The Programmed Death-1 Immune-Suppressive Pathway: Barrier to Antitumor Immunity

The programmed death-1 (PD-1) pathway is essential for maintaining peripheral T cell tolerance and is critical for attenuating autoimmunity and maintaining T cell homeostasis. However, this pathway is also a deterrent to antitumor immunity. Advanced cancer patients who have failed all other therapies have impressive responses when treated with mAbs that block this pathway, either as monotherapy or in combination with mAbs that block signaling through CTLA-4 (1–4). The PD-1 pathway includes the receptor, PD-1 (CD279) and two ligands: programmed death ligand-1 (PD-L1; also named B7 homolog 1 or CD274) and programmed death ligand-2 (B7-DC or CD273). The receptor and its ligands are type 1 transmembrane proteins and are members of the B7/CD28 family of ligands and receptors that includes both costimulatory (CD28) and coinhibitory (PD-1, CTLA-4) receptors. The ligands PD-L1 and programmed death ligand-2 are coinhibitory, whereas CD80 is costimulatory when bound to CD28 but coinhibitory when bound to CTLA-4 (Fig. 1A). PD-1 consists of a single extracellular IgV domain, a transmembrane region, and a cytoplasmic domain that includes an ITIM and immunoreceptor tyrosine-based switch motif (5, 6). PD-L1 consists of extracellular IgV and IgC domains, a transmembrane region, and an intracellular domain (7) (Fig. 1B). Because PD-L1 is an established impediment to antitumor immunity and is either constitutively expressed or induced on most carcinoma cells and also can be expressed by immune cells relevant in tumor immunity (e.g., dendritic cells [DCs], myeloid cells, and T cells), this review focuses on the role of the PD-1 pathway in antitumor immunity.

The role of PD-1 in programmed cell death (apoptosis) was first recognized in the early 1990s (5). It was subsequently shown that PD-1 expression on activated T cells results in T cell death, and it was proposed that the autoimmunity observed in PD-1–knockout mice was due to a breakdown of tolerance to self-Ags (8). PD-1 and the receptors CD28 and CTLA-4 share structural and functional characteristics, suggesting that the ligand for PD-1 might be similar to the ligands for CD28 and CTLA-4: CD80 (B7.1) and CD86 (B7.2). By screening human and mouse databases for genes with sequence homology to CD80, both the human and mouse ligands for PD-1 were identified (9, 10). Soon after its discovery, PD-L1 was recognized as a cancer immunotherapy target because of its widespread expression on many cancer cells and because blockade of the PD-1 pathway reduced tumor progression, whereas overexpression of PD-L1 promoted tumor progression in mice (11–14).

Because the PD-1 pathway plays a central role in downregulating activated T cells in the periphery, it is important during infection and autoimmunity, as well as in tumor immunity. Multiple studies with PD-1–deficient mice demonstrated its critical role in dampening down T cell responses after the clearance of pathogens and in preventing autoimmunity. In contrast to CTLA-4, which predominantly regulates the early stages of T cell activation, PD-1 acts on activated T cells (reviewed in Ref. 15). PD-1 itself is a marker of activated T cells, because its expression is induced only after T cell activation. The pathway appears to affect the ability of activated T cells to kill tumor cells (16), as well as the survival of activated T cells (17).

Many human tumor cells either constitutively express or are induced to express PD-L1. These include cervical, pancreatic, urothelial, gastric, esophageal, renal cell, hepatocellular, head and neck squamous cell, ovarian, breast, non-small cell lung, and bladder carcinomas, as well as cutaneous and uveal melanoma, various leukemias, multiple myeloma, and glioma. PD-L1 is present in the cytoplasm and plasma membrane of both mouse and human tumors; however, not all tumors or all cells within a tumor express PD-L1 (14, 18). Activated immune cells, including DCs, NK cells, macrophages, monocytes, B cells, T cells, and nonhematopoietic cells, also may express PD-L1 (9, 10, 13, 19). PD-L1 expression is induced by multiple proinflammatory molecules, including types I and II IFNs, TNF-α, LPS, GM-CSF, and VEGF, as well as the cytokines IL-10 and IL-4, with IFN-γ being the most potent inducer (20, 21). Because IFN-γ and TNF-α are produced by activated type 1 T cells, and GM-CSF and VEGF are produced by a variety of tumor stromal cells, the tumor microenvironment upregulates PD-L1 expression and, thereby, promotes immune suppression. This latter effect is called “adaptive immune resistance,” because the tumor protects itself by inducing PD-L1 in response to IFN-γ produced by activated T cells (18).

PD-L1 is regulated at both the transcriptional and translational levels. PI3K activation of Akt is essential for transcription of PD-L1 mRNA (22), and inhibition of the PI3K pathway reduces the immune resistance of PD-L1⁺ tumor cells (23) (Fig. 1A). PD-L1 is also regulated by anaplastic lymphoma kinase via the STAT3 transcription factor (24). In addition, the PD-L1 promoter region contains an NF-κB responsive element, and pharmacological inhibition of NF-κB activation inhibits PD-L1 expression (21). At the translational level, PD-L1 expression is suppressed by the PTEN gene. Cancer cells frequently contain mutated PTEN, which activates the S6K1 gene. The S6K1 gene, in turn, shifts PD-L1 mRNA to polysomes and, hence, increases the translation of PD-L1 mRNA and increases plasma membrane expression of PD-L1 (22). Micro-RNAs also translationally regulate PD-L1 expression. miRNA-513 is complementary to the 3′ untranslated region of PD-L1 and prevents PD-L1 mRNA translation. Treatment with IFN-γ downregulates miRNA-513 and, thereby, facilitates PD-L1 mRNA translation (25).

PD-L1 and PD-1 suppress antitumor immunity and promote tumor progression by inactivating T cells, protecting tumor cells, and activating tumor-suppressive cell populations (Fig. 2). Activated NKT cells (26), B cells (27), and DCs (28) also can express PD-1 and be suppressed through the PD-1 pathway. Early studies demonstrated that PD-L1⁺ murine and human tumor cells induce apoptosis of activated T cells and that Ab blocking of PD-L1 facilitates antitumor immunity (13, 14). PD-L1⁺ APCs anergize autoreactive T cells (29), and the anergy is reversed by mAbs to PD-L1 (30). It is likely that the persistence of PD-L1⁺ tumor cells similarly anergizes, functionally inactivates, and/or tolerizes PD-1⁺ tumor-reactive and tumor-infiltrating T cells (TILs). Blockade of PD-1 and T cell Ig mucin-3, a marker of T cell exhaustion, reverses exhaustion and reduces tumor growth (31).

Ligand binding to PD-1 blocks the downstream signaling events triggered by Ag/MHC engagement of the TCR and costimulation through CD28, resulting in impaired T cell activation and IL-2 production (Fig. 1A). Signaling through the TCR requires phosphorylation of the tyrosine kinase ZAP70. Once docked on phosphorylated tyrosine residues of the TCR-associated CD3ζ chain, phosphorylated ZAP70 activates downstream adapter molecules and enzymes. PD-1 engagement reduces the phosphorylation of ZAP70 and, hence, inhibits downstream signaling events. PD-1 ligation also prevents phosphorylation of PKC-theta, which is essential for IL-2 production (32), and arrests T cells in the G₁ phase, blocking proliferation. PD-1 mediates this effect by activating Smad3, a factor that arrests cycling (33).

PD-1 also reduces T cell survival by impacting apoptotic genes. During T cell activation, CD28 ligation sustains T cell survival by driving expression of the antiapoptotic gene Bcl-x_L. PD-1 prevents Bcl-x_L expression by inhibiting PI3K activation, which is essential for upregulation of Bcl-x_L. Because CD28-mediated upregulation of Bcl-x_L is relatively unaffected by CTLA-4 ligation, these experiments demonstrated that PD-1 and CTLA-4 suppress via distinct mechanisms (34). Signaling through PD-1 also prevents the conversion of functional CD8⁺ T effector memory cells into CD8⁺ central memory cells (35) and, thus, reduces long-term immune memory that might protect against future metastatic disease. This effect occurs because PD-1 ligation decreases T cell survival by upregulating the proapoptotic factor Bim (36).

PD-1 ligation also facilitates downmodulation of the TCR. The TCR is downregulated as a part of a normal T cell response and is controlled, in part, by the E3 ubiquitin ligase Cbl-b (37). PD-L1 binding to PD-1 increases Cbl-b expression and, therefore, increases TCR downregulation (38), which may serve as another method that tumor cells use to escape T cell immunity. mAb blockade of PD-1 and PD-L1 leads to a decrease in Cbl-b expression, which abrogates TCR internalization.

Microscopy studies revealed some of the morphological events that occur during PD-1–mediated T cell suppression. Successful T cell activation involves the accumulation of naive T cells around APCs. Stable engagement of PD-1 by PD-L1 restricts this homing by limiting T cell mobility and, thus, prevents efficient Ag presentation (39). During T cell activation, involved receptors redistribute in the T cell plasma membrane and form peripheral and central supramolecular activation clusters. T cell activation involves binding of receptors within the microclusters with ligands on the APCs. When bound by ligand, PD-1 translocates to microclusters within the central supramolecular activation clusters. The phosphatase SHP2 is then transiently bound to the immunoreceptor tyrosine-based switch motif of PD-1 and blocks activation of downstream mediators (40), including the kinases Akt and PI3K, which are activated through CD28 (32).

PD-L1 also promotes tumor progression by protecting PD-L1⁺ tumor cells from CTL-mediated and Fas-mediated lysis (“molecular shield”) (41), as well as by reverse signaling through CD80 into T cells. Reverse signaling results in tolerizing CD80⁺ T cells in the periphery and is due to the binding of PD-L1 to T cell–expressed CD80 (42). CD80–PD-L1 interactions restrain self-reactive T cells in an autoimmune setting (43); therefore, their inhibition may facilitate antitumor immunity.

The PD-1 pathway is also involved in generating immune-suppressive regulatory T cells (Tregs). APC-expressed PD-L1 induces natural Tregs (nTregs) in the thymus and converts peripheral naive CD4⁺ T cells to inducible Tregs. The latter conversion involves TGF-β and is the result of blocking the Akt/mTOR pathway. PD-L1 simultaneously sustains the survival and increases the suppressive activity of inducible Tregs by maintaining and increasing their expression of Foxp3 (44). Engagement of PD-1 also converts mature Th1⁺ CD4 T cells to Foxp3⁺ Tregs (45, 46). PD-1–induced expansion of nTregs is amplified by IFN-γ–driven upregulation of PD-L1 and reverse signaling through nTreg CD80 (47).

Many immunohistochemical studies assessed whether tumor cell expression of PD-L1 is associated with tumor progression and/or survival (48-54). The majority of these studies concluded that higher expression of PD-L1 in the tumor microenvironment correlates with poor prognosis and reduced survival time and/or tumor grade. These studies used a variety of anti–PD-L1 Abs, some of which are not optimal for histology. Most of these studies considered cells PD-L1⁺ if they expressed either cytoplasmic or plasma membrane staining. Because cell surface, but not cytoplasmic, PD-L1 is physiologically relevant, the validity of these reports is controversial. In studies of melanoma (18) and Merkel cell carcinoma (55) patients using the well-validated 5H1 anti–PD-L1 Ab and plasma membrane PD-L1 expression as the criterion, tumor cell expression of PD-L1 positively correlated with improved survival. Using the same Ab and assessment criteria, melanoma and non-small cell lung cancer PD-L1 expression positively correlated with improved survival following PD-1 Ab therapy (2, 56). These latter findings were interpreted as confirming the concept that increased PD-L1 was due to the presence of IFN-γ–secreting tumor-reactive TILs. However, studies using the same assessment criteria and anti–PD-L1 Ab revealed an inverse correlation between renal cell carcinoma expression of PD-L1 and survival (57), and they identified PD-L1 as a prognostic indicator of poor survival for urothelial carcinoma patients receiving bacillus Calmette–Guérin therapy (58). The discrepancies between these reports could be due to differences between various types of cancer. Additional studies are needed to resolve this issue.

PD-L1 is also present in some cancer patients in a soluble form. Soluble PD-L1 in the serum of some renal cell carcinoma patients induced apoptosis of PD-1⁺ T cells and correlated with tumors of advanced stage and grade and increased risk for death (59). Similarly, soluble PD-L1 in the plasma of patients with large B cell lymphoma was identified as a biomarker of poor survival (60).

The first report demonstrating that blockade of the PD-1 pathway facilitated tumor rejection used Abs to PD-L1 and PD-1–deficient mice (11). Several studies performed immediately thereafter confirmed these initial findings (13, 61), and many subsequent studies expanded the efficacy of PD-1 pathway blockade to multiple types of tumors. For example, adoptive transfer of ex vivo–activated tumor-specific T cells, followed by in vivo treatment with mAbs to PD-L1, delayed tumor progression and increased survival of mice with squamous cell carcinoma (62) and acute myelogenous leukemia (46). In mice with B16 melanoma, simultaneous inhibition of CTLA-4 and PD-1, combined with administration of a Flt3 ligand–transduced tumor cell vaccine, dramatically increased survival rates and viable TILs that produced higher levels of IFN-γ and TNF-α, as well as increased the ratio of T effector cells to both Tregs and myeloid-derived suppressor cells (63). Similar findings were observed for a combination therapy of mAbs to PD-L1 or PD-1 with a GM-CSF vaccine in mice with CT26 colon carcinoma or ID8 ovarian carcinoma (64). In a murine rhabdomyosarcoma system, monotherapy with anti–PD-1 mAbs had limited therapeutic effect; however, inclusion of mAbs against CXCR2, a receptor expressed by myeloid-derived suppressor cells, synergized with Ab therapy (65).

PD-1 pathway mAb treatment also has been combined with mAb therapies aimed at facilitating T cell activation by triggering costimulatory molecules. Treatment with mAbs to the costimulatory molecule 4-1BB (CD137) has therapeutic efficacy. However, in mice with PD-L1⁺ tumor cells, this therapy is not effective, and concomitant treatment with mAbs to PD-1 or PD-L1 is needed (66). CD137 therapy also synergizes with mAbs to PD-1 or PD-L1, as well as with GM-CSF cancer vaccine or Flt3 ligand–transduced tumor cell vaccine (67). The latter studies are particularly noteworthy because late-stage tumors were responsive.

Exhausted/dysfunctional TILs of cancer patients express PD-1 (68), as well as other receptors that inhibit T cell activation. LAG-3 (CD223) and T cell Ig mucin-3 are two such additional coinhibitory TCRs. Combination therapy with mAbs to PD-1 plus LAG-3 reduced tumor progression of MC38 colon carcinoma and Sa1N fibrosarcoma and was significantly more effective than either mAb alone. Treatment of melanomas containing few TILs with immunostimulatory RNA polyinosinic-polycytidylic acid increased type I IFN, which, in turn, upregulated PD-L1. Although polyinosinic-polycytidylic acid by itself temporarily restrained tumor progression, addition of PD-1 blockade extended survival time (69). These studies should be viewed with caution because therapy usually was initiated early during tumor growth when stromal cells and vasculature were likely to be minimal; however, they do indicate that PD-1 pathway inhibition is most effective when combined with treatments that activate the immune system.

PD-1 pathway blockade also has been combined with chemotherapy or radiotherapy. Early treatment of mice with Panc02 adenocarcinoma with gemcitabine, the standard of care for pancreatic cancer, combined with PD-1 blockade increased CD8⁺ TILs and intratumoral immune activation and improved survival (70). Combination of irradiation with mAbs to PD-L1 is much more effective than either therapy alone, and it rendered mice with MC38 colon carcinoma or TUBO breast carcinoma resistant to rechallenge with tumor. The combination therapy is most likely effective because irradiation upregulates PD-L1 expression in the tumor microenvironment, and concomitant administration of mAbs to PD-L1 minimizes this suppression (71).

The critical role of the PD-1 pathway in human cancer was firmly established in recent clinical trials in which 17–28% of patients with advanced cancers had partial or complete remissions following treatment with mAbs to PD-L1 or PD-1 (1, 2). Follow-up studies with a subset of patients demonstrated that posttreatment responses lasted as long as 3 y and that a partial responder who had tumor recurrence responded to reinduction therapy (72). PD-1 mAb therapy also was effective in 28% of advanced melanoma patients who were nonresponsive to ipilimumab therapy (anti–CTLA-4 mAbs) (73), and ipilimumab plus anti–PD-1 mAb therapy yielded an objective response rate of 53% (3). These responses are impressive because they occurred in advanced cancer patients who were nonresponsive to conventional therapies. However, the response rates also indicate that inhibition of the PD-1 pathway by itself is insufficient for the treatment of all patients.

Immunotherapy strategies that inhibit the PD-1 pathway are based on the hypothesis that eliminating PD-1–mediated suppression permits the natural development of tumoricidal T cells. These strategies will be effective provided the tumor is sufficiently immunogenic and activated APCs can present Ag. However, it is likely that not all cancer patients will have spontaneously activated tumor-reactive T cells and that strategies aimed at directly activating T cells will be needed for optimal antitumor immunity. This concept is supported by murine studies that demonstrated that concurrently blocking the PD-1 pathway and using a vaccine or other means to activate tumor-reactive T cells is more efficacious than PD-1 monotherapy (63, 64, 67). These findings led to the development of strategies that combine PD-1 pathway blockade and enhancement of Ag presentation in a single reagent.

The finding that CD80 binds to PD-L1 with a binding affinity approximately equal to its affinity for CD28 (74) suggested that CD80 could bind to PD-L1 and sterically block PD-1–PD-L1 interactions. Studies with human and mouse tumor cells demonstrated that soluble CD80 (CD80-Fc consisting of the extracellular CD80 IgV and IgC regions fused to an IgG Fc domain) and membrane-bound CD80 prevented PD-1 from binding to PD-L1⁺ tumor cells and maintained IFN-γ production by activated PD-1⁺ CD4⁺ and CD8⁺ T cells (75, 76).

Surprisingly, soluble CD80 was more effective than were mAbs to PD-L1 or PD-1 in maintaining T cell activation, suggesting that CD80-Fc prevented PD-1:PD-L1 binding, as well as mediated other effects (76). Studies with CD28-deficient mice and with Abs that prevented CD80:CD28 binding demonstrated that CD80 also costimulated T cells via CD28 (77). Although not shown experimentally, soluble CD80 may also prevent T cell apoptosis by reverse signaling of PD-L1 into CD80⁺ T cells, because the binding of soluble CD80 to PD-L1 could block PD-L1 from binding to T cell–expressed CD80. PD-1 mAb cannot prevent reverse signaling; however, PD-L1 mAbs could, provided that they block the PD-L1 binding site for CD80.

There are potential drawbacks to using soluble CD80 as an immunotherapeutic. Although soluble CD80 has the potential to suppress via CTLA-4, sustained production of IFN-γ and blocking studies with mAbs to CTLA-4 in the presence of CD80-Fc suggested that this suppression does not occur (L.A. Horn and S. Ostrand-Rosenberg, unpublished observations). The absence of CTLA-4–mediated suppression may be due to the inability of CD80-Fc to cross-link CTLA-4 or because PD-L1 blocks the CTLA-4 binding site on CD80. If the latter is occurring, then it is interesting that the CTLA-4, but not the CD28, binding site on CD80 is affected because the two sites were considered to overlap (78). Fig. 3 summarizes the known ligand–receptor interactions that soluble CD80 impacts, as well as additional interactions that soluble CD80 has the potential to impact.

Clinical trials and animal studies established PD-1 and PD-L1 as major deterrents to T cell–mediated antitumor immunity. Other coinhibitory receptor–ligand systems also may contribute to T cell dysfunction in individuals with cancer. However, the PD-1 pathway may play a key role because PD-L1 is expressed by most tumor cells, as well as by immune cells and other host cells in the tumor microenvironment. Preclinical and ongoing clinical trials are demonstrating that treatments combining PD-1 pathway blockade with inhibition of other coinhibitory pathways, as well as combination treatments that enhance Ag presentation, may yield positive responses in even more patients. The clinical findings are very encouraging and have raised intense interest in cancer immunotherapy. Although much is known about the PD-1 pathway in tumor immunity, many issues remain unexplored. For example, does therapy with soluble CD80 or with mAbs to PD-1 or PD-L1 inhibit tolerance induction by reverse signaling by PD-L1 into CD80⁺ T cells? Does mAb or soluble CD80 therapy generate memory T cells capable of protecting against recurrent disease? Will combining immunotherapies be additive, synergistic, or redundant? Which immune therapies, when combined with inhibition of the PD-1 pathway, will be most efficacious? Does blockade of the PD-1 pathway result in activation of only tumor-reactive T cells or also T cells that are autoreactive to nontumor Ags and that could cause undesirable autoimmunity? Will immunotherapies such as soluble CD80 that combine blockade of the PD-1 pathway with costimulation of tumor-reactive T cells in a single reagent be more effective than combinations of monotherapies? Would a soluble CD80 engineered to lack the CTLA-4 binding site provide enhanced T cell activation? Addressing these questions will provide mechanistic information on how blockade of the PD-1 pathway facilitates antitumor immunity and is likely to provide new insights for designing better immunotherapeutics for the treatment of cancer. Thus, additional basic immunology studies are an essential complement to ongoing clinical trials.

The authors have no financial conflicts of interest.