Cancer-induced heterogeneous immunosuppressive tumor microenvironments and their personalized modulation

Abstract

Although recent cancer immunotherapy strategies, including immune-checkpoint blockade (i.e. blocking PD-1, PD-L1 or CTLA-4), have shown durable clinical effects in some (but not all) patients with various advanced cancers, further understanding of human immunopathology, particularly in tumor microenvironments, is essential to improve this type of therapy. The major hurdle for immunotherapy is the immunosuppression that is found in cancer patients. There are two types of immunosuppression: one is induced by gene alterations in cancer; the other is local adaptive immunosuppression, triggered by tumor-specific T cells in tumors. The former is caused by multiple mechanisms via various immunosuppressive molecules and via cells triggered by gene alterations, including activated oncogenes, in cancer cells. The various immunosuppressive mechanisms involve signaling cascades that vary among cancer types, subsets within cancer types and individual cancers. Therefore, personalized immune-interventions are necessary to appropriately target oncogene-induced signaling that modulates anti-cancer immune responses, on the basis of genetic and immunological analysis of each patient. Further understanding of human cancer immunopathology may lead to real improvement of current cancer immunotherapies.

Introduction

Cancers that are seen in clinics have already evaded the patients’ immune-defense systems, for example through immune-editing mechanisms, whereby immune responses trigger changes in the immunogenicity of the tumor and acquired immune-resistance. There have been debates about the possibility of immunotherapeutically attacking cancer cells that have already evaded immune responses by stimulating the autologous immune system or by administration of artificially generated anti-tumor T cells through the introduction of genes encoding receptors for tumor antigens. Indeed, recent successful immunotherapies, including immune-checkpoint blockade and T-cell-based adoptive cellular therapy (ACT), have demonstrated that immune responses can occur against cancer cells and can eliminate cancer cells that were resistant to other conventional cancer therapies, including chemotherapy.

Recent immune-checkpoint blockade therapy has revealed basically two types of immunosuppression mechanisms (Fig. 1). One is immunosuppression that is induced by gene alterations in cancer cells and is mediated through various immunosuppressive mechanisms: firstly, through the production of secreted suppressive molecules such as TGF-β, IL-10, IL-6, vascular endothelial growth factor (VEGF), prostaglandin E2 (PGE2), soluble IL-2 receptor α chain (sIL2Ra) and soluble MHC class I-related gene A (sMICA); secondly, through various immunosuppressive cells such as regulatory T cells (Tregs), myeloid-derived suppressor cells (MDSCs), tumor-associated macrophages (TAMs), tolerogeneic dendritic cells (tDCs) and mesenchymal stem cells (MSCs) that are partly induced by the secreted factors mentioned above; and, thirdly, via intrinsically induced inhibitory immune-checkpoint molecules such as PD-L1 and its ligand (PD-1). Mechanisms for the autochthonous PD-L1 expression were reported in lung cancer, Hodgkin’s lymphoma and adult T cell leukemia (1–3). Interestingly PD-L1 overexpression through structural variation of the 3′-UTR of the PD-L1 gene was identified by genomic analysis using supercomputers (3).

Fig. 1.

Cancer gene alteration-induced immunosuppression and T-cell-triggered adaptive resistance. There are two types of immunosuppression: one is caused by various immunosuppressive molecules and cells triggered by activated oncogene signaling pathways mainly resulting from gene alterations in cancer; the other is local adaptive immunosuppression, triggered by tumor-specific T cells in tumors. Cytokines such as IFN-γ produced from activated tumor-specific CTLs induce immunosuppressive molecules such as PD-L1, CD155 and IDO in cancer cells. The main immunosuppression mechanisms vary among patients. Personalized therapies appropriately using small-molecule inhibitors that target oncogenic signals, using specific inhibitors and using specific blocking antibodies can reverse the immunosuppression in cancer patients.

The other immunosuppression mechanism (adaptive immune-resistance) is triggered by tumor-specific CTLs, whereby cytokines such as IFN-γ produced by anti-tumor T cells induce immune-checkpoint molecules such as PD-L1, induce immunosuppressive indoleamine (IDO), which catabolizes tryptophan, or induce chemokines such as CCL22 that recruit CCR4⁺ Tregs.

Here, we mainly discuss recent findings of the multiple mechanisms for immunosuppression induced by alterations in cancer genes that vary among cancer patients and the personalized modulation of these mechanisms by utilizing low-molecular-weight compounds, including drugs targeted at specific molecules and by utilizing antibodies targeted at immune-regulating molecules. Since there are still many nonresponders and partial responders to current immunotherapies, further understanding of human tumor immune-microenvironments and their appropriate modulation for each patient to improve cancer therapy is one of the major topics in cancer immunology.

The immunosuppressive immune-status of tumor microenvironments

Tumor microenvironments include tumor tissues in which many immunosuppressive mechanisms are present and generate DCs loaded with tumor antigens as well as various suppressive molecules and cells, for example TGF-β, tDCs and Tregs. These molecules and cells are supplied to lymph nodes (e.g. sentinel lymph nodes) where anti-tumor T cells are induced, but are often immunologically suppressed in cancer patients. Bone marrow, which is the source for both anti-tumor memory T cells and various immunosuppressive cells (e.g. MDSCs and MSCs) (Fig. 2) is also an important component in tumor-associated microenvironments. Overproduction of immunosuppressive molecules such as TGF-β induces immunosuppression, not only in tumors but also in lymph nodes via various mechanisms including Treg and MDSC induction, and subsequent T-cell suppression (4). Tumor tissues and sentinel lymph nodes observed in clinics are actually under immunosuppressive and tumor-promoting conditions.

Fig. 2.

Tumor immune-microenvironments. In tumor microenvironments, not only DCs taking up antigens but also various immunosuppressive cells and molecules are generated and migrate into the nearest (sentinel) lymph nodes, where anti-tumor T cell responses are induced, but are often immunologically suppressed. The migration of anti-tumor T cells in response to chemokines is sometimes suppressed, whereas that of immunosuppressive cells such as Tregs is enhanced. The bone marrow is an important component of tumor-associated microenvironments since it works as the source of anti-tumor memory T cells and various immunosuppressive cells.

Differences in the immune-status of tumor microenvironments correlate with prognosis after various cancer therapies

In various cancers (e.g. colon cancer, lung cancer, head and neck cancer, ovarian cancer and cervical cancer), T-cell infiltration of tumors before treatment was reported to correlate with prognosis after conventional therapies including surgery (5). In colon cancer patients, tumor infiltration by T cells (e.g. CD3⁺, CD8⁺ or FOXP3⁺ T cells) and B cells (e.g. CD20⁺ cells) correlates with prognosis after curative surgery. Among them, CD3⁺ and CD8⁺ T-cell infiltration (measured using the ‘Immunoscore’) was confirmed to be significantly correlated with prognosis after curative surgery in an international collaborative study (International Immunoscore validation) (6). The inclusion of immunological status into the current tumor, nodes, metastasis (TNM) staging classification may improve the clinical management of colon cancer patients. Some of the mechanisms for T-cell infiltration were reported, including loss of immune-related genes encoding CXCL13 and IL-15 in colon cancer cells (7, 8).

Different from other types of cancers, we found that high infiltration of FOXP3⁺ T cells strongly correlates with favorable prognosis after surgery in colon cancer. Some of the FOXP3⁺ T cells appear to be helper T cells (9). We can classify at least six subpopulations even in patients at the same stage (Stage II) of colon cancer, and they correlated with overall survival. In some of the subsets, relatively high CD8⁺ T cell response and IFN-γ responses were observed. One of the CD8-high subsets was found to have tumors that were positive for microsatellite instability (MSI⁺) possibly due to decreased gene expression of the DNA mismatch-repair (MMR) enzyme hMLH1; MSI is a form of genetic hypermutability that results from MMR, and increases DNA mutation-derived neo- antigens. Interestingly, there is significant correlation between such sporadic MSI⁺ colon cancers and a high presence of fusobacterium in the colon.

We have previously reported that MSI⁺ colon cancer contains abundant CD8⁺ T cells in the tumor and that autologous immune responses occur against tumor-specific peptides in which frameshift-changes are caused by dysfunctions in DNA MMR enzymes, so we predicted that MSI⁺ cancer may be susceptible to immunotherapies (10).

Recently, anti-PD-1 antibody treatment showed strong anti-tumor effects on patients with MSI, in not only colon cancer but also other types of cancers including endometrial cancer and pancreatic cancer. On the other hand, PD-1 blockade was known to be ineffective in colon cancer showing microsatellite stability, even with T-cell tumor infiltration and PD-L1 expression (11). One possibility for this unresponsiveness is antigen loss through relatively strong immune-editing (12). Another possibility is an immunosuppressive mechanism other than PD-1–PD-L1. We found other immune-checkpoint molecules such as lymphocyte-activation gene 3 (LAG3) and T cell immunoglobulin and ITIM domain (TIGIT) may be involved in the relatively T-cell-rich subset without MSI.

Melanoma was the first cancer for which anti-PD-1 antibody therapy resulted in durable clinical responses. Subsequent analysis revealed that activation of CD8⁺ T cells present in peri- and intra-tumor locations at pretreatment is responsible for melanoma elimination (13). The CD8⁺ T-cell infiltration status correlated with response to anti-PD-1 antibody therapy.

We have previously identified various melanoma antigens recognized by tumor-infiltrating T cells, for example melanocyte-specific antigens, cancer-testis antigens and DNA mutation-derived antigens (neo-antigens) (14–16). Recent studies suggested the important role of tumor-infiltrating T cells specific for neo-antigens, particularly derived from DNA mutations generated in early tumor development, to eradicate cancer cells after immune-checkpoint blockade (17). Interestingly, the CD8⁺ T cell tumor infiltration appears to be partly regulated by activation of melanoma β-catenin signaling (18), which was also shown to induce immunosuppression at both induction and effector levels in our study (19), rather than simple numbers of DNA mutations and associated neo-antigens, which are one of the targets recognized by tumor-infiltrating T cells. Signaling via TGF-β–Smad–Snail, which induced epithelial to mesenchymal transition (EMT), was also found to be involved in immunosuppression (20) and resistance to PD-1 blockade in melanoma (21).

Adaptive immune-resistance other than that mediated through PD-1–PD-L1 may also be present in melanoma. We found that the inhibitory co-stimulatory molecule TIGIT expressed in tumor-infiltrating lymphocytes (TILs) is involved in immune-resistance of melanoma that constitutively expressed the corresponding ligand CD155 in some patients (22). Synergistic in vivo anti-tumor effects of anti-TIGIT antibody and anti-PD-L1 antibody were reported in a murine colon-cancer model (23).

In human papilloma virus (HPV)-induced cervical cancer, we found that CD8⁺ T-cell infiltration in tumors correlated with a favorable prognosis, but FOXP3⁺ Treg infiltration correlated with poor prognosis after chemotherapy and radiation therapy (unpublished data). Cultured TILs recognized tumor-specific HPV-E6 and HPV-E7 proteins and we are planning to perform TIL therapy. Administration of the cultured TILs has been reported to show clinical effects in patients with chemo-resistant cervical cancer (24). In serous ovarian cancer, T-cell infiltration in tumors was reported to correlate with prognosis after treatment (25). We found that tumor infiltration by T cells was relatively weak in ovarian clear cell carcinoma (OCCC), which are ~25% of Japanese ovarian cancers. We found the OCCC produced very high amounts of IL-6 and IL-8 in an NF-κB dependent manner, and inhibited DC functions and enhanced immunosuppressive myeloid cells such as MDSCs (26), which might cause immunosuppression and reduce subsequent T-cell infiltration in tumors.

These results with various types of cancers indicate that pretreatment T cell status defined by multiple mechanisms such as cancer gene alterations is correlated with prognosis after various therapies including surgery, chemotherapy and radiation therapy. Therefore, further analysis of cancer gene-induced immunosuppression mechanisms and their appropriate modulation depending on the immune-evading mechanism may further improve current immunotherapy.

Immunosuppression caused by cancer cell oncogenes and signal activation

One of the main inducers for immunosuppression in the cancer microenvironment are suppressive molecules such as cytokines and chemokines produced from cancer cells. We and other groups have recently reported that oncogenic signals like the MAPK signaling pathway are frequently activated in cancer cells and are involved in the production of these immunosuppressive molecules (Fig. 1).

Importantly, even in the same cancer type, the oncogene signaling status can be different. For example in melanoma, MAPK signal activation mainly due to the BRAF^V600E mutation, PI3K/Akt signal activation mainly due to PTEN loss or Akt3 amplification and mutation, β-catenin signal activation and STAT3 activation have been reported. Each of these signals contributes to the malignant phenotypes of melanoma including immunogenicity and immunosuppression as well as cell proliferation and invasion. The pattern and intensity of the oncogenic signal activation varies among patients. Thus, personalized modulation that targets the responsible oncogenic pathways for each patient is very important. Here, we summarize the contribution of oncogenic pathways in cancer-induced immunosuppression.

The MAPK signaling pathway

The MAPK signaling pathways, which are frequently activated by mutated constitutively activated BRAF^V600E in human melanoma cells, are involved not only in their proliferation and invasion but also in the production of immune suppressive cytokines such as IL-6, IL-10 and VEGF, which inhibit the ability of DCs to stimulate T cells (27, 28). Treatment of melanoma cells with BRAF^V600E-specific RNA interference (RNAi) or MEK inhibitors suppressed the cytokine production and restored DC function (28). Activated MAPK signaling pathways in melanoma cells are also reported to be involved in the production of IL-1α and IL-1β, which acted on cancer-associated fibroblasts and enhanced their ability to suppress melanoma-specific T cells via the production of cyclooxygenase 2 (COX-2), PD-L1 and PD-L2 (29).

As well as being involved in the production of these immunosuppressive molecules, activated MAPK signaling pathways are reported to be involved in cancer immune evasion by the downregulation of tumor antigen expression and MHC expression (30, 31). MEK inhibitors increase the susceptibility of melanoma cells to cytotoxic T cell (CTL) lysis partly because of increased expression of melanocyte differentiation antigens such as MART-1/melan-A and gp100 (30). Vemurafenib, an inhibitor of mutant BRAF, enhances IFN-γ and IFNα2b-induced MHC expression on melanoma cells (31).

These in vitro studies suggest that cancer-induced immunosuppression can be reversed by targeting activated MAPK signaling pathways in cancers using a MEK inhibitor or selective inhibitors of mutant BRAF, vemurafenib and dabrafenib, which have recently demonstrated strong clinical responses. In mouse models, BRAF inhibitors are reported to improve the anti-tumor activity of ACT (32). However, another study (33) shows a decrease of tumor-infiltrating immune cells including T cells, NK cells and macrophages after use of BRAF inhibitors and no increase of anti-tumor effects when combined with anti-CTLA-4 antibody in BRAF^V600E/PTEN^−/− gene-engineered mice spontaneously developing melanoma.

Although any synergistic effects of combined therapies of immune therapies and BRAF inhibitors are still controversial in the above pre-clinical mouse models, interesting observations were reported in the analysis of human melanoma samples of BRAF inhibitor-treated melanoma patients. The numbers of tumor-infiltrating granzyme⁺ CD8⁺ T cells are increased in regressing tumors with necrosis, but not in progressing tumors (34). This T-cell infiltration appears to be not a simple scavenger phenomenon, rather an active induction and recruitment of tumor-antigen-specific T cells.

Considering those studies and observations, BRAF inhibitors might enhance the anti-tumor immune responses not only by the release of multiple endogenous tumor antigens by direct tumor destruction (immunogenic cancer cell death) but also by decreased production of immunosuppressive cytokines from cancer cells and an increase of tumor antigen expression and MHC expression. Therefore, combined use of BRAF inhibitors may enhance the current immunotherapy for melanoma patients, including cancer vaccines, immune-checkpoint blockers and T-cell based ACT. For the last 2 or 3 years, many clinical trials for combined therapies of BRAF inhibitors and immunotherapies have started. Although the initial phase I study combining vemurafenib (a BRAF inhibitor) with ipilimumab (an antibody that blocks CTLA-4) in patients with advanced melanoma was terminated prematurely due to significant hepatotoxicity, another phase I trial using dabrafenib (a BRAF inhibitor) and ipilimumab is currently ongoing without significant hepatotoxicity (35).

The JAK–STAT3 signaling pathway

The above-mentioned immunosuppressive cytokines produced from melanoma cells are also regulated by STAT3 signaling. We have shown that knockdown of STAT3 by RNAi resulted in a decrease of IL-6, IL-10 and VEGF production from human melanoma cells (28). In this study, inhibition of MAPK pathways had little effect on STAT3 activity. Thus, the MAPK pathway and STAT3 seem to regulate the production of these immunosuppressive cytokines independently.

Interestingly, STAT3 plays an immunosuppressive role not only in cancer cells but also in various immune cells such as DCs, MDSCs and Tregs in cancer microenvironments. STAT3 activation in DCs polarizes DC differentiation toward a tolerogenic phenotype (Fig. 1), with high IL-10 and low IL-12 production, in vitro. These DCs have less T-cell stimulatory activity and an enhanced Treg-inducing ability. In a tumor-bearing mouse model, STAT3 depletion in DCs enhanced their T-cell stimulatory activity (36). Moreover, intratumoral vaccine therapies using STAT3-depleted DCs induced better anti-tumor effects accompanied by tumor antigen-specific T_h1 responses featuring high IFN-γ production (36). STAT3 was also involved in expansion of MDSCs (37) and suppressive activity of Tregs (38).

These observations indicate that constitutive activation of STAT3 is an upstream event for induction of immunosuppression not only in cancer cells but also tumor-promoting immune cells such as tolerogenic DCs, MDSCs and Tregs (39). Therefore, STAT3 inhibitors can be used for reversal of the cancer-induced immunosuppression by acting on both cancer cells and various immune cells.

Several STAT3-specific inhibitors are currently under development; however none of them has succeeded in clinical trials. In addition to specific STAT3 inhibitors, a variety of existing drugs have been reported to have STAT3 inhibition activity. For example administration of a multikinase inhibitor, Sunitinib, which also suppresses downstream STAT3 signaling, decreased MDSCs and Tregs along with an increase of IFN-γ-producing T cells in the peripheral blood of patients with kidney cancer (40). We have found that curcumin, a natural compound contained in Japanese traditional medicines, inhibits STAT3, resulting in the reduction of immunosuppressive cytokines from cancer cells, activating DCs, and synergistic enhancement of anti-tumor effects of anti-PD-1 antibody (unpublished data).

The NF-κB signaling pathway

Serum IL-6 and IL-8 levels are known to be prognostic factors for poor responses by cancer patients, particularly those treated with immunotherapies. OCCC, the second most common subtype of ovarian cancer in Japan, is associated with high production of IL-6 and IL-8, which were correlated with poor prognoses.

We found the number of tumor-infiltrating T cells in OCCC was significantly fewer than those in other types of cancers including serous ovarian cancers. The activation of NF-κB and STAT3 was found to be a possible upstream event to induce immunosuppression by IL-6 and IL-8. Correlation between NF-κB p65 and IL-6 expression in human ovarian tissues was observed by immunohistochemical analysis. NF-κB inhibitors such as DHMEQ inhibited production of these immunosuppressive cytokines (26).

In nude mice implanted with human OCCC cell lines, impairment of the T-cell stimulatory activity of murine DCs and accumulation of murine MDSCs in spleen and tumors were observed partly due to an increase of mouse-compatible human IL-6 produced by human OCCC. Administration of DHMEQ reversed these immunosuppressive effects accompanied by a decrease of human IL-6. Although NF-κB plays a very important role in variety of immune cells such as DCs for the stimulation of immune responses, our study suggests that appropriate doses of NF-κB inhibitors can be used for reversal of cancer-induced immunosuppression.

Recently, various existing drugs have been found to have activity to inhibit NF-κB signaling. Using these drugs, anti-tumor effects of current immunotherapies such as immune-checkpoint blockade therapies can be improved especially in NF-κB-activating cancers such as OCCC.

The Wnt–β-catenin signaling pathway

Activation of the Wnt–β-catenin signaling pathway is frequently involved in cancer development and aggressive phenotypes in several types of human cancers. One third of melanoma tissues show activation of the Wnt–β-catenin pathway (41). However, the contribution of the Wnt–β-catenin pathway to the aggressive behavior in melanoma cells is controversial, because activated β-catenin signaling was not always associated with malignant phenotypes of melanoma.

Some studies reported better prognoses and lower proliferation in melanoma cells activated via Wnt–β-catenin (42, 43). We have reported that activation of the Wnt–β-catenin pathways in melanoma cells was involved in cancer-induced immunosuppression via the ectopic production of IL-10 from melanoma cells, which could induce DCs showing tolerogenic phenotypes such as high IL-10 production, low IL-12 production, and less T cell stimulatory ability (19). IL-10 expression is directly regulated by the binding of β-catenin/TCF to its promoter in melanoma cells. In our study, activated Wnt–β-catenin was also involved in the resistance of melanoma cells to melanoma-specific CTLs by an unknown mechanism.

Recently, activated Wnt–β-catenin pathways in melanoma were reported to be associated with lack of T-cell infiltration into tumor tissues (18). Comparing the gene expression profiles between ‘T cell inflamed’ patients and ‘non T cell inflamed’ patients, the β-catenin pathway was found to be more activated in ‘non T cell inflamed’ patients. In the BRAF^V600E/PTEN^−/− gene-engineered mouse model of spontaneously developing melanoma, activation of β-catenin in melanoma resulted in T cell exclusion by repressing the CCL4 expression and subsequent infiltration of CD103⁺CD8α⁺ DCs (18). Interestingly, β-catenin was also reported to be involved in the generation of regulatory DCs and survival and suppressive function of Tregs (44–45). Therefore, β-catenin inhibitors may also be useful for the reversal of the immunosuppression by targeting both on cancer cells and on immune cells.

Concluding remarks

Understanding the mechanisms underlining anti-tumor effects and resistance (e.g. cancer cell-induced immunosuppression and T-cell-triggered adaptive resistance) of cancer immunotherapy in patients is a key to improve current cancer immunotherapies. One of the major mechanisms of the immunosuppression is gene alteration and signal activation in cancer cells as described here. Further studies utilizing multi-omics analyses possibly combined with bioinformatics analysis by using self-learning super-computers may lead to development of precision-medicine-based, really effective personalized, combination cancer immunotherapy in the future.

Funding

This work was supported by Grants-in-aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan (26221005), the Project for Development of Innovative Research on Cancer Therapeutics (P-DIRECT) and the Project for Cancer Research And Therapeutic Evolution (P-CREATE) from Japan Agency for Medical Research and Development (AMED), and a grant from Tokyo Biochemical Research Foundation.

Conflict of Interest statement: The authors declare no conflict of interest.