Research ArticleIMMUNOTHERAPY

Platelets subvert T cell immunity against cancer via GARP-TGFβ axis

See allHide authors and affiliations

Science Immunology  05 May 2017:
Vol. 2, Issue 11, eaai7911
DOI: 10.1126/sciimmunol.aai7911

Cancer immunotherapy according to GARP

Cancer, like microbes, can adapt to a single therapy, making combination therapies the approach of choice. Complementary therapies that decrease immunosuppression may boost the efficacy of immunotherapies. Now, Rachidi et al. report that targeting platelets improves adoptive T cell therapy of multiple cancers in mice. They found that transforming growth factor β (TGFβ) from platelets decrease T cell function largely through the expression of the TGFβ-docking receptor glycoprotein A repetitions predominant (GARP). These data suggest that combining immunotherapy with platelet inhibitors may be a complementary approach to cancer therapy.

Abstract

Cancer-associated thrombocytosis has long been linked to poor clinical outcome, but the underlying mechanism is enigmatic. We hypothesized that platelets promote malignancy and resistance to therapy by dampening host immunity. We show that genetic targeting of platelets enhances adoptive T cell therapy of cancer. An unbiased biochemical and structural biology approach established transforming growth factor β (TGFβ) and lactate as major platelet-derived soluble factors to obliterate CD4+ and CD8+ T cell functions. Moreover, we found that platelets are the dominant source of functional TGFβ systemically as well as in the tumor microenvironment through constitutive expression of the TGFβ-docking receptor glycoprotein A repetitions predominant (GARP) rather than secretion of TGFβ per se. Platelet-specific deletion of the GARP-encoding gene Lrrc32 blunted TGFβ activity at the tumor site and potentiated protective immunity against both melanoma and colon cancer. Last, this study shows that T cell therapy of cancer can be substantially improved by concurrent treatment with readily available antiplatelet agents. We conclude that platelets constrain T cell immunity through a GARP-TGFβ axis and suggest a combination of immunotherapy and platelet inhibitors as a therapeutic strategy against cancer.

INTRODUCTION

Platelets, or thrombocytes, play essential roles in hemostasis (1). Increasingly, they have emerged to possess other regulatory functions in physiology such as angiogenesis, wound healing, and immunomodulation (24). Intriguingly, cancer-associated thrombocytosis is an independent poor prognostic factor in multiple cancer types (5, 6), via enhancing invasiveness of cancer cells (7), promoting cancer motility (4, 8), and inducing epithelial-mesenchymal cell transition (9). Despite knowledge of platelet cross-talk with natural killer cells (10), neutrophils (11), macrophages (12), dendritic cells (1315), and T lymphocytes (14), the direct impact of thrombocytes on T cell immunity against cancer and the underlying molecular mechanisms have yet to be fully elucidated.

Platelets are bioactive, anuclear cellular fragments that are shed out of megakaryocytes in the bone marrow vasculature (16). They are the smallest cellular component of the hematopoietic system and are second only to red blood cells in number. Platelets express a number of cell surface receptors for adhesion and aggregation (1, 17), such as glycoprotein (GP) Ib-IX-V complex, which serves as a receptor for von Willebrand factor, and GPIIb-IIIa integrin, which binds to fibrinogen and fibronectin. They also express other activation receptors, including the thromboxane A2 receptor, the adenosine diphosphate (ADP) receptors P2Y1 and P2Y12, and the protease-activated receptors PAR1 and PAR4, the latter of which can be activated by thrombin (18). Platelets have been found to constitutively express the nonsignaling transforming growth factor β (TGFβ)–docking receptor glycoprotein A repetitions predominant (GARP) (19), encoded by the leucine-rich repeat-containing protein 32 (Lrrc32) gene. The role of GARP is to increase the activation of latent TGFβ (LTGFβ) and thus its biological function in the close proximity of GARP-expressing cells. The other cells that express GARP are regulatory T (Treg) cells, which do so only after activation via T cell receptor (TCR) (20). Both GPIb-IX-V complex and GARP depend on the molecular chaperone gp96 for folding and cell surface expression (21, 22). Genetic deletion of Hsp90b1 (encoding gp96) from platelets results in significant thrombocytopenia and impaired platelet function (21). Last, there are cytoplasmic granules in platelets containing a variety of molecules such as TGFβ, ADP, serotonin, and proteases, which are released upon platelet activation and degranulation to exert their functions (23, 24).

The key unresolved questions are how platelets affect the adaptive immunity in cancer and what are the underlying molecular mechanisms for such an action. With regard to TGFβ, it is completely unknown what the physiological function of platelet-specific cell surface GARP-TGFβ is in host immunity. In addition, GARP-TGFβ complex on platelets could be formed intracellularly, during the de novo biogenesis, or extracellularly, where GARP snatches LTGFβ in the extracellular matrix from nonplatelet sources and binds to it. However, it is unclear which source of the GARP-TGFβ complex is critical in regulating the host immunity against cancer in vivo. In this study, we systematically probed the effect of platelets on the effector function of antitumor T cells. We also took an unbiased approach to identify platelet-derived soluble immunoregulatory factors in blunting T cell function. This study not only uncovers mechanisms of platelet-mediated T cell suppression but also demonstrates the validity of the combination therapy of cancer with immunotherapeutics and antiplatelet (AP) agents in clinically relevant mouse models.

RESULTS

Genetic inhibition of platelets enhances adoptive T cell therapy of cancer

Using bone marrow chimeric mice, we previously demonstrated that genetic deletion of Hsp90b1 from the hematopoietic system resulted in macrothrombocytopenia coupled with dysfunctional platelets due to the loss of cell surface GPIb-IX-V, the receptor for von Willebrand factor (21). To probe the immunological function of platelets, we generated a megakaryocyte-specific Hsp90b1 knockout (KO) mouse model in this study. As expected, KO mice had significantly lower platelet counts in the blood compared with wild-type (WT) mice (Fig. 1A). The dysfunction of platelets was evidenced by prolonged bleeding time (Fig. 1B). Extensive phenotypical analysis showed no obvious abnormalities in other cellular lineages, including T and B cells in the hematopoietic system of KO mice (fig. S1, A to D). The ability of CD8+ and CD4+ cells from the KO mice to produce interferon-γ (IFNγ) in response to polyclonal activation was also unaffected (fig. S1, E and F).

Fig. 1 Targeting platelets genetically potently enhances ACT of cancer.

(A) Platelet counts from the peripheral blood of WT and Pf4-cre-Hsp90b1flox/flox (KO) mice (n = 15 per group). (B) Bleeding time was measured in WT and KO mice by pricking the lateral tail vein (n = 5 per group). (C) B16-F1 melanoma tumors were established in WT and Pf4-cre-Hsp90b1flox/flox mice, followed by adoptive transfer of activated Thy1.1+ Pmel cells on day 11 plus IL-2–anti–IL-2 antibody complexes on days 11, 13, 15, and 17. Average tumor growth curves (n = 7 to 9 per group) are shown. (D) Same as in (C) except that mice did not receive T cells (n = 5 to 6 per group). (E and F) Pmel cells from the tumor-draining lymph nodes of the adoptively transferred mice in (D) were stimulated with hgp100 peptide for 4 hours, followed by intracellular staining for IFNγ (E) and TNFα (F). MFI, mean fluorescence intensity. Repeated-measures two-way ANOVA was used to compare the tumor growth curves. Two-tailed independent Student’s t test was used in (A), (B), (E), and (F). Data are means ± SEM.

Adoptive T cell therapy (ACT) was next used to determine whether platelet dysfunction in the host affects the ability of transferred donor T cells to control cancer. ACT is a process of transferring preactivated antigen-specific T cells for the treatment of established cancers (25). Melanoma was chosen because (i) immunotherapy of unresectable melanoma has been increasingly implemented over the past few years with encouraging efficacies (26) and (ii) TCR transgenic mice of both CD4+ (TRP1) and CD8+ (Pmel) lineages (27, 28) against melanoma antigens permit studying tumor-reactive T cells in our mouse models. B16-F1 melanomas were therefore established in C57BL/6 mice after subcutaneous injection on day 0, followed by infusion of ex vivo primed Pmel cells on day 11, along with interleukin-2 (IL-2)–anti–IL-2 antibody complex (29). Transferred Pmel cells had better antitumor activity in the Hsp90b1 KO recipients compared with WT ones (Fig. 1C), whereas no difference in tumor growth was observed between the two groups without ACT (Fig. 1D). The improved ACT efficacy in the KO mice was associated with increased production of IFNγ (Fig. 1E) and tumor necrosis factor–α (TNFα) (Fig. 1F) by the donor T cells. These results suggest that platelet function in tumor-bearing mice constrains T cell–mediated cancer immunotherapy.

Platelet releasate suppresses T cell activation and function

We next focused on understanding the molecular mechanisms of T cell suppression by platelets. Given that platelets are not usually found in the lymphatic system and the T/B cell zone of the lymphoid organs, we reasoned that activated platelets exert their immunosuppressive function via releasing soluble factors. Purified platelets were suspended at 108 platelets/ml (several times lower than physiological platelet concentration) and activated with thrombin to generate platelet releasate (PR). The suppressive capacity was then measured in vitro using a standard polyclonal T cell activation assay (fig. S2A). Soluble factors in PR, but not platelet microvesicles (MVs), completely blocked T cell proliferation, blastogenesis, and IFNγ production (Fig. 2 and fig. S2, B to G). The activity of PR was not species-specific because both mouse (Fig. 2A) and human (Fig. 2B) PRs suppressed the activation and effector function of T cells from either source. Several parameters related to PR-mediated T cell suppression were also examined (fig. S3, A to J). The negative effect of PR could not be rescued by high-dose IL-2 (fig. S3A). Kinetic studies showed that the inhibitory effect was most pronounced during the first 2 days of T cell activation and was irreversible (fig. S3, B to G). Supernatant from unstimulated platelets had minimal effects (fig. S3H). PR had no direct effect on the proliferation of nonlymphocytes such as fibroblasts and B16-F1 melanoma (fig. S3, I and J). In addition, PR-treated T cells displayed a naïve phenotype expressing more CD62L and less CD44, programmed death–1 (PD-1), glucocorticoid-induced TNF receptor–related protein (GITR), and CD25 (Fig. 2, C and D). Moreover, the presence of PR during in vitro activation of CD4+ TRP1 transgenic T helper 17 (TH17) cells (27) abolished their activity upon adoptive transfer against B16-F10 melanoma (Fig. 3A) and reduced their persistence in the recipient mice (Fig. 3B). Similarly, CD8+ Pmel cells also lost their antitumor activity in the adoptive transfer setting after exposure to PR during ex vivo activation (Fig. 3C), which correlated with poor donor cell persistence (Fig. 3D).

Fig. 2 Platelet-derived soluble factors suppress T cell function.

(A) Naïve splenic CD8+ T lymphocytes were activated polyclonally with or without mouse PR for 3 days, followed by measuring forward scatter profile (FSC) and multiple intracellular molecules. Data from multiple experiments (n > 5 times) were summarized at the bottom of the panel. GZMB, granzyme B. (B) Human peripheral blood CD4+ or CD8+ cells were activated polyclonally for 7 days with or without human PR, followed by measuring the indicated markers (representative of two experiments). Correlation between % PR and cytokine quantity was established using Spearman’s correlation coefficient. (C and D) Naive CD8+ T cells were polyclonally activated in the presence or absence of PR for 3 days, followed by flow cytometry. Numbers represent percentage of cells in the corresponding quadrants (n = 3 to 4 per group). Comparisons in (A), (C), and (D) were performed using two-tailed independent Student’s t tests.

Fig. 3 Priming of antigen-specific T cells in the PR abrogates their antitumor immunity in vivo.

(A) TRP1 transgenic CD4+ T cells were primed under TH17-differentiating conditions, in control media or human PR (n = 6 per group). They were then adoptively transferred on day +10 to B16-F10–bearing mice that also received sublethal dose of total body irradiation on day +9. (B) Percentages of TRP1 cells in the peripheral blood on day +37 (n = 6 to 7 per group). Mann-Whitney test for non-normal distribution was used to compare the two groups. WBC, white blood cell. (C) Pmel T cells were primed with IL-12 and hgp100 peptide, in control media or human PR. They were then adoptively transferred on day +8 to B16-F10–bearing mice that also received cyclophosphamide on day +7 (n = 4 to 8 per group). (D) Percentages of Pmel cells in the peripheral blood on day +33 (n = 4 to 8 per group). Tumor growth curves were compared using repeated-measures two-way ANOVA. Percentages of T cells were compared using two-tailed independent Student’s t tests. Data are means ± SEM.

Platelet TGFβ and lactate mediate T cell suppression

Next, we took an unbiased approach to identify active T cell–suppressive molecule(s) in the PR. Human PR was fractionated by size exclusion chromatography, followed by quantifying individual fractions for their suppressive activity. Two major peaks with suppressive activity were resolved (Fig. 4A). Fraction A (>150 kDa) was further subfractionated by anion exchange chromatography (Fig. 4B). Coomassie blue staining of a reducing SDS–polyacrylamide gel electrophoresis (SDS-PAGE) of the most active subfractions showed prominent bands corresponding to 150 to 250 kDa and 10 to 50 kDa (Fig. 4B). Mass spectrometry identified these proteins to be mature TGFβ (mTGFβ), latency-associated peptide (LAP), LTGFβ-binding protein 1 (LTBP1), and thrombospondin-1 (TSP1), indicating the presence of an mTGFβ-LAP-LTBP1-TSP1 complex (30). Immunoblot confirmed the existence of mature (12.5 kDa) and latent (44 kDa) TGFβ, LTBP1 (180 kDa), and TSP1 (110 to 180 kDa) in the whole PR, as well as in fraction A (Fig. 4B). Neutralizing TGFβ with the combination of an inhibitor for activin receptor–like kinase 5 (ALK5) (also known as type I TGFβ receptor) and anti-TGFβ antibody in fraction A completely rescued T cell function (Fig. 4C). We thus defined TGFβ as a major T cell–suppressive factor in the PR.

Fig. 4 TGFβ and lactate contribute to platelet-mediated T cell suppression.

(A) Human PR was fractionated by size exclusion chromatography, followed by quantifying individual fractions for their suppressive activity. T cell suppression indices of all fractions are shown. Suppression index of media (percent of undivided CD8 T cells) was set as 1 and is indicated. (B) Fraction A was further resolved by DEAE column, and subfractions were assayed for T cell suppression. The most active fractions (boxed) were resolved by SDS-PAGE and Coomassie blue stain. The protein identities of the major bands were determined by mass spectrometry and immunoblot. (C) CFSE-labeled naïve CD8+ T cells were stimulated with anti-CD3ε/CD28 antibodies and IL-2 for 3 days in media or fraction A (Fr. A) with or without TGFβ blockade. Cells were assayed for CFSE dilution, granzyme B, and IFNγ production by CD8+ cells (n = 3 per group). Representative histograms are shown. Corresponding quantifications are displayed below the flow cytometry data. (D) Suppressive activity of various subfractions B1 to B8. (E) Upfield 600-MHz 1H excitation sculpting NMR spectra of fractions B2 to B5 (bottom to top). Solid and dashed boxes highlight varying concentrations of lactate methyl and methine proton resonances, respectively. Spectra are normalized to the defined concentration of TSP. (F) Purified CD8+ T cells were cultured for 3 days in the presence of agonistic CD3ε and CD28 antibodies and IL-2. Cell surface expression of CD25 and CD69 and blastogenesis (FSC) were assayed by flow cytometry. Correlation between LA concentration and cytokine production was established using Spearman’s correlation coefficient. Difference between groups in (C) was tested by two-tailed independent Student’s t test.

However, blocking TGFβ in the whole mouse PR only partially rescued T cell activity in vitro (fig. S4), indicating the presence of other TGFβ-independent factors. Furthermore, the activity of fraction B in both human and mouse PRs was heat-stable, proteinase K–resistant, and smaller than 1.0 kDa in size (fig. S5). Fraction B was then subfractionated by an anion exchange column to obtain subfractions B1 to B8 and used nuclear magnetic resonance (NMR) spectroscopy to delineate the metabolite composition (Fig. 4, D and E). Targeted profiling using the Chenomx NMR Suite software identified lactate as the most abundant metabolite (~3.4 mM) in the most suppressive B3 and B4 subfractions (Fig. 4E). The immunoregulatory roles of lactate in T cells (31) and macrophages (32) have been reported. The concentration of lactate in the whole PR was ~5.7 mM. Lactate efficiently suppressed T cell activation with concentrations as low as 2.5 mM (Fig. 4F). Blocking both TGFβ (by a neutralizing antibody) and lactic acid (LA) (by inhibiting monocarboxylate transporter I with α-cyano-4-hydroxycinnamic acid) (32) in the whole human PR almost completely rescued IFNγ production, CD25 expression, and blastogenesis of CD8+ T cells (fig. S6). Thus, we conclude that the suppressive activity of PR primarily resides in TGFβ and lactate.

To further address the suppressive components in the PR, we performed an in vitro Treg induction assay. Splenocytes were activated with anti-CD3/28 antibody in the presence of PR, TGFβ, and lactate for 3 days. Consistent with earlier findings, PR attenuated T cell blastogenesis (fig. S7, A and D), and this was partially recapitulated by each of TGFβ and LA. A proportion of CD4+ cells cultured with PR differentiated into the Treg lineage (fig. S7, B and D), and this was accompanied by up-regulation of phosphorylated Smad2/3 (p-Smad2/3) (fig. S7, C and D). Expectedly, TGFβ, but not LA, also induced Treg differentiation and p-Smad2/3.

We then investigated whether TGFβ and/or LA can independently recapitulate the inhibitory effects of PR on tumor-reactive T cells. B16-F1 melanoma tumors were established in C57BL/6 mice, which were subsequently treated with Pmel CD8 T cells, similar to the experiment described in Fig. 3C. T cells were primed ex vivo with human gp100 (hgp100) and IL-12 in control media (Pmel-12), PR, TGFβ, and/or LA (fig. S8). T cells primed in the presence of PR or TGFβ (650 pg/ml, the concentration of TGFβ present in PR from 1 × 108 platelets/ml) failed to control melanoma progression and to persist in peripheral blood (fig. S8, A to C). This poor in vivo persistence is likely explained by the failure of Pmel cells to up-regulate receptors of the homeostatic cytokines IL-2 and IL-7 under these conditions (fig. S8D). IL-7 is crucial for Pmel cell persistence in vivo in the Pmel tumor model (33). In turn, LA in the priming phase had no effect on the subsequent antitumor activity of Pmel T cells. This suggests that platelet-derived TGFβ is likely a more relevant target in immunotherapy.

Platelet-intrinsic GARP plays critical roles in generating active TGFβ

Platelets not only produce and store high levels of TGFβ intracellularly (34) but also are the only cellular entity known so far that constitutively expresses the cell surface–docking receptor GARP for TGFβ (19). Thus, platelets may contribute to the systemic levels of TGFβ via active secretion, as well as GARP-mediated capturing from other cells or the extracellular matrix (9, 3537). We next addressed to what extent and how platelets contribute to the physiological TGFβ pool. Baseline sera were obtained from WT mice followed by administration of a platelet-depleting antibody. These mice were sequentially bled, and serum TGFβ was quantified by enzyme-linked immunosorbent assay (ELISA). Depletion of platelets resulted in a complete loss of active and total TGFβ, which rebounded effectively as soon as platelet count recovered (Fig. 5A). These experiments demonstrate that platelets contribute dominantly to the circulating TGFβ level. By comparison, serum LA with or without depletion showed no significant changes (fig. S9), arguing that platelet-derived TGFβ, but not LA, is a more relevant platelet-derived immunosuppressive molecule in vivo.

Fig. 5 Platelet-intrinsic GARP plays critical roles in generating active TGFβ.

(A) Baseline serum was collected from WT C57BL/6 mice, followed by a single dose of anti-mouse thrombocyte sera (n = 7). Serum was collected 24, 48, and 72 hours after injection. Ab, antibody. (B) Representative flow cytometry plots. Platelet-specific marker CD41+ population was gated on and analyzed for the expression of cell surface GARP and LTGFβ. Numbers represent percentages of the gated population over all CD41+ events. (C and D) Graphical representation of flow cytometry data from (B) (n = 4 to 9 per group). (E) Serum and plasma levels of active TGFβ from indicated mice (n = 5 per group). (F) Serum and plasma levels of total TGFβ from indicated mice (n = 5 per group). Comparison was performed using two-tailed independent Student’s t test. Data are means ± SEM.

The biology of platelet-derived TGFβ in cancer immunity was addressed next, focusing on the role of platelet GARP in the production of active TGFβ.In addition to platelet-specific Hsp90b1 KO mice, two additional mouse models were generated: one with selective deletion of GARP in platelets (Pf4-cre-Lrrc32flox/flox or Plt-GARPKO) and another with platelet-restricted KO of TGFβ1 (Pf4-cre-Tgfb1flox/flox or Plt-TGFβ1KO) (Fig. 5B). Because gp96 is also an obligate chaperone for GARP (22), platelets from neither Plt-gp96KO mice nor Plt-GARPKO mice expressed cell surface GARP-TGFβ complex. However, platelets from Plt-TGFβ1KO mice expressed similar levels of surface GARP-TGFβ1 complex when compared with WT platelets (Fig. 5, B to D), indicating that the GARP-TGFβ1 complex can be formed without autocrine TGFβ1.

The levels of active TGFβ and LTGFβ were then measured in the plasma and sera of WT and KO mice (Fig. 5, E and F). In WT mice, active TGFβ was elevated in serum compared with plasma, indicating a role for platelets and/or the coagulation cascade in TGFβ activation (Fig. 5E). Plt-gp96KO and Plt-GARPKO mice had very little active TGFβ in their sera, confirming the importance of platelet-intrinsic GARP in converting LTGFβ to the active form. In contrast, the serum level of active TGFβ in Plt-TGFβ1KO mice was comparable with that of WT mice (Fig. 5E), indicating that platelets are capable of activating TGFβ from nonplatelet sources in a trans fashion. Significantly, the total LTGFβ level in the serum is only reduced in Plt-TGFβ1KO mice but not Plt-gp96KO or Plt-GARPKO mice (Fig. 5F). Collectively, these data indicate that platelet-intrinsic GARP is the most important mechanism in the activation of TGFβ systemically. This experiment also categorically confirmed that serum but not plasma level of active TGFβ reflects platelet activation.

Platelet GARP-TGFβ complex blunts antitumor T cell immunity

So far, we have shown that TGFβ is a major T cell suppressor molecule from PR and that platelet-specific deletion of gp96 (which functionally deletes GARP) improves ACT of cancer. These fortuitous observations suggest that platelet-specific GARP plays critically negative roles in antitumor T cell immunity. This hypothesis was next addressed by comparing the efficacy of ACT of melanoma in WT, Plt-TGFβ1KO, and Plt-GARPKO recipient mice (Fig. 6). B16-F1 melanomas were established in either WT or KO mice, followed by lymphodepletion with cyclophosphamide (Cy) on day 9, and the infusion of ex vivo activated Pmel T cells on day 10 (Fig. 6A). Tumors were controlled much more efficiently in the Plt-GARPKO mice compared with WT mice (Fig. 6A). This was associated with enhanced persistence (Fig. 6B) and functionality of Pmel cells in the peripheral blood of Plt-GARPKO mice (Fig. 6, C and D). In contrast, Plt-TGFβ1KO mice, whose platelets express GARP and remain capable of activating TGFβ, did not have improved control of tumors (Fig. 6D). The generality of these findings was next studied in the MC38 colon carcinoma system, given that the growth of this transplantable tumor in syngeneic mice undergoes both CD4- and CD8-mediated immune pressure (38, 39). The growth of MC38 was significantly diminished in Plt-GARPKO mice compared with WT mice (Fig. 7, A to C). The MC38-bearing Plt-GARPKO mice had reduced serum levels of active TGFβ (Fig. 7D). Staining for p-Smad2/3 in MC38 tumor sections demonstrated a remarkable attenuation of TGFβ signaling in MC38 cells in Plt-GARPKO mice (Fig. 7, E and F). This was associated with reduction of both systemic myeloid-derived suppressor cells (Fig. 7G) and tumor-infiltrating Treg cells in Plt-GARPKO mice (Fig. 7H). Together, this demonstrates that platelets are the commanding source of TGFβ activity in the tumor microenvironment and that they exert potent immunosuppressive effects on antitumor immunity via GARP-TGFβ.

Fig. 6 Platelet-derived GARP-TGFβ complex blunts ACT of melanoma.

(A to C) B16-F1–bearing WT and Plt-GARPKO mice (n = 7 to 8 per group) expressing congenic marker Thy1.2 were given Cy on day 9, followed by adoptive transfer of activated Thy1.1+ Pmel cells on day 10. (A) Tumor growth curves. (B) The frequency of Pmel cells in mice was enumerated 3 weeks after adoptive transfer of T cells by flow cytometry in the peripheral blood (CD8+Thy1.1+/total CD8+). (C) IFNγ-producing ability of antigen-specific donor T cells (Pmel) from indicated mice 3 weeks after T cell transfer. (D) B16-F1 melanoma tumors were established in WT and Plt-Tgfβ KO mice, followed by adoptive transfer of activated Thy1.1+ Pmel cells on day 11 plus IL-2–anti–IL-2 antibody complexes on days 11, 13, 15, and 17. Average tumor growth curves (n = 4 to 9 per group) are shown. Repeated-measures ANOVA was used in (A) and (D). Two-tailed independent Student’s t test was used in (B) and (C). Data are means ± SEM.

Fig. 7 Targeting platelet-derived GARP-TGFβ complex results in reduction of TGFβ activity in the tumor microenvironment and protection against colon cancer.

(A) WT or Plt-GARPKO mice (n = 5 per group) were injected in the right flank with 1 × 106 MC38 colon cancer cells. Tumor size was measured every 3 days with digital vernier caliper. (B) Kaplan-Meier survival curve in MC38-bearing mice (n = 5 per group). (C) In a separate experiment, 6 weeks after MC38 injection, mice were sacrificed and the primary tumors were resected and weighed. The inset shows the images of primary tumors resected from mice 6 weeks after injection. (D) Serum was obtained from mice 6 weeks after MC38 injection, and active TGFβ1 was measured by ELISA. (E) IHC for p-Smad2/3 in MC38 tumors from indicated mice; representative images are shown. Scale bar, 12.5 μm. (F) Independent histopathological quantification of p-Smad2/3 staining intensity from (E) (n = 4 per group). (G) Flow cytometric analysis of peripheral blood myeloid–derived suppressor cells. (H) Percentage of Treg cells (CD25+ Foxp3+) in the CD4+ tumor-infiltrating lymphocytes from the indicated mice. Repeated-measures two-way ANOVA was used in (A); Kaplan-Meier curves and log-rank tests were used in (B). Two-tailed independent Student’s t test was used in (C), (D), (F), (G), and (H). Data are means ± SEM.

Antiplatelet pharmacological agents potentiate ACT of cancer

To establish the clinical relevance of the suppressive effect of platelets on antitumor immunity, we sought to inhibit platelets pharmacologically. The results so far suggest that antiplatelet (AP) pharmacological agents can be exploited for enhancing cancer immunotherapy. This possibility was addressed using Pmel adoptive therapy of B16 melanoma (4042). B16-F1 melanomas were established in C57BL/6 mice after subcutaneous injection on day 0, followed by lymphodepletion with Cy on day 7, and infusion of ex vivo primed Pmel cells on day 8 (29), along with AP agents: aspirin and clopidogrel (43). Aspirin and clopidogrel inhibit platelet activation by blocking cyclooxygenase and ADP receptors, respectively. Cy alone failed to control tumors, and the additional AP also had no antitumor effects in this model (Fig. 8A, left). Melanoma was controlled well with T cells plus Cy for about 1 month, but most mice eventually relapsed. In contrast, AP agents plus adoptive T cell transfer were highly effective against B16-F1, with relapse-free survival of most mice beyond 3 months (Fig. 8A, right). As a further proof, antigen-specific T cells were sustained at higher numbers in the blood, inguinal lymph nodes (ILNs), and spleens of mice receiving concurrent AP therapy and ACT (Fig. 8B). AP agents conferred no benefit when the transferred T cells lacked IFNγ (Fig. 8C) or when anti-IFNγ neutralization antibodies were administered (Fig. 8D).

Fig. 8 Pharmacological inhibition of platelets enhances ACT of cancer.

(A) C57BL/6 mice were inoculated with B16-F1 subcutaneously on day 0, given Cy on day 7, followed by adoptive therapy with activated Pmel cells on day 8 (Cy + T; n = 7 to 8 per group) or not (Cy; n = 4 to 5 per group). Each of the above groups of mice received concurrent aspirin and clopidogrel or water. Left: Average tumor growth curves. Right: Recurrence-free survival. (B) Pmel cells in the peripheral blood (day +62 after tumor inoculation), ILNs, and spleens (upon sacrifice) of mice in different treatment groups were enumerated by flow cytometry. (C) C57BL/6 mice were inoculated with B16-F1 subcutaneously on day 0. Mice were lymphodepleted with Cy on day 9, followed by adoptive transfer of either activated Pmel WT cells or IFNγ−/− Pmel cells on day 10. AP were given as described in (A) (TWT, WT Pmel cells; TKO, IFNγ KO Pmel cells; n = 4 to 10 per group). (D) C57BL/6 mice were inoculated with B16-F1 subcutaneously on day 0. Mice were lymphodepleted with Cy on day 9, followed by adoptive transfer of activated Pmel cells on day 10. IFNγ-neutralizing antibody (clone XMG1.2, BioXCell) was delivered intraperitoneally at 100 μg per mouse every other day starting on day 11 until sacrifice. AP were given as described in (A) (n = 4 to 8 per group). Repeated-measures two-way ANOVA was used to compare the tumor growth curves in (A), (C), and (D). Kaplan-Meier curves and log-rank tests were used for relapse-free survival analysis. Two-tailed independent Student’s t test was used in (B). Data are means ± SEM.

DISCUSSION

The role of platelets in promoting cancer invasion has been previously observed (44, 45). Multiple mechanisms have been attributed to this phenomenon including the promotion of angiogenesis (46) and stimulating epithelial-mesenchymal cell transition (9). However, the direct contribution of platelets to anticancer immunity has not been well described despite the emerging appreciation of the cross-talk between platelets and the host immunity. The current study uncovers that platelets directly dampen T cell function both in vitro and in vivo. Furthermore, we demonstrated that the PR suppresses both CD4+ and CD8+ T cells mostly via TGFβ and, to a lesser extent, through lactate. It is intriguing that both lactate and TGFβ are enriched in the tumor microenvironment, whose source so far has been attributed mostly to cancer cells and other stromal cells (32, 47). This study revealed that platelet-related TGFβ activation contributes dominantly to this immunosuppressive pool in cancer via cell surface TGFβ-docking receptor GARP. This conclusion is supported by enhanced tumor-specific T cell immunity in mice with platelet-specific deletion of GARP or its critical molecular chaperone gp96.

Platelets are known to respond to tissue injury and infection. Upon activation, platelets self-aggregate and release a variety of soluble factors to promote tissue homeostasis (48). Multiple molecules in the PR have immunomodulatory properties (2, 23). We identified TGFβ and lactate to be the major mediators. Platelet contribution to extracellular TGFβ can be accomplished through the release of prestored TGFβ in the cytoplasmic granules or via the ability of surface GARP on platelets to snatch and bind TGFβ from nonplatelet sources. Our study demonstrates that platelet-intrinsic GARP plays the most dominant role in activating TGFβ and thus likely contributes significantly to the immunosuppressive molecular hallmarks in the cancer microenvironment. Platelets are known to express GARP constitutively and to up-regulate its expression upon activation. The other cells that are known to express GARP are Treg cells. We found that conditional deletion of GARP from Treg cells is not as effective as platelet-specific KO of GARP in supporting ACT (fig. S10). Future studies are necessary to understand the roles and mechanisms of platelets broadly and the GARP-TGFβ axis specifically in regulating the biology of endogenous T cells in the tumor microenvironment such as differentiation and functionality.

There have been inconsistent reports on the systemic TGFβ level as a reliable biomarker for cancer, inflammation, and other conditions (4951). Consistent with the literature, we found that active TGFβ level is low in the plasma; however, after platelet activation and coagulation, serum active TGFβ level increased significantly. It has been unclear where active TGFβ in the serum comes from and what is the underlying mechanism of activation. The current study has resolved these long-standing puzzles. By genetically deleting GARP or TGFβ1 from platelets selectively, we have now reached several important conclusions: (i) Platelet-specific GARP is responsible for TGFβ activation because little active TGFβ can be detected in the sera of mice with platelet-specific deletion of either GARP or gp96; (ii) LTGFβ in the blood (both serum and plasma) is primarily supplied by platelets as revealed by platelet-specific TGFβ1 KO mice and platelet depletion studies; and (iii) the serum level of active TGFβ depends on the cell surface GARP-TGFβ complex, not the total level of soluble LTGFβ. Such evidence derived from the fact that, although platelet-specific TGFβ1 KO mice have drastically reduced soluble LTGFβ1 in the serum, they remain capable of making a comparable level of active TGFβ.

Consistent with the genetic studies, pharmacological platelet inhibitors were found to be effective in potentiating ACT of melanoma. It is also possible that platelet inhibitors may have other antitumor mechanisms, such as blocking angiogenesis and immunosuppressive prostaglandins (52), which contribute to their antitumor activity. Notwithstanding, our work demonstrated that the AP agents alone do not have significant antitumor activity in our model. In addition, the improved antitumor effect was abolished when IFNγ was removed from the system, demonstrating that platelet inhibition promotes antitumor efficacy via an immunologically based mechanism. Given the clinical availability of multiple platelet inhibitors targeting distinct pathways of platelet activation, we hope that our study will catalyze a systematic effort to optimize cancer immunotherapy by simultaneously blocking platelets and immune checkpoint molecules in prospective clinical trials.

Platelets have also been shown to play positive roles in the homing of T cells to sites of inflammation, to mediate a positive feedback loop of T cell recruitment through T cell activation via platelet CD40 (14), and to promote liver cancer induced by dysfunctional liver-directed T cell responses (53). The complexity of the roles of platelets in the tumor microenvironment is also illustrated by the finding that platelets can be extensively educated by tumor cells to uptake tumor-associated biomolecules such as RNAs (54). However, our work strongly indicates that the net effect of platelets in cancer patients is to promote immune evasion of cancer. Physiologically, cancer represents a chronic nonhealing wound, whose progression and metastasis are inevitably accompanied by vascular endothelial damage and local exposure to multiple platelet activators (1, 48). Our study thus suggests that cancer hijacks the tissue-repairing and hemostatic functions of platelets to suppress antitumor T cell immunity. A combination therapy with AP agents and immunotherapeutical modalities may thus represent a new paradigm for rational treatment of cancer in the future.

Last, although this study uncovers major suppressive molecules in the PR in an unbiased approach, such molecules were identified using in vitro experiments and translated to in vivo models using a hypothesis-driven approach. For example, TGFβ was identified as a major suppressive molecule secreted by platelets and this was validated in tumor mouse models. Although TGFβ was shown to be of biological relevance, it remains possible that, in vivo, the PR has a different composition and molecules other than TGFβ could play stronger roles. Furthermore, standard-of-care immunological therapies for melanoma at this point in time are mostly based on checkpoint inhibition {PD ligand–1 [PD(L)-1] and cytotoxic T lymphocyte–associated antigen 4 (CTLA-4)}. Various forms of adoptive T cell transfer have shown very promising results in clinical trials but are still under investigation and not yet standard of care. Our mouse models are based on adoptive T cell transfer and not checkpoint inhibition, so understanding these differences in immunological therapies is important when designing future studies.

MATERIALS AND METHODS

Study design

In this study, an unbiased approach was used to identify the major T cell suppressors in the PR. This was achieved by fractionating the releasate as described below, screening for the active fractions, and subsequently identifying the active molecules. End points for in vitro experiments included T cell proliferation, blastogenesis, and cytokine production and activation markers. The clinical relevance of the in vitro findings was investigated in vivo. For in vivo experiments, each group contained between 4 and 10 mice; this provided enough power and validity to detect biologically relevant phenomena while ensuring the use of minimal numbers of mice necessary as per the guidelines of the Medical University of South Carolina (MUSC) Institutional Animal Care and Use Committee. Commercially obtained mice were randomly assigned to different groups in each experiment. For in house–bred, genetically engineered mice, littermates were assigned for comparison groups. Efficacy end points for in vivo experiments were tumor size, T cell engraftment, and cytokine secretion. Mice were sacrificed when they showed signs of severe moribund disease. Measurement techniques for in vitro and in vivo experiments are indicated accordingly for each experiment. All experiments were performed at least two times. Numbers of key experiments were indicated in the figure legend. Blinding was not feasible for most of the in vitro experiments and was not critical because data collection was mostly through objective measures such as flow cytometry. The surgical pathologist scoring immunohistochemical (IHC) stains was blinded to the identity of the samples. For in vivo experiments, researchers were blinded because genetically modified mice are not readily identifiable, as their outer appearances are comparable and they share the same cages. In Fig. 3, blinding was maintained for the CD4 TH17 adoptive transfer experiment, but not the CD8 transfer one.

Mice

Platelet-specific Hsp90b1 KO mice were generated by crossing Pf4-cre mice (55) with Hsp90b1flox/flox mice (21, 56). Lrrc32flox/flox mice were obtained from Riken (Japan) (20). Tgfb1flox/flox, Foxp3eGFP-CreERT2, Pmel-1, and TRP1 mice were purchased from the Jackson Laboratory. Ifng−/− Pmel-1 mice were a gift from S. Mehrotra (MUSC). All animal experiments were conducted under approved protocols by the Institutional Animal Care and Use Committee at MUSC.

Ex vivo stimulation of tumor-draining lymph nodes

ILNs from tumor-bearing mice were isolated, mashed in cold phosphate-buffered saline (PBS), and filtered. One million cells per well were cultured in 96-well plates for 4 hours in the presence of phorbol 12-myristate 13-acetate (PMA) (500 ng/ml) and ionomycin (1.5 μg/ml), or hgp100 peptide (Lys-Val-Pro-Arg-Asn-Gln-Asp-Trp-Leu, 25–33) (5 μg/ml) for melanoma-draining lymph nodes. Brefeldin A (BD Biosciences) was added to the cells in all experiments.

Preparation of releasate from activated platelets

Mice were anesthetized, and blood was withdrawn to a 5-ml tube containing another 0.5 ml of acid citrate dextrose (ACD) buffer [39 mM citric acid, 75 mM sodium citrate, 135 mM dextrose, and prostaglandin E1 (1 μg/ml) (pH 7.4)]. Samples were centrifuged for 10 min at 100g, and the upper layer of platelet-rich plasma was collected. Platelets were washed twice with citrate washing buffer [128 mM NaCl, 11 mM glucose, 7.5 mM Na2HPO4, 4.8 mM sodium citrate, 4.3 mM NaH2PO4, 2.4 mM citric acid, 0.35% bovine serum albumin, and prostaglandin E1 (50 ng/ml) (pH 6.5)], then resuspended in RPMI 1640, enumerated by a blood cell counter, and diluted to a final concentration of 1 × 108/ml. Purified platelets were incubated with thrombin (1 IU/ml) for 45 min at 37°C (125 rpm). Stimulated platelets were sedimented by centrifugation at 3200g for 15 min, and supernatant was collected. Microvesicles (MVs) were collected by centrifugation at 25,000g and resuspension in PBS in the same volume as the original PR volume. For human platelet secretome, platelets from healthy donors were obtained from the blood bank at MUSC, resuspended in ACD buffer at room temperature, and then activated as above.

Size-based and anion exchange fractionation

PR fractionation was carried out using a Pharmacia Akta–fast performance liquid chromatography system, and columns were purchased from GE Healthcare. The first fractionation was based on size by loading PR onto a Superdex 200 column and eluting with PBS or RPMI 1640 (Gibco). The active fractions were pooled and dialyzed with phosphate buffer (20 mM; pH 7.2). The resulted material was then loaded onto a diethylaminoethyl (DEAE) column and eluted with a linear gradient of NaCl from 0 to 1 M.

NMR-based metabolite profiling

Human PR was prepared at concentrations of 2 × 109/ml and fractionated by size exclusion chromatography, followed by anion exchange chromatography. Fractions were eluted with sodium phosphate buffer (pH 7.0; 20 mM final concentration) containing sodium 3-trimethylsilyl-2,2,3,3-d4-propionate (TSP; 0.1 mM final concentration) and 10% D2O. NMR data were collected at 298 K on a Bruker Avance III 600 MHz NMR spectrometer (Bruker BioSpin Inc.) equipped with a 5-mm cryogenically cooled QCI-inverse probe. Solvent suppression was achieved using the excitation sculpting scheme (57). Typically, one-dimensional (1D) 1H NMR spectra with a 7-s recycle delay were acquired with a total of 128 transients in addition to four dummy scans. Real data points (32,768) were collected across a spectral width of 12 parts per million (ppm) (acquisition time, 2.27 s). Data were zero-filled to twice the original data set size, manually phased, and automatically baseline-corrected using TopSpin 3.1 software (Bruker BioSpin Inc., Billerica, MA), and a 1.0-Hz line-broadening apodization was applied before spectral analysis. The singlet produced by the known quantity of the TSP methyl groups was used as an internal standard for chemical shift referencing (set to 0 ppm) and quantification.

Metabolite assignments were established after comparison of chemical shifts and spin-spin couplings with reference spectra as implemented in the Chenomx NMR Suite (Chenomx Inc., Edmonton, Alberta, Canada) profiling software (version 7.72). Specifically, quantification was achieved using the Chenomx 600 MHz metabolite library (version 8). Confirmatory 1D 31P and 2D 1H,13C-multiplicity–edited heteronuclear single quantum correlation spectra with adiabatic 13C inversion, refocusing, and decoupling were recorded for selected PR fractions to enhance metabolite identification by comparison of 13C chemical shifts with the biological magnetic resonance data bank. Concentration of LA was then quantified by an l-lactate assay kit (Eton Bioscience).

In vitro T cell culture suppression assay of PR

CD4+ and CD8+ T cells were purified using magnetic beads to a purity of ≥95%. Cells (1 × 105) were cultured in 96-well plates precoated with anti-CD3ε antibody (3 μg/ml) in the presence of IL-2 (100 U/ml) and soluble anti-CD28 (2 μg/ml), together with either media or PR. On day 3 of culture, T cells were stimulated with PMA (500 ng/ml)/ionomycin (1.5 μg/ml) or hgp100 peptide 25–33 (5 μg/ml) for 4 hours in the presence of GolgiPlug (BD Biosciences), followed by staining for relevant markers. TGFβ receptor signaling was blocked using the combination of an ALK5 inhibitor (SB431542, Selleckchem) at 20 μM and anti-TGFβ antibody (clone MAB1835, R&D Systems) at 2 μg/ml. LA activity was inhibited by blocking the monocarboxylate transporter using α-cyano-4-hydroxycinnamic acid (C2020, Sigma-Aldrich). For carboxyfluorescein diacetate succinimidyl ester (CFSE) dilution assays, cells were labeled with 5 μM CFSE for 10 min at room temperature before culture on day 0. Flow cytometry was then performed, and the data were analyzed and displayed with FlowJo software. Suppression index by PR was calculated as percentage of undivided cells treated with a given fraction of PR/percentage of undivided cells in the control media.

Adoptive T cell therapy

Treatment of B16-F1 melanoma by adoptive transfer of ex vivo activated Pmel T cells was done as described previously (29). To test the effect of AP agents on ACT, we administered clopidogrel by oral gavage 3 days after T cell transfer and then every 48 hours until day 25. Aspirin was administered through drinking water (150 mg/liter) starting 2 days before T cell transfer and was replaced every 48 hours afterward. For TRP1 TH17 T cell therapy, single-cell suspensions of splenocytes from Rag1−/− TRP1 mice (27) were seeded with irradiated C57BL/6 splenocytes pulsed with TRP1 106–130 peptide (SGHNCGTCRPGWRGAACNQKILTVR; American Peptide). To obtain TH17-polarized cells, we added recombinant human IL-6 [100 ng/ml; National Institutes of Health (NIH)], TGFβ1 (30 ng/ml; BioLegend), IL-1β (10 ng/ml; Shenandoah), and anti–IL-4 and anti-IFNγ antibodies (10 μg/ml; BioXCell) to the cultures. On the second day of culture, complete medium containing recombinant human IL-2 and IL-23 (40 ng/ml; PeproTech) was added. Where appropriate, human PR was added at 100% on days 0, 2, and 4. Cells were cultured for 5 days before experimentation. C57BL/6 mice were injected subcutaneously with 4 × 105 B16-F10 melanoma cells and treated 10 days later with TRP1-specific CD4+ T cells. Recipient mice were lymphodepleted using 5-gray total body irradiation on the day before cell transfer. Tumor growth was measured using calipers, and the products of the perpendicular diameters were recorded. In some experiments, IFNγ-neutralizing antibody (clone XMG1.2, BioXCell) or isotype control antibody was administered via intraperitoneal injection at 100 μg per mouse every other day starting on day 12 until sacrifice (58). The adoptive T cell transfer experiment with Ifng KO Pmel-1 cells was done identically as above except that Pmel cells were isolated from Ifng KO Pmel-1 mice (59). For some experiments, adoptive transfer was supplemented with exogenous IL-2–anti–IL-2 complexes on days 0, 2, 4, and 6 after transfer in the absence of lymphodepletion (60). Briefly, 1.5 μg of human IL-2 (National Cancer Institute Biological Resources Branch Preclinical Repository) was mixed with 7.5 μg of anti–IL-2 monoclonal antibody (clone 5355, R&D Systems) for 15 min at room temperature. Cytokine complexes were administered via intraperitoneal injections.

Platelet depletion and serum collection for TGFβ ELISA

WT mice were given one dose of rabbit anti-mouse thrombocyte polyclonal sera (1:40, Cedarlane) at 500 μl per mouse intraperitoneally in sterile-filtered PBS. Blood was collected at 0, 24, 48, and 72 hours. Serum was harvested via coagulation and centrifugation (12,000g).

Measurement of TGFβ via ELISA

Mouse serum or plasma samples were collected by pricking the lateral tail vein. Capture ELISA for TGFβ1 was performed according to the manufacturer’s instructions (BioLegend). Total TGFβ1 was measured after acidic activation using 1 M HCl for 10 min at room temperature.

Activation of human T cells

Human samples were isolated from buffy coats (Pennsylvania Plasma). CD8+ T cells were positively enriched from human peripheral blood mononuclear cells, followed by negative isolation of CD4+ T cells using magnetic isolation kits (Invitrogen). Cells were then stimulated with anti-CD3/anti-ICOS beads (Dynal) at 10:1 T cell/bead ratio for 4 days. IL-2 (100 IU/ml; NIH) was added to the T cell cultures. Human PR was added on day 0 and every other day thereafter. Cells were assayed on day 7 for cytokine production and phenotype.

MC38 model

MC38 tumor cells were obtained from Y.-X. Fu (University of Texas Southwestern Medical Center). WT or Plt-GARPKO mice were injected in the right flank with 1 × 106 MC38 colon cancer cells. Tumor size was measured with digital caliper kinetically. Tumor-infiltrating lymphocytes were isolated from fresh primary tumors by density gradient after single-cell suspensions were made with mechanical dissociation and enzymatic digestion (deoxyribonuclease and collagenase).

Immunohistochemistry

For p-Smad2/3 stain on fresh-frozen MC38 tumors, 5-μm tumor sections were fixed with 4% paraformaldehyde followed by incubation with 3% H2O2. To minimize nonspecific staining, we incubated the sections with the appropriate animal serum for 20 min at room temperature, followed by incubation with primary anti–p-Smad2/3 antibody (EP823Y, Abcam) overnight at 4°C. Staining with secondary antibodies (Vectastain ABC Kit) was then performed before development using DAB substrate (SK-4100, Vector Labs). The staining intensity of p-Smad2/3 was graded as follows, with the sample identity blinded: 0, negative; 1, faint; 2, moderate; 3, strong but less intense than 4; 4, intense.

Statistical analysis

Two-sided, two-sample Student’s t tests were used for all comparisons involving continuous dependent variables and categorical independent variables using Excel software. The variances were compared between groups using an F test. The Student’s t test was then implemented assuming equal or unequal variances (i.e., if the F test P value was less than 0.05, then unequal variances were assumed). For tumor curves, two-way repeated-measures analysis of variance (ANOVA) was used. For dose-response correlations between continuous dependent and independent variables in Figs. 2 and 4, Spearman’s rank-order correlation test was used to determine rho (ρ). Kaplan-Meier curves were compared using log-rank tests. Error bars represent SEM. NS denotes statistically nonsignificant difference.

SUPPLEMENTARY MATERIALS

immunology.sciencemag.org/cgi/content/full/2/11/eaai7911/DC1

Table S1. Source data for all the figure panels with small n (n < 20).

Fig. S1. Pf4-cre-Hsp90b1flox/flox (KO) mice show no noticeable immune dysfunction at baseline.

Fig. S2. PR, but not MVs, directly suppresses T cell proliferation and differentiation in vitro.

Fig. S3. Immune suppression by PR is independent of TCR signaling and specific to lymphocytes.

Fig. S4. T cell–suppressive function of the whole PR is significantly, but not completely, neutralized by blocking TGFβ pathway.

Fig. S5. A small–molecular weight, heat-stable, proteinase K–resistant T cell–suppressive fraction is shared between human and mouse PRs.

Fig. S6. TGFβ and LA in the PR are the major suppressors of CD8+ T cell activation.

Fig. S7. TGFβ contained in the PR drives Foxp3 expression and up-regulates p-Smad2/3.

Fig. S8. TGFβ1, but not LA, abrogates CD8-mediated tumor control.

Fig. S9. Platelet depletion has no effect on serum LA concentration.

Fig. S10. Inducible deletion of GARP from Foxp3+ Treg cells does not improve ACT of melanoma.

Fig. S11. Sample staining and isotype controls for flow cytometry.

REFERENCES AND NOTES

Acknowledgments: We received technical help from Y. Zhang, F. Hong, E. Ansa-Addo, S. Mehrotra, J. Thaxton, C. Morales, J. Bowers, S. Suriano, C. Cloud, and T. Benton. We thank Y.-T. Hsu, L. Ball, R. Drake, and J. Bielawski for their initial assistance with PR fractionation. Funding: This work was supported by multiple NIH grants CA186866, CA188419, AI070603, and AI077283 (to Z.L.); CA175061 and CA208514 (to C.M.P.); UL1 TR001450 and NIH–National Center for Advancing Translational Sciences Grant TL1 TR001451 (to C.W. and B.R.); and Hollings Cancer Center Cancer Center Support Grant P30CA138313. Author contributions: Z.L. and S.R. conceived the idea, designed the study, and wrote the manuscript. S.R., A.M., B.R., B.X.W., M.H.N., C.W., C.M.P., M.P.R., M.H., and D.W.B. performed the experiments. Z.L., B.L., and Y.Y. supervised the study. E.G.-M. assisted in data analysis, statistics, and interpretation. All authors provided critical comments on the manuscript. Competing interests: A provisional patent application has been filed to target GARP and platelets for cancer immunotherapy. The authors declare no other competing interests.
View Abstract

Navigate This Article