Tumor hypoxia drives suppressor function in exhausted T cells limiting antitumor immunity

While CD8 + cytotoxic T cells are clearly critical for identification and elimination of cancer cells, 15 factors concentrated within the tumor microenvironment drive their altered differentiation to a 16 hypofunctional, short-lived state termed exhaustion 1 . Exhaustion is a progressive lineage, and it 17 is now clear that the most terminally exhausted T (tT exh ) cells are not the targets of checkpoint 18 blockade immunotherapy but rather serve as factors that limit immunotherapeutic efficacy 2–4 . Compared directly, tumor-infiltrating CD8 + tT exh cells bear notable phenotypic similarity to CD4 + Foxp3 + regulatory T (T reg ) cells, suggesting beyond loss of proinflammatory function, tT exh 21 cells may be directly anti-functional and constrain tumor specific immunity. Here we show, when 22 isolated from the same tumor microenvironment, terminally exhausted CD8 + T cells carried 23 similar capacity to suppress T cell proliferation as CD4 + Foxp3 + T reg cells. Unlike T reg cells, tT exh cells suppress solely via the ectonucleotidase CD39, which serves to deplete extracellular ATP (eATP) and produce an adenosinergic environment. CD39 expression in tT exh cells is driven by 26 exposure to tumor hypoxia, such that therapeutic targeting of hypoxia limits tT exh suppression

T cell exhaustion is a terminally differentiated state defined by low potential for self-renewal, 37 loss of polyfunctionality, and an absence of progression to memory 1 . Terminal exhaustion is 38 driven by continuous T cell receptor (TCR) signaling and metabolic stress and is characterized 39 by expression of the transcription factors (Tox, Blimp, and Eomesodermin) and co-inhibitory 40 receptors (PD-1, Lag-3, Tim-3, and Tigit) 5 . While it was once hypothesized that checkpoint 41 blockade therapy could restore function to terminally exhausted CD8 + T (tT exh ) cells, we now 42 understand PD-1 blockade preferentially acts upon less differentiated progenitor exhausted T 43 (pT exh ) cells (PD-1 + Tim-3 -) 2-4,6-8 and improves their functionality. Numerous reports have 44 2 associated tT exh cell enrichment with poor patient outcomes and predictive of resistance to 45 immunotherapy 9,10 . In immunotherapy (IMT)-resistant murine tumor models (Fig. 1a) and 46 treatment naïve melanoma samples from patients (Fig. 1b), PD-1 + Tim-3 + tT exh cells are 47 numerically enriched, and significantly outnumber other T cell populations, including regulatory 48 T cells (Fig. 1c-d; Extended Fig. 1a). Notably, tT exh cells express several lineage-defining 49 factors in common with T reg cells ( Fig. 1e; Extended Fig. 1b). While murine tT exh cells do not 50 significantly upregulate Foxp3 ( Fig. 1e; Extended Fig. 1b) Fig. 1c). These putative suppressor populations were 66 then co-cultured at diminishing ratios with proliferation dye-labeled, congenically mismatched 67 naïve CD4 + T cells (responder T cells) and T cell-depleted splenocytes coated with anti-CD3.

68
Strikingly, on a per-cell basis, tT exh cells and T reg cells equivalently inhibit responding T cell 69 proliferation in ex vivo microsuppression assays (Fig. 1f) whereas CD8 + PD-1 + Tim-3 -pT exh cells 70 ( Fig. 1f) and CD8 + PD-1 -T cells (data not shown) did not, suggesting tT exh cells harbored a 71 unique regulatory functionality among CD8 + TIL. To validate these findings, we repeated these 72 data in another melanoma line, clone 24, derived from the autochthonous melanoma mouse 73 model, Pten f/f Braf LSL.V600E Tyr2 Cre.ER 12,13 . Clone 24, like B16-F10, fails to respond to single agent 74 checkpoint blockade therapy and aggressively progresses in vivo. Clone 24-infiltrating tT exh cells 75 and T reg cells, like those sorted from B16-F10 tumors, were equivalently suppressive ex vivo, 76 while the more stem-like pT exh cells were not (Extended Fig. 1d). These findings of suppressor 77 functions in tT exh cells were also repeated in size-matched anti-PD-1 therapy-resistant head and 78 neck carcinoma, MEER PD1res , which we have recently reported to produce highly suppressive 79 T reg cells ( Fig. 1g; Extended Fig. 1c) 14 . However, when isolated from the well-characterized 80 anti-PD-1-sensitive colorectal carcinoma line MC38, tT exh cells failed to control responding T cell 81 effector functions ( Fig. 1h; Extended Fig. 2e). Notably, T reg cells infiltrating MC38 tumors 82 retained suppressor capacity (Fig. 1h). These data suggest co-inhibitory receptor expression is CD4 + Foxp3 + T reg cells are potent suppressors that employ multiple, sometimes redundant 91 means to suppress immune function in the local environment. Thus, we sought to dissect which 92 inhibitory mechanism(s) were employed by terminally exhausted CD8 + T cells. Several reports 93 have noted CD8 + T cell-derived IL-10 promotes autoimmune disease 15 , chronic viral infection 6 , 94 and cancer 16,17 . Therefore, we initially interrogated the contribution of tT exh cell-secreted IL-10 by 95 directly comparing B16-F10-derived tT exh cells from either wild-type or IL-10 deficient C57/BL6 96 mice (Il10 -/-). Though baseline tumor growth was unaffected by the germline absence of IL-10 97 (Extended Fig. 2a), TIL from IL-10-deficient mice displayed a less exhausted signature 98 (Extended Fig. 2c), consistent with previous reports that have shown IL-10 can enforce T cell 99 exhaustion 13,17 . Despite this, purified IL-10-deficient PD-1 + Tim-3 + tT exh cells retained significant 100 suppressor functionality compared to their corresponding pT exh cell counterparts ( Fig. 2a-b). In 101 concurrent experiments, titration of a well characterized IL-10 neutralizing antibody 18 failed to 102 prevent suppression by tT exh cells (Extended Fig. 2c).

104
We next investigated whether tT exh cell cytotoxicity could explain their suppressor activity. Host 105 antigen specific CD8 + T cells have been shown to induce tolerance in autoimmune disease 106 through granzyme-and perforin-mediated lysis of autoreactive CD4 T cells 19,20 . Indeed, 107 cytotoxicity is also a critical feature of CD4 + Foxp3 + T reg cell suppression 21,22 . Tumor-resident 108 tT exh cells maintain elevated expression of multiple cytolytic effector molecules-granzymes, 109 perforins, and FasL. Thus, we interrogated the role for tT exh cell cytolysis by deriving the 110 responder pool from the well-established model of apoptosis-resistance via Bcl2 overexpression 111 (Bcl2 tg ). Bcl2 overexpression did not prevent responder T cell suppression by tT exh cells, in fact, 112 by improving cell viability (Extended Fig. 2d), Bcl2 tg T cells were more sensitive to tT exh cell 113 suppression (Fig. 2c). These data suggest tT exh suppression was not mediated through direct 114 cytotoxicity of the responding T cells. However, it is well established that tT exh cells are 115 themselves apoptosis prone 3,5,6 and indeed suppressor tT exh cells from B16-F10 have higher ex

119
Evidence suggests CD4 + Foxp3 + T reg cells not only maintain suppressive capabilities as they 120 undergo cell death but suppression can be augmented by the process 23 . Thus, tT exh cells may 121 not suppress local immunity by killing, but rather further exert their regulatory capabilities by 122 dying.

124
Apoptosis augments T reg cell suppressive function through the release and subsequent 125 CD39/CD73-mediated hydrolysis of extracellular ATP (eATP) to adenosine 23-26 . CD39 has 126 previously been shown to faithfully mark the most terminally exhausted CD8 + T cells [27][28][29] , and in 127 our system, higher CD39 expression corresponds to increased ex vivo suppression (9.64% 128 increase per 0.1 log CD39 MFI; ***p < 0.001) (Fig. 2e). However, unlike regulatory T cells, CD8 + 129 tT exh cells do not co-express CD73, and during differentiation to exhaustion 'switch' from being 130 CD73 + to CD39 + (Fig. 2f, Extended Fig. 3a). Indeed, CD39 expression in peripheral and tumor-131 infiltrating CD8 + T cells is restricted to tT exh cells (Extended Fig. 3b). Beyond the   132   4   immunoregulatory effects of adenosine, CD39 may also restrict T cell function by limiting   133   available eATP, aids efficient T cell activation 30-32 . Thus, while tT exh cells may not be capable of   134   producing adenosine in isolation, as they are CD73 lo , elevated CD39 expression may contribute   135   to immune suppression either by limiting eATP or by supporting an adenosinergic environment   136 with CD73 + cells. Indeed, tT exh failed to mediate suppressor activity when the suppression assay 137 was conducted using stimulatory magnetic beads rather than live or fixed antigen presenting 138 cells (Extended Fig. 3c), suggesting interaction with local immune cells may be critical to tT exh 139 cell suppression.

141
To determine whether local CD73 expression facilitates tT exh cell suppression, we assayed 142 responder T cell and APC populations from CD73-deficient (Nt5e -/-) mice in microsuppression 143 experiments. Indeed, CD73-deficient T cells resist tT exh cell-mediated suppression, supporting 144 the notion that local adenosine production is at least partially responsible for the observed 145 regulatory role of tT exh cells (Extended Fig. 3d). Concurrently, we treated our microsuppression 146 assays with a small molecule inhibitor specific to adenosine receptors, A2AR and A2BR 166 167

169
We next sought to better understand the mechanism by which CD39 may be limiting anti-tumor 170 immunity with a gain-of-function approach. We retrovirally overexpressed CD39 on polyclonal,  Fig. 2i-j). Metabolic reprogramming to glycolysis 40 (Extended Fig. 4c-f) and calcium flux 180 (Extended Fig. 4g) were also diminished in CD39-overexpressing cells, suggesting local CD39 181 expression was impacting TCR signaling. Further, when restimulated by anti-CD3/anti-CD28 182 antibody-bound microbeads, CD39-overexpresing T cells secreted significantly less interferon 183 gamma (IFNγ) and tumor necrosis factor (TNF) (Extended Fig. 4h). Notably, CD39-184 overexpressing T cells were not inherently deficient in cytokine production, as TCR signaling-185 independent activation of these cells by phorbol 12-myristate 13-acetate (PMA) and ionomycin 186 produced equivalent cytokine production as to controls (Extended Fig. 4i).  (Fig. 2m). However, wild-type T cells cultured in the presence of CD39-194 overexpressing T cells also suffered a loss of cytokine and proliferation following TCR 195 stimulation ( Fig. 2k-m). These data suggest CD39 is sufficient to limit effector functions both 196 within the cell expressing CD39, but also those in the local environment. response, resulting in 43% partial response and 14% complete remission and a prolonging 7 survival of 13 days (improvement of 72%) from baseline ( Fig. 4a-c). To understand the 264 functional consequence of CD39 deletion in this system, we isolated TIL from B16-F10 tumors 265 after three consecutive days of TAM treatment (day 8 post-tumor inoculation), when tumors 266 were still similarly sized (~5x5 mm 2 ). At this timepoint, CD39 expression was significantly 267 depleted in E8i Cre.ERT2 Entpd1 f/f TIL as compared to controls (Fig. 4d) and consequentially 268 displayed a less exhausted (Fig. 4e,f) and more effector-like signature (Fig. 4g). Indeed, even 269 at this early timepoint, E8i Cre.ERT2 Entpd1 f/f TIL secreted significantly more proinflammatory 270 cytokines, IL-2 and interferon gamma (IFNγ) upon ex vivo restimulation (Fig. 4g)

344
Tissue Culture and T cell Isolation