TALEN-Mediated Inactivation of PD-1 in Tumor-Reactive Lymphocytes Promotes Intratumoral T-cell Persistence and Rejection of Established Tumors

Adoptive cell therapy (ACT) using autologous tumor-infiltrating lymphocytes (TIL) or T cells redirected to the tumor by chimeric antigen receptors (CAR) can mediate substantial regression of human cancers (1). Although multiple early-phase studies using ACT have demonstrated durable responses that correlate with persistence of the transferred T cells (2, 3), their in vivo activity and potency can be thwarted by the complex immunosuppressive nature of the tumor microenvironment (4). Combination protocols incorporating checkpoint inhibitors (CPI) such as anti–CTLA-4 or anti–PD-L1/PD-1 are currently being considered (NCT02408861; NCT02210117; NCT02374242; NCT02453620), but the potential exacerbation of the toxicities associated with CPIs may limit their implementation. Advanced genetic engineering of the ACT product offers potential solution to the toxicity issues by specifically targeting tumor-reactive T cells rather than inducing systemic checkpoint blockade. The protocols used for the generation of ACT are compatible with genetic manipulation of cells before transfer, which can be used to increase the proliferative capacity of lymphocytes (5), prolong their in vivo persistence (6), improve tumor infiltration (7), and bypass tumor immune suppression.

Programmed death-1 (PD-1, CD279) is one of the key coinhibitory immune receptors expressed by activated T cells. Initially described on a T-cell hybridoma undergoing cell death (8), PD-1 is a critical regulator of T-cell responses and essential to the maintenance of peripheral self-tolerance (9). In mouse and human cancers, high levels of PD-1 are found on TILs, although its ligand (PD-L1) is upregulated by tumor and stromal cells in response to inflammatory cytokines (e.g., IFNγ; refs. 10, 11). The PDL-1–PD-1 axis plays a major role in tumor growth and immune escape through inhibition of T-cell proliferation, effector function (12), and of T-cell survival by promotion of apoptosis (13). The administration of blocking anti–PD-L1 or PDL-1 antibodies in vivo results in enhanced CD8⁺ T cell responses and decreases tumor burden in mouse models of cancer (14, 15). In humans, expression of PD-1 by TILs is associated with impaired effector function and/or poor outcome in several tumor types (16, 17). Direct evidence supporting the importance of this pathway in modulating antitumor immunity comes from recent clinical trials in melanoma and non–small cell lung cancer, where anti–PD-1 antibodies have improved both progression-free and overall survival (18, 19). Of relevance, systemic immunotherapy with CPIs can also induce immune-related adverse events (irAE). Specific ablation of negative regulators of T-cell function on tumor-reactive T cells used for ACT offers the potential to exploit the antitumor activity while reducing systemic toxicity.

Efficient and rapid inactivation of genes in primary antigen-specific T cells is achievable using the transient expression of transcription activator-like effector nucleases (TALEN) mRNA, which offers several safety advantages for clinical applications (20). Using two different mouse models of cancer, we defined the impact of TALEN-mediated PD-1 gene inactivation in adoptively transferred tumor-reactive CD4⁺ and CD8⁺ T cells. Our data support a model in which PD-1 gene inactivation increases persistence of T cells at the tumor site, enhancing tumor control of B16 melanoma by gp-100-reactive CD8⁺ T cells and mediating complete rejection of established MCA205 fibrosarcoma by endogenous CD4⁺ and CD8⁺ T cells. This work demonstrates both the feasibility and the potential potency of strategies incorporating immune checkpoint editing of tumor-reactive T cells for use in the context of adoptive T-cell therapies for cancer.

Six- to 8-week-old pmel-1/SJL TCR-transgenic, C57BL/6, and C57BL/6-SJL mice were purchased from Charles River Laboratories. All animal use was in accordance with the Home Office guidelines. The B16.BL6 melanoma cell line was obtained from Prof. James Allison (MD Anderson Cancer Center, Houston, TX), the MCA205 fibrosarcoma cell line was provided by Prof. Guido Kroemer (Institut Gustave Roussy, Villejuif, South Paris, France), EL4 thymoma cells were provided by Prof. Henning Walczak (UCL Cancer Institute, London, United Kingdom), and authentication was performed by the ATCC using short tandem repeat DNA profiles. The cells were grown in RPMI-1640 supplemented with 10% FCS, l-glutamine, penicillin/streptomycin (Sigma-Aldrich).

Three pairs of TALEN targeting the PD-1 murine gene were produced by Cellectis using the solid phase assembly method (21). Their activity was evaluated in 5 × 10⁶ EL4 cells, cotransfected with 10 μg of in vitro transcribed mRNA (mMESSAGE mMACHINE T7 Kit; Ambion) from each half TALEN (Supplementary Fig. S1). Three days after transfection, locus-specific PCRs were performed on genomic DNA (Supplementary Table S1), mutations were assessed by mismatches digestion (T7) and Miseq analysis (20).

Splenocytes and lymph nodes cells from pmel-1/SJL transgenic mice were cultured in complete RPMI with gp100 peptide (Pepceuticals) and rhIL-2 (Preprotech) 100 UI/mL for 24 hours before selection of CD8⁺ T cells using the Dynabeads FlowComp Mouse CD8 Kit. TRL were selected from 100 to 150 mm³ tumors using Dynabeads FlowComp Mouse Pan T (CD90.2) and maintained in complete RPMI with rhIL-2 (1,000 UI/mL). A total of 5 × 10⁶ Pmel-1 CD8⁺ T cells and MCA205 TRL/180 μL of BTX medium T were mixed with 20 μg of IVT mRNA, before electroporation using an Agile Pulse BTX system (Harvard Apparatus). C57BL/6 mice were subcutaneously implanted with either 2.5 × 10⁵ B16 cells or 4 ×10⁵ MCA205 cells (day 0). On day 6 for the pmel-1 model and on day 4 for the MCA205 model, total body irradiation (TBI) was performed (5 Gy) and 1 × 10⁶ pmel-1 T cells or 1 × 10⁶ MCA205 TRL were adoptively transferred into tumor-bearing mice (n = 5/group), followed by intraperitoneal (i.p.) injection of rhIL-2 (1.2 × 10⁶ IU) for 3 days. Mice were sacrificed when the tumors exceeded 15 mm in diameter. Mice were treated with the anti-mouse CD8a 2.43–depleting antibody (BioXcell) and the anti-MCH class II M5/114 (BioXcell), intraperitoneally on day 0, 2, 4, 6, and 8 after ACT (10 mg/kg).

Tumors were weighed, digested in Liberase D (Roche) and mechanically disaggregated through 70-μm filters. Lymph nodes cells and lymphocytes from tumor samples, enriched on a Ficoll gradient were incubated with murine antibodies: anti–CD4-V500 (BD Horizon), anti–CD8-BV650, anti–PD-1-PE-Cy7 (Biolegend), anti–CD45.1-APC-Cy7, anti–FoxP3-AF700, anti–IFNγ-FITC (restimulation with gp100-pulsed dendritic cells for pmel cells or PMA/ionomycin for TRL), anti–KI67-PerCP-eFluor 710, Fixable Viability Dye eFluor 450 (eBioscience), anti-FLICA (Vybrant FAM Poly Caspases Assay Kit; ThermoFisher), Annexin-V–PE Apoptosis Detection Kit II (BD Pharmingen), Cell Sorting Set-up Beads (Life Technologies), and anti-human Granzyme B-APC (Invitrogen) in a blocking solution: 5% mouse serum, 5% rat serum (Thermo Scientific), 2% FCS, and 2% anti-FcR 2.4G2 (BioXcell). Samples were permeabilized using a Fixation/Permeabilization kit (eBioscience), acquired on a LSRII Fortessa and analyzed with FlowJo software (Tree Star).

Proteins were separated and subjected to the following antibodies: anti-TALEN (1 mg/mL), anti-GAPDH (D16H11; Cell Signaling Technology; 1:2,000), HRP-coupled secondary antibody anti-Rabbit IgG1 HRP-linked (Cell Signaling Technology).

One-way ANOVA, two-way ANOVA, or t tests with P < 0.05 were performed using Prism 6.0 software (GraphPad). Multiple comparisons were corrected with the Bonferroni coefficient and Kaplan–Meier survival curves were compared with the log-rank test.

For additional information, see Supplementary Data Section.

Three pairs of TALENs targeting the murine PD-1 gene were tested in EL4 cells using an optimized mRNA–electroporation protocol. Only the pair targeting the exon 2 sequence (Fig. 1A) caused detectable mismatch-identified mutagenesis (up to 27%) in the PD-1 gene, correlating with loss of PD-1 expression in up to 15% of electroporated EL4 cells (Supplementary Fig. S1). To investigate whether PD-1 inactivation of primary tumor-reactive T cells provided superior antitumor activity against the poorly immunogenic mouse melanoma B16 model, we used CD8⁺ T-cell receptor transgenic (TCR Tg) cells specific to the melanoma differentiation antigen gp-100 (pmel-1, Supplementary Fig. S2A; ref. 22). In vitro activated pmel-1 T cells were electroporated with control GFP mRNA (CD8^wt) or PD-1–targeting TALEN mRNA (CD8^PD-1Ex2), delivering a high transfection efficiency of >85% as confirmed by GFP expression (Supplementary Fig. S2B). Western blot analysis confirmed transient TALEN expression one day (d1) after transfection (Fig. 1B) and Miseq analysis of the targeted sequence showed a high frequency of mutations (53% non-homologous end joining, NHEJ; Fig. 1C). Three days after transfection, PD-1–negative cells were enriched using magnetic beads (Supplementary Fig. S2C) before ACT of 1 × 10⁶ CD8^wt or CD8^PD-1Ex2 pmel-1 cells and rhIL-2 into B16 tumor-bearing mice. Significantly, enhanced tumor control was consistently observed in mice receiving CD8^PD-1Ex2 cells compared with those treated with CD8^wt cells (Fig. 1D). Six days after transfer in vivo, a significant enrichment of tumor-infiltrating PD-1–negative pmel-1 cells was observed in mice treated with CD8^PD-1Ex2 T cells (48.2%; Fig. 1E, top), which contributed to a 2-fold increase in the total number of tumor-infiltrating pmel-1 cells (Fig. 1E, bottom). In contrast, the absolute number of T cells in draining lymph nodes (DLN) was similar between the two groups, suggesting that accumulation of CD8^PD-1Ex2 T cells at the tumor site was not due to differences in engraftment nor infiltration from the DLNs (Supplementary Fig. S2D). Whereas use of RNA and electroporation to deliver TALENs is compatible with future clinical translation, it does not allow incorporation of a reporter gene to distinguish PD-1–edited cells. Nevertheless, the rate of mutagenesis, investigated ex vivo 7 days after ACT demonstrates an effective PD-1 gene editing (70%) in the PD-1–negative fraction from the CD8^PD-1Ex2 group (Supplementary Fig. S2E). To investigate the mechanisms underpinning enhanced tumor control by pmel-1 cells in the CD8^PD-1Ex2 group, we characterized the proliferation and effector function of CD8^wt and CD8^PD-1Ex2 T cells in both PD-1⁺ and PD-1⁻ tumor-infiltrating pmel-1 populations. Contrary to previous work suggesting that PD-1 signaling reduces effector activity at the tumor site (23), no differences were observed in KI67, granzyme B (GZB), or IFNγ expression between PD-1⁺ and PD-1⁻ cells. When we compared the adjusted percentages (percentage of cells in PD-1⁺ and PD-1⁻ subpopulations) of tumor-infiltrating KI67⁺, GZB⁺, and IFNγ⁺ cells, we observed that the TALEN-induced PD-1 gene inactivation did not impact the proliferative or functional characteristics of pmel-1 CD8⁺T cells in vivo (Supplementary Fig. S2F–S2H). Instead, we consistently observed the appearance of a PD-1⁻ pmel-1 population in the CD8^PD-1Ex2 group, which contributed to an increase in the total number of KI67-, GZB-, and IFNγ-expressing pmel-1 cells in tumors from mice receiving CD8^PD-1Ex2 T cells (Fig. 1F and G). Together, these data strongly suggest that, in the context of ACT, PD-1 expression primarily controls number of tumor-reactive T cells at the tumor site rather than their proliferative or functional status.

To define the therapeutic potential and functional impact of PD-1 gene inactivation in a more physiologically relevant T cell population, we performed PD-1 gene editing in polyclonal tumor-reactive lymphocytes (TRL) from C57BL/6^SJL mice challenged with MCA205 cells (24). CD4⁺ and CD8⁺ T cells were isolated from the tumor (TILs) and tumor-draining lymph node (TDLN) via positive selection. After electroporation of the donor TRL, we observed a transfection efficiency >65% and we identified TALEN-induced PD-1 gene mutations by Miseq analysis in 26% of CD4⁺ T cells and almost 40% of CD8⁺ T cells (Supplementary Fig. S3A and S3B).

In parallel to the generation of the donor mock-transfected (TRL^wt) and PD-1–edited TRL cells (TRL^PD-1Ex2), we challenged a separate group of C57BL/6 mice with MCA205 tumors. Four days later, tumor-bearing mice were treated with TBI followed by ACT with TRL^wt or TRL^PD-1Ex2 cells and rhIL-2. Six days after transfer, analysis of PD-1 expression on TILs from mice receiving TRL^wt or TRL^PD-1Ex2 revealed a 15% to 20% average increase in PD-1⁻ CD4⁺ and PD-1⁻ CD8⁺ T cells (Fig. 2A and B). Consistent with our findings in the TCR Tg model, a significant accumulation of PD-1⁻ cells was observed only at the tumor site (Fig. 2C; Supplementary Fig. S3C). Although PD-1 gene inactivation did not alter TRL proliferation (KI67) or function (GZB and IFNγ; Supplementary Fig. S3D–S3F), the absolute number of PD1⁻ TRL–expressing KI67, GZB, and IFNγ was significantly higher in mice receiving TRL^PD-1Ex2 (Fig. 2D–F).

Given that the increase in the number of PD-1⁻ cells in mice receiving TRL^PD-1Ex2 cells was not associated with increased proliferation, we evaluated whether this could be explained by a reduced rate of PD-1–induced cell death. Using Annexin-V and a fluorescent caspase substrate kit (FLICA), we observed an overall reduction in the frequency of intratumoral apoptotic cells in PD-1⁻ compared with PD-1⁺ CD4⁺ and CD8⁺ T cells (Fig. 3A–D). Whereas we observed no difference in the frequency of apoptotic cells between mice treated with TRL^wt and TRL^PD-1Ex2, a significant increase in the absolute number of Annexin-V⁻ and FLICA⁻ TRL^PD-1Ex2cells per mg of tumor was reported (Fig. 3E and F). The data suggest that instead of increasing T-cell reactivity against tumors, TRL^PD-1Ex2 cells are rendered resistant to PD-1–mediated cell death at the tumor site.

PD-1 gene inactivation significantly increased the activity of our ACT protocol (P = 0.0052), and antitumor activity was dependent on both CD8⁺ and CD4⁺ T cells as administration of depleting anti-CD8 or MHC-II–blocking antibodies ablated antitumor responses (Fig. 4A and B). No evidence of tumor recurrence was observed after MCA205 rechallenge on day 50 in mice that had previously rejected tumors (data not shown). The development of long-term memory, as demonstrated by protection against a subsequent tumor challenge, is an important feature of successful T-cell–adoptive immunotherapy (25). Our data show high levels of KI67, consistent with high proliferative capacity for both TRL and TRL^PD-1Ex2 cells, which also displayed a T-bet^hi CD62L^lowCD44^hi phenotype consistent with effector memory T cells (Supplementary Fig. S4A and S4B; ref. 26). Finally, ex vivo stimulation of circulating T cells from surviving mice (40 days after tumor rejection) demonstrated high levels of IFNγ production in both TRL and TRL^PD-1Ex2 groups arguing against terminal differentiation (Supplementary Fig. S4C; ref. 27).

In summary, our data demonstrate the feasibility and potential clinical relevance of gene-editing approaches, conferring superior in vivo activity in the context of adoptive cell therapy protocols. Of relevance, we observed no evidence of toxicity (weight and physiologic alterations) after ACT. To the best of our knowledge, this is the first proof-of-concept study illustrating enhancement and persistence of antitumor responses using targeted genome editing of primary tumor reactive T cells. Our data indicate that the primary mechanism by which PD-1 gene inactivation affects tumor-reactive T cells is by regulating their ability to survive rather than increasing their activity on a per cell basis, which aligns with the original description of the role of PD-1 signaling in T-cell apoptosis (8). Although we investigated PD-1 as the initial proof-of-concept target, the technology and approach described potentially allow the permanent disruption of other inhibitory checkpoints (one or more), considerably advancing the design of the next generation of cancer immunotherapies.

M. Pule is a CSO at Autolus, reports receiving a commercial research grant from Cellectis, has received speakers bureau honoraria from Amgen and Roche, and has ownership interest (including patents) in Autolus. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Menger, L. Poirot, M. Pule, K.S. Peggs, S.A. Quezada

Development of methodology: L. Menger, A. Sledzinska, F. Arce Vargas, L. Poirot, M. Pule

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Menger, A. Sledzinska, K. Bergerhoff, K.S. Peggs

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Menger, J. Herrero, K.S. Peggs

Writing, review, and/or revision of the manuscript: L. Menger, A. Sledzinska, L. Poirot, M. Pule, J. Herrero, K.S. Peggs, S.A. Quezada

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Menger, F. Arce Vargas, J. Smith

Study supervision: K.S. Peggs, S.A. Quezada

S.A. Quezada is funded by a CRUK Career Development Fellowship and a Cancer Research Institute Investigator Award. K.S. Peggs receives funding from CRUK, Leukemia and Lymphoma Research. This work was funded in part by an EU FP7 grant and undertaken at UCL Hospitals/UCL with support from the Department of Health and CRUK funding schemes for National Institute for Health Research Biomedical Research Centers and Experimental Cancer Medicine Centers. The research leading to these results has received funding from the European Union Seventh Framework Programme under grant amendment no 602239 ATECT.