PD-1 expression on dendritic cells suppresses CD8+ T cell function and antitumor immunity

ABSTRACT Programmed death one (PD-1) is a well-established co-inhibitory regulator that suppresses proliferation and cytokine production of T cells. Despite remarkable progress in delineating the functional roles of PD-1 on T lymphocytes, little is known about the regulatory role of PD-1 expressed on myeloid cells such as dendritic cells (DCs). Here, we show that CD8+ T cells can be more potently activated to secrete IL-2 and IFNγ by PD-1-deficient DCs compared to wild-type DCs. Adoptive transfer of PD-1-deficient DCs demonstrated their superior capabilities in inducing antigen-specific CD8+ T cell proliferation in vivo. In addition, we provide first evidence demonstrating the existence of peripheral blood DCs and CD11c+ tumor-infiltrating myeloid cells that co-express PD-1 in patients with hepatocellular carcinoma (HCC). The existence of PD-1-expressing HCC-infiltrating DCs (HIDCs) was further supported in a mouse model of HCC. Intratumoral transfer of PD-1-deficient DCs rendered recipient mice resistant to the growth of HCC by promoting tumor-infiltrating CD8+ effector T cells to secrete perforin and granzyme B. This novel finding provides a deeper understanding of the role of PD-1 in immune regulation and has significant implications for cancer immunotherapies targeting PD-1.


Introduction
PD-1, a member of the immunoglobulin superfamily, is inducibly expressed on activated T and B lymphocytes. 1,2 The critical role of PD-1 in the control of peripheral tolerance and lymphocyte activation is supported by the development of various autoimmune phenotypes in PD-1-deficient mice. [3][4][5][6][7][8] Co-inhibitory signals in lymphocytes are initiated by binding of PD-1 to its ligands PD-L1 (B7-H1) and PD-L2 (B7-DC). 9,10 Engagement of PD-1 results in the recruitment of the tyrosine phosphatase SHP-2, which dephosphorylates and inactivates proximal effector molecules such as Zap70 in T cells. [11][12][13][14] The immediate outcomes of PD-1-mediated signaling in T cells are the inhibition of both T cell proliferation and secretion of cytokines such as IFNg and IL-2.
PD-1 acts as a rheostat to modulate immunity by increasing the threshold of antigen responses and reducing the cytotoxic lymphocyte activity of CD8 C T cells. 15 Such unique properties endow PD-1 with immunoregulatory functions during chronic viral infections [16][17][18][19] and tumor progression. A critical observation that the blockade of PD-1:PD-L1 interactions can restore the capacity of exhausted CD8 C T cells to attack tumors 20,21 led to clinical trials using antibodies targeting PD-1 22 and PD-L1. 23 These studies have produced impressive results, reporting significant and durable responses in several types of cancer. Thus, PD-1 blockade is considered a promising target for cancer immunotherapy, 24,25 with a number of therapeutic antibodies targeting PD-1, including nivolumab (Opdivo; BristolMyers Squibb/Ono Pharmaceuticals) and pembrolizumab (Keytruda; Merck & Co.), have recently been approved by the FDA for the treatment of metastatic melanoma.
Although PD-1 is known to be expressed on T lymphocytes, evidence has emerged indicating that other non-lymphoid innate cells also express PD-1, suggesting that it may play additional roles in immune regulation that are mediated by these cells. 1 A number of recent studies demonstrated that expression of PD-1 on DCs impedes DC survival, 26 reduces pro-inflammatory cytokine production, 27 and diminishes innate immunity against bacterial infection. 28 Whether expression of PD-1 on DCs also regulates CD8 C T cell function or plays a role in immune surveillance remains to be seen.
In this study, we show that expression of PD-1 on DCs plays a crucial role in the suppression of IL-2 and IFNg secretion in vitro and of antigen-specific CD8 C T cell proliferation in vivo. Our study provides the first evidence demonstrating the existence of PD-1-expressing DCs in peripheral blood and CD11c C myeloid cells in cancerous tissues derived from patients with HCC. Intratumoral transfer of PD-1-deficient DCs was beneficial in suppressing tumor growth in a mouse model of HCC, and deficiency of PD-1 expression in DCs rendered mice resistant to tumor progression through increased CD8 C T cell function. Our findings provide a deeper conceptual understanding of the regulation of CD8 C T cell function and antitumor immunity by DC-expressed PD-1.

PD-1 expression on DCs suppresses T cells responses in vitro and in vivo
To evaluate whether T cell activation was influenced by expression of PD-1 on DCs, we measured cytokine production from OT-I T cells stimulated with WT or PD-1 KO bone marrowderived dendritic cells (BMDCs), pre-pulsed with antigenic OVA peptide (SIINFEKL) at concentrations ranging from 1 pg to 10 ng/mL. We observed significantly higher production of IL-2 ( Fig. 1A) and IFNg (Fig. 1B) from T cells activated by antigen-pulsed PD-1 KO BMDCs compared with WT BMDCs (p < 0.05). Induction of IL-2 and IFNg secretion required the presence of antigen-presenting DCs, as DCs or T cells alone failed to stimulate cytokine responses (Fig. 1A, 1B). Consistent with previous reports, 28 we observed that WT BMDCs do not upregulate surface PD-1 expression following Poly I:C stimulation (Fig 1C), unlike splenic DCs (Fig. S1). However, the differential regulation of T cell activation by WT and PD-1 KO BMDCs indicates that PD-1 is functional in this system. We hypothesized that the induction of PD-1 expression in BMDCs is crucially dependent on their intercellular contacts with T cells. Indeed, PD-1 expression was significantly upregulated in WT BMDCs (gated on DAPI -, CD3e -, TCR-Va2 -, CD11c C , MHC-II C cell population), following co-culture with OT-I T cells (p D 0.018) (Fig. 1C, 1D). BMDCs from WT and PD-1 KO mice displayed similar levels of peptide-linked MHC complexes at the cell surface (Fig. S2). Furthermore, the expression of several other cell surface signaling molecules, such as CD40, CD80 and CD86 were expressed at similar levels in WT and PD-1 KO BMDCs (Fig. S3), indicating that these factors did not contribute to the differential capacity of PD-1 KO BMDCs to modulate T cell function.
To establish the role of PD-1 on DCs in regulating T cell proliferation in vivo, we stimulated intravenously transferred CFSE-labeled, CD45.2 C OT-I T cells by adoptive transfer of SIINFEKL-pulsed WT or PD-1 KO BMDCs into the footpad of WT CD45.1 C recipient mice ( Fig. 2A). To minimize the variation of T cell responses triggered by different recipient mice, adoptive transfer of BMDCs was performed into different footpads of the same recipient mice. At three days post adoptive T cell transfer, a significantly higher level of T cell proliferation (gated on DAPI -, CD45.2 C , CD3e C , CD8a C , and CFSE low cell population) was measured in the draining lymph node in the presence of PD-1 KO BMDCs compared with WT BMDCs (71.7 § 7.7% vs. 50.9 § 9.5%, N D 5, p D 0.0052) (Fig. 2B). These results indicate that DCs are able to elicit antigen-specific T cell proliferation to a greater extent in the absence of PD-1 expression.

Human peripheral blood dendritic cells co-express PD-1
To establish the relevance of DC expression of PD-1 in human disease, we characterized PD-1 expression in DC subsets from human peripheral blood using flow cytometry (Fig. 3A). In the steady state, a low level of PD-1 was expressed in CD141 hi (or BDCA3 hi ) myeloid DCs (mDCs), but not in CD1c C (or BDCA1 C ) mDCs and CD123 C plasmacytoid DCs (pDCs) in healthy donors (Fig. 3B, 3D). However, expression of PD-1 was found in all DC subsets (mDCs and pDCs) in the peripheral blood of HCC (HCC) patients. CD14 C monocytes from both healthy donors and HCC patients were negative for PD-1 expression (Fig. 3B, 3C).
To investigate if PD-1 expression could be induced upon activation, we cultured human blood DC subsets and CD14 C monocytes from healthy donors in medium with Poly I:C or PBS for 18 h (red boxes, Fig. 3A). PD-1 expression in CD141 hi DCs, which exhibited low levels of PD-1 expression ex vivo, was unchanged by Poly I:C stimulation. However, we observed that PD-1 expression in CD1c C DCs and CD123 C pDCs could be induced by Poly I:C stimulation. CD14 C monocytes remained negative under all test conditions (Fig. 3D).
Human and murine liver cancer-infiltrating CD11c C myeloid cells co-express PD-1 It was previously reported that tumor-associated myeloid cells express PD-L1, which transmits co-inhibitory signals to PD-1expressing tumor-infiltrating T lymphocytes (TILs), thus suppressing the antitumor activity of TILs. 20,29 However, little is known about the functional role of PD-1 expressed on myeloid cells in the regulation of antitumor immunity. To assess whether PD-1 was present on liver cancer-associated myeloid cells, we performed immunofluorescence staining on cancerous tissues derived from patients with HCC. Consistent with previous findings, a large fraction of HCC-infiltrating CD3 C T cells expressed PD-1. Notably, we also observed expression of PD-1 on CD3 -CD11c C myeloid cells in the same tissue (Fig. 4A). Given that CD11c is expressed on all conventional myeloid DCs, we hypothesized that these cells represented HIDCs that expressed PD-1 in the tumor microenvironment.

Effects of PD-1 on DCs in regulating antitumor effects on primary established HCC
Although expression of PD-L1 on tumor-infiltrating DCs is known to regulate immune suppression, the function of PD-1 expression on these cells remains elusive. To investigate the role of PD-1 expressed on DCs in T cell suppression in the tumor microenvironment, we performed an adoptive transfer of WT or PD-1 KO BMDCs into Hepa1-6 transplantable tumor-bearing mice. Although intratumoral transfer of DCs from both WT and PD-1 KO mice resulted in a significant suppression of growth of primary established subcutaneous HCC (as early as 3-7 d post DC transfer), PD-1 KO DCs suppressed tumor growth more efficiently than WT DCs, suggesting that immune surveillance against tumors is tightly regulated by PD-1 expression on DCs (Fig. 5A).

Discussion
The co-inhibitory functions of T cell-expressed PD-1 have been studied extensively over the last decade, but the functional role of PD-1 expressed on myeloid cells, including DCs, was unclear. Using PD-1 deficient mice, 4 we demonstrate that   PD-1 expression is regulated by distinct pathways in T cells and DCs. In T cells, PD-1 expression is induced by TCR triggering; 30 however in murine splenic DCs, it is regulated by TLR signaling instead. 28 Consistent with previous findings, 28 PD-1 expression was induced in murine splenic DCs after activation by Poly I:C in vivo (Fig. S1), but was not inducible in activated BMDCs per se. Poly I:C upregulated surface expression of other co-stimulatory molecules, including CD40, CD80, and CD86, in BMDCs (Fig. S3), indicating that appropriate test conditions were used. The inducible PD-1 expression in WT BMDCs was crucially dependent on their intercellular contacts with T cells, but was not due to the acquirement of PD-1 molecules from neighboring T cells through trogocytosis, 31,32 since PD-1 surface expression level was not detected in PD-1 KO BMDCs in the same DC:T co-culture system.
The ability of PD-1 in BMDCs to modulate T cell function was not related to differences in antigen presentation or the expression of other cell surface signaling molecules because these were unchanged between WT and PD-1 KO BMDCs. The suppression of T cell proliferation mediated by PD-1 on BMDCs may be caused by ligation of PD-1 to PD-L1, which is also expressed on the surface of T cells. 1 Alternatively, T cell inhibition may be caused by reverse signaling in DCs triggered by PD-1 engagement, which reduces DC survival signals 26 and pro-inflammatory cytokine production. 27 In addition, antibodies targeting PD-1 in co-culture experiments with T cells and PD-1 KO DCs significantly promoted IL-2 and IFNg cytokine secretion, confirming the role of PD-1 expression on T cells in suppressing cytokine responses (Fig. S4). Taken together, these results suggest that both PD-1 and PD-L1 on DCs are capable of delivering bidirectional and synergistic signaling to inhibit T cell responses.
Although the co-inhibitory functions of PD-1 expressed on tumor-infiltrating T cells have been studied extensively, the functional role of PD-1 on tumor-infiltrating myeloid cells, the presence of which we report for the first time in HCC patients, requires further investigation. In an ovarian tumor mouse model, it was recently shown that PD-1-expressing tumor-associated DCs inhibited T cell proliferation and activation through downregulation of NF-kB activation, cytokine production, and co-stimulatory molecule expression. 27 However, the ability of PD-1 expressed on DCs to regulate tumor growth in vivo has not been demonstrated. Selective deletion of PD-1 on DCs in mouse tumor models allows a precise evaluation of the roles of PD-1 in antitumor immunity. Unfortunately, DC-specific PD-1-deficient mice are not yet available. Nevertheless, our in vivo DC transfer experiments provide the first functional insights into the role of PD-1 expression in DCs in modulating T cell responses directly in the tumor microenvironment. Using a mouse model of HCC, we show that tumor growth can be effectively suppressed following the transfer of PD-1-deficient DCs. This is accompanied by an expansion of perforin-and granzyme B-secreting tumor-infiltrating CD8 C T cells, which are crucial in suppressing tumor growth. Thus, in addition to the previously understood functions of PD-1 in inhibiting innate immunity, 28 cytokine production 27 and survival signals 26 in DCs, our results suggest a novel role for PD-1 expression in DCs in regulating effector functions of CD8 C TILs and antitumor immunity.
Cancer immunotherapy targeting the PD-L1:PD-1 pathway aims to restore the function of exhausted T cells in the tumor microenvironment. 25,33 Therapeutic antibodies targeting PD-1 22 for cancer immunotherapy have shown remarkable results in clinical trials, with two agents (nivolumab and pembrolizumab) recently gaining FDA approval. It is well established that immunosuppressive signals can be transmitted via PD-1 on T cells to inhibit their antitumor activity. 20,29 Thus, one of the predicted mechanisms of action of these agents is to block PD-1:PD-L1 interactions by directly binding to PD-1 on TILs in the tumor microenvironment. However, many studies have suggested that PD-1 blockade could potentially invoke other mechanisms of action. 34 In this study, we provide the first evidence of the existence of blood-circulating DCs and CD3 -CD11c C tumor-infiltrating myeloid cells that co-express PD-1 in HCC patients, suggesting that a clinical blockade targeting PD-1 may have an important functional impact directly on circulating DCs and tumor-infiltrating myeloid cells in cancer patients.
Our study provides novel functional insights into the regulation of immune responses in the tumor microenvironment by Perforin-and granzyme B-secreting CD8 C T cells were evaluated by intracellular staining followed by flow cytometry. All cells were gated on the CD45 C , CD3e C , and CD8 C population. Quadrants were established with reference to the isotype controls. A representative density plot from one mouse is shown. (C) Percentages of perforinand granzyme B-expressing CD8 C T cells and total tumor-infiltrating CD8 C T cells 7 d post intratumoral transfer of WT and PD-1 KO BMDCs. Cell percentages are relative to total CD45 C live singlet T cells; Ã p < 0.05, unpaired t-test, N 3 per group.
expression of PD-1 on DCs, and will contribute to a better understanding of the consequences of cancer immunotherapy targeting PD-1.

Materials and methods
Mice C57BL/6 (CD45.2 or CD45.1 background) mice were purchased from the Biological Resource Center (BRC, A Ã STAR, Singapore). OT-I.Rag1-/-mice 35,36 were provided by Taconic from the National Institute of Allergy and Infectious Diseases Exchange Program (# 004175; Bethesda, MD). PD-1 KO (C57BL/6-Pdcd1-/-) mice 4 were kindly provided by Prof. Tasuku Honjo (Kyoto University) and were imported from the RIKEN Bioresource Center, Japan. All mice were maintained in the SPF animal facility of the Biological Resource Center (BRC) of Biopolis in Singapore.

Ethics statement
Mice studies were carried out in strict accordance with the recommendations of the Institutional Animal Care and Use Committee (IACUC) of the Biological Resource Center (BRC) of Biopolis in Singapore. The BRC IACUC protocol was approved by the National Advisory Committee for Laboratory Animal Research in Singapore (Permit Numbers: 110626 and 120773). Peripheral blood and resected tumor samples were obtained with Ethics Committee approval from HCC patients who underwent curative resection. Formalin-fixed paraffin-embedded tissue sections were used for immunofluorescence staining.

Cell isolation and culture
Splenic CD8 C T cells were purified from OT-I mice by negative selection using the CD8a C T cells isolation kit II (Miltenyi Biotec). Splenic DCs were purified from WT or PD-1 KO mice by positive selection using CD11c C microbeads (Miltenyi Biotec). All cell isolations were performed according to the manufacturer's protocols. OVA peptide (SIINFEKL) was purchased from Anaspec Corporation. Freshly isolated splenic T cells were maintained at 37 C in RPMI-1640 medium containing 10% fetal bovine serum, 100 IU/mL Penicillin, 100 mg/mL Streptomycin, 2 mM Glutamine, and 50 mM b-mercaptoethanol (all from GIBCO). Bone marrow-derived DCs (BMDCs) from 8-to 12-week-old WT and PD-1 KO mice were prepared as previously described. 37 After 7-10 d of culture, BMDCs were analyzed for CD11c expression and used in assays when > 85% were CD11c positive. The Hepa1-6 38 HCC cell line was prepared and maintained as described previously. PBMCs were isolated from blood using Ficoll-Paque PREMIUM (GE Healthcare). Human DC subsets were sorted by flow cytometry and cultured at 37 C in X-Vivo 15 medium (Lonza) with 2% human serum (Off the Clot; Gemini), 100 IU/mL Penicillin (GIBCO) and 100 mg/mL Streptomycin (GIBCO).

Flow cytometry
BMDCs, splenic DCs and OT-I T cells were stained with antibodies against cell surface molecules, and were analyzed using a BD FACS LSR flow cytometer. BMDCs were co-cultured with or without CD8 C OT-I T cells in the presence of 0.1 mg/mL OVA peptide (SIINFEKL) and 10 mg/mL Poly I:C (Invivogen) in 96-well plates (5 £ 10 4 BMDC and 5 £ 10 4 OT-I T cells per well) for 2 d before FACS analysis. Human DC sorting was performed using a BD FACS Aria flow cytometer. Intracellular staining was performed using BD GolgiPlug according to the manufacturer's protocol. Dead cells were discriminated from live cells using DAPI (Sigma-Aldrich) or the LIVE/DEAD® Fixable Blue Dead Cell Stain Kit (Molecular Probes). Antibodies purchased from eBioscience were as follows: hamster anti-

ELISA cytokine assay
BMDCs were pre-pulsed for 4 h with different concentrations of OVA peptide (SIINFEKL) in the presence or absence of 10 mg/mL Poly I:C (Invivogen) and were then co-cultured with OT-I T cells in 96-well plates (5 £ 10 4 DC and 5 £ 10 4 T cells per well) without washing. IL-2 and IFNg secretion were measured by ELISA 48 h later. For blocking experiments, 10 mg/ mL of purified antibodies (functional grade) against PD-1/ CD279 (RMP1-30) (all from eBiosciences) or their corresponding isotype controls were added into DC-T co-cultures.

In vivo T cell proliferation assay
BMDCs from WT or PD-1 KO mice were pre-pulsed with or without 1 mg OVA peptide (SIINFEKL) in medium containing 10 mg/mL Poly I:C for 4 h and washed twice with PBS, before injecting them together with 50 mg Poly I:C into the footpads of CD45.1 C WT congenic recipient mice (3 £ 10 6 BMDCs per footpad). OT-I T cells (CD45.2 C background) were labeled with 5 mM CellTrace TM CFSE (Molecular Probes, Invitrogen) and transferred intravenously into recipient mice (10 6 T cells per mouse). After three days, draining popliteal lymph nodes (LNs) were collected and T cell proliferation was analyzed by flow cytometry.

Transplanted tumor mouse model
The Hepa1-6 transplanted tumor mouse model was established as described previously. 38 Briefly, 5 £ 10 6 Hepa1-6 cells were subcutaneously injected into each flank of 6-to 8-week-old WT C57BL/6 mice (day 7). One week post Hepa1-6 tumor transplantation (day 0), tumor-bearing mice were injected intratumorally with 100 mL PBS containing 1 £ 10 6 WT or PD-1 KO BMDCs or 100 mL PBS only (control with no DC transfer). Tumor growth was monitored on the indicated days by palpation or using a digital caliper from day 1 to 14 before the tumors were harvested and analyzed by flow cytometry.

Spontaneous liver tumor mouse model
Mice expressing the hepatitis virus B surface antigen in their hepatocytes to promote chronic inflammation 39 and the liverspecific Sleeping Beauty transposon to drive spontaneous mutagenesis 40 were generated. Aged mice (11-to 18-month-old) with liver tumors were sacrificed and livers were harvested for analysis by flow cytometry.

Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.