Obstacles posed by the TME to ICB therapy
TME is composed of blood vessels, marrow-derived suppressor cells (MDSCs), APCs, lymphocytes, neutrophils, tumor-associated macrophages (TAMs) and fibroblasts, and the extracellular matrix composed of collagen and proteoglycans, and soluble factors (e.g., cytokines and growth factors), all of which may assist or hinder antitumor immune responses [
38]. It is now increasingly accepted that cancer cells, rather than working alone, develop close interactions with the extracellular matrix, stromal cells, and immune cells that together form the TME, facilitating a chronic inflammatory, immunosuppressive, and proangiogenic intratumoral environment in which tumor cells could adapt and grow with a lower likelihood of detection and eradication by host immunosurveillance [
38]. The importance of the immune system in protecting the body against internal threats (e.g., malignant cells) has been described as the cancer-immunity cycle [
39,
40]. The cycle comprises the release of neoantigens created by oncogenesis, their release and capture by APCs for processing, antigen presentation to T cells at secondary lymphoid organs, and activation of effector cytotoxic T lymphocytes (CTLs) that then migrate and infiltrate the tumor, recognizing and killing cancer cells. One or more of the above steps required for T cell immunosurveillance are often compromised in developing tumors, leading to evasion of immune-mediate tumor control. Thus, given that the efficacy of ICB therapy is driven by T cells, this effective immune evasion can ultimately lead to failures in ICB treatment. Immune tolerance can result from suppression at any point in cancer-immunity cycle [
39]. A suboptimal immune response can result from limited antigen uptake and presentation. T cells capable of responding to specific tumor antigens may be significantly reduced because of immunoediting. The ability of tumor-specific lymphocytes to be fully activated and to proliferate may be limited by a lack of effective co-stimulatory signals. Even if a robust immune response is generated, it may not last long enough to induce tumor regression. Activated T cells need to efficiently migrate to and accumulate at the tumor site, and then they also need to resist exhaustion and immunosuppression in the TME. Multiple mechanisms used by tumor cells, including alteration of the antigen presentation machinery, secretion of immunosuppressive factors that can induce apoptosis of lymphocytes, or activate negative regulatory pathways, could induce tolerance and limit the effectiveness of the immune response [
41]. Tumor cells that either directly or indirectly enhance immune tolerance have a selective survival advantage thereby resulting in their outgrowth.
Given the heterogeneity in the expression levels of PD-1 ligands and their potential relevance as biomarkers for blockade of the PD-1 pathway, it is important to understand the signals that induce the expression of PD-1 ligands on tumor cells and hematopoietic cells within the TME [
28,
42,
43]. Two general mechanisms for the regulation of PD-L1 by tumor cells have emerged: innate immune resistance and adaptive immune resistance. These mechanisms are not mutually exclusive and may co-exist in the same TME. Innate immune resistance refers to the constitutive expression of PD-L1 by tumor cells caused by genetic alterations or activation of certain signaling pathways. For some tumors, such as glioblastomas, it has been shown that PD-L1 expression is driven by constitutive oncogenic signaling pathways in the tumor cell. The expression on glioblastomas is enhanced upon deletion or silencing of
PTEN, which implicates the involvement of the PI3K pathway. Similarly, constitutive ALK signaling, which is observed in certain lymphomas and occasionally in lung cancer, has been reported to drive PD-L1 expression by STAT3 signaling [
44,
45]. The alternative mechanism for PD-L1 upregulation on tumors that has emerged from both clinical and pre-clinical studies reflects their adaptation to endogenous tumor-specific immune responses, known as adaptive immune resistance [
46,
47]. In adaptive immune resistance, the tumor uses the natural physiology of PD-1 ligand induction that normally occurs to protect a tissue from infection-induced immune-mediated damage to protect itself from antitumor immunity. Expression of PD-L1 as an adaptive response to endogenous antitumor immunity occurs because PD-L1 is induced on most tumor cells in response to IFNs, predominantly IFN-γ. This induction also occurs in epithelial and stromal cells in normal tissues. IFN-γ is known to have dichotomous immunological properties. It can induce apoptosis of tumor cells, blood vessel disruption, and upregulation of MHC-expression on the one hand. On the other hand, IFN-γ can also promote the expression of immunosuppressive molecules such as indolaimine-2,3-deoxygenase (IDO), which inhibits immunity locally via conversion of tryptophan to kynurenines and can contribute to peripheral tolerance and can have a direct negative effect on T
eff cell function in coordination with upregulated PD-L1 [
46,
47]. Understanding the mechanisms contributing to an effective response and resistance are of utmost importance to optimize treatment with ICBs. In this context, novel CMTM6/4 transmembrane proteins, considered as PD-L1 regulators by decreasing ubiquitination and stabilizing PD-L1, have been recently discovered in maintaining antitumor immunity [
48,
49].
Resistance to ICB within TME involves components other than tumor cells, including T
reg cells, MDSCs, γδT cells, TAMs, and other inhibitory immune checkpoints, which may all contribute to inhibition of antitumor immune responses. Humans that lack a functional T
reg cell population, characterized by their expression of the
Foxp3, develop a lethal autoimmune disorder, which can be recapitulated in mice via
Foxp3 deletion [
50]. While T
reg cells are required to limit autoimmunity, maintain immune homeostasis, and prevent excessive tissue damage, they can be deleterious in tumor through suppression of antitumor immunity [
51,
52]. Indeed, high numbers of T
reg cells and T
reg cells to T
eff cells ratio are considered poor prognostic factors for many tumor types, including melanoma, ovarian cancer, and colorectal carcinoma [
53‐
55]. T
reg cells are known to suppress T
eff cell responses via secretion of certain inhibitory cytokines (e.g., IL-10, IL-35, and TGF-β) or via direct cell contact [
56‐
60]. Multiple studies obtained from murine models have revealed that the depletion of T
reg cells within TME could enhance or restore antitumor immunity [
61‐
63]. Therapeutic mAbs that target co-inhibitory receptor pathways (e.g., CTLA-4 or PD-1/PD-L1) limit T cell exhaustion, enhance CD8
+ T cell antitumor activity, and increase T
eff cells to T
reg cells ratio in the tumors [
64]. In murine models, response to CTLA-4 mAb therapy was shown to be correlated with an increase in the ratio of T
eff cells to T
reg cells [
65]. This shift in the ratio of T
eff cells to T
reg cells has been found to be a result of both an increase in T
eff cells and depletion of T
reg cells in a murine tumor model, suggesting that tumors for which immunotherapy cannot increase T
eff cells and/or deplete T
reg cells to enhance the ratio of T
eff cells to T
reg cells are likely to be resistant to treatment, either initially or during the relapsed disease setting [
61]. However, it is possible that tumor-infiltrating T
reg cells might co-exist with other immune cells, reflecting a potentially immunogenic “hot” TME. One study of patients treated with CTLA-4 mAb showed that a high baseline expression of Foxp3
+ T
reg cells in the tumor was correlated with better clinical outcomes [
66]. T cell exhaustion is a primary limiting factor affecting the efficacy of current cancer modalities, including CAR T cell therapies [
67]. However, the promising antitumor effects noted in humans with PD-1 blockade alone offers substantial potential for reversing T cell exhaustion and improving the clinical outcome of next-generation immunotherapies [
64]. Reversal of CD8
+ T cell exhaustion and efficient control of viral load was noted following dual blockade of T
reg cells and PD-L1 [
68], or IL-10 and PD-L1 [
57], or following inhibition of TGF-β signaling [
56]. Thus, there is a clear role for T
reg cells and its derived inhibitory cytokines in mediating T cell exhaustion, even if the precise mechanisms remain to be defined. Additional studies are ongoing to determine the impact of tumor-infiltrating T
reg cells on clinical outcomes for patients who receive treatment with immunotherapy agents.
MDSCs, which were initially defined in murine models, have emerged as major regulators of immune responses in various pathological conditions, including tumors. Mouse MDSCs were classified as CD11b
+Gr-1
+ and could be further sub-divided into the monocytic-CD11b
+Ly6C
+Ly6G
− population and the polymorphonuclear-CD11b
+Ly6G
+Ly6C
lo population [
69]. Human MDSCs are classified as CD11b
+CD33
+HLA-DR
−, which may co-express with other markers such as CD15, CD14, CD115, and/or CD124 [
70‐
72]. MDSCs represent 30% of cells in the bone marrow and 2–4% cells in the spleen in normal mice. MDSCs normally differentiate into granulocytes, macrophages, or dendritic cells. However, under pathological conditions such as cancer, MDSCs become activated, rapidly expand, but remain undifferentiated. Moreover, clinical data have shown that the presence of MDSCs associates with reduced survival in several human tumors, including colorectal cancer, and breast cancer [
73]. Growing evidence also suggest that heavy tumor infiltration by MDSCs correlated with poor prognosis and decreased efficacy of immunotherapies, including ICB therapy [
74], adoptive T cell therapy (ACT) [
75], and DCs vaccines [
76]. Thus, eradicating or reprogramming MDSCs could enhance clinical responses to immunotherapy. Indeed, in multiple mouse tumor models, selective inactivation of tumor-associated myeloid cells PI3Kγ synergized with ICBs to promote tumor regression and increase survival, suggesting a critical role of suppressive myeloid cells in ICB resistance and a therapeutic potential of PI3Kγ inhibitors when combined with ICB therapy in cancer patients [
77,
78]. Moreover, MDSCs have been also used to predict response to ICB [
79]. Intriguingly, in 126 patients with metastatic melanoma treated with PD-1 blockade, pre-treatment MDSC numbers in the peripheral blood are correlated with response to treatment, with high MDSCs associated with reduced overall survival [
80]. Analysis of peripheral blood of 59 melanoma patients treated with CTLA-4 inhibitor showed that the baseline monocytic MDSCs, neutrophils, and monocytes were more abundant in non-responders when compared to responders, which also experienced increased serum concentrations of MDSC attractants [
81]. Thus, patients with existing immunosuppressive TME are poor responders to immunotherapy, and react to ICB by potentiating these immunosuppressive mechanisms. Additionally, the Fas/Fas ligand-mediated cell death pathway represents typical apoptotic signaling in many cell types, including tumor-infiltrating T cells (TILs) [
82]. Tumors with apoptotic TILs, which are triggered by polymorphonuclear MDSCs in TiRP tumors, which express high level of Fas-ligand, resist immunotherapy based on ICB, cancer vaccines, or ACT [
83]. Apoptosis of TILs can be prevented by interrupting the Fas/Fas-ligand axis, which enhances the antitumor efficacy of ACT in TiRP tumors, and increases the efficacy of ICB in transplanted tumors [
83]. Thus, TILs apoptosis is a relevant mechanism of immunotherapy resistance, which could be blocked by interfering with the Fas/Fas-ligand axis.
γδT cells, which are innate-like T lymphocytes characterized by TCRs composed of γ and δ chains, are widely distributed in the peripheral blood and mucosal tissues. γδT cells are also a conserved population of innate lymphocytes with diverse structural and functional heterogeneity, possessing multi-functional capacities in the repair of host tissue pathogen clearance, and tumor surveillance [
84]. γδT cells are important for immunosurveillance by exerting direct cytotoxicity, strong cytokine production, and indirect antitumor immune responses [
85]. However, accumulating evidence suggests that certain γδT cell subsets unexpectedly drive tumor development and progression by (i) inducing an immunosuppressive TME and angiogenesis via cytokine production, (ii) interfering with DC effector function, and (iii) inhibiting antitumor adaptive T cell immunity via the PD-1/PD-L1 pathway (reviewed in ref. [
86]). For example, certain γδT cell subsets also contribute to tumor progression by facilitating tumor-related inflammation and immunosuppression, with suppressive γδT cells producing IL-10 and TGF-β. TGF-β plays important roles in angiogenesis and immunosuppression by stimulating T
reg cells [
87]. Furthermore, CD39
+ γδT
reg cells are the predominant T
reg cells in human colorectal cancer, which have more potent immunosuppressive activity than CD4
+ or CD8
+ T
reg cells through the adenosine-mediated pathway but independent of TGF-β or IL-10. CD39
+ γδT
reg cells also secrete cytokines including IL17A and GM-CSF, which may attract MDSCs, thus establishing an immunosuppressive network [
88]. Additionally, an indirect regulatory role of γδT cells has been reported in colorectal cancer, whereby activated γδT17 cells in the TME also secreted other cytokines including IL-8, TNF-α, and GM-CSF, which might help support immunosuppressive MDSC [
89]. Many of the immunosuppressive subsets, including γδT cells, can express inhibitory ligands, such as PD-L1, which interferes with the antitumor activity of T cells expressing the PD-1 receptor. Blockade of PD-L1 in γδT cells could enhance CD4
+ and CD8
+ T cell infiltration and immunogenicity in pancreatic ductal adenocarcinoma (PDA), suggesting γδT cells as central regulators of T
eff cells activation in cancer via novel crosstalk [
90]. γδT cells are not APCs and thus not likely to present antigen to T cells, suggesting inhibition by PD-L1 expression on γδT cells and potentially other TME cells is occurring in
trans.
TME contributes to T cell suppression via both direct contact and secretion of soluble factors. Stromal cells can limit T cell trafficking within the TME, promote T
reg cell development, and inhibit T cell proliferation [
91]. Macrophages can be classified as pro-inflammatory and anti-inflammatory, also known as classic (M1) and alternative (M2). TAMs are another subset of cells that seem to affect responses to immunotherapy and are key coordinators of tumor-promoting angiogenesis, fibrous stroma deposition, and metastasis formation [
92,
93]. Skewing or depleting TAMs could therefore affect multiple critical steps in oncogenesis and abrogate different modes of immune resistance. Different TAMs could be distinguished based on the differential expression of transcription factors, surface molecules, and the disparities in their cytokine profile and metabolism [
94]. TAMs display an alternatively activated M2 phenotype known to be critical in controlling tissue homeostasis and wound healing; however, in the tumor, this phenotype is undesirable since it enables potent T cell inhibition via cytokines (e.g., IL-10), depletion of key metabolites (expression of arginase, IDO), or by contact inhibition (e.g., via PD-L1) [
95]. Clinical studies have demonstrated an association between higher frequencies of TAMs and poor prognosis in human cancers [
96]. Previous results suggested that macrophages could directly suppress T cell responses via PD-L1 in hepatocellular carcinoma [
97], and B7-H4 in ovarian carcinoma [
98]. The M2 phenotype is also critical in determining ICB efficacy as an innate wound healing and immune-suppressive gene signature was found to optimally predict non-responders prior to PD-1 mAb treatment. New findings reveal that TAMs are also important when targeting the PD-1/PD-L1 axis. Pittet and colleagues show that TAMs can capture PD-1 targeting antibodies on the T cell surface thereby considerably limiting the duration of drug efficacy [
99], whereas in another paper, Weissman and collaborators reveal that TAMs also express PD-1 on their surface, which impairs their phagocytic activity [
100]. To overcome the potential resistance mechanisms of macrophages, blockade of CSF1R, a receptor of macrophage-colony stimulating growth factor, in a murine model of pancreatic cancer showed decreased frequencies of TAMs, with subsequent increase in IFN production and restrained tumor progression. Tellingly, neither PD-1 nor CTLA-4 blockade could significantly reduce tumor growth in the murine model, which was similar to findings from single-agent studies in patients with pancreatic cancer [
101]. However, CSF1R blockade in combination with either an antibody against PD-1 or CTLA-4, except for gemcitabine, led to improved tumor regression [
101], suggesting that CSF1R blockade induced reduction of TAMs and enabled response to ICB.
Tumor-cell-intrinsic barriers of ICB therapy
Tumor-cell-intrinsic factors that contribute to cancer immunotherapy resistance include expression or repression of certain genes and pathways in tumor cells that compromise the function of TILs in TME. Constitutive WNT signaling via the stabilization of β-catenin was shown to be associated with T cell exclusion in melanoma [
102]. Active β-catenin signaling in melanoma has been previously reported to correlate with more aggressive disease [
103]. The role of β-catenin signaling as an immune escape mechanism was demonstrated in genetically engineered mice developing autochthonous melanoma [
103]. As in humans, activation of the oncogenic WNT-β-catenin signaling pathway in melanoma cells correlates with the absence of T cells and reduced infiltration of a subset of DCs, known as CD103
+ DCs, due to decreased expression of CCL4 that is responsible for DCs recruitment into the TME. Thus, lack of DCs limited tumor-specific T cell priming, leading to development of resistance to PD-L1 and CTLA-4 blockers-based therapies in experimental murine tumor models [
102]. Since the WNT/β-catenin oncogenic pathway has been found activated in several tumor types, this mechanism of resistance might apply to other tumors.
Oncogenic signaling pathways, such as the PI3K pathway, have been proved to associate with primary resistance to PD-1/PD-L1 blockers as well. Signaling via the PI3K-AKT-mTOR pathway contributes to tumorigenesis by impacts on a multitude of cellular processes. This pathway is commonly activated through loss of expression of the tumor suppressor PTEN, a lipid phosphatase suppressing the activity of PI3K signaling [
104], which is a common phenomenon across several cancers, including 30% of melanomas, and has been found to be correlated with resistance to ICB therapy [
105]. PTEN loss in melanomas is associated with significantly decreased gene expression of IFN-γ and granzyme B, with reduced infiltration of CD8
+ T cells, and inferior outcomes after anti-PD-1 therapy. More importantly, the frequency of PTEN deletions and mutations was higher in non-T cell-inflamed tumors as compared to T cell-inflamed tumors. In murine models, the effectiveness of either anti-PD-1 or anti-CTLA4 antibodies is enhanced by treatment with a selective PI3Kβ inhibitor [
105]. Similarly, oncogenic signaling through the MAPK pathway results in the production of VEGF and IL-8, among many other secreted proteins, which have known inhibitory effects on T cell recruitment and function. Combination of targeting the MAPK pathway by selective BRAF and MEK inhibitors with immunotherapy is proposed to improve the long-term outcomes of patients. More importantly, additional PI3K inhibition could be an option for BRAF plus MEK inhibitor resistant patients that receive targeted therapy in combination with ICBs [
106]. Oncogenic signaling pathways are so common in many other tumors where many studies are exploring the clinical benefit of ICBs that such researches may shed some lights on additional strategies to enhance the efficacy of ICBs.
Type I and type II IFNs responses play predominant roles during distinct phases of antitumor immunity. In tumors, secretion of the type I IFNs (IFN-α and IFN-β) facilitates DC maturation that is necessary for the generation of T
eff cells, which return to the tumor to secrete the type II IFN (IFN-γ) to cause vascular destruction and to sensitize tumors to CTLs. TMEs with a significant lack of type I IFN-producing DCs will naturally result in limited antitumor T cell priming and thus a limited pool of useful T cells for ICB therapy to reactivate. Moreover, type I IFN activation allowed for prolonged survival when the PD-1/PD-L1 axis was subsequently targeted. Tumor cells could escape the effects of type II IFN (IFN-γ) by downregulating or mutating molecules involved in the IFNs signaling pathway, which goes by the IFN receptor chains Jak1/Jak2 and STATs [
107]. Jak1/2 are tyrosine kinases essential for intracellular signaling in response to IFNs. IFN-γ released by TILs induces the expression of several IFN-stimulated genes, eventually leading to direct tumor growth arrest and apoptosis, as well as increased antigen processing and presentation, production of chemokines that attract T cells, and upregulation of PD-L1 [
108]. As direct consequence of loss-of-function in Jak1/2, tumors in these patients were completely devoid of T cell infiltrates. Tumors carrying homozygous loss-of-function mutations in Jak1/2 were resistant to anti-PD-1 treatment, despite the presence of a high mutational burden [
109]. Thus, Jak1/2 loss-of-function could be incorporated in the genetic screening of candidates that can be subjected to ICB therapy. Additionally, loss of JAK1 and JAK2 expression might also derive from epigenetic silencing, as it has been described in prostate cancer cell lines [
110]. On the same line, genetic defects in the IFN-γ pathway have been shown to reduce the chance of response to antibodies targeting CTLA-4 in melanoma patients [
111]. Analysis of tumors in patients who did not respond to therapy with the CTLA-4 blockade revealed an enriched frequency of mutations in IFNGR1 and IFNGR2, JAK2, and interferon regulatory factor 1 (IRF1). Any of these mutations would prevent signaling in response to IFN-γ and give an advantage to the tumor cells in escaping from T cell attack, thereby resulting in primary resistance to anti-CTLA-4 therapy. Loss of antigen display by tumor cells leading to acquired resistance to cancer immunotherapy may be due to mutations in the antigen-processing machinery or proteins involved in antigen presentation can lack of recognition by CD8
+ T cells following immunotherapy [
112,
113]. Such mutations were recently detected in patients who relapsed following anti-PD-1 therapy. While several other signatures might stem from the ongoing genomic, methylomic, and transcriptomic analyses in pre-existing samples from candidates to ICB, it is rather clear that high mutational burden and high CD8
+ T cell infiltrate do not necessarily predict sensitivity to ICBs.
Tumors are a major disturbance to tissue homeostasis, creating metabolically demanding environments that encroach on the stroma and infiltrating immune cells. The unrestrained cell growth seen in cancer is often supported by aerobic glycolysis, the same metabolic pathway needed to fuel optimal effector functions in many immune cells [
114]. Altered nutrient availability in tumors affects metabolic reprogramming of T cells, resulting in impaired effector functions and differentiation toward suppressive phenotypes. TILs are exposed to low extracellular glucose and glutamine due to high nutrient uptake by tumor cells [
115]. Importantly, glucose is a critical substrate for the antitumor functions of T
eff cells and M1 macrophages, which both require engagement of aerobic glycolysis for their activation and full effector functions [
116,
117]. Augmented aerobic glycolysis in tumor cells and endothelial cells places immune cells and their neighbors at odds. Glucose deprivation represses Ca2
+ signaling, IFN-γ production, cytotoxicity, and motility in T cells and pro-inflammatory functions in macrophages [
118‐
120]. Cytosolic Ca2
+ concentration serves as a metabolic threshold, allowing activation of the family of transcription factors collectively named nuclear factor of activated T cells (NFAT) [
121]. Consequently, glucose deprivation results in a dose-dependent decrease in IFN-γ, mediated at the translational level by decreased mTOR activity. Recently, several studies have shown that the glycolytic activities of tumor cells may restrict glucose utilization by TILs, thereby impairing antitumor immunity [
119,
121]. Amino acid deprivation in the TME serves as another metabolic checkpoint regulating antitumor immunity. Glutaminolysis in tumor cells is critical to replenish metabolites by anaplerotic reactions [
122], which could result in competition between tumor cells and TILs for glutamine that controls mTOR activation in T cells and macrophages. Glutamine is also a key substrate for protein
O-GlcNAcylation and synthesis of S-2HG that regulate T
eff cell function and differentiation [
123]. TAMs, MDSCs, and DCs could suppress TILs via expression of essential amino acid-degrading enzymes (i.e., ARG1 and IDO) [
124,
125]. Indeed, inhibitors of ARG1 and IDO are under investigation as therapeutic agents in clinical trials [
126]. Bioactive lipids, modified lipoproteins, and cholesterol metabolism within the tumor are also important mediators of immune cell function. DCs in the tumor can accumulate oxidized lipoproteins through scavenger receptor-mediated internalization and formation of lipid droplets, which can ultimately impair their ability to cross-present tumor antigens and activate T cells [
127]. As TILs adapt to the tumor microenvironment, they progressively lose their ability to respond to TCR stimuli, produce effector cytokines, and proliferate—a process termed functional exhaustion or hyporesponsiveness. This is in part due to the upregulation of several inhibitory receptors like PD-1, LAG3, TIGIT, and CTLA-4 that desensitize T cells to tumor antigens [
128]. Intriguingly, both chronic exposure to antigen and environmental triggers (i.e., glucose deprivation) could upregulate PD-1 [
118,
128], which not only suppresses TCR, PI3K, and mTOR signaling in T cells but also dampens glycolysis and promotes fatty acid oxidation-features that may enhance the accumulation of suppressive T
reg cells in tumors [
129,
130]. Extracellular adenosine, a by-product of altered tumor metabolism, induces expression of both CTLA-4 and PD-1 on T cells. Indeed, blockade of PD-1 re-energizes anabolic metabolism and glycolysis in exhausted T cells in an mTORC1-dependent manner [
119,
131], breathing caution into the types of drug combinations one may consider with PD-1/PD-L1 blockades or other forms of immunotherapy. Metabolic interventions, such as the use of mTOR inhibitors, must be targeted specifically to avoid unintended intervention of immune cell function. Signaling through PD-L1 also has direct metabolic effects on cancer cells. In response to PD-L1 blockade, glucose uptake and lactate extrusion are decreased, suggesting that pathological expression of PD-L1 by cancer cells not only impairs T cell metabolism but also benefits cancer cell metabolism.