Transforming growth factor-β (TGF-β) exists in the extracellular matrix as latent precursors with prodomain, and the transformation from latent pro-TGF-β molecule to active TGF-β is a multiple-step process (1). Firstly, pro-TGF-β contains a long signal sequence, a long N-terminal sequence named latency-associated peptide (LAP), and a short C-terminal, which is the mature cytokine (2). Then, dimerized pro-TGF-β is cleaved by Furin (a protease) in Golgi complex. As a result, the bioactive TGF-β moieties are linked with LAP homodimer through disulfide bonds. The LAP encircles bioactive TGF-β moiety and hampers the binding of TGF-β with its receptor. After secretion, The LAP homodimer could anchor to Glycoprotein A repetitions predominant (GARP) on the cell surface or crosslink with the extracellular matrix by latent TGF-β binding proteins (LTBPs). Then, active TGF-β is released by integrin-transmitted forces when cell contraction ( Figure 1 ) (4).

TGF-β signaling is triggered by the interaction of TGF-β ligands with TGF-β type II receptors (TGFβRII) (5). Following the recruitment and phosphorylation of TGF-β type I receptors (TGFβRI) by TGFβRII, SMAD2 and SMAD3 are phosphorylated and further assembled into trimeric complexes with SMAD4 (6). The SMAD complexes could translocate into cell nucleus and regulate the expression of TGF-β-targeted genes, including TWIST1, SNAI1, and SNAI2 (7). Besides canonical SMAD signaling, TGF-β can initiate non-SMAD signalings, such as PI3K-AKT, MAPK, and RHO-ROCK pathways (8–10). TGF-β signaling plays a vital role in embryonic development and homeostasis by controlling cell proliferation, apoptosis, survival, differentiation, and stem-cell self-renewal (11).

TGF-β is a bifunctional cytokine in cancer, acting as tumor promoter and suppressor (12). For healthy cells and early-stage cancer cells, TGF-β inhibits tumorigenesis by inducing cell-cycle arrest (13). However, for late-stage cancers, cancer cells could bypass TGF-β-mediated apoptosis by mutating core components of TGF-β pathway (14). Contrarily, TGF-β promotes tumorigenesis by inducing epithelial-to-mesenchymal transition (EMT), eventually contributing to enhanced metastasis and chemoresistance (15–17). Besides, TGF-β also supports tumor progression by improving angiogenesis and immune evasion (4, 18). This transformation of TGF-β from tumor suppressor to tumor promoter is an important biological characteristic for advanced cancers (19).

The discovery of immune checkpoints and the development of drugs represented by programmed cell death protein 1/programmed cell death ligand 1 (PD-1/PD-L1) monoclonal antibodies are landmark events in cancer immunotherapy (20–24). Anti-PD-1/PD-L1 treatments have shown potent and sustained antitumor effects in patients across multiple cancer types (25–32). However, the low response rate is a crucial drawback of anti-PD-1/PD-L1 therapies, and ideal molecular markers are unavailable to select patients (33–35). The classical cancer-immunity cycle model describes antitumor immunity as a cascade of multistep cascade responses (36). PD-1/PD-L1 axis in the tumor is not the only immunosuppressive pathway (37). It has been shown that hyperactive TGF-β signaling in the tumor microenvironment (TME) can broadly modulate multiple immune cell activities, reshape the TME, and collectively participate in tumor cell immune escape (3). The TGF-β and PD-1/PD-L1 pathways are independent of and complementary to each other. Recent studies have shown that TGF-β is a determinant for anti-PD-1/PD-L1 therapies, which could effectively predict treatment efficacy (38–40). Therefore, constructing TGF-β-involved predictive biomarkers and exploring TGF-β-targeted therapies are valuable to cancer immunotherapy.

Given that TGF-β contributes to cancer immune evasion and immunotherapy resistance, blocking TGF-β could overcome immunotherapy resistance by reprogramming the TME. At present, TGF-β signaling has been a hot therapeutic target for cancer investigators, and enormous efforts have been expended on the development of TGF-β-targeted agents (41). TGF-β blockade strategies, including monoclonal antibodies (containing bispecific antibodies), ligand traps (containing bi-functional proteins), receptor kinase inhibitors, vaccines, and antisense oligonucleotides, are under clinical evaluation ( Table 1 and Figure 2 ) (42).

PD-1/PD-L1 is an important signaling pathway to suppress immune responses and maintain autoimmune homeostasis (94, 95). However, in the TME, the hyperactive PD-1/PD-L1 pathway inhibits immune surveillance. It is traditionally believed that PD-L1, which is highly expressed on tumor cells, binds to PD-1 on the surface of T cells and suppresses the activity of T cells (96). PD-1/PD-L1 monoclonal antibody rescues T cells and restores antitumor immunity by blocking this negative immunomodulatory signal (97, 98). Recent studies have found that anti-PD-L1 antibodies also activate dendritic cells (DC) (99) and natural killer (NK) cells (100). Although PD-1/PD-L1 monoclonal antibodies are approved for the treatment of various cancers and have shown promising results in some patients, the problem of low objective response rates has not been effectively addressed (82, 101, 102). Therefore, screening for molecular biomarkers adapted to PD-1/PD-L1 therapy is an urgent issue at the present stage.

In terms of clinical efficacy, PD-L1 expression could not predict patient outcomes well, and even some patients whose tumors do not express PD-L1 can benefit from anti-PD-1/PD-L1 treatment (103–105). Apart from PD-L1 level, other predictive biomarkers have been identified, including tumor mutational burden (TMB) (106), mismatch repair (MMR) deficiency (107), the status of tumor-infiltrating lymphocyte (TIL) (108), immunosuppressive cell populations (109), oncogenic driver mutations (110–112), neoantigen repertoire (113), gut microbiota (114–116), inflammation-related genes (117, 118), extracellular vesicles (119), and patient’s clinical characteristics (120).

High TGF-β in tumor tissues is mainly produced by tumor cells and mesenchymal cells. TGF-β promotes EMT of tumor cells and acts on various immune cells in the TME, inducing the generation of a suppressive immune microenvironment (121). On the one hand, TGF-β inhibits the cytotoxic activity of CD8⁺ T cells, CD4⁺ T cells, and NK cells. On the other hand, TGF-β increases the proportion of regulatory T cells (Treg) and myeloid-derived suppressor cells (MDSC) (122–125). Moreover, recent studies have found that TGF-β modulates the activity of tumor-associated fibroblast (CAF) and increases the content of collagen fibers in the tumor stroma (126). The thickened collagen fibers surrounding the tumor tissue are detrimental to lymphocyte infiltration, resulting in an immune-excluded tumor type (126). It is generally believed that this type of tumor does not respond to anti-PD-1/PD-L1 therapy, while antagonizing the TGF-β signaling pathway significantly improves anti-PD-1/PD-L1 therapeutic resistance and enhances the effect of antitumor immunotherapy (53, 127). Actually, although CAF was broadly classified into myofibroblastic (myCAF) and inflammatory and growth factor-enriched subgroups, some specific phenotypes are validated to participate in tumor progression as well (128). Besides, Grauel et al. found that TGF-β blockade induced the differentiation of IFN-licensed CAF, enhanced T cell recruitment and infiltration, and improved the effect of anti-PD-1 (129). Moreover, Krishnamurty identified a TGF-β-dependent CAF cluster with highly expressed LRRC15, which could support tumor progression by limiting T cell activity. Abrogating LRRC15⁺ CAF also significantly enhanced the efficacy of anti-PD-1 in mouse models (130).

Microsatellite-stable (MSS) colorectal cancer (CRC) is generally regarded as the cold tumor with poor immunogenicity and scare immune cell infiltration, which is unlikely to benefit from anti-PD-1/PD-L1 (131). However, this type of CRC could be conquered by the combination of anti-TGF-β and anti-PD-1/PD-L1 (132). Tauriello et al. established a metastatic CRC model by genetically engineering Apc, Kras, Tgfbr2, and Trp53 quadruple mutant mice (132). Metastatic cancer tissues display characteristics of human MSS CRC: low mutation burden, T cell depletion, and TGF-β activation (132). Normal intestinal mucosa and adenoma had T cell infiltration in the mesenchyme, but not in adjacent cancer tissue (132). Anti-PD-1/PD-L1 treatment had limited effects on these tumors, while TGF-β inhibitors increased the sensitivity of anti-PD-1/PD-L1 treatment (132). Further investigations showed that combination therapy upregulated T-bet and IFN-γ levels in CD4⁺ Th1 cells and increased GZMB generation in CTLs, eventually eradicating metastases and prolonging survival (132). The results support that the TME with hyperactive TGF-β signaling caused T cell depletion and a decrease in Th1 effector cells, leading to cancer immune escape (132).

Besides, Mariathasan et al. analyzed cancer tissues from patients with metastatic urothelial carcinoma receiving anti-PD-L1 treatment (126). The responders were characterized by high PD-L1 expression, high tumor mutation burden/neoantigen, and CD8⁺ effector T cells (126). The non-responders had tumor tissue containing dense mesenchymal stroma, CAF with high TGF-β activity, and T cell deficiency (126). The mouse breast cancer EMT-6 model mimicked the phenotype of epithelial carcinoma, where blocking either PD-L1 or TGF-β alone was ineffective (126). Combined inhibition of TGF-β and PD-1 signaling reduces TGF-β activity in stromal cells, promotes T cell infiltration into the tumor, stimulates a robust immune response, and leads to tumor regression (126). In conclusion, several studies have shown that TGF-β pathway activity is hyperactivated in anti-PD-1/PD-L1-resistant tumor tissues (3). The high expression of TGF-β in the TME suppresses the antitumor immune response (3). The immunosuppressive mechanisms of TGF-β and PD-1/PD-L1 pathways on tumors are independent and complementary, promoting the escape from immune surveillance (36).

In parallel with the immunosuppressive role of TGF-β in cancer immunology, the predictive value of TGF-β signaling in anti-PD-1/PD-L1 therapies has been well documented in multiple clinical studies. In the single-arm phase 2 study NCT02662309, 95 muscle-invasive urothelial cancer patients were recruited and received anti-PD-L1 treatment before cystectomy (38). In this study, the presence of preexisting activated CD8⁺ T cells (dual CD8 and GZMB positive staining) in the tumor was closely correlated with patient outcomes. Moreover, FAP, the surrogate biomarker of CAF, was upregulated in relapsing tumor tissues but was downregulated in responders (38). Notably, the signatures of cytotoxic T cell and TGF-β signaling could also effectively predict treatment response to atezolizumab (38). In addition, in the single-institutional phase 2 trial NCT02658019 for advanced hepatocellular carcinoma (HCC), patients with low plasma TGF-β (< 200 pg/ml) at baseline had improved OS and PFS after anti-PD-1 treatment (39). Also, in non-small cell lung cancer (NSCLC), TGF-β concentration in the plasma collected seven days after anti-PD-1 treatment effectively predicted patient outcomes (133).

Transcriptomic data of microsatellite instability-high/mismatch repair-deficient gastrointestinal tumors showed TGF-β, EMT, Wnt/β-catenin, angiogenesis, hypoxia, KRAS, mTORC1, and metabolism-associated pathways were enriched in non-responders after PD-1 treatment (40). Similarly, the transcriptomic profile of metastatic bone and soft tissue sarcomas demonstrated that TGF-β signaling enrichment was negatively correlated with the efficacy of anti-PD-1 (134). Furthermore, the TGF-β signature (based on mRNA levels of BMPR2, FKBP1A, SLC20A1, SKIL, TGFBR1, and XIAP) predicted anti-PD-1/PD-L1 resistance in gynecologic cancer (135). The high TGF-β score was associated with shorter progression-free survival after immunotherapy (8.1 vs. 2.8 months, P < 0.05) (135). Additionally, for triple-negative breast cancer receiving Durvalumab with Nab-Paclitaxel, RNA-seq data showed that EMT, TGF-β, and extracellular matrix pathways were enriched in patients with residual disease (136).

Given the negative role of TGF-β signaling in cancer immunology and immunotherapy, it is rational to enhance ICI efficacy by blocking TGF-β. In preclinical explorations and clinical practice, combination therapies of TGF-β inhibitor and anti-PD-1/PD-L1, as well as anti-PD-L1/TGF-β bispecific antibodies/fusion proteins, have made rapid progress (137).

TGF-β is a paradoxical regulator in cancer progression, which acts as a suppressor in early-stage cancer but as a promoter in advanced cancer. The negative effects of TGF-β on cancer immune surveillance have been well studied, including impairing immune infiltration, inducing the differentiation toward MDSC/M2-like macrophage/Treg, limiting the cytotoxicity of T cell and NK cell, and undermining the antigen presentation capability of DC. Accumulating evidence shows that TGF-β not only promotes cancer immune evasion but also predicts the efficacy of immune checkpoint inhibitors. Increased TGF-β level at baseline is commonly associated with a poor response to anti-PD-1/PD-L1 therapy. Blocking TGF-β could improve response to anti-PD-1/PD-L1 and patient outcomes. At present, dual PD-1/PD-L1 and TGF-β blockade have made a breakthrough, especially by anti-PD-L1/TGF-β bispecific antibody or bi-functional protein. This updated immune checkpoint inhibitor might alter the therapeutic paradigm for cancer in the future.

MY and TL performed the selection of literature, drafted the manuscript and prepared the figures. MN and YW collected the related references and participated in discussion. KW and ZZ designed this review and revised the manuscript. All authors contributed to the article and approved the submitted version.