all the assessed cytokines and was chosen to measure responses of Vδ subsets to tumor cells (Figure 4B). Co-culture of polyclonal aAPC-propagated/activated γδT cells with cancer cells resulted in a hierarchy of IFNγ production following Vδ2>Vδ1>Vδ1^negVδ^neg as shown by MFI of 855 ± 475, 242 ± 178, and 194 ± 182 (mean ± SD; n=4), respectively. IFNγ production by Vδ1, Vδ2, and Vδ1^negVδ2^neg subsets was inhibited by pan-TCRγδantibody when added to of γδT cell/tumor co-cultures indicating that response to the tumor in each subset was dependent upon activation through TCRγδ(Figure 4C). This observation supported the premise that a single cancer cell could be targeted by discrete γδTCRs. Thus, a multivalent pro-inflammatory response to the tumor cell was achieved by polyclonal γδT cells.

Figure 4. Dependence on TCRγδ for IFNγ secretion in response to tumor cells.

At Day 22 of co-culture on γ-irradiated aAPC (clone #4) with IL-2 and IL-21, T cells were incubated with CM (mock) or leukocyte activation cocktail (LAC; PMA/Ionomycin) for 6 hours at 37 °C. Tissue culture supernatants were interrogated using 27-Plex Luminex array to detect presence of (A) T_H1, T_H2, and T_H17 cytokines and selected chemokines (from left to right). Data are mean ± SD pooled from 4 donors in 2 independent experiments where each donor had triplicate experimental wells pooled prior to multiplex analysis. Student’s one-tailed t-test performed for statistical analysis between mock and LAC groups. *p<0.05, **p<0.01, and ***p<0.001 (B) Polyclonal γδT cells were incubated for 1 hour prior to and during 6 hour tumor cell co-culture with normal mouse serum or neutralizing TCRγδantibody (clone IM). Cells were stained for TCRδ1, TCRδ2, CD3, and IFNγ to gate T-cell subsets and assess IFNγ production. Comparisons of histograms detailing Vδ1, Vδ2, and Vδ1^negVδ^neg gates (from left to right) co-cultured with CAOV3 ovarian cancer cells and treated with serum (open) or TCRγδ(shaded). Numbers next to histograms are MFI. Flow plots are representative of 1 of 3 PB donors co-cultured with CAOV3 cells in 2 independent experiments. (C) Percent inhibition of IFNγ secretion in response to CAOV3 cells was calculated for each Vδ T-cell subset based on the following equation: Inhibition (%) = 100 − 100 × [(MFI_TUMOR + T_CELL − MFI_{T CELL ONLY})_TCRγδ/(MFI_{TUMOR + T CELL} − MFI_{T CELL ONLY})_Serum]. Data are mean ± SD (n=3) pooled from 2 independent experiments.

Polyclonal γδT cells lyse a broad range of tumor cells through combination of DNAM1, NKG2D, and TCRγδ

After establishing that propagated γδT cells could be activated to produce pro-inflammatory cytokines, we examined their ability to specifically lyse a panel of tumor cell lines. Polyclonal γδT cells demonstrated a range of cytolysis against solid and hematological cancer-cell lines without a clear preference towards a particular tumor histology or grade (Figure 5 and Supplemental Figure 11). We previously established that B-cell acute lymphoblastic leukemia (ALL) cell line NALM-6 was largely resistant to lysis by γδT cells, which required a CD19-specific CAR to acquire significant killing capability (27). In this study it was also observed that autologous and allogeneic normal B cells were spared from cytolysis (Figure 5A), and that B-ALL cell line cALL-2 and murine T cell lymphoma cell line EL4 were lysed poorly by polyclonal γδT cells, which indicated that some cells were resistant and/or not recognized by polyclonal γδT cells. In contrast, T-ALL cell line Jurkat and B-ALL cell lines RCH-ACV were both killed efficiently by polyclonal γδT cells (Figure 5B), indicating that γδT cells could be used to target some B-cell and T-cell malignancies. Kasumi-3 is a CD33⁺CD34⁺ undifferentiated leukemia cell line that was lysed at intermediate levels by γδT cells. Chronic myelogenous leukemia (CML) cell line K562 and K562-derived clone#4 aAPC were killed by polyclonal γδT cells, which corroborated the notion that these cells could serve as a proliferative substrate. Pancreatic cancer cell lines BxPc-3, MiaPaCa-2, and Su8686, were lysed by γδT cells, as was the colon carcinoma cell line HCT-116 (Figure 5C). Ovarian cell lines were killed by polyclonal γδT cells in the following order of decreasing sensitivity: CAOV3 > EFO21 > UPN251 > IGROV1 > OC314 > Hey > A2780 > OVCAR3 > OAW42 > EFO27. Each of the separated Vδ subsets lysed hematological (Jurkat and K562) and solid (OC314 and CAOV3) tumor cell lines, which showed that polyclonal γδT cells could direct a multivalent response against common targets (Supplemental Figure 12). The strength of cytolysis followed the hierarchy of TCR usage (Vδ2>Vδ1^negVδ2^neg>Vδ1) that was consistent with the premise that a propensity to be triggered for effector function would increase with T-cell differentiation (Supplemental Figure 13). Lysis by polyclonal populations was apparently not due to one specific Vδ subtype but rather from contributions of multiple γδT-cell subsets, because it was observed that (1) a number of tumor cell lines were equivalently killed by polyclonal γδT cells containing different frequencies of Vδ1, Vδ2, and Vδ1^negVδ2^neg cells and (2) a polyclonal population was not identified with dominant cytolysis. We also sought to determine which surface molecules were responsible for cytolysis by blocking immunoreceptors with antibodies (Figure 5D). Our experimental approach also took into account that γδT cells co-express DNAM1 (97.7% ± 0.9%; mean ± SD; n=4) and NKG2D (40.1% ± 16.5%) which can activate both T cells and NK cells for killing (32, 33). Addition of individual antibodies did not reduce lysis, except for TCRγδin 2 of 3 cell lines tested. In contrast, a pool of antibodies binding NKG2D, DNAM1, TCRγδresulted in significant inhibition, in a dose-dependent manner, of γδT-cell mediated cytolysis against all 3 targets. Collectively, these data established that ex vivo-propagated γδT cells have broad anti-tumor capabilities likely mediated by activation though DNAM1, NKG2D, and TCRγδ.

Figure 5. Specific lysis of tumor-cell panel by polyclonal γδT cells.

(A-C) Standard 4-hour CRA was performed with increasing effector (polyclonal γδT cells; each shape represents a different donor) to target (E:T) ratios against (A) healthy B cells from an allogeneic donor (one of four representative donors), (B) hematological tumor cell lines derived from B-ALL: RCH-ACV, T-ALL: Jurkat, and CML: K562, (C) solid tumor cell lines derived from pancreatic cancer: BxPc-3, colon cancer: HCT-116, and ovarian cancer: OC314 and CAOV3. Data are mean ± SD (n=3 wells per assay) from 2 independent experiments. (D) Neutralizing antibodies to NKG2D (squares), DNAM1 (triangles), TCRγδ(inverted triangles), or a pool (diamonds) of all three antibodies were used to block killing of Jurkat (left), IGROV1 (middle), or OC314 (right) tumor targets antibodies at 0.3, 1, and 3 μg/mL and an E:T ratio of 12:1 in standard 4-hour CRA. Normal mouse serum (circles) served as control for addition of antibody and wells without antibody were used for normalization purposes. Specific lysis was normalized to wells without antibody to yield relative cytolysis as defined by: Relative cytolysis (%) = (Specific Lysis)_{With Antibody}/(Specific Lysis)_{Without Antibody} × 100. Data are mean ± SD (n=4 donors) from triplicates pooled and normalized from 2 independent experiments. Repeated-measures Two-way ANOVA was used for statistical analysis between antibody treatments. **p<0.01 and ***p<0.001

Established ovarian cancer xenografts are eliminated by adoptive transfer of γδT cells

To test whether polyclonal γδT cells were effective in targeting and killing tumors in vivo, we created a xenograft model for ovarian cancer in immunocompromised mice. NSG mice were injected intraperitoneally with CAOV3-effLuc-mKate ovarian cancer cells and then randomized into five treatment groups. Following eight days of tumor engraftment, either PBS (vehicle/mock), Vδ1, Vδ2, Vδ1^negVδ2^neg, or polyclonal γδT cells were administered in escalating doses (Figure 6). Tumor burden and biodistribution were serially measured by non-invasive bioluminescence imaging. Established tumors continued to grow in vehicle-treated mice, but tumor bioburden was significantly reduced (p≤0.001) in mice receiving γδT-cell treatments at day 72, relative to their initial tumor burden (Figure 6A-B). Adoptive transfer of polyclonal γδT cells, Vδ1, and Vδ1^negVδ^neg T cells significantly (p≤0.01), and Vδ2 almost significantly (p=0.055), increased long-term survival compared to mock-treated mice. This corresponded to overall survival following polyclonal>Vδ1>Vδ1^negVδ^neg>Vδ2 (Figure 6C). This is the first time that three Vδ subsets have been compared for their ability to target tumor in vivo and is the first display of in vivo anti-tumor activity by Vδ1^negVδ^neg cells. In sum, activated and propagated γδT cells were effective in treating cancer in vivo and thus represent an attractive approach to adoptive immunotherapy.

Figure 6. In vivo clearance of ovarian cancer upon adoptive transfer of polyclonal γδT cells and γδT-cell subsets propagated/activated on aAPC with IL-2 and IL-21.

CAOV3-effLuc-mKate tumor cells were injected into NSG mice at Day -8 and engrafted until Day 0 when treatment was started with either PBS (vehicle/mock) or γδT cells. Four T-cell doses were administered in weekly escalating doses. (A) BLI images at Day 0 (top panels) or Day 72 (bottom panels) in PBS, Vδ1, Vδ2, Vδ1^negVδ^neg, and polyclonal γδT-cell treatment groups. Images are representative of 6-14 mice from 2 independent experiments. (B) BLI measurements of mice at Day 0 (white) and Day 72 (gray) pooled from 2 independent experiments. Box-and-whiskers plots display 25% and 75% percentiles where lines represent maximum, mean, and minimum from top to bottom (n = 6-14). Student’s paired, two-tailed t-tests were used for statistical analysis between time points. (C) Overall survival of mice treated with PBS (dashed), polyclonal (black), Vδ1 (red), Vδ2 (blue), or Vδ1^negVδ2^neg (green) γδT cells. Log-rank (Mantel-Cox) test was used to calculate p values. *p<0.05, **p<0.01, and ***p≤0.001

DISCUSSION

This study establishes our aAPC clone #4 as a cellular platform for the sustained proliferation of multiple γδT cell populations that demonstrate extensive reactivity against hematologic and solid malignancies. T cells expressing defined Vδ TCRs have been associated with clinical responses against cancer. For example, the Vδ1 subset correlated with complete responses observed in patients with ALL and acute myelogenous leukemia (AML) after αβ T cell-depleted haploidentical HSCT (15-17). Vδ1 cells were also shown to kill glioblastoma independent of cytomegalovirus (CMV) status (34). However, Vδ1 cells have not been directly administered. Our data establish that such cells could mediate anti-tumor immunity and supports the adoptive transfer Vδ1 T cells for cancer therapy. In contrast to Vδ1 and Vδ1^negVδ^neg cells, T cells expressing Vδ2 TCR have been directly infused and elicited responses against solid and hematological tumors (9, 35). Little is known about Vδ1^negVδ^neg T cells, but these lymphocytes have displayed recognition of the non-classical MHC molecule CD1d with corresponding NKT-like functions and have also been correlated with immunity to human immunodeficiency virus (HIV) and CMV (36-39). Our results are the first to directly show that Vδ1^negVδ^neg cells exhibit anti-tumor activities, and given their propensity to engage both viruses and cancer the add-back of this subset could especially benefit immunocompromised cancer patients. Because aAPC with IL-2 and IL-21 can propagate polyclonal γδT cells, mAbs can now be raised against Vδ3, Vδ5, Vδ7, and Vδ8 isotypes to help elucidate their potential roles in clearance of pathogens and cancer. In aggregate, our data support the adoptive transfer of γδT cells that maintain expression of multiple Vδ TCR types as investigational treatment for cancer.

The molecules on aAPC that activate γδT cells for numeric expansion are not well known. K562-derived aAPC express endogenous MHC Class-I chain-related protein A and B (MICA/B) which are ligands for both Vδ1 and NKG2D (6, 40). Indeed, NKG2D was observed on polyclonal γδT cells that also predominantly expressed Vδ1 TCR (Figure 1A). Polyclonal γδT cells also demonstrate expression for activating receptors typically found on NK cells (NKp30, NKp44, and NKp46; collectively expressed at 26% ± 7%), and future studies will examine their contribution to γδT-cell effector function. Some malignant cells were recognized poorly by γδT cells, e.g., EL4, EFO27, OAW42, cALL-2, and NALM-6, which provides an opportunity to further interrogate the mechanism by which γδT cells recognize and kill tumor cells. Given that inhibition of cytolysis was maximized by neutralizing DNAM1, NKG2D, and TCRγδreceptors simultaneously, it may be that sensitivity of a tumor cell resides on the expression of ligand combinations that can bind these receptors. Two ligands recognized by Vδ2 TCR are surface mitochondrial F₁-ATPase and phospho-antigens, both of which are found in K562 cells (41, 42). Enhanced responses of T cells expressing Vγ9Vδ2 were observed when K562 cells were treated with aminobisphosphonates (41) and a similar strategy could be employed upon co-culture with aAPC clone #4 to increase the abundance of T cells bearing Vδ2 TCR (18). Future studies will evaluate additional TCRγδligands that naturally occur in these aAPC.

We enforced expression of co-stimulatory molecules to ascertain and improve the capability of K562-derived aAPC to propagate γδT cells expressing a diversity of TCR. Indeed, CD137L was the dominant co-stimulatory proliferative signal on aAPC for expansion of γδT cells with broad tumor-reactivity (Figure 1C), and its receptor, CD137, has been used to enrich tumor-reactive αβ T cells following antigen exposure and presumably TCR stimulation (43-45). CD137 was not expressed on resting γδT cells prior to expansion, suggesting that the importance of CD137L co-stimulation by aAPC followed TCR stimulation by the aAPC and expression of CD137 on the γδT cell surface. CD27⁺ and CD27^neg γδT cells have been shown to produce IFNγ and IL-17 (46), respectively; therefore, CD27 could be used as a marker for isolating γδT cells with a preferred cytokine output. ICOS-ligand in absence of CD86 was shown to polarize CD4⁺ αβ T cells to produce IL-17 instead of IFNγ (47), and current studies are investigating whether combinations of co-stimulatory molecules can selectively propagate cytokine-producing sub-populations of γδT cells. Thus, the aAPC co-culture system in the context of desired cytokines provides a clinically-relevant methodology to tailor the type of therapeutic γδT cell produced for adoptive immunotherapy.

Our data have implications for the design and interpretation of clinical trials. Expression of IL-15 was important for the maintenance of transferred γδT cells in vivo (48), supporting the use of IL-15 on aAPC, and future studies could inform on other molecules that could be introduced to maximize the cell therapy product. Correlative studies are enhanced by our observation that TCRγδmAb can be used to readily distinguish the three (Vδ1, Vδ2, and Vδ1^negVδ2^neg) T-cell subsets based on MFI of TCRγδexpression (Figure 3B). Given that γδT cells are not thought to recognize ligands in the context of MHC (17), there is potential to infuse allogeneic, including 3^rd party, γδT cells in lymphodepleted hosts to achieve an anti-tumor effect while mitigating the risk of graft-versus-host disease. Restoration of lymphopoiesis may result in graft rejection, but a therapeutic window could be established whereby tumors are directly killed by infused γδT cells, which may result in desired bystander effects as conserved or neo-antigens are presented to other lymphocytes. Indeed, γδT cells have been shown to lyse cancer cells, cross-present tumor-specific antigens to αβ T cells, and license them to kill tumors (49, 50). The aAPC clone #4 has been produced as a master cell bank in compliance with current good manufacturing practice and provides a clear path to generating clinical-grade γδT cells for human application. Human trials can now, for the first time, test the efficacy of adoptive transfer of T cells with polyclonal TCRγδrepertoire for treatment of solid and hematological tumors.

TRANSLATIONAL RELEVANCE.

γδT cells have anti-cancer activity, but only one subset, Vγ9Vδ2, has been harnessed for immunotherapy. Our study establishes that artificial antigen presenting cells (aAPC), IL-2, and IL-21 can activate and propagate γδT cells with polyclonal TCR repertoire to clinical scale. The heterogeneous population of γδT cells produced from ex vivo culture secreted pro-inflammatory cytokines, lysed a broad range of malignancies, and improved survival in an ovarian cancer xenograft model. Given that γδT cells are not thought to recognize ligands in the context of MHC, there is limited risk of graft-versus host disease in an allogeneic setting. Thus, 3^rd party γδT cells from an unrelated (healthy) donor could be produced in bulk and be administered as an off-the-shelf investigational therapy for hematologic and solid tumors. The aAPC are already available as a clinical reagent, which will facilitate the human application of polyclonal γδT cells.