B cells are abundantly present in the TME.57 58 A recently conducted single-cell RNA-sequencing study on human lung cancer stromal cells demonstrated that B cells were the most enriched cell type in tumor stroma.59 However, the precise role of B cells in tumor immunity is debatable, as paradoxical roles have been reported. On one hand, tumor-infiltrating B cells have been reported to promote tumor growth by producing inhibitory cytokines, including IL-10 and TGF-β60 61 and by interacting with either immune cells via PD-L160 or tumor cells through CD40/CD154 signaling, as shown in hepatocellular carcinoma (HCC).62 On the other hand, tumor-infiltrating B cells reportedly delay tumor growth by secreting antibodies against the tumor,63 producing proinflammatory cytokines, including IL-12,64 antigen presentation,65 or inducible T cell costimulator ligand (ICOSL) / inducible T cell costimulator (ICOS interaction.66 The contrasting roles of tumor-infiltrating B cells may be caused by their high heterogeneity.59 Interestingly, recent studies suggest the involvement of B cells in PD-1 blockade therapy response,67–69 showing clonal expansion, and a unique functional state of B cells as responders to PD-1 blockade therapy.68 Although the underlying mechanism remains largely unknown, these studies suggest a potential role of PD-1 signaling in controlling the functions of tumor-infiltrating B cells.

In addition to T cells, B cells have also been reported to express PD-1.70 71 The role of PD-1 on B cells was first observed in PD-1^−/− mice, which exhibited moderate splenomegaly and increased B cell proliferation in response to anti-IgM antibody stimulation.72 Further, resting human B cells express PD-1 at a basal level, while its expression is rapidly induced on stimulation of the toll-like receptor 9 (TLR9) pathway by CpG-B.70 Subsequent PD-1/PD-L1 interaction inhibits the B cell receptor signaling pathway, resulting in the attenuation of cytokine production and proliferation of B cells.71 Recently, a novel tumor-promoting subset of B cells has been identified that expresses high levels of PD-1 in human HCC.25 These PD-1^high B cells constitute approximately 10% of the total B cell population in advanced HCC. Unlike conventional regulatory B cells (Bregs), PD-1^high B cells exhibit a CD5^highCD24^−/−CD27^high/+CD38^dim phenotype. PD-1 expression is highly induced by hyaluronan fragments and TLR agonists, including Pam3CysSK4, lipopolysaccharide (LPS), and oligodeoxynucleotides containing CpG motifs.25 TLR4-induced BCL6 upregulation plays a dominant role in the induction of PD-1 on B cells. The interaction between PD-L1 and PD-1^high B cells induces the expression of IL-10, a well-defined immune-suppressive cytokine. These data suggest that the contribution of B cells in PD-1-based immunotherapy should be critically evaluated and PD-1 signaling must be further investigated in different B cell subsets (figure 3). Besides, follicular T-helper cell (Tfh), which exhibit high PD-1 expression, plays a critical role in promoting maturation of B cells.73 Given PD-1 signaling is essential for positioning and function of Tfh, it is reasonable to speculate that PD-1 blockade therapy could affect B cells in an indirect manner.74 In this regard, the role of Tfh on B cells in PD-1 needs blockade therapy which need further study.

NK cells were first identified in 1975.75 76 Since that time, they have been classified as lymphocytes based on their morphology, phenotype, and origin, and are considered as a part of innate immunity due to the absence of antigen-specific receptors.77 The functional status of NK cells is mediated by the cumulative effect of multiple activating and inhibitory receptors. When activated, NK cells kill virus-infected cells and malignant tumor cells in an major histocompatibility complex (MHC) non-restricted manner.78 NK cells also secrete various inflammatory cytokines with antitumor effects, including IFN-γ and tumor necrosis factor (TNF)-α. Indeed, the presence of NK cells in solid tumors has been described as a good prognostic factor.79–82

Studies have suggested three potential contributions made by PD-1 signaling in NK cells during PD-1 blockade therapy: (1) HLA-I is generally not expressed in human cancer cells, resulting in no interaction between HLA-I and CD8 T cells83 84; however, PD-1-based immunotherapy is effective in these tumors. For instance, 79% (85/108) of the classical Hodgkin’s lymphomas (cHL) exhibit low to no expression of MHC class I molecules.19 Meanwhile, PD-1 blockade therapy is highly effective in the treatment of relapsed or refractory cHL.18 (2) Tumor cells with high tumor mutation burden (TMB) are more likely to be recognized by CD8 T cells, and patients carrying tumors with TMB, such as malignant melanoma85 and non-small cell lung cancer,86 are more likely to benefit from PD-1 blockade therapy.2 However, even in tumors with low TMB, including malignant melanoma and non-small cell lung cancer, a small fraction of patients are responsive to PD-1 blockade therapy.85 86 (3) Human NK cells express PD-1 in various cancers, including Hodgkin’s lymphoma.87–89 Hsu et al27 used different mouse models to investigate PD-1 signaling in NK cells and its role in PD-1-based immunotherapy. Interestingly, PD-1 was found to be expressed by tumor-infiltrating NK cells with an activated but exhausted phenotype. Furthermore, NK cells contributed to the antitumor effect of PD-1-based immunotherapy.27 However, this study had several limitations. For instance, although PD-1 expression on human NK cells has been well characterized, its expression on mouse NK cells remains debatable, as they do not express PD-1 even under robust activation conditions, such as cytomegalovirus infection.90 This study concluded the PD-1 expression on NK cells by flow cytometry, while a recent study suggested that dying immune cells express a nuclear antigen, which cross-reacts with mouse anti-PD-1 monoclonal antibody, leading to a false positive PD-1 staining on NK cells.91 Second, the molecular mechanisms underlying the effect of PD-1-based immunotherapy on NK cells remains unclear. Conversely, a recent study challenged the expression of PD-1 on mouse and human NK cells under diverse conditions.92 Overall, the expression and effect of PD-1 on NK cells remains unclear. However, combination therapies including PD-1 blockade and NK cell activation strategies are currently in clinical trials.93 Further studies are urgently needed to address the expression of PD-1 as well as its precise role in NK cells.

Lastly, considering that NK cells are a small minority of cells in the TME of multiple cancer types, other cell types, such as CD4 T cells94 95 and macrophages96 may function cooperatively with NK cells to exert antitumor effects during PD-1 blockade therapy. In a study of patients with relapsed or refractory cHL, the authors demonstrated that PD-L1 expression and MHC class II positivity on Hodgkin Reed-Sternberg cells are favorable prognostic biomarkers of PD-1 blockade therapy, which potentiate an alternative CD4 T cell-mediated mechanism of response to PD-1 blockade therapy.97 Accordingly, a study suggest that cytotoxic CD4 T cells are essential to the efficacy of PD-1 blockade therapy on MHC class II-expressing tumors.98 Intriguingly, a recent study suggested a tumor-promoting role for PD-L1 reverse signaling in promoting tumor cell growth, proliferation and metabolism in cHL, which may also participate in PD-1 blockade therapy.99

Macrophages that infiltrate solid tumors are referred to as tumor-associated macrophages (TAMs).100 High expression of TAM markers, especially M2 markers, is generally associated with poor prognosis of patients with cancer.100 In most cases, bone marrow monocytes differentiate into TAMs following stimulation by TME factors, including cytokines and hypoxia.101 However, TAMs are also reportedly derived from myeloid progenitors present in the yolk sac.102 103 Additionally, monocyte-derived macrophages can be polarized following stimulation by cytokines and other environmental factors.

Macrophages are categorized as M1 or M2 based on their polarization. M1 refers to a proinflammatory state, which is generally driven by IFN-γ/LPS; while M2 refers to an anti-inflammatory state, generally mediated by IL-4 or IL-13.100 104 However, the M1/M2 model is an oversimplification and cannot precisely describe the polarized state of TAMs.105 In fact, TAMs express both M1 and M2 related markers concurrently.26 106 Therefore, provided the complexity of their origin and polarization state, TAMs exhibit high heterogeneity.107 Generally speaking, TAMs play a dominant role in promoting cancer progression by modulating nearly every aspect of tumor biology, including angiogenesis,108 metastasis,109 proliferation,110 immune suppression,111 112 inflammation113 and stem cell maintenance.114

Macrophages have been recently described as expressing PD-1 under specific conditions, such as tuberculosis,115 sepsis116 and zymosan-induced inflammation.117 However, its expression in macrophages is induced by TLR signaling, whereas in T cells, it is driven primarily by TCR signaling.117 PD-1 signaling in macrophages inhibits M1 polarization in-vitro via attenuation of STAT1/NF-κB phosphorylation.117 Furthermore, PD-1^−/− mice are markedly protected from lethal sepsis in vivo; however, the bactericidal effect is reversed, when macrophages are depleted by clodronate liposomes.116 Moreover, PD-1 blockade augments phagocytosis and intracellular killing activity of macrophages against BCG.115 Although these results reveal a potential inhibitory effect for PD-1 signaling on macrophages, the specific role of macrophage-associated PD-1 in tumor immunity remains unclear. In 2017, the research group of Sydney R. Gordon was the first to evaluate the expression and function of PD-1 on TAMs.26 Using flow cytometry and immunofluorescence, they reported that both human and murine TAMs express PD-1. Specifically, PD-1⁺ TAMs exhibited M2-like phenotype and accumulated in TME over time. Furthermore, in vitro and in vivo studies showed that PD-1⁺ TAMs had lower phagocytic properties, which were abrogated on PD-L1-knockout in mice. These results suggest that PD-1 is a functionally important M2-like marker. In fact, PD-1⁺ TAM infiltration correlates with poor prognosis in patients with human gastric cancer.118 Furthermore, PD-1 expression has been reported in murine tumorous myeloid cells using flow cytometry (antibody clone: RMP1-30). Intriguingly, PD-1 deletion in myeloid cells (PD-1^f/fLysMcre mice) effectively delays tumor growth, similar to that observed during global deletion of PD-1, whereas deletion of PD-1 in T cells (PD-1^f/fCD4cre mice) is less effective.119 These results indicate a crucial role for PD-1 in inhibiting the antitumor immunity of myeloid cells. Similarly, PD-1 expression was also reported in granulocyte/macrophage progenitors (GMPs), which increases during emergency hematopoiesis facilitating differentiation into myeloid-derived suppressor cells (MDSCs). Meanwhile, PD-1 deletion in myeloid cells (PD-1^f/fLysMcre mice) inhibits the accumulation of GMPs and MDSCs.119 Further, activation of ERK1/2, as well as the mTOR1 kinase complex by granulocyte colony-stimulating factor in myeloid cells, is inhibited by PD-1 expression. mTOR participates in the regulation of myeloid progenitor cell differentiation, while ERK1/2 controls the differentiation of APCs.119 Taken together, these data indicate that the effect of PD-1 blockade on TAMs should not be neglected, although it remains unclear whether patients with cancer with high PD-1⁺ TAM infiltration would also benefit from PD-1 blockade therapy.

In spite of their paucity, DCs play a central role in the initiation of antigen-specific immunity and tolerance.120 Similar to other tumor-infiltrating stromal cells, DCs exhibit high heterogeneity.120 The functional specificity of DC subpopulations results from the expression of different receptors, including PD-1.28 29 121 122 Furthermore, various inflammatory factors induce PD-1 expression on DCs, which then suppresses innate immunity against bacterial infections by inhibiting IL-12 and TNF-α,122 and promotes apoptosis of activated DCs.123 Since PD-1 expression on DCs is induced by inflammatory factors, it is possible that tumor-infiltrating DCs also express PD-1, driven by chronic inflammation, a hallmark of cancer. By using flow cytometry and immunofluorescence, James et al observed a similar phenomenon of PD-1 expression in tumor-infiltrating DCs using an implanted tumor model of ovarian cancer. In murine species, immature PD-1⁺ DCs exhibit a classical DC phenotype (CD11c⁺ CD11b⁺ CD8^-) that is suppressive, and respond weakly to danger signals. PD-1 signaling in mice has also been reported to inhibit NF-κB, a crucial signaling molecule involved in the maturation and activation of DCs.124 Similarly, PD-1⁺ DCs were also observed in human ovarian cancer, as determined by flow cytometry.28 Similar to its role in murine species, PD-1 expression on human DCs suppresses NF-κB-dependent cytokine release in a SHP-2-dependent manner. Further, PD-1 expression on DCs is induced by IL-10, a well-established suppressive cytokine, while PD-1 blockade leads to increased IL-10 production by DCs. In this regard, combined PD-1 blockade therapy with IL-10 neutralization induced a synergistic antitumor effect.29 These data are in agreement with that of another study, which suggested that PD-1 signaling in monocytes of HIV-infected individuals inhibits IL-10 production.125 Meanwhile, a recent study employed flow cytometry (antibody clone: EH12.2H7) to demonstrate that human tumorous DCs express PD-1. Moreover, PD-1 on DCs neutralizes PD-L1 in cis to inhibit canonical PD-1 signaling in T cells.126 However, these findings were exclusively based on in vitro experiments and, thus, may not reflect actual tumor-infiltrating DCs. Therefore, in vivo studies are warranted to examine the cis effect of PD-1 on DCs to modulate cancer immunity.

Until 2010, the accepted dogma was that PD-1 is specifically expressed by cells of the hemopoietic lineage. However, by using flow cytometry and immunofluorescence, Tobias et al reported that a subpopulation of melanoma cells express PD-1, and that PD-1⁺ cancer cells are responsible for tumor initiation.127 In this study, PD-1 was primarily examined as a biomarker, enriched in ABCB5⁺ malignant melanoma-initiating cells. Meanwhile, the function of PD-1 signaling in melanoma cells remained largely unknown until 2015. Tobias et al further demonstrated that human and murine melanoma cells contain PD-1-expressing subpopulations by using flow cytometry, immunofluorescence, RT-PCR and western blotting.22 In melanoma cells, intrinsic PD-1 signaling plays a key role in tumor initiation in an immunosuppression-independent manner. By interacting with its cognate ligand, melanoma-PD-1 triggers the activation of downstream effectors (eg, ribosomal protein S6) of mTOR signaling, which further accelerate tumor growth. Interestingly, intrinsic PD-1 signaling in melanoma cells activates downstream mTOR signaling in a PI3K/AKT-independent manner, distinct from that observed in canonical PD-1 signaling. In T cells, interaction of PD-1 with PD-L1/PD-L2 inhibits TCR signaling via SHP-2 tyrosine phosphatase.14 SHP-2 expression is reported in a number of cancer cell types, including melanoma,128 breast cancer129 and glioblastoma.130 In cancer cells, SHP-2 may activate mTOR signaling130; hence, the effect of PD-1 signaling on mTOR activation might be tissue specific. Consistent with this, both murine and human HCCs were reported to express PD-1.131 HCC cell-PD-1 promotes phosphorylation of eukaryotic initiation factor 4E (eIF4E) and ribosomal protein S6 (S6), leading to enhanced tumor cell proliferation.131 In addition, intrinsic PD-1 signaling in pancreatic cancer cells promotes tumor proliferation by decreasing the phosphorylation of MOB1, a central component of the Hippo signaling pathway.132 These data demonstrate that tumor cell-intrinsic PD-1 signaling exerts a protumor effect by activating mTOR in melanoma, HCC and pancreatic cancer cells (figure 4).

On the contrary, tumor cell-intrinsic PD-1 has been reported as a tumor suppressor gene in non-small cell lung carcinoma (NSCLC) and colon cancer.21 133 A previous study reported a 61-year-old woman diagnosed with stage IV NSCLC who was unresponsive to several chemotherapies and experienced rapid disease progression after receiving radiotherapy combined with pembrolizumab.133 PD-1 and PD-L1 expression in the irradiated tumor tissue biopsies obtained prior to pembrolizumab treatment were assessed by immunohistochemistry (IHC). Unexpectedly, the cancer cells displayed diffuse PD-1 staining. Meanwhile, RNA-sequencing data from lung cancer cell lines validated PD-1 expression in only 7 of 236 cell lines considered. Further, treatment with PD-1 antibodies (clone RMP1-14, rat IgG2a) accelerated the growth of M109 murine NSCLC in vitro and in vivo. However, the underlying mechanisms associated with these phenomena remain unclear. In line with this, a recent study shows that tumor cell-intrinsic PD-1 expression suppresses the canonical signaling pathways, including AKT and ERK1/2 in NSCLCs and colon cancer, while mTOR signaling remains unaffected.21 Treatment with either nivolumab (anti-PD-1 antibody) or ateolizumab (anti-PD-L1 antibody) enhances the growth of NCI-H1299 transplanted tumors in NOD-SCID IL-2 receptor gamma null mouse, one of the most immunodeficient mouse strains (figure 4). However, further studies are needed to explore the underlying molecular mechanisms responsible for these contradictory effects mediated by tumor cell-intrinsic PD-1 in melanoma, HCC and NSCLCs.

Furthermore, a recent study reported the potential for obtaining false-positive PD-1 staining in melanoma cells, including B16F1091 due to cross reaction of antibodies with a nuclear antigen. These results raised the question regarding whether epithelial tumor cells truly express PD-1. Although evidence of PD-1 expression in tumor cells has been provided at the mRNA (qRT-PCR and RNA-sequencing), and protein (IHC, immunofluorescence, western blotting, and FACS) levels in previous studies,21 22 127 131 133 we re-evaluated the expression of PDCD1 using the Cancer Cell Line Encyclopedia (Broad, 2019) in lung cancer (NSCLC and SCLC), colorectal, breast, melanoma, kidney, stomach and liver cancer cell lines. Our data suggest that majority of epithelium-derived cancer cells express PDCD1 at extremely low, to undetectable, levels (RPKM <1; an RPKM of 1 is considered as a threshold and is equivalent to one mRNA copy per cell134) (figure 5). The low to undetectable expression of PDCD1 in cancer cell lines may be attributable to its expression only in a subpopulation of cancer cells.21 22 Therefore, single-cell RNA-sequencing of cancer cells might be helpful in validating the expression of PDCD1 in specific cancer cell subpopulations and characterizing the signaling involved in the regulation of PDCD1 in cancer cells.

PD-1-based immunotherapy has achieved great clinical success and has been approved for the treatment of various cancers.16 17 135 136 However, only a fraction of the patient population benefits from PD-1 blockade therapy.137 Even with the help of validated biomarkers, including PD-L1 and TMB, the response rate remains low (eg, approximately 40% in patients with NSCLC with PD-L1 expression in at least 50% of the cancer cells16). Furthermore, a small fraction of patients with cancer receiving PD-1-based immunotherapy succumbs to HPD.20 Hence, there is a critical need to determine the cause for the beneficial or deleterious effects of PD-1 blockade therapy in patients.

Note, although treatment with anti-PD-1 antibodies affects all PD-1-expressing cells, since conventional CD4 T and CD8 T cells often account for the dominant cells expressing PD-1, effects of PD-1 blockade on other PD-1-expressing cells have been neglected. In fact, the antitumor effect induced by PD-1 blockade therapy is a cumulative effect of its influence on all PD-1-expressing cells. In this regard, cells associated with non-canonical PD-1 signaling should also be evaluated as biomarkers for combinatorial therapeutic strategies.

To date, most of the validated biomarkers related to PD-1 blockade therapy, including PD-L1,138 139 TMB86 136 and CD8 T cell infiltration,140 are based on the mechanisms underlying CD8 T cell-driven antitumor activity. However, even with the help of these biomarkers, it remains difficult to identify responders and non-responders of PD-1 blockade therapy.11 For example, PD-L1 expression does not always correlate with clinical outcomes. Although patients with advanced NSCLC with PD-L1⁺ tumors (PD-L1 expression on at least 50% of tumor cells) have been reported to respond well to PD-1 blockade therapy, only 44.8% of patients benefit from this therapy.15 Similar results have been reported for TMB as a biomarker for PD-1 blockade therapy, for which a group of patients with low TMB responded well to PD-1 blockade therapy.85 As mentioned earlier, since all PD-1-expressing cells are affected by PD-1 blockade therapy, it is helpful to consider the effect of PD-1 antibodies on these additional PD-1-expressing cells when developing novel biomarkers. For example, Hodgkin’s lymphoma lacks MHC-I expression and has low TMB, making it challenging to generate effective antitumor CD8 T cell responses. Nevertheless, it responds well to PD-1 immunotherapy. These results suggest that other PD-1-expressing cell types, such as B cells and TAMs, may serve as novel biomarkers for PD-1 blockade therapy (figure 6). Interestingly, a few recent studies reveal that B cells are the strongest prognostic factor in patients with melanoma,67 68 sarcoma69 and renal cell carcinoma68 receiving PD-1 blockade therapy, even though the tumors have low level of CD8 T cell infiltration.69 Moreover, patients having tumors with high Treg infiltration should avoid anti-PD-1 antibody monotherapy, as it may lead to HPD due to increased expansion and immunosuppressive activity of Tregs.23 Further, the effect of PD-1 blockade therapy in such patients can be determined using patient derived xenograft models. Lastly, given a recent study demonstrated that PD-1 expression balance of T cells could predict efficacy of PD-1 blockade therapy,52 it would be valuable to determine the association between PD-1 expression balance of all PD-1-expressing cell subsets and clinical response of patients receiving PD-1 blockade therapy.

As the response rate to PD-1 blockade monotherapy remains low, combination therapy has emerged as a recent trend in cancer treatment.141 In this context, targeting non-canonical PD-1 signaling may be considered as a novel strategy. For example, expansion of PD-1-expressing Tregs during PD-1 blockade therapy serves as a potential cause of HPD. Hence, targeting Tregs could be an important strategy for the prevention of HPD, while enhancing the efficacy of PD-1 blockade therapy. In fact, combination of nivolumab (anti-PD-1 antibody) and ipilimumab (anti-CTLA4 antibody) has resulted in higher response rates and longer progression-free survival in malignant melanoma patients than nivolumab or ipilimumab alone.135 Further, the HPD rate was also reduced in the combination group,135 which was likely due to the tumorous Treg-depleting effect of ipilimumab.142 In addition, a previous study reported that CCR4 is expressed specifically on the surface of tumorous effector Tregs, and an anti-CCR4 mAb (mogamulizumab) effectively depletes tumorous Tregs.42 Further, a recent phase I clinical trial reported an acceptable safety profile for mogamulizumab and nivolumab combination therapy.143