Background

In recent decades, the field of cancer immunotherapy has expanded greatly, leading to improved outcomes for many patients. T cell-based immunotherapy in particular has yielded significant benefits, with great success delivered by immune checkpoint inhibitors, now used widely to treat numerous cancers.1 Despite these successes, not all patients benefit, with only a small proportion experiencing durable response. Multiple challenges preventing universal success are presented by the immunosuppressive tumor microenvironment (TME; reviewed in Munn and Bronte2). In recent years, it has been well established that a major obstacle to successful cancer immunotherapy is T cell exhaustion. This refers to a state of dysfunction in T cells in response to persistent stimulation with antigen in chronic viral infections and cancer, characterized by transcriptional and epigenetic changes leading to progressive loss of effector function, impaired persistence, and coexpression of multiple inhibitory receptors (IRs).3

Adoptive cell therapy (ACT), in which autologous peripheral blood or tumor infiltrating T lymphocytes (TILs) are manipulated ex vivo before expansion and infusion to the patient, provides an opportunity to surmount some of these barriers. Chimeric antigen receptor (CAR) expressing T cells are an example of ACT that has demonstrated remarkable results in patients with hematological malignancies refractory to conventional therapies.4 However, treatment of solid tumors with CAR T cells is yet to yield such promising results, in part because CARs recognize intact tumor cell surface proteins, while most solid tumor antigens are intracellularly derived peptides presented on the cell surface by MHC. Genetic engineering of autologous T cells to express αβ T cell receptors (TCRs) specific to such antigens is therefore an alternative approach for ACT in solid tumors. This approach also allows for additional modifications aimed at improving T cell effector function and longevity in tumors. It has been demonstrated previously that the autoimmunity-associated tyrosine phosphatase PTPN22 restrains T cell responses to weak affinity and/or self-antigens,5 and that systemic inhibition or deletion of PTPN22 improves responses to tumors.6 7 PTPN22 deletion restricted to T cells is sufficient to induce enhanced tumor control, as demonstrated by tumor rejection after adoptive transfer of Ptpn22KO naïve, effector or memory T cells into tumor bearing wild type host mice.8 9

An important factor in developing a T cell product for optimal anti-cancer protection is the choice of T cell phenotype. Effector cytotoxic T lymphocytes (CTLs) are straightforward to generate and expand to large numbers in vitro and have the greatest cytotoxicity, but their restricted longevity has raised questions about their ability to provide long-term tumor control. Prior investigation in in vivo tumor models has shown superior tumor control from adoptive transfer of Ptpn22KO CTL when administered in high numbers.8 9 However, systematic evaluation of the longer-term fate of Ptpn22KO CTL in the face of persisting tumor challenge has not been carried out. Here, we use in vitro and in vivo models of chronic antigen exposure and show that despite Ptpn22KO CTL out-performing Ptpn22-sufficient (Ptpn22WT) CTL in their initial responses, CTL lacking PTPN22 more rapidly acquire an exhausted phenotype. In consequence, low numbers of Ptpn22KO CTL transferred to tumor-bearing mice were less effective than Ptpn22WT CTL at controlling established tumor growth and showed enhanced expression of IRs, such as PD-1, and in particular TIM-3. Inhibition of PD-1 improved tumor control by Ptpn22KO CTL, supporting a previous report.7 In contrast, deletion or blockade of TIM-3 was detrimental to Ptpn22KO CTL function, suggesting TIM-3 does not function as a conventional IR in Ptpn22KO CTL and, therefore, is not a good target for reversing exhaustion in Ptpn22KO cells. Together, our findings illustrate that strategies aiming to optimize T cell effector function, such as PTPN22 deletion or inhibition, need to carefully consider selection of T cell phenotype for ACT in order to balance enhanced short-term effector function with susceptibility to exhaustion so as to optimize long-term tumor control.

ResultsPtpn22KO CTLs have enhanced effector function, but control tumors less effectively following adoptive transfer

To understand the impact of PTPN22 deletion in CTL for long-term tumor control, we employed the Class I H-2Kb-restricted OT-I TCR transgenic system, which allows TCR stimulus strength to be altered through use of cognate OVA peptides with varying affinities.10 CTLs were generated from naïve OT-I Rag1KO T cells (hereafter referred to as OT-I cells), which were either Ptpn22WT or Ptpn22KO, by activating in vitro with the strong agonist peptide, SIINFEKL (N4), for 2 days prior to expansion in IL-2 for 4 days.5 We have shown previously that proliferation in vitro of Ptpn22WT or Ptpn22KO T cells is equivalent in response to strong agonist N4 peptide5 8 and in keeping with previous work,5 9 Ptpn22KO OT-I CTL produced more cytokine than Ptpn22WT OT-1 cells in response to 4 hours of restimulation with weak OVA peptide SIITFEKL (T4, figure 1A). In addition, Ptpn22KO OT-I CTL were ~3 fold more cytotoxic than Ptpn22WT CTL against MC38 tumor cells expressing T4 (MC38-T4; figure 1B).

Figure 1

Ptpn22KO CTL are more effector-like, but control tumors less efficiently following adoptive transfer. (A) Cytokine production from Ptpn22WT and Ptpn22KO CTL following re-stimulation for 4 hours with 10nM T4 (dot plots) or T4 at concentrations shown (graphs). Dot plots are gated on live, single cells. Numbers in dot plot quadrants are percentages. Data are representative of 4 independent experiments. (B) Cytotoxicity of Ptpn22WT and Ptpn22KO CTL against MC38 tumor cells expressing T4 antigen, at ratios indicated. Data are representative of four independent experiments. (C) MC38-T4 tumor growth (left) and time to tumors reaching 200 mm3 volume in Rag1KO hosts following ACT with Ptpn22WT and Ptpn22KO CTL, or no ACT. 0.5×106 tumor cells were injected at day 0, followed by ACT with 1×106 CTL at d4. n=5–8 per group in the experiment shown. Data are representative of three independent experiments (total n=27 mice per group for Ptpn22WT and Ptpn22KO CTL; 10 per group for no ACT). (D) Tumor infiltration by adoptively transferred Ptpn22WT and Ptpn22KO CTL. Mice were culled once tumors reached humane end points, and tumors dissociated to obtain single cell suspensions for flow cytometric analysis. Data are pooled from two independent experiments. n=9–12 per group.(E) Cytokine production by Ptpn22WT and Ptpn22KO TIL. Mice were injected with brefeldin A 4 hours before culling. Data are pooled from two independent experiments. Data were excluded from tumors with insufficient (<150) numbers of CD8+TIL. n=9–12 per group. (F) Inhibitory receptor expression on Ptpn22WT and Ptpn22KO TIL. Data are representative of two independent experiments. All bars on graphs represent mean±SEM. P values as determined by two-way ANOVA with Šidák correction for multiple comparisons (A,B), or Student’s t-test (D,E,F). ACT, adoptive cell therapy; ANOVA, analysis of variance; MFI; median fluorescence intensity.

Control of tumor growth by ACT is effectively a competition between tumor bulk and efficacy of the transferred T cells. Previous studies showed that transfer of large numbers (107) of Ptpn22KO CTL provided better control of EL4 lymphoma and ID8 ovarian carcinoma than Ptpn22WT CTL.8 9 We asked whether this would also be the case if fewer cells were adoptively transferred, thus taking longer to gain control of the tumors, which might be more in keeping with a therapeutic scenario in which tumors are more established at the time of commencing treatment. MC38 tumor cells expressing the low affinity peptide, T4, were inoculated subcutaneously (s.c.) into Rag1KO mice, providing a model system in which tumor rejection would be mediated by the transferred CTL alone without contribution from the Ptpn22WT host T cell response. 106Ptpn22KO or Ptpn22WT CTL per recipient were injected intravenously once tumors were palpable, and tumor growth was followed. Unexpectedly, although better than no ACT, Ptpn22KO ACT controlled tumor growth less efficiently compared with mice receiving Ptpn22WT ACT (figure 1C). To understand the reduced efficacy of Ptpn22KO CTL in this situation, we analyzed tumors from the mice and found that both genotypes were equal in their ability to infiltrate and persist in tumors (figure 1D). Furthermore, assessment of cytokine production in vivo following brefeldin A administration intravenously showed more Ptpn22KO than Ptpn22WT TILs stained positively for cytokines (figure 1E). However, expression of IRs such as PD-1 and LAG-3 was marginally higher on Ptpn22KO TILs, while expression of TIM-3 was significantly higher when compared with Ptpn22WT TILs (figure 1F), suggesting that heightened negative regulatory signals in Ptpn22KO TILs may play a role in their impaired control of tumors. Collectively, these results show that lack of PTPN22 in CTL enhances effector function in response to antigen, but if administered to tumor bearing hosts in numbers insufficient to rapidly control tumor growth, Ptpn22KO CTL are more prone to exhaustion and fail to give prolonged protection against tumors.

Ptpn22KO CTLs become more dysfunctional on chronic TCR stimulation

To understand the reasons underlying the impaired tumor control by lower numbers of Ptpn22KO CTL, we modeled in vitro the chronic TCR stimulation experienced by TILs by repeatedly culturing CTL with antigen-pulsed tumor cells (figure 2A). As expected, d6 Ptpn22KO CTLs were more cytotoxic and produced more cytokine in response to 4 hours restimulation with fresh T4-pulsed tumor cells (figure 2B). However, the functionality of both Ptpn22WT and Ptpn22KO CTL diminished on repeated antigen exposure and by d9 of culture both genotypes lost the capacity to produce cytokines (figure 2B). Tumor target cell killing was still demonstrable up to d15 of culture by both Ptpn22KO and Ptpn22WT CTL, however, the cytotoxic advantage exhibited by Ptpn22KO CTL at d6 (figure 1B) was lost in favor of Ptpn22WT CTL at the later timepoint (figure 2C). While LAG-3 was highly expressed at similar levels by both cell genotypes, there was higher expression of IRs PD-1, TIGIT and most strikingly TIM-3 in Ptpn22KO compared with Ptpn22WT CTL at d6, even in the absence of antigen re-exposure (figure 2D, closed histograms); this was further boosted by 4 hours restimulation (figure 2D, open histograms). Over the time course of chronic re-stimulation PD-1 and TIGIT expression was consistently and significantly higher on PTPN22KO CTL, while TIM-3 expression was strikingly elevated on Ptpn22KO CTL throughout, yet its expression on Ptpn22WT cells remained low (figure 2E).

Figure 2

Ptpn22KO CTLs become more dysfunctional on chronic TCR stimulation. (A) Schematic of experimental design. Naïve OT-I Ptpn22WT and Ptpn22KO T cells were activated with N4 (10 nM) for 48 hours, then cultured in IL-2 (20 ng/mL) for 4 days to expand them and induce differentiation to CTL. CTL on day 6 were cultured with antigen-pulsed MC38 tumor cells, which were replenished every 3 days. Cytokine production was assessed on cells either non-stimulated or restimulated for 4 hours with peptide as indicated. (B) Cytokine production by Ptpn22WT and Ptpn22KO CTL following chronic restimulation with antigen. CTLs were cultured for the indicated periods of time with MC38 tumor cells pulsed with N4 antigen. At each time point, CTLs were restimulated with MC38 pulsed with T4 (100 uM) for 4 hours and cytokine production in response was measured by intracellular staining for flow cytometry. Numbers in dot plot quadrants are proportions; gates are based on non-restimulated CTL. Graph shows TNF as representative cytokine. Data are representative of three independent experiments. (C) Killing of luciferase expressing MC38-T4 cells by Ptpn22WT and Ptpn22KO CTL after chronic (day 15) restimulation with antigen-bearing tumor cells. Data are representative of three independent experiments. (D) Inhibitory receptor expression on Ptpn22WT and Ptpn22KO CTL, resting or restimulated for 4 hours with MC38 tumor cells pulsed with T4 (100 uM). Data are representative of three independent experiments. (E) Inhibitory receptor expression on Ptpn22WT and Ptpn22KO CTL after chronic restimulation with antigen. Data are pooled from two independent experiments. (F) Eomes and TCF-1 expression by Ptpn22WT and Ptpn22KO over chronic Ag exposure. Representative dot plots of two independent experiments. (G) TCF-1, Eomes and Tbet expression in Ptpn22WT and Ptpn22KO CTL at the indicated time points. Graphs show pooled data from two independent experiments. All bars on graphs represent mean±SEM. P values as determined by two-way ANOVA with Šidák correction for multiple comparisons (E,G). ANOVA, analysis of variance; MFI, median fluorescence intensity.

Impaired function was accompanied by a change in abundance of certain transcription factors (TFs), in particular, T cell factor 1 (TCF-1; encoded by Tcf7) and Eomesodermin (Eomes), both of which are essential in T cell fate decisions, such as the generation of long-lived memory T cells11–13 and these TF are characteristically changed in exhausted T cells.14 15 TCF-1 is essential for generation and maintenance of the CD8+ T cell memory response,11 13 as well as being associated with a stem-like population of exhausted cells in models of chronic infection and cancer.16–20 We found that expression of TCF-1 diminished with increasing tumor antigen exposure in keeping with reducing stemness and memory formation (figure 2F,G). Similarly Eomes, another TF that is upregulated in exhausted T cells and which correlates with high IR expression15 was increased following repeated Ag exposure (figure 2F,G). Expression of these TFs and of Tbet (figure 2G) was not significantly different between Ptpn22WT and Ptpn22KO cells at any time point, suggesting either a combined effect of several TFs rather than the action of one specifically, or a different TF was responsible for inducing the more dysfunctional and exhausted phenotype of cells lacking PTPN22.

Ptpn22KO CTLs become more dysfunctional in the TME

To understand how the response of Ptpn22KO CD8+ T cells developed in vivo we used the faster growing p53KO version of the ID8 ovarian carcinoma21 transfected with N4 peptide (ID8-N4) which allows tumor responsive CD8+ T cells to be readily retrieved from the peritoneal exudate (PE). To eliminate any confounding effect of differing host environments we cotransferred a mix of naïve Ptpn22WT (CD45.1) and Ptpn22KO (CD45.1/2) OT-I T cells to C57BL/6 CD45.2 wild-type recipient mice with established tumors. Mice were culled at weekly intervals and CD8+ T cells in the PE and mesenteric lymph nodes (mLN) were analyzed (figure 3A). The presence of tumors was monitored in individual animals using in vivo bioluminescence imaging so that mice that rejected tumors could be excluded from the analysis, thus ensuring that only CD8+ T cells that had been continually exposed to tumor antigens were analyzed. Transferred T cell were readily detected in both PE and mLN at early time points (day 4 and day 12 post-transfer of T cells) but their numbers diminished with increasing time from transfer. Despite starting as 35% of the initial injection mix, Ptpn22KO T cell recovery was greater than that of Ptpn22WT T cells in PE at d4 and significantly more in both PE and mLN at d12 (figure 3B,C), indicating greater expansion in the initial response to antigen. However, by d19 the numbers of recovered Ptpn22WT and Ptpn22KO cells had equalized and by d26 Ptpn22KO cells were barely detectable, while low numbers of Ptpn22WT cells could still be identified in both mLN and PE. Notably, Ptpn22WT and Ptpn22KO T cells that were transferred into control mice without tumors were maintained and readily retrievable in the mLN at the later d26 time point (figure 3B,C). These findings show that T cells lacking PTPN22 initially undergo greater expansion in response to tumor antigens, however, when tumors are not cleared and thus antigen persists, Ptpn22KO cells that are chronically exposed to antigen in vivo fail to sustain proliferation and/or die more rapidly. This is specifically antigen dependent, since Ptpn22KO T cell survival was not impaired in tumor-free hosts.

Figure 3

Ptpn22KO CTL become more dysfunctional in the tumor microenvironment. (A) Experimental design. C57BL/6 hosts were injected i.p. with 1×106 ID8-N4 tumor cells on day −7. IVIS imaging was carried out on day −1, prior to adoptive transfer of a mix of 1×106 each of naïve Ptpn22WT and Ptpn22KO OT-I T cells. Tumor presence was confirmed with IVIS on the day before analysis of peritoneal exudate T cells at the indicated time points. Mice that spontaneously rejected tumors were excluded from analysis. n=6 mice at each time point; n=3 control mice (no tumor). (B,C) Transferred Ptpn22WT and Ptpn22KO T cells in peritoneal exudate and mLN. (B) Representative dot plots from each time point, as indicated. Gated on single, live, CD8+, CD45.1+ cells. Numbers are proportions of total donor cells. Control mice received T cells but no tumors. Recovered numbers of transferred cells per mouse in experiment shown: d4, PE=99–806, mLN=1892–4163; d12, PE=60–588, mLN=276–1600; d19, PE=128–869, mLN=51–776; d26, PE=61, mLN=65; d26 controls, mLN=535–948. PE or mLN with fewer than 50 recovered cells were excluded from analysis. (C) Ptpn22WT and Ptpn22KO cells in peritoneal exudate and mLN at indicated time points, as a proportion of total donor (CD45.1+) cells. Data are representative of two independent experiments. P values as determined by two-way ANOVA with Šidák correction for multiple comparisons. (D) UMAP embedding analysis of flow cytometry data showing donor Ptpn22WT and Ptpn22KO T cells in peritoneal exudate (gated on single, live, CD8+ CD45.1+ cells). Data from donor (CD45.1+) cells in all six mice at day 4 were concatenated for UMAP analysis. Plots show expression of indicated proteins. (E) Representative contour plots showing PD-1 and Slamf6 expression on Ptpn22WT and Ptpn22KO T cells isolated from peritoneal exudate of mice at the indicated time points (gated on single, live, CD8+ cells). Numbers in gates represent proportions. ANOVA, analysis of variance.

In order to further characterize differences between Ptpn22WT and Ptpn22KO cells in TMEs, we performed uniform manifold approximation and projection (UMAP)22 on flow cytometry data from Ptpn22WT and Ptpn22KO cells recovered from PE of mice with ID8 tumors at day 4, when phenotypic changes were likely to be established. Ptpn22KO and Ptpn22WT cells occupied largely similar regions of phenotypic space but there were small distinct regions that were unique. In particular, we identified a subpopulation of Ptpn22KO cells that separated from the bulk of the cells (figure 3D, indicated by arrow). This subpopulation expressed highly markers associated with terminal effector differentiation and exhaustion, such as PD-1 and Granzyme B. In addition, IFNγ was reduced suggesting early loss of effector function. In contrast the expression of markers associated with stemness or memory formation and longevity, including Slamf6 and CD27, was low. The subpopulation was absent in Ptpn22WT cells at day 4, suggesting that Ptpn22KO cells undergo terminal effector differentiation and exhaustion more readily than Ptpn22WT cells, potentially at the expense of memory cell differentiation.

We followed expression of Slamf6 and PD-1 on Ptpn22WT and Ptpn22KO cells from PE over the entire time-course of tumor exposure as indicators of stemness and terminal effector/exhaustion, respectively. These markers have been used to differentiate progenitor (Slamf6hi PD-1int) and terminally exhausted (Slamf6- PD-1hi) T cell populations in chronic viral infection and tumor models.14 23 Both Ptpn22WT and Ptpn22KO cells displayed a Slamf6hiPD-1- phenotype at early time points consistent with a less differentiated state. However, Slamf6 expression was progressively lost and PD-1 acquired at later time points, in keeping with increasing terminal differentiation and development of exhaustion. Importantly, this occurred earlier in Ptpn22KO cells than in Ptpn22WT cells, with most Ptpn22KO cells identified as Slamf6- PD-1hi by d19 (figure 3E). These data are consistent with a model in which deletion of PTPN22 in T cells leads to acutely enhanced effector differentiation and function but this is at the expense of memory formation, and that in the setting of persisting antigen (such as in tumors) Ptpn22KO T cells ultimately exhaust more quickly than equivalent Ptpn22WT cells. This process is cell intrinsic and occurs even when both genotypes are exposed to the same environmental factors.

Dysfunctional Ptpn22KO CTL can be rescued by PD-1 blockade

Given that IRs such as PD-1 were significantly elevated on Ptpn22KO CTL in vitro prior to adoptive transfer, we sought to test the hypothesis that this could be driving their impaired function. Previous studies have shown a synergistic effect of combining PTPN22 deletion or pharmacological inhibition with blockade of the PD-1/PD-L1 axis,6 7 but not the extent to which this treatment was due to a T cell-intrinsic effect excluding contributions from other hematopoietic PTPN22-expressing lineages. To determine the effect of PD-1 inhibition on Ptpn22KO CTL specifically, we combined our Rag1KO adoptive transfer model with PD-1 blocking mAb treatment (figure 4A).

Figure 4

Dysfunctional Ptpn22KO CTL can be rescued by PD-1 blockade. (A) Schematic of experiment. 0.5×106 MC38-T4 tumor cells were injected at d0, followed by ACT with 1×106Ptpn22WT or Ptpn22KO CTL at day 4, then i.p. injection of anti-PD-1 or isotype control Ab (200 mg per mouse) on day 7, day 14, and day 21. (B) Tumor growth after ACT of Ptpn22WT or Ptpn22KO CTL±anti-PD-1 mAb. Data are representative of three independent experiments. n=7 mice per group in the experiment shown (total n across all experiments=19–24). (C) Survival to exponential tumor growth or ulceration. (D) PD-1 and TIM-3 expression on Ptpn22KO TIL. Data are representative of three independent experiments. Tumors with fewer than 100 CD8+ TIL were excluded from analysis. n=6–9 per group in each experiment. (E) Tumor infiltration by adoptively transferred Ptpn22KO CTL with or without anti-PD-1 mAb. Mice were culled once tumors reached humane end points, and tumors dissociated to obtain single cell suspensions for flow cytometric analysis. Data are pooled from two independent experiments. n=6–9 per group in each experiment. Bars on graphs represent mean±SEM. P values as determined by pairwise survival analysis (B), or Student’s t-test (D, E). ACT, adoptive cell therapy; MFI, median fluorescence intensity.

As before, adoptively transferred Ptpn22KO CTL provided less effective control of tumor growth, and this was significantly improved by anti-PD-1 treatment (figure 4B,C). PD-1 blockade had no significant impact on tumor control by Ptpn22WT CTL, which were anyway effectively controlling tumor growth in these experiments. Further analysis of TILs from these mice indicated that the predominant effect of PD-1 mAb on Ptpn22KO TILs was to reduce surface PD-1 expression without significantly altering expression of other IRs such as TIM-3 (figure 4D) or infiltration and persistence in tumors (figure 4E). These data together provide further evidence that Ptpn22KO CTLs become dysfunctional secondary to chronic antigen exposure in the TME, and that this dysfunction is reversible, although partly, by blockade of the PD-1 axis.

TIM-3 inhibition worsens Ptpn22KO CTL tumor control

Alongside PD-1 expression, we noted that TIM-3 abundance was markedly increased on Ptpn22KO CTL (figures 1F and 2D,E). TIM-3 (encoded by Havcr2) is postulated to be an IR, although its precise role and signaling is less well defined than that of more classic IRs such as PD-1. The majority of evidence, particularly from human tumor data, points to an inhibitory role for TIM-3,24–28 and it is well documented that TIM-3 marks terminally exhausted cells in chronic viral infections and cancer.16 19 20 23 With this in mind and given the significant improvement by PD-1 blockade of Ptpn22KO CTL tumor control, we investigated the consequence of TIM-3 inhibition in Ptpn22KO cells. Ptpn22WT cells were not included in these analyses as they expressed very little TIM-3.

Electronic gating of FACS plots showed that Ptpn22KO CTL with highest TIM-3 expression had highest cytokine production in vitro, consistent with greater effector potential (figure 5A,B). We next sought to block TIM-3 signaling to investigate its role in Ptpn22KO CTL. We used CRISPR-Cas9 to knockout Havcr2 in activated Ptpn22KO cells before differentiating them to CTL and performing functional assays in vitro (figure 5C). Two independent CRISPR guides were tested which gave similar KO efficiencies (figure 5D). In response to restimulation with antigen for 4 hours, we found that loss of TIM-3 did not significantly alter CTL function in terms of cytotoxicity (figure 5E) or cytokine production (figure 5F).

Figure 5

TIM-3 expression does not impair PTPN22KO CTL function (A) TNF and IFNγ production by Ptpn22KO CTL following 4 hours restimulation with 10nM T4 antigen. Cells were gated on single live cells. Numbers in dot plot quadrants are proportions. Data are representative of three independent experiments. (B) Cytokine production by Ptpn22KO CTL following 4 hours restimulation with 10nM T4 antigen. Data are pooled from three independent experiments. P values as determined by paired t-test. (C) Schematic of experimental design. Naïve OT-I Ptpn22KO T cells were activated with N4 (10 nM) for 48 hours, then CRISPR-Cas9 was used to delete Havcr2, before cells were cultured in IL-2 (20 ng/mL) for 4 days to induce differentiation to CTL. (D) Representative histograms showing TIM-3 expression on non-restimulated Ptpn22KO CTL on day 6. Two independent guide RNA were tested (KO1 and KO2) in separate populations of cells. Untr; untransfected. (E) Cytotoxicity of Havcr2KO KO or mock transfected Ptpn22KO CTL against MC38 tumor cells expressing T4 antigen, at ratios indicated. Data are representative of two independent experiments. Bars represent mean±SEM. (F) Cytokine production from Ptpn22KO Havcr2KO and Ptpn22KO untransfected (Havcr2WT) CTL after 4 hours restimulation with T4 at indicated concentrations. Data are representative of three independent experiments. Bars represent mean±SD. (G) Representative dot plots showing cytokine production by Ptpn22KO Havcr2+/+ and Ptpn22KO Havcr2KO CTL after 4 hours restimulation at day 6 and chronic (d15) restimulation. Following Havcr2 KO on day 2 (figure 5C), cells were differentiated to CTL and repeatedly restimulated with antigen-bearing tumor cells as in figure 2A. Cells were gated on single, live, CD8+. Numbers in quadrants represent proportions. Data are representative of two independent experiments. (H) Schematic of experiment. 0.5×106 MC38-T4 tumor cells were injected at d0, followed by ACT with 1×106Ptpn22KO control or Ptpn22KO Havcr2KO CTL at day 4. Groups of mice that had received control T cells were given anti-TIM-3 or isotype control Ab on d7, d14, and d21. (I) Tumor growth in groups as in (H). n=5–10 mice per group in experiment shown. Data are representative of two independent experiments (total=10–20 per group). ACT, adoptive cell therapy; MFI, median fluorescence intensity.

We asked whether an impact of TIM-3 signaling might only be revealed over a more prolonged period or in response to chronic TCR stimulation, such as in the TME. First, we tested Havcr2KO CTL in chronic restimulation assays in vitro (as indicated in figure 2A). However, Havcr2KO/Ptpn22KO and Havcr2WT/Ptpn22KO CTL developed an equivalent loss of function following chronic antigen exposure (figure 5G). Second, to establish whether there was an influence of TIM-3 on Ptpn22KO CTL function in tumors, we adoptively transferred Havcr2WT or Havcr2KO Ptpn22KO CTL into mice with MC38-T4 tumors. Additionally, for a group of mice that received control (Havcr2WT) cells, we administered TIM-3 blocking or isotype control monoclonal antibody (figure 5H). Strikingly, loss of TIM-3 either via knockout or mAb-mediated blockade impaired control of tumors by Ptpn22KO CTL (figure 5I), suggesting a positive benefit from the presence of TIM-3 on these dysfunctional cells in preserving some control against tumors. Together these data suggest that the high expression of TIM-3 on Ptpn22KO CTL is no