Antitumor T-cell Homeostatic Activation Is Uncoupled from Homeostatic Inhibition by Checkpoint Blockade

Lymphodepletive therapies, such as autologous bone marrow transplant (BMT) following high-dose chemotherapy, improve survival in patients with lymphoma and are standard therapy for relapsed and/or refractory disease (1). Despite this, aggressive lymphomas are incurable in ∼10,000 people annually, indicating the need for novel therapies. Although some lymphomas are sensitive to T-cell killing, as exemplified by the unprecedented efficacy of PD-1 blockade in Hodgkin lymphoma (2), checkpoint blockade has largely been ineffective in non-Hodgkin lymphomas (NHL). Even anti–PD-1/anti-CTLA4 dual checkpoint blockade (dCB) has yielded limited efficacy (complete response rate 0%, median survival < 3 months; ref. 3) despite higher intratumoral expression of response predictors (e.g., PD-L1, CD8, and IFNγ) in NHL compared with responsive tumor types (4). This inefficacy may be due to insufficient T-cell activation. Previously, we improved checkpoint blockade efficacy by cross-priming anti-lymphoma T cells (5) to enhance their T-cell receptor (TCR)–induced activation.

Homeostatic activation is triggered in mature T cells upon their transfer into a lymphodepleted recipient and their increased access to common gamma chain cytokine family members IL7 and IL15, promoting antitumor responses (6). In contrast to TCR-mediated activation, which occurs via ZAP70/LCK/LAT/MAPK/ERK signaling (7), homeostatic activation is mediated by cytokine receptor signaling via JAK/STAT (8), resulting in distinct states of activation and exhaustion (9). Lymphodepletion is necessary for the efficacy of cellular therapies and has been incorporated into T-cell transfer therapies, including chimeric antigen receptor T-cell (CAR-T; ref. 10), tumor-infiltrating lymphocyte (11), and transgenic TCR therapies (12) as well as allogeneic BMT (13).

Both canonical TCR-mediated activation and homeostatic activation of T cells result in proliferation and memory induction (14). Canonical TCR-mediated activation and homeostatic activation share some common features but also numerous differences, e.g., homeostatic activation does not induce upregulation of CD25, CD69, and CD71 (15, 16) or downregulation of CD49d (17), CD62L (16), or CD127 (15). Although TCR-mediated activation increases PD-1 expression, other T-cell activators such as IL12 actually decrease PD-1 expression (18).

The effects of homeostatic activation on expression of checkpoint molecules and response to checkpoint ligands have not been well studied. Broadly, the regulation of homeostatic activation, the expression of checkpoint molecules on transferred T cells, and the effect of checkpoint blockade in this setting are poorly understood. Here, we show that transfer of both murine and human CD8⁺ T cells into lymphodepleted recipients induces not only their activation but also expression of functional checkpoint molecules, including CLTA4 and PD-1—we term this homeostatic inhibition.

Common gamma chain (γ_c) cytokine family members IL7 and IL15 are known to induce homeostatic proliferation (19, 20); we further show that homeostatic inhibition is induced by IL7 and IL15 and JAK/STAT signaling. Although homeostatic activation and inhibition are causally linked by the γ_c cytokines, we hypothesized that they could be effectively uncoupled by dCB. To test our hypothesis, we investigated the unconventional use of dCB during BMT. Our studies indicate that dCB protects transplanted T cells from inhibitory signaling, and potentiates their homeostatic activation. This increased T-cell activation from the combination of BMT and dCB—termed “immunotransplant”—significantly amplified antitumor immune responses, yielding durable tumor regressions in lymphoma and solid tumor models, even when dCB alone yielded no apparent antitumor effect. These findings both reveal a novel T-cell regulatory mechanism and suggest a therapeutic approach for checkpoint-refractory tumors, now being studied in patients with aggressive NHL (NCT03305445).

Standard autologous BMT reinfuses patients' peripheral blood mononuclear cells (PBMC) with mobilized stem cells after high-dose chemotherapy. To determine the effects of lymphodepletion on patient T cells, PBMCs were isolated from patients prior to receiving BEAM (BCNU, etoposide, cytarabine, melphalan) chemotherapy (i.e., pretransplant) and 7 to 10 days after autologous BMT (i.e., post-transplant). As expected, transfer of T cells into lymphodepleted recipients led to activation and proliferation of the CD4⁺ and CD8⁺ T cells, as indicated by increased expression of the homeostatic activation markers CD44 and CD122/IL2Rβ (21)—and staining of the proliferation marker Ki-67 (Fig. 1A; Supplementary Fig. S1A). Interestingly, there were also 2- to 6-fold higher levels of PD-1 and CTLA4 after reinfusion and expansion after lymphodepletion (Fig. 1B; Supplementary Fig. S1B).

Expression of PD-1 and CTLA4 on T cells can mark activated or dysfunctional cells. To determine whether the increased checkpoint expression on post-transplant T cells was functionally suppressive, we assessed T-cell stimulation with concurrent ligation of PD-1 or CTLA4 by plate-bound PD-L1 or CD80. In the presence of checkpoint ligands, the stimulation of post-transplant CD8⁺ T cells, per their expression of activation markers and cytokines, was diminished as compared with matched pretransplant samples (Fig. 1C and D). We termed this transfer-induced, checkpoint-mediated suppression “homeostatic inhibition.” Interestingly, transferred CD4⁺ T cells did not show the same diminished activation in the presence of PD-L1 or CD80 despite upregulation of PD-1 and CTLA4 (Supplementary Fig. S1C and S1D).

As an alternate source of checkpoint ligands than plate-bound PD-L1 and CD80, we developed a cell-based assay using staphylococcal enterotoxin B (SEB) to aggregate antigen-presenting cells (APC) with T cells and activate the latter. Here, SEB binding to T cell–expressed TCR and APC-expressed MHC molecules colocalizes T-cell checkpoints with their ligands (e.g., PD-L1 and CD80; ref. 22). Pre- and posttransplant PBMCs were incubated with dCB, then stimulated with SEB and assessed for activation per their expression of IFNγ and CD25. As previously shown, homeostatic activation lowers the threshold for TCR activation (23). Post-transplant CD8⁺ T cells were significantly more activated (CD25⁺) by SEB than pretransplant CD8⁺ T cells (Fig. 1E left, pre- vs. post-transplant SEB, 27% vs. 80%).

Surprisingly, although pretransplant CD8⁺ T cells were activated with SEB, checkpoint blockade did not enhance this activation (Fig. 1E, left, SEB vs. SEB⁺dCB, 27% vs. 29%). By contrast, post-transplant CD8⁺ T-cell activation by SEB was markedly enhanced by dCB (Fig. 1E, left, 80% to 90% activated and 7% to 23% highly activated, IFNγ⁺), with aggregate data showing this benefit across the cohort (Fig. 1E, right, paired t test, P < 0.05). The greater activation of post-transplant CD8⁺ T cells with checkpoint blockade is consistent with the above-described increased expression of checkpoints and increased sensitivity to checkpoint ligation. Additionally, neither pre- nor post-transplant CD4⁺ T-cell activation significantly increased in response to dCB (Supplementary Fig. S1E).

Taken together, these data suggest that—similar to canonical TCR-mediated activation—homeostatic activation is coupled to checkpoint-mediated homeostatic inhibition, but these can be effectively uncoupled by checkpoint blockade. Furthermore, these data suggest that dCB therapy may be significantly more (or exclusively) effective in the setting of T-cell transfer into lymphodepleted recipients than as a stand-alone therapy.

To further determine the relationship between homeostatic activation and homeostatic inhibition, we used a mouse model of syngeneic bone marrow and splenocyte transfer into lymphodepleted recipients. Mice received total body irradiation (TBI) followed by cell transfer. Control recipients received syngeneic splenocytes without TBI (denoted “No Rx”). As observed in the patients, transfer into lymphodepleted recipients induced a phenotype indicating both homeostatic activation and homeostatic inhibition of the CD8⁺ T cells (Fig. 2A and B). This was indicated by simultaneous upregulation of CD44, IL2Rβ, and Ki-67 along with the inhibitory checkpoints PD-1 and CTLA4. Because only CD8⁺ T cells evidenced transfer-induced homeostatic inhibition that could be reversed by dCB, we focused subsequent experiments on CD8⁺ T cells.

To determine functional effects of increased checkpoint expression on post-transplant T cells, we assessed T-cell stimulation with concurrent ligation of PD-1 or CTLA4 by plate-bound PD-L1 or CTLA4. As in our patients, the stimulation of post-BMT CD8⁺ T cells was enhanced as compared with the control cohort (Fig. 2C, solid red vs. solid black). PD-L1 only inhibited CD8⁺ T-cell cytokine production from BMT recipients (Fig. 2C, solid red vs. checkered red). Similarly, activation of CD8⁺ T cells from BMT-treated mice was inhibited in the presence of CD80 (Fig. 2C, right). Because PD-1 induces BIM-mediated T-cell apoptosis (24), we assessed activation-induced cell death and observed that only post-BMT CD8⁺ T cells increased transition into early apoptosis in the presence of PD-L1 (Supplementary Fig. S2A, red solid vs. red checkered).

To test the ability of dCB to protect against homeostatic inhibition, mice were treated in vivo with dCB before (donor) and after (recipient) BMT—a combination approach we termed immunotransplant (IT). CD8⁺ T cells from IT recipients were not suppressed by exposure to PD-L1 and did not undergo apoptosis (Fig. 2C; Supplementary Fig. S2A). In vivo dCB treatment also prevented suppression and partially restored activation in the presence of plate-bound CD80 (Fig. 2C, right). These data suggest that checkpoint blockade before and after transfer into lymphodepleted recipients (IT) protects transferred CD8⁺ T cells from checkpoint-induced suppression. By uncoupling homeostatic inhibition from homeostatic activation, IT significantly increased the latter, as CD8⁺ T cells expressed higher levels of CD44, IL2Rβ, and Ki-67 post-IT versus post-BMT (Supplementary Fig. S2B). Thus, IT reveals the potential magnitude of uninhibited homeostatic activation.

Next, we determined whether IT would also enhance transfer of antigen-specific CD8⁺ T cells into a lymphodepleted host. IT was performed using donor mice expressing TCR specific for GFP; wild-type (WT) mice were irradiated and subsequently received syngeneic GFP-specific T cells, which recognize the immunodominant epitope of GFP (GFP_200-208) presented on H-2Kd (25, 26) either alone or with dCB. Recipient splenocytes were stimulated in vitro with GFP-expressing A20 lymphoma cells and assessed for IFNγ production, shown to be necessary for tumor cytolysis (27). Antigen-specific CD8⁺ T-cell activation was significantly greater in IT recipients than in mice receiving BMT or dCB alone (Fig. 2D). Mirroring the poor response to dCB seen in human NHL, dCB treatment in the absence of transfer into lymphodepleted mice (“dCB only”) did not increase the constitutive or antigen-specific activation of CD8⁺ T cells (Supplementary Fig. S2B and Fig. 2D). These data establish that in vivo, dCB is especially beneficial to CD8⁺ T cells upon transfer to the lymphodepleted recipient, as seen in vitro with patient samples.

To determine whether the increased CD8⁺ T-cell activation induced by IT could increase antitumor immunity, we developed an IT tumor model. A20 tumor–bearing donor mice were treated with dCB before splenocytes and bone marrow were harvested, splenocytes were transferred to A20 tumor–bearing TBI-treated syngeneic recipient mice, and recipients received additional dCB after transfer. This therapeutic IT was compared with BMT—the same procedure without the dCB treatment—as well as “dCB only” and “No Rx” cohorts, which received no irradiation followed by bone marrow and splenocyte transfer.

IT therapy induced tumor regressions in the majority of recipients, whereas BMT or dCB alone induced only transient growth delay (Fig. 3A, top). Although dCB and BMT yielded durable remissions in 10% of recipients, IT significantly prolonged survival and yielded durable remissions in 40% (Fig. 3A, bottom), i.e., without evidence of disease past 60 days. These data demonstrate that IT possesses additive or synergistic antitumor efficacy over its components, BMT and dCB, with kinetics suggesting that the effect is mediated by transferred immune cells.

To measure tumor-specific T-cell responses, recipient peripheral blood lymphocytes were stimulated in vitro with A20 lymphoma cells and assessed for activation. The proportion of tumor-specific CD8⁺ T cells (IFNγ production in A20 coculture) was significantly higher in IT-treated mice than in other cohorts and ∼10-fold higher than in mice receiving dCB alone (Fig. 3B).

Tumors from each cohort were harvested and tumor infiltrates were assessed by immunofluorescence, revealing marked, uniform infiltration of intratumoral CD8⁺ T cells in IT mice (Fig. 3C, top). To better understand the intratumoral immune repertoire, we performed time-of-flight mass cytometry (CyTOF) using a broad multiplex panel focused on T-cell activation state. Compared with dCB alone, both BMT and IT recipients demonstrated a ∼6- to 7-fold increase in the proportion of intratumoral PD-1^hiCD8⁺ T cells, a surrogate of tumor antigen–specific T cells (ref. 28; Fig. 3C, bottom).

Although BMT and IT recipients had similar increases in intratumoral PD-1^hiCD8⁺ T cells, there were marked differences in the activation state between these cohorts. In BMT recipients, these T cells had an exhausted phenotype, per their lower expression of CD122/IL2Rβ (29) and higher expression of TIM3 (ref. 30; Fig. 3D, right); however, their exhausted phenotype is distinct from conventional antigen/TCR-driven terminally exhausted T cells. These “progenitor-like” T-bet^hi cells (31) are more responsive to checkpoint blockade (32). In IT recipients, we observed reversal of exhaustion, as PD-1^hiCD8⁺ T-cell proliferation (quantified by Ki-67) markedly increased, a correlate of antitumor efficacy in preclinical models and patients (33). This T-cell subset in IT recipients also expressed significantly higher levels of T-bet, denoting residual potential for reinvigoration (34), proliferation (31), and salvage by checkpoint blockade (35).

The efficacy of IT was tested in additional lymphoid and solid tumor models. Mice bearing B16 melanoma (Fig. 3E, top), KLN205 lung squamous cell carcinoma (Fig. 3E, middle), or EL4 T-cell lymphoma tumors (Fig. 3E, bottom) were treated as described above. IT induced regression and improved survival in all tumor models more effectively than dCB or BMT. In some models, such as lung carcinoma, IT yielded durable remissions of established tumors. As noted by others (36), even dCB therapy was minimally effective in all tumor models tested. Therefore, the significantly improved antitumor effect of IT in treating these established tumors appears to be greater than additive, suggesting a cooperative immune effect in combining dCB and T-cell transfer into lymphodepleted recipients.

To determine whether the antitumor efficacy of IT requires blockade of both PD-1 and CTLA4, cohorts were treated with BMT and IT as in Fig. 3 with additional cohorts in which IT was performed using only anti-CTLA4 (IT αCTLA4) or anti–PD-1 (IT αPD-1). Mice that received the dCB and transfer (IT) showed significantly greater tumor regressions and improved survival than in either of the monotherapies (Supplementary Fig. S3A). Furthermore, recipient peripheral blood lymphocytes from each cohort were stimulated in vitro with A20 lymphoma cells as described above. Recipients of dCB-treated IT had a 3- to 4-fold greater proportion of tumor-specific CD8⁺ T cells than recipients of IT with either single checkpoint blockade (Supplementary Fig. S3B; IT compared with BMT or IT αPD-1). These data show that dCB is necessary for the superior antitumor immunity observed with IT.

Checkpoint blockade therapy, especially dCB, can induce high-grade autoimmune toxicities in both patients and mice (37), including hepatoxicity, which has been among the most common of these (38). Because IT increased antitumor immune responses, we assessed whether it also increased these toxicities. Sera from mice from each treatment cohort were assessed for liver-enzyme levels. All three treatment cohorts (dCB, BMT, and IT) showed mild enzyme elevations, though results in the IT cohort were similar to or better than those in the dCB cohort and similar to described effects in combination immunotherapy studies (Supplementary Fig. S3C). Likewise, although weight loss was seen in BMT and IT cohorts, results with IT were similar to those with the BMT cohort (Supplementary Fig. S3D) and similar to the effects described in syngeneic BMT studies (39) attributed to TBI.

To determine whether the primary role of TBI in the IT model is its lymphodepletive effect and not its direct antitumoreffects, treatments were repeated but with recipient tumor challenge 2 days after irradiation, such that tumors were not exposed to TBI. In this format, IT recipients experienced transient tumor growth followed by regression, and ∼80% experienced durable remissions (i.e., without evidence of disease past 60 days). By contrast, there was no significant survival prolongation with either BMT or dCB alone (Fig. 4A). Notably, the tumor regressions observed in IT recipients occurred in a time course consistent with the homeostatic proliferation/activation observed in earlier experiments (Fig. 2A; Supplementary Fig. S2B). These data confirm that the IT antitumor effect is not a result of increased tumor radiosensitivity but is primarily immune-mediated.

To determine cellular and molecular effectors of the antitumor immune response, IT recipients were depleted of CD8⁺ T cells (IT + αCD8), CD4⁺ T cells (IT + αCD4), or IFNγ (IT + αIFNγ) using depleting antibodies before transfer and continuously thereafter. IT antitumor immunity was diminished by CD8⁺ T-cell and IFNγ depletion but not CD4⁺ T-cell depletion (Fig. 4B). Because recipient lymphodepletion induced homeostatic activation of transferred CD8⁺ T cells (Fig. 2), we also confirmed that lymphodepletion was necessary for the observed antitumor effect. Mice were treated with IT, but recipient mice were not irradiated (IT + 0 Gy), which completely abrogated the antitumor effect (Fig. 4C; “IT + 0 Gy” orange line n.s. vs. “No Rx” black line). The effectiveness of IT was also tested by substituting lymphodepleting chemotherapy for myeloablative TBI, using fludarabine and cyclophosphamide (Flu/Cy) as in clinical cellular therapies. There was no significant difference between Flu/Cy and TBI in IT (Fig. 4C; “IT + Flu/Cy” brown line n.s. vs. “IT” purple line), suggesting that IT antitumor immunity may depend more on recipient lymphodepletion than on other effects of TBI, e.g., Toll-like receptor (TLR) agonism related to intestinal irradiation (40).

Due to the rapidity of tumor regressions noted in IT recipients (Fig. 3A), we hypothesized that the tumor-specific immune response required priming in the donor. To test this, donor mice with or without tumors were treated with dCB as described above prior to transfer. The antitumor immunity of IT was abolished by using tumor-free donors (Fig. 4C; “Donor No Tumor” blue line n.s. vs. “No Rx” black line).

To determine the durability of the antitumor immune response, IT recipients that had cleared the tumor were rechallenged with A20 lymphoma or 4T1 breast cancer cells. Recipients were completely protected from A20 rechallenge (Fig. 4D; IT + A20) but not from 4T1 (Fig. 4D; IT + 4T1), indicative of a tumor-specific memory immune response. Taken together, these data show that the antitumor immunity of IT is durable, CD8- and IFNγ-dependent, and requires priming of a donor antitumor response and recipient lymphodepletion.

As IT appears to modulate the peripheral and intratumoral CD8⁺ T-cell activation state, and because a broad range of cytokines regulate T-cell activation, we determined whether the increased homeostatic and tumor-specific activation might be cytokine-regulated. To evaluate differences in cytokine levels, serum concentration of 30 cytokines was assessed using a bead-based multiplex assay in mice of each treatment cohort. Of the 30 tested (Supplementary Fig. S4A), the only cytokines that differed significantly between IT and other cohorts were the γ_c family members IL2, IL7, and IL15, as well as TNFα (Supplementary Fig S4B and Fig. 5A). Remarkably, the increase in serum γ_c concentrations in the IT cohort was greater than additive relative to the effects of dCB plus BMT. For example, the mean serum concentration of IL2 was 4 pg/mL with dCB and 12.5 pg/mL with BMT but more than 98 pg/mL in the IT cohort. Serum γ_c in patients receiving BMT were consistent with (or even several-fold higher than; ref. 41) those observed in mice receiving BMT alone, suggesting that our observations may translate well to the clinic.

IL2, IL7, and IL15 induce signaling through a shared γ_c receptor and their respective alpha-receptor components, with IL2 and IL15 sharing a third component, IL2Rβ/CD122. As such, we examined differences in the expression of γ_c receptor subunits by flow cytometry. IT and BMT were performed using congenic donors; in these experiments, the dCB and No Rx control cohorts were also recipient mice that received congenic splenocyte transfer but without TBI. IT treatment induced significantly higher expression of IL15Rα, IL2Rβ, and γ_c on transferred CD8⁺ T cells as compared with the other cohorts, but not IL2Rα or IL7Rα (Fig. 5B).

Next, we determined whether increased γ_c receptor subunit expression correlated with downstream signaling and increased sensitivity to receptor ligation. IL2, IL7, and IL15 signal through the JAK–STAT pathway, primarily JAK1/3 and STAT5 (42). Recipient splenocytes from each cohort were stimulated with recombinant IL2, IL7, and IL15 in vitro, and signaling was measured per STAT5 phosphorylation. CD8⁺ T cells from IT recipients showed slightly higher basal levels of pSTAT5 and significantly higher levels of inducible pSTAT5 in response to all γ_c cytokines tested as compared with other cohorts (Fig. 5C). To determine whether increased γ_c cytokine signaling could potentiate TCR signaling, recipient splenocytes from each cohort were stimulated with suboptimal doses of αCD3ϵ plus individual γ_c cytokines in vitro (43). CD8⁺ T cells from IT-treated mice proliferated more robustly in response to IL2 and IL15 than in other cohorts (Fig. 5D). Of note, BMT recipients showed changes similar to, though lesser than, IT mice in these metrics (Fig. 5B–D). These data suggest that IT increases not only γ_c cytokine production, but also γ_c subunit receptors on CD8⁺ T cells and sensitivity to receptor ligation.

To determine whether the increased JAK/STAT signaling in IT recipients potentiates physiologic TCR activation (as opposed to CD3ϵ ligation), IT was performed as previously, using GFP-reactive CD8⁺ T-cell donors and in vitro stimulation with GFP-A20 lymphoma cells. As above, IT induced the greatest proportion of tumor-specific CD8⁺ T cells (Figs. 2D and Fig. 5E). However, when these CD8⁺ T cells were pretreated in vitro with the JAK3-specific inhibitor tofacitinib (44), the increased tumor antigen–specific response was abrogated (Fig. 5E). Conversely, when recipient splenocytes from each cohort were treated with a mix of IL2, IL7, and IL15, antigen-specific cytokine production increased significantly for CD8⁺ T cells from the other cohort groups, and those from the BMT cohort became comparable to IT (Fig. 5E). These results suggest that the amplified tumor-specific CD8⁺ T-cell responses observed with IT depend upon γ_c cytokine signaling.

To determine the importance of IL7 or IL15 signaling in enhancing the function of antitumor T cells induced by IT, IL7Rα^−/− or IL15Rα^−/− donor mice (and WT recipients) were used for IT, as previously described. If donor splenocytes lacked IL7Rα or IL15Rα, IT antitumor immunity was greatly reduced or abolished (Fig. 5F). These data confirm that γ_c cytokine signaling is critical in augmenting the antitumor immune response observed with IT.

To determine if molecular mechanisms critical to the antitumor effect of the preclinical IT model could be observed in patients, we measured γ_c receptor expression and function in our autologous BMT patient cohort (Fig. 5B–D). Common gamma chain receptor expression was assessed by flow cytometry on pre- and post-transplant PBMCs. Similar to BMT-treated mice, post-transplant CD8⁺ T cells expressed higher levels of IL15Rα, IL2Rβ, and γ_c receptors but not IL2Rα or IL7Rα (Fig. 6A).

Next, pre- and post-transplant PBMCs were stimulated with in vitro IL2, IL7, and IL15, and pSTAT5 was measured by flow cytometry. As seen in mice, post-transplant CD8⁺ T cells were more sensitive and responsive to γ_c cytokines, inducing significantly increased levels of pSTAT5 (Fig. 6B). Furthermore, post-transplant CD8⁺ T cells showed greater proliferation in response to IL2 and IL15 stimulation, with a similar trend for IL7 (Fig. 6C). These data confirm similar changes in γ_c cytokine and responses in patients and mice treated with BMT.

Given our initial finding of increased checkpoint molecules on post-BMT patient CD8⁺ T cells, we returned to more broadly assess their activating and inhibitory phenotype by mass cytometry. We observed that homeostatic inhibition may extend beyond PD-1/CTLA4 as marked upregulation of other CD8⁺ T-cell inhibitory checkpoints (e.g., TIGIT, TIM3, LAG3, and CD38) was also observed post-BMT (Fig. 6D, left). Though patient and murine T-cell panels differed, overall similar phenotypic changes were seen, with BMT inducing upregulation of TIGIT and CXCR3 and downregulation of CD103 on both patient and murine T cells (Fig. 6D, right), thus suggesting the similarities between our mouse model and patients if generally conserved.

To determine if the upregulation of PD-1 and CTLA4 that we observed is γ_c cytokine–dependent, we took advantage of genetic knockouts of the cytokine receptors. When splenocytes from IL7Rα^−/− and IL15Rα^−/− mice were transferred into TBI-treated WT recipient mice, upregulation of PD-1 and CTLA4 was significantly reduced on CD8⁺ T cells post-transplant (Fig. 7A). Furthermore, disruption of these receptors also abrogated CD8⁺ T-cell homeostatic proliferation and activation (Fig. 7B), suggesting that homeostatic activation and the concomitant homeostatic inhibition are γ_c cytokine–dependent.

Next, we further assessed mechanisms of this finding by using γ_c cytokine–treated T cells as a surrogate for lymphopenia-exposed T cells. Using healthy donor human PBMCs or naïve mouse splenocytes, we confirmed that in vitro treatment with IL2, IL7, or IL15 induces PD-1 expression on human and mouse CD8⁺ T cells (Fig. 7C), whereas IL2 and IL15 also induced CTLA4 expression on CD8⁺ T cells (Fig. 7C). Additionally, IL2 and IL15 induced expression of the homeostatic activation markers Ki-67, IL2Rβ, and CD44 on human CD8⁺ T cells, whereas all three γ_c cytokines induced Ki-67 and IL2Rβ expression on mouse CD8⁺ T cells (Supplementary Fig. S5A).

Signaling of γ_c cytokines is mediated primarily through JAK1/3, and it has been shown that JAK3 inhibition is sufficient to abrogate downstream STAT5 signaling in immune cell subsets (45). When γ_c cytokine–treated healthy human PBMCs or naïve mouse splenocytes were preincubated with the JAK3 inhibitor tofactinib, γ_c cytokine–induced PD-1 and CTLA4 upregulation were significantly inhibited (Fig. 7D), as was homoeostatic activation (Supplementary Fig. S5B).

To determine whether the PD-1 and CTLA4 induced on healthy human PBMCs by IL15 were functional, the cells were activated by ligating CD3 and CD28 with or without plate-bound PD-L1. The γ_c cytokine–induced PD-1^hiCD8⁺T cells were more apoptotic in the presence of PD-L1 (Supplementary Fig S5C). Similarly, in the presence of PD-L1 or CD80, the high PD-1– and CTLA4-expressing CD8⁺ T cells were less activated (Supplementary Fig. S5D). Taken together, these data suggest that transfer-induced homeostatic activation and homeostatic inhibition are largely dependent upon—and regulated by—γ_c cytokine signaling.

Homeostatic activation of T cells transferred into lymphodepleted recipients is routinely exploited in the lab and the clinic to amplify the antitumor efficacy of T-cell therapies (10–13); however, the concurrent homeostatic inhibition has not been well recognized or addressed. Here, we demonstrate that transfer into a lymphodepleted recipient induces functional PD-1 and CTLA4 (among numerous checkpoint molecules) expression in both patient and murine CD8⁺ T cells. This homeostatic inhibition can be reversed with dCB to reveal the underlying magnitude of homeostatic T-cell activation per improved T-cell proliferation, tumor-specific IFNγ production, and tumoricidal activity. Homeostatic activation combined with dCB yielded a synergistic increase in patient and murine T-cell function. dCB alone did not increase IFNγ production upon TCR ligation but did so only in the context of homeostatic activation. These data predict the in vivo results in lymphoma, melanoma, and lung cancer models, in which dCB yielded no significant benefit but IT induced tumor regressions and improved survival.

Homeostatic inhibition is—at least partly—regulated by γ_c cytokines and thereby mechanistically coupled to homeostatic activation (19, 20), i.e., IL7R/IL15R expression on donor T cells is required for PD-1 and CTLA4 upregulation and for the antitumor effects of IT. These data are consistent with prior results showing that IL7/IL15 increase after lymphodepletion (46) and mediate checkpoint blockade therapy antitumor efficacy (47).

Although it is well established that T-cell activation induces T cell–inhibitory signals, prior to this work the specific mechanisms of how homeostatic activation induces inhibitory signals were not well understood. Mechanistic coupling by γ_c cytokine signaling could not be presupposed. Because TBI also modulates transferred T cells by IL7/IL15-independent mechanisms (e.g., microbial activation of enteric TLR and host dendritic cell–produced IL12; ref. 48), it might have been the case that homeostatic activation is mediated by γ_c cytokines and homeostatic inhibition mediated by TLR agonism. Conversely, uncoupling of these processes by PD-1 and CTLA4 blockade could not be presupposed; numerous checkpoints are upregulated post-BMT including TIGIT, TIM3, LAG3, and CD38 (Fig. 6D), any of which might mediate T-cell inhibition. Although PD-1/CTLA4 blockade uncouples the competing effects of IL7/IL15 on transferred T cells, blocking these other checkpoints might further reverse homeostatic inhibition, one avenue of ongoing investigation.

Downstream, γ_c receptors signal through JAK1/3 and STAT5, known mediators of T-cell antitumor efficacy (49). Although IL7/IL15 can signal through JAK1 (50), notably, both the activating (Fig. 5E) and inhibitory (Fig. 7D) effects of γ_c cytokine signaling appeared to be JAK3-dependent, another level of mechanistic coupling. Further downstream, because JAK3 signals largely through STAT5 (51), it is noteworthy that both PD-1 and CTLA4 signaling have been implicated in STAT5 inhibition and sequestration (52, 53). Reversal of such sequestration is a potential mechanism for enhanced STAT5 signaling in IT (vs. BMT) recipients. We observed increased basal STAT5 phosphorylation in the T cells of IT recipients as well as increased γ_c cytokine sensitivity, indicating the interaction of checkpoint signaling with the JAK/STAT pathway. STAT5 regulates expression of Eomes and T-bet in T cells and thereby reversal of exhaustion (54) and cytotoxic function (55). It is likely that atypical “progenitor-like” T-bet^hi exhausted cells observed post-BMT are the product of a mechanistically distinct exhaustion downstream from different means of NFAT activation. In contrast to TCR-mediated exhaustion which results from calcineurin-dependent NFAT signaling (56), instead common γ-chain cytokines induce calcineurin-independent NFAT by means of a novel Y371 phosphorylation (57, 58). The favorable T-bet^hiEomes^lo phenotype (55) and markedly increased proliferative rate of intratumoral T cells in IT recipients suggest that, ultimately, using dCB to uncouple homeostatic inhibition from activation potentiates antitumor T-cell responses.

We propose a model (Supplementary Fig. S5E) whereby, on transfer to BMT recipients, IL7 and IL15 activate T cells, but also induce mediators (e.g., NFAT) that increase the transcription of PD-1, CTLA4, and other checkpoints. PD-1 and CTLA4 ligation promotes SHP2 inhibitory signaling. The resulting inhibitory signals prevent maximal homeostatic activation. Though blocking these γ_c cytokines would obviate both their activating and inhibitory effects, blocking PD-1/CTLA4 further downstream uncouples these effects. Checkpoint blockade in this setting (IT) relieves inhibition of both TCR signaling (e.g., decreasing SHP2 activity) as well as JAK/STAT signaling (e.g., decreasing STAT5 sequestration). The resulting JAK/STAT signaling likely induces T-bet (59) and thereby expression of homeostatic activation markers (IL2Rβ and CD44), cytokine (IL2 and IFNγ) production, and antitumor effects.

Our results connect prior preclinical and clinical observations. Kinter and colleagues observed increased T-cell PD-1 expression upon in vitro γ_c cytokine exposure (60), whereas Fortner and colleagues noted increased PD-1 and LAG3 transcripts in murine T cells after transfer into lymphocyte-deficient mice (61); these results were confirmed and expanded upon by our studies, but these studies did not determine the potential of checkpoint blockade in these settings. Kearl and colleagues used an approach comparable with our IT model but without CTLA4 blockade, myeloablative TBI, donor priming or T-cell transfer. Still, Kearl and colleagues' data demonstrated significant antitumor effects, though their therapy was insufficient for benefit in B16 or EL4 models (62), both of which were effectively treated in our IT model. Although Jing and colleagues included TBI in a dCB-based regimen, they did not evaluate the necessity of TBI or endeavor to understand the mechanisms that underlie the benefits of combining TBI and checkpoint blockade (63). Finally, Vu and colleagues observed increased expression of costimulatory molecules—OX40, ICOS, and CD137—on homeostatically activated T cells (64), complementing our observations of increased checkpoint molecule expression after homeostatic activation. Indeed, the basic concepts underlying our observations are not entirely novel; increased checkpoint expression post-BMT has previously been observed (65), and early studies of checkpoint blockade in this context have been initiated (ref. 66; NCT02362997, NCT02771197, NCT02681302, and NCT02331368). The observations herein add mechanistic insight and additional data on the activation state of tumor-specific T cells induced by each component (dCB and BMT) of the combined therapy.

The patient and preclinical data predict that IT may provide clinical benefit in patients with aggressive lymphoma, despite the limited efficacy of dCB in these patients (3); this is currently being evaluated in an ongoing trial (NCT03305445). Using IT to enhance the efficacy of dCB could be broadly significant; dCB is a standard therapy for melanoma, renal cell cancer, and subsets of colorectal cancer (67) and lung cancer (38), and is in numerous ongoing trials for other cancers. Even for settings in which dCB proves ineffective, our data suggest that dCB efficacy may be “rescued” by IT. Conversely, as homeostatic activation might improve dCB therapy, the addition of checkpoint blockade may improve T-cell therapies already exploiting lymphodepletion, such as CAR-T therapy. Anecdotal evidence of that approach is promising (68), and larger studies are under way (NCT02926833).

Despite the importance of γ_c cytokines for T-cell activation, their development as cancer therapies has been elusive. Recent improvements in γ_c cytokine formulation and dosing (69) have yielded promising preliminary results, prompting combination trials with checkpoint blockade (69). Our IT approach, similarly, provides transferred T cells with increased access to (endogenous) γ_c cytokines. A notable distinction is that recombinant γ_c cytokine therapy induces tachyphylaxis with continued exposure (70, 71), whereas the IT approach actually increases γ_c cytokine receptor expression and signaling. Regardless of the optimal approach for providing γ_c cytokines clinically, assessing homeostatic inhibition with these therapies will be important. Rather than presumptively combining all such therapies with PD-1 blockade, the specific inhibitory checkpoints elicited by each therapy should be determined, e.g., one γ_c cytokine therapy may increase T-cell expression of CTLA4 and LAG3, whereas another increases PD-1 and TIGIT.

Addressing homeostatic inhibition is not limited to the blockade of PD-1 and CTLA4. Our data revealed significant increases in CD8 T-cell expression of TIGIT, TIM3, LAG3, and CD38 post-BMT, all of which are currently being targeted in clinical trials. Further preclinical study of these and other checkpoint targets could expand our understanding of homeostatic inhibition and could translate into novel rational combinatorial approaches for patients, particularly those with tumor types in which immunotherapy has yet to offer significant benefit.

Protocols for the treatment of patients, as well as human sample collection and analysis, were approved by the Mount Sinai Institutional Review Board, and written informed consent was obtained from all patients in accordance with the Declaration of Helsinki. All experiments including human specimens were performed in compliance with the relevant ethical regulations.

A20, B16, EL4, and KLN205 cells were purchased from ATCC. Cells were cultured as described by ATCC, except EL4 cells were cultured in complete RPMI-1640 media plus 1 mmol/L sodium pyruvate. Heat-inactivated FBS was purchased from Gemini Bio-Products. All cells were tested and confirmed Mycoplasma negative with the Universal Mycoplasma Detection Kit (ATCC) before in vitro culturing and again before in vivo injection. A20 cells were tested in 2014 and 2016, B16 in 2014, KLN205 in 2017, and EL4 in 2017. No cell authentication was performed. Cells were cultured at 37°C in 5% CO₂. Cells were cultured 7 days after thawing before in vitro and in vivo use.

Six- to 10-week-old male BALB/c, C57BL/6, DB2/A, B6129SF2/J; and breeding pairs of B6.129S7-Il7r^tm1Imx/J [IL7rα knockout (KO)] and B6; 129 × 1-Il15ra^tm1Ama/J (IL15rα KO) mice were ordered from The Jackson Laboratory. Mice were maintained in barrier conditions. All experiments involving live mice were performed in compliance with ethical regulations approved by the Institutional Animal Care and Use Committee of the Icahn School of Medicine at Mount Sinai (New York, NY).

Seven- to 9-week-old donor and recipient male BALB/c mice were inoculated with 1 × 10⁶ A20 cells subcutaneously in the right hind flank. One week after tumor inoculation, the donor was treated with 100 μg of anti–PD-1 (RMP1-4; Bio X Cell) and anti-CTLA4 (9H10; Bio X Cell) per mouse 3 times 3 days apart. Three days after the final injection, recipient mice were irradiated with 900 cGy of TBI in a Rad Source RS-2000 Biological Research Irradiator. The spleens and bone marrow of donor mice were harvested. Irradiated recipients were injected intravenously with 0.3 mL in Hank's Balanced Salt Solution in the tail vein admixed with splenocytes and bone marrow cells. Splenocyte dose was one spleen per recipient and 5 × 10⁶ bone marrow cells per mouse.

Beginning the day of ablative therapy and for 2 weeks after, drinking water was supplemented with 1 mg/mL neomycin and 1,000 U/mL polymyxin sulfate B (Sigma-Aldrich) for gut decontamination.

IT was tested in other tumor models as described, with the following modifications:

• 5 × 10⁵ B16 cells were inoculated subcutaneously in the right hind flank of male C57BL/6 donor and recipient mice. Recipient mice were treated 6 times with 100 μg of anti–PD-1 and CTLA4.

• 2 × 10⁵ EL4 cells were inoculated subcutaneously in the right hind flank of the male C57BL/6 donor and recipient mice. Recipient mice received 600 cGy of TBI.

• Recipient mice were treated 6 times with 100 μg of anti–PD-1 and CTLA4.

• 1 × 10⁶ KLN205 cells were inoculated subcutaneously in the right hind flank of the male C57BL/6 donor and recipient mice.

• 5 × 10⁵ B16 cells were inoculated subcutaneously in the right hind flank of the male IL7Rα KO donor and male C57BL/6 recipient mice. Splenocyte dose was five IL7Rα KO donor spleen per one C57BL/6 WT recipient and 5 × 10⁶ bone marrow (BM) cells per mouse. 2 × 10⁵ EL4 cells were injected subcutaneously in the right hind flank of the female IL15Rα KO donor and female B6129SF2/J WT recipient mice. Recipient mice were irradiated with 600 cGy of TBI. Splenocyte dose was two IL15Rα KO donor spleen per one B6129SF2/J WT recipient and 5 × 10⁶ BM cells per mouse.

The following antibodies were used during this study, at manufacturer's recommended concentration.

PD-1 (EH12.1), IL2Rα/CD25 (M-A251), CD8 (RPA-T8), CD3 (SK7), CD4 (RPA-T4), CTLA4 (BNI3), γ_c (TUGh4), IL2Rβ (TU27), IL7Rα (A019D5), IL15Rα (JM7A4), Ki-67 (Ki-67; BioLegend), CD44 (BT18), IFNγ (4S.B3), TNFα (Mab11), and pSTAT5 (pY694; clone 47).

B220 (RA3-6B2), CTLA4 (UC10-4B9), CD127 (SB/1999), IFNγ (XMG1.2), CD69 (H1.2F3), TCRβ (H57-597), IL2 (JES6-5H4), TNFα (MP6-XT22), CD45.1 (A20), CD45.2 (104), IL2Rα/CD25 (PC61), CD3 (145-SC11), CD8 (53-6.7), CD4 (RM4-5), PD-1 (J43), CD8β (eBioH35-17.2), pSTAT5 (pY694; Clone 47), H2K^d HALSTQSAL pentamer (ProImmune), IL2Rβ (TM-Beta 1), IL15Rα (DNT1Ra), Thy1.1 (OX-7), and Thy1.2 (53.21).

Patient PBMCs were thawed and incubated for 30 minutes at 37°C and then cultured with recombinant human carrier-free IL2 (R&D Systems), IL7, and IL15 (both PeproTech) with Fixable Viability Stain for 15 minutes at 37°C.

Murine splenocyte single-cell suspensions were cultured with 100 ng recombinant carrier free of mouse IL2, IL7, and IL15 (Gemini Bioproducts) with BG Horizon Fixable Viability Stain (BD Biosciences) for 20 minutes at 37°C.

Signaling was stopped by IC Fixation Buffer (Thermo-Fisher). Extracellular and nuclear staining was performed as described in the BD Perm IV manual (BD Biosciences).

Flat-bottom 96-well plates were coated with the described amounts of anti-CD3 (OTK3; eBioscience) and human recombinant PD-L1–Fc (BioLegend), CD80-Fc (BioLegend), or IgG1-Fc (Sino Biological) diluted in PBS overnight at 4°C. Mixture was aspirated, and PBMCs were cultured with soluble anti-CD28 (CD28.2; BD Biosciences) for the described amounts of time.

Single-cell suspensions of splenocytes were CD8⁺ T cell–enriched with the MagniSort mouse CD8⁺ T-cell Enrichment Kit (Invitrogen). Flat-bottom 96-well plates were coated with the described amounts of anti-CD3ϵ (145-2C11; BD Biosciences) and mouse recombinant PD-L1–Fc (BioLegend), CD80-Fc (BioLegend), or IgG1-Fc (Sino Biological) diluted in PBS overnight at 4°C. The mixture was aspirated, and CD8⁺ T cells were cultured with soluble anti-CD28 (37.51; BD Biosciences) for the described amounts of time.

PBMCs were cultured with 10 μg of ipilimumab and nivolumab for 48 hours. Cells were washed and replenished with fresh ipilimumab, nivolumab, and media. They were then cocultured with 100 ng of SEB (Toxin Technology) for another 72 hours.

IT-treated recipient mice were depleted of CD8⁺ T cells using with 250 μg anti-CD8 (2.43), and of CD4⁺ T cells using anti-CD4 (GK1.5). IFNγ was depleted using 500 μg of anti-IFNγ (R4-6A2). Recipient mice were treated with antibody 3 days before transplantation, the day of transplantation, and every 3 days until the study's completion.

Mice were retro-orbitally bled, and blood was anticoagulated with 2 mmol/L EDTA in PBS, then diluted 1:1 with Dextran T500 (Pharmacosmos) 2% in PBS and incubated at 37°C for 45 minutes to precipitate red cells. PBMCs were then cocultured with 5 × 10⁵ irradiated A20 cells for 48 hours with 1 μg mouse anti-CD28 and in the presence of brefeldin A (Invitrogen) for the last 5 hours at 37°C.

GFP-specific donor mice (B10.D2 mouse strain background) were bred with BALB/c mice. F1 pups were tested by PCR for recombined TCR alpha and beta chains and confirmed to be GFP-reactive. F1 mice were treated with 100 μg anti-CTLA4 and anti–PD-1 3 times 3 days apart. Splenocytes were intravenously injected into congenic, irradiated F1 B10.D2-BALB/c mice. Recipient mice were also treated with 100 μg anti-CTLA4 and anti–PD-1 3 times 3 days apart. Recipient mice were harvested 7 days post-transfer. At a ratio of 20:1, splenocytes were cultured with GFP-overexpressing A20 in vitro for 24 hours, with the last 8 hours in the presence of brefeldin A.

Donor BALB/c mice were treated with 100 μg anti-CTLA4 and anti–PD-1 3 times 3 days apart. Three days after the final treatment, donor splenocytes were harvested and intravenously injected into congenic, irradiated BALB/c recipient mice. Recipient mice received one treatment of 100 μg anti-CTLA4 and anti–PD-1. Three days post transfer, splenocytes of recipient mice were harvested.

Tumors were excised after 14 days and placed in 10% formalin (Fisher Scientific) at room temperature overnight. Tumors were submerged in 15% sucrose in PBS for 6 hours, then 30% sucrose in PBS for 6 hours. Tumors were embedded in a base mold (Fisher Scientific) filled with OCT (Fisher Scientific). Cryosections were cut at a thickness of 10 μm using a Leica cryostat and mounted on Superfrost Plus slides (Fisher Scientific) and permeabilized for 20 minutes in 0.5% Triton X-100 in PBS at room temperature. The primary antibodies were stained. The sample slides were counterstained with 4′6′-diamidino-2-phenylindole (DAPI) and cover-slipped with ProLong Gold Antifade Reagent (Life Technologies). Imaging was performed using a Zeiss LSM 780 confocal microscope. Images were collected using Zeiss ZEN software.

Tumor cell suspensions were then depleted of tumor cells by magnetic separation using CD19 nanobeads according to the manufacturer's protocol (MojoSort, BioLegend).

Viability staining was performed using Cell-ID Intercalator-103Rh (Fluidigm). Barcoding was performed by staining each sample with a unique combination of up to 4 different metal conjugations of a monoclonal antibody against CD45. Samples were then combined for further staining. Surface staining was performed in staining buffer containing FcX (BioLegend) and metal-conjugated antibodies. For intracellular staining, cells were incubated in Cytofix/Cytoperm (BD Biosciences) and stained with metal-conjugated antibodies in Perm/Wash buffer with heparin. Cells were washed and stained with Cell-ID Intercalator-Ir (Fluidigm) to label all nucleated cells. Acquisition was performed on a CyTOF 2 cytometer, and data were analyzed with Cytobank.

The data sets supporting the findings presented in this study are available from the corresponding author upon reasonable request. All requests for data and materials will be promptly reviewed by the Icahn School of Medicine at Mount Sinai to verify whether the request is subject to any intellectual property or confidentiality obligations. Any data that can be shared will be released via a Material Transfer Agreement.

T.U. Marron reports receiving a commercial research grant from Regeneron and is a consultant/advisory board member for the same. M. Merad reports receiving commercial research grants from Regeneron, Boerhinger Ingelheim, Inc., Takeda, and Innate Pharma, has received honoraria from the speakers bureaus of Compugen Inc., Innate Pharma, Inc., Dynavax, and Pioneer, Inc., has ownership interest (including stock, patents, etc.) in Compugen Inc., Dynavax, Inc., Pionyr, Inc., and Myeloid Therapeutics, and is a consultant/advisory board member for Compugen, Inc., Innate Pharma, Inc., Dynavax, Pionyr, Inc., and Myeloid Therapeutics. J.D. Brody reports receiving commercial research grants from Merck, Genentech, Acerta Pharma, and Celldex Therapeutics, reports receiving other commercial research support from Seattle Genetics, and is a consultant/advisory board member for Bayer, Gilead, Janssen, and Amgen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N. Marshall, T.U. Marron, J.D. Brody

Development of methodology: N. Marshall, T.U. Marron, M. Aleynick, L. Hammerich, J.D. Brody

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Marshall, K. Hutchinson, T.U. Marron, M. Aleynick, L. Hammerich, R. Upadhyay, J. Svensson-Arvelund, J.D. Brody

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Marshall, K. Hutchinson, T.U. Marron, L. Hammerich, J. Svensson-Arvelund, B.D. Brown, M. Merad, J.D. Brody

Writing, review, and/or revision of the manuscript: N. Marshall, K. Hutchinson, T.U. Marron, L. Hammerich, J. Svensson-Arvelund, B.D. Brown, J.D. Brody

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Upadhyay, J.D. Brody

Study supervision: N. Marshall, T.U. Marron, J.D. Brody

We thank Dr. Peter Heeger for critically reading and editing this manuscript. We also thank the core facilities of Flow Cytometry CORE, the Center for Comparative Medicine and Surgery, and the Human Immune Monitoring Center at Icahn School of Medicine, Mount Sinai. Research reported in this article was supported by the Tisch Cancer Institute through the NCI Cancer Center Support Grant (P30 CA196521). N. Marshall is funded by the NIH T32 Transplant 5T32AI078892 and T32 Immunology 5T32AI007605 grants. B.D. Brown is funded by a grant from the Cancer Research Institute and by NIH R21HD091461. M. Merad is funded by NIH R01 CA154947 and R01 CA190400. J.D. Brody is funded by a grant from the Damon Runyon Cancer Research Foundation and by NIH 5R00CA140728.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.