TGFβ control of immune responses in cancer: a holistic immuno-oncology perspective

Abstract

The immune system responds to cancer in two main ways. First, there are prewired responses involving myeloid cells, innate lymphocytes and innate-like adaptive lymphocytes that either reside in premalignant tissues or traffic directly to tumours, and second, there are antigen priming-dependent [Au:OK?] responses, in which adaptive lymphocytes are primed in secondary lymphoid organs before homing to tumours. Transforming growth factor-β (TGFβ) — one of the most potent and pleiotropic regulatory cytokines — controls almost every stage of the tumour-elicited immune response, from leukocyte development in primary lymphoid organs, to their priming in secondary lymphoid organs and their effector functions in the tumour itself. The complexity of TGFβ-regulated immune cell circuitries, as well as the contextual roles of TGFβ signalling in cancer cells and tumour stromal cells, necessitate the use of rigorous experimental systems that closely recapitulate human cancer to uncover the underlying immunobiology. The diverse functions of TGFβ in healthy tissues further complicate the search for effective and safe cancer therapeutics targeting the TGFβ pathway. Here, we discuss the contextual complexity of TGFβ signalling in tumour-elicited immune responses and explain how understanding this may guide the development of mechanism-based cancer immunotherapy. [Au:OK?]

[H1] Introduction

Members of the haematopoietic cell lineage make up the most diverse and dynamic cell populations in the cancer environment. Frequent infiltration of leukocytes in a growing tumour was observed as early as 1863, which prompted Rudolf Virchow to propose sites of chronic inflammation as origins of neoplastic malignancy. At the turn of the 20^th century, Paul Ehrlich postulated an anti-cancer function for the immune system, proposing it eliminates aberrant cells generated during the course of fetal and post-fetal development. With an increasing understanding of leukocyte differentiation and regulatory mechanisms, the anti-cancer and pro-cancer functions of the immune system have begun to be elucidated. Based on how leukocytes are activated and recruited to the tumour, cancer-associated immune responses are either prewired or are priming-dependent (Box 1). The prewired immune response involves myeloid cells as well as innate lymphocytes and innate-like adaptive lymphocytes that are derived from progenitors generated in primary lymphoid organs. By contrast, the priming-dependent immune response mobilizes adaptive lymphocytes that are activated by antigens in secondary lymphoid organs before homing to cancerous tissues.

Box 1: A holistic immuno-oncology perspective and implications for preclinical cancer model choice.

The immune-recognition characteristics of leukocytes are used to define innate immunity and adaptive immunity, two arms of the immune system that mediate host defence against challenges such as cancer. Immune responses to cancer can also be defined by the behavioural features of leukocytes, which include pre-wired immunity and priming-dependent immunity (see Box Figure, part (a)). Pre-wired immunity is mediated by innate myeloid cells as well as innate lymphocytes and innate-like adaptive lymphocytes, which are pre-programmed to exert their effector functions and can traffic directly from the bone marrow or thymus to peripheral organs and tumours. Priming-dependent immunity is mediated by adaptive lymphocytes that require an antigenic activation [Au:OK?] step in secondary lymphoid organs, such as the lymph node, before they can mediate effector functions in the tumour. Crosstalk between the tumour and its draining lymph node involves the trafficking of immune cells, such as the migration of dendritic cells carrying antigen from the tumour to the draining lymph node and the trafficking of primed lymphocytes from the lymph node to the tumour. [Au: Edit OK?]

Many mouse models are employed to investigate cancer immunobiology (see Box Figure, part (b)). These include transplantable tumour models (top panel), where cancer cell lines are propagated in vitro and inoculated into target tissues of recipient mice. In this context, epithelial cancer cell lines may have become more mesenchymal-like during propagation⁷⁶, and the growing tumour may not be integrated into the organ of interest, which alters the composition and function of tumour-infiltrating leukocytes. Autochthonous tumour models (bottom panel) involve tumour initiation by cells of an endogenous organ in an intact animal via processes such as oncogene expression or carcinogen exposure. Compared with transplantable models, autochthonous models more accurately recapitulate the endogenous tumour microenvironment, including cues of tissue homeostatic disruption²⁰², and thus provide a better model of human disease.

In this Review, we focus on transforming growth factor β (TGFβ)-mediated control of immune cell responses in cancer, a topic of substantial recent discussion^1–3. We cover the pleiotropic effects of TGFβ on prewired and priming-dependent immune cell responses and consider how these are regulated by cross-communications between cancer cells and the endogenous cancer environment — such responses are best recapitulated in autochthonous tumour models in mice, though we also discuss relevant data from transplantable models (Box 1). Finally, we consider how our growing understanding of these pathways can inform therapeutic targeting of the multi-functional TGFβ pathway for cancer immunotherapy. [Au: Edits OK?]

[H1] TGFβ and cellular regulation

The TGFβ signalling pathway is an intercellular communication pathway with pleiotropic effects on almost all metazoan cell lineages⁴. There are three mammalian TGFβ family members (TGFβ1, TGFβ2 and TGFβ3) with TGFβ1 playing a major role in immune regulation⁵ and TGFβ2 and TGFβ3 functioning in other cellular contexts, including fibrosis⁶. TGFβ is synthesized and assembled as a latent complex made of two copies of the N-terminal latency-associated peptide (LAP) and the C-terminal active cytokine, which can be covalently linked to latent TGFβ binding proteins (LTBPs) or leucine-rich repeat-containing proteins including LRRC32 (also known as GARP) and LRRC33. The LTBP-bound LAP–TGFβ complex is targeted to the extracellular matrix (ECM) by binding to fibrillin or fibronectin, while the plasma membrane-localized GARP or LRRC33 anchors LAP–TGFβ to the plasma membrane (Figure 1).

Figure 1. Molecular modalities of TGFβ activation and signaling.

TGFβ is produced in a latent complex involving two copies of latency-associated peptide (LAP) and the active cytokine. This complex can be covalently linked to latent TGFβ-binding proteins (LTBPs) which can be found in the extracellular matrix (ECM), or to leucine-rich repeat-containing proteins such as LRRC32 (also known as GARP), which can be found on the cell surface. The integrin αvβ6, which interacts with the cytoskeleton, and integrin αvβ8 can liberate and expose the receptor-binding site of active TGFβ. TGFβ binds to the heterotetrameric TGFβ type II and type I serine/threonine kinase receptors (TGFBR2 and TGFBR1), triggering TGFBR2 phosphorylation of TGFBR1, which then phosphorylates SMAD2 and SMAD3. These two phosphorylated proteins can modulate gene expression in two ways: (1) through forming a heterotrimeric complex with SMAD4 and translocating to the nucleus, controlling target gene expression, and (2) through binding the SMAD4–SKI–SKIL complex, which normally represses transcription where it binds, and leads to its degradation, thus alleviating transcriptional repression. Phosphorylated SMAD3 also has transcription-independent cell signalling functions including liberation of protein kinase A (PKA) from an inactive PKA complex. Only the mediators of TGFβ signalling discussed in the text are displayed.

The arm and ‘straightjacket’ domains of LAP form a ring around TGFβ, masking its receptor interaction sites⁷, with TGFβ activation primarily mediated by integrin molecules. The LAP domains of TGFβ1 and TGFβ3 contain an integrin-binding RGD (Arg-Gly-Asp) motif, and substitution of RGD with RGE (Arg-Gly-Glu) abolishes integrin binding. Mice harbouring RGE mutant alleles of Tgfb1 manifest an autoimmune phenotype indistinguishable from that of TGFβ1-deficient mice⁸. In addition, mice devoid of the αv family integrins αvβ6 and αvβ8 phenocopy TGFβ1- and TGFβ3-deficient mice⁹. αvβ6 liberates ECM-associated TGFβ through physical force generated from its association with the cytoskeleton¹⁰. αvβ8 does not interact with the cytoskeleton, but its interaction with the GARP-associated LAP–TGFβ complex effectively exposes the receptor-binding site of TGFβ¹¹ (Figure 1).

Active TGFβ mediates its biological functions by binding to heterotetrameric TGFβ type II and type I serine/threonine kinase receptors (TGFBR2 and TGFBR1), which triggers TGFBR2-mediated phosphorylation of TGFBR1 that in turn phosphorylates the SMAD2 and SMAD3 proteins (Figure 1). Phosphorylated SMAD2 and SMAD3 form heterotrimeric complexes with SMAD4, and translocate to the nucleus to control target gene expression⁴. SMAD4 can also constitutively bind to target gene loci in complex with SKI and SKIL to suppress gene transcription, and binding of phosphorylated SMAD2 and SMAD3 to the SMAD4–SKI–SKIL complex induces SKI–SKIL degradation^12,13. Clinically relevant SKI mutations abolish SKI interaction with phosphorylated SMAD2 and SMAD3, but not SMAD4, thereby preventing SKI degradation and markedly blunting TGFβ-regulated gene expression¹⁴. Thus, the TGFβ receptor-activated SMAD2–SMAD3 complex can engage two opposing modes of transcriptional regulation in terms of SMAD4 function: a cooperative mode that involves the SMAD2–SMAD3–SMAD4 complex directly controlling gene transcription, and an antagonistic mode in which SMAD2–SMAD3 relieves transcriptional repression of TGFβ target genes by the SMAD4–SKI–SKIL complex by promoting SKI–SKIL degradation [Au: Edit OK?] (Figure 1). In addition, SMAD complexes exhibit non-transcriptional functions including activation of protein kinase A (PKA)¹⁵ (Figure 1). The TGFβ signalling pathway also involves negative regulators. SMAD7 works to suppress TGFβ signalling by binding to TGFBR1 and blocking SMAD2–SMAD3 from being activated, blocking SMAD2–SMAD3 from complexing with SMAD4, and triggering ubiquitin-mediated degradation of TGFBR1¹⁶. TGFβ signalling induces expression of SMAD7 and SKIL, and these negative feedback loops ensure TGFβ responses are tightly regulated.

[H1] TGFβ in adaptive lymphocytes in cancer

The adaptive immune system consists of T cells and B cells that express antigen-specific T cell receptors (TCRs) and B cell receptors (BCRs) on their cell surfaces, with BCRs also secreted as antibodies. TGFβ critically regulates the development, activation and differentiation of these lymphocytes with pleiotropic effects in cancer.

[H2] TGFβ in adaptive T cell responses in cancer.

In the thymus during T cell development, αβ lineage CD4⁺ and CD8⁺ T cells are positively selected by low affinity self-peptide antigens presented by MHC class II and MHC class I molecules, respectively, while a fraction of highly self-reactive T cells differentiate into CD4⁺ regulatory T (Treg) cells and CD8⁺ memory phenotype (MP) T cells [Au: Reference this statement?]. Following priming in secondary lymphoid organs by antigen-presenting cells (APCs) bearing their cognate antigens, naive CD4⁺ and CD8⁺ T cells as well as Treg cells and MP CD8⁺ T cells can exert effector and regulatory functions in cancer.

[H2] T helper cells.

TGFβ signalling promotes IL-7 receptor α (IL-7Rα) expression in low-affinity thymic CD4⁺ T cells to support their maintenance in peripheral tissues¹⁷, while dampening agonistic antigen-driven negative selection of CD4⁺ T cells¹⁸. In addition, TGFβ inhibits activation, proliferation and effector differentiation of peripheral autoreactive T cells¹⁹, and in mice conditional ablation of TGFBR2 or SMAD2 and SMAD3 at the CD4⁺CD8⁺ stage of T cell development causes lethal autoimmunity[Au:OK?]^20–23. The TGFβ-induced SMAD3–SMAD4 complex activates PKA to trigger C-terminal SRC kinase (CSK)-mediated suppression of proximal TCR signalling to prevent inadvertent T cell priming^24,25. TGFβ also downregulates the transcription factors T-bet and GATA3 to suppress CD4⁺ T cell differentiation into T helper 1 (Th1) and Th2 cells, respectively^26–28. In contrast, TGFβ synergizes with IL-6 to induce the differentiation of IL-17A-producing Th17 cells^29,30, driven in part by STAT3 attenuation of the SMAD3–SMAD4-mediated suppression of TCR signalling²⁵, and SMAD2–SMAD3 reversal of the SMAD4–SKI–SKIL-mediated transcriptional repression of RORγt³¹. Thus, TGFβ functions as a pivotal regulator of CD4⁺ T cell activation and proliferation as well as guiding fate specification of the three major Th cell subsets (Figure 2).

Figure 2. TGFβ control of priming-dependent lymphocyte responses in cancer.

a, In secondary lymphoid organs, TGFβ inhibits naive CD4⁺ T cell, regulatory T (Treg) cell, naïve CD8⁺ T cell and memory-phenotype (MP) CD8⁺ T cell priming by dendritic cells (DCs). b, TGFβ also inhibits Th1 and Th2 cell, but promotes Th17 cell and peripheral Treg (pTreg) cell, differentiation by regulating expression of the transcription factors T-bet, GATA3, RORγt, and FOXP3, while its effect on T follicular helper (Tfh) cell differentiation is contextual. Among CD8⁺ T cells, TGFβ promotes differentiation of regulatory CD8⁺ T cells, in part by promoting expression of the transcription factor Helios. These cells reside in the B cell follicle where they can inhibit Tfh cell responses. In addition, TGFβ promotes tissue resident memory (Trm)-like cytotoxic T lymphocytes (CTLs) through expression of the integrin CD103, while also repressing CTL differentiation by suppressing expression of T-bet and Eomes, IFNγ and granzyme B, and promoting expression of the inhibitory receptor PD-1.

c, TGFβ inhibits B cell proliferation, and promotes both IgA class-switching and migration from the light zone (LZ) and the dark zone (DZ) in the germinal centre.

d-e, In the tumour, TGFβ, in part regulated by Treg cells, inhibits circulating T effector memory (Tem)-like CTL responses against mesenchymal phenotype cancer cells (d), while promoting Trm-like CTL responses, including cytotoxicity against epithelial cancer cells through CD103 interaction with E-cadherin (e).

f-g, TGFβ also inhibits both Th1 and Th2 effector states. Should this inhibition be removed, IFNγ-producing Th1 cells can impact angiogenesis (f), ‘M1-like’ macrophage polarization and CTL function (g), while IL-4-producing Th2 cells can influence tissue-level vascularization and tumour tissue healing (f), ‘M2-like’ macrophage polarization and eosinophil responses (g), which collectively suppress tumor development by targeting cancer cells (g) and the cancer environment (f).

TGFβ-mediated control of peripheral Th cell responses in cancer has recently been investigated in a transgenic model of breast cancer, where TGFBR2 was depleted in mature CD4⁺ T cells³². Blockade of TGFβ signalling results in enhanced CD4⁺ T cell activation and differentiation into Th1 and Th2 cells in tumour-draining lymph nodes³². Of note, TGFBR2-deficient CD4⁺ T cells predominantly localize in the tumour stroma, and they promote tumour tissue healing and blood vasculature reconfiguration to trigger hypoxia and starvation-induced cancer cell death³². Ablation of IL-4 reverses vasculature remodelling and tumour suppression in this system³², supporting the idea that a tissue-level cancer defence mechanism is mediated by Th2 cells in this model. Amplified Th1 cell responses following TGFβ blockade has also been shown to suppress cancer progression. Metastasis of prostate cancer to bone is associated with enhanced Th17 cell, but not Th1 cell, differentiation in patients treated with the immune checkpoint inhibitor anti-CTLA4³³. In a mouse model, blocking TGFβ along with anti-CTLA-4 and anti-PD-1 treatment diminishes Th17 cell, but enhances Th1 cell, differentiation in association with clonal expansion of CD8⁺ T cells and this leads to regression of prostate cancer bone metastasis³³. These and other studies have started to reveal that TGFβ is a major negative regulator of Th1 and Th2 cell-driven anti-tumour responses in different cancer settings³⁴ (Figure 2).

In contrast, Th17 cells mostly exhibit pro-tumour activities³⁴. In patients with colon cancer, a high proportion of IL-22-expressing Th17 cells in the tumour tissue is associated with increased TGFβ1 expression³⁵. TGFβ1 induces IL-22 expression in Th17 cells under conditions of strong co-stimulation³⁵. In an azoxymethane (AOM) and dextran sulfate sodium (DSS)-induced mouse model of colon cancer, ablation of TGFBR2 in Th17 cells suppresses IL-17A and IL-22 expression and leads to diminished tumour growth³⁵. Similarly, overexpression of SMAD7 in T cells blocks TGFβ signalling, causing enhanced Th1 cell, but diminished Th17 cell, differentiation and tumour inhibition in AOM/DSS and transplantable colon cancer models^36,37. Taken together, TGFβ generally inhibits both CD4⁺ T cell priming and differentiation of anti-tumour Th1 and Th2 cells while fostering pro-tumour Th17 cell responses, although type 1 and type 2 cytokines can also have pro-tumour functions^38–41.

T follicular helper (Tfh) cells also exhibit potent anti-tumour functions in models expressing B cell neoantigens⁴². Tfh cells are preferentially differentiated from IL-2-producing CD4⁺ T cells that have undergone strong TCR stimulation during priming⁴³. In this context, TGFβ may limit Tfh cell generation by inhibiting TCR signalling. Alternatively, TGFβ-mediated inhibition of IL-2R expression can insulate CD4⁺ T cells from IL-2 signalling, and thus promote Tfh cell generation⁴⁴. Thus, TGFβ can exhibit opposing roles in control of Tfh cell differentiation (Figure 2), and future studies will be needed to unravel how it affects Tfh cell responses in the context of cancer.

[H2] Regulatory T cells.

Thymic Treg (tTreg) cell differentiation is promoted by TGFβ signalling⁴⁵. TGFβ-mediated support of tTreg cell generation does not involve induction of the lineage-defining transcription factor FOXP3^46,47, but rather acts indirectly via attenuation of agonistic antigen-triggered T cell clonal deletion¹⁸. Peripheral Treg (pTreg) cells can also differentiate from naïve CD4⁺ T cells in the presence of TGFβ⁴⁸, in part via SMAD3-mediated induction of Foxp3 transcription⁴⁹, but to what degree the induction of pTreg cells by TGFβ contributes to tumour-associated immune regulation is incompletely understood. Studies in mouse models as well as in human melanoma, gastrointestinal and ovarian cancers have revealed that Treg cells and other CD4⁺ T cells exhibit a largely non-overlapping TCR repertoire in different cancer settings^50–52. This suggests that pTreg cell conversion does not occur in the tumour, although substantial TCR overlap was observed between intratumoral Treg cells and other intratumoral CD4⁺ T cells in patients with breast cancer⁵³. TGFβ plays a more prominent role in controlling Th cell differentiation than in driving pTreg cell differentiation, and this is likely because a higher level of TGFβ is needed to induce FOXP3 expression than to regulate Th cell lineage-specifying transcription factors⁵⁴. Beyond influencing pTreg cell differentiation, TGFβ can suppress Treg cell expansion in an autocrine manner, as demonstrated by inducible depletion of TGFBR2 in CD4⁺ T cells or Treg cell-specific ablation of TGFβ1, which both result in enhanced Treg cell proliferation^55,56. Depletion of TGFBR2 or TGFβ1 in Treg cells does not impact tumour growth in transgenic models of breast and prostate cancers^32,57. Thus, although TGFβ induction of Foxp3 expression is a widely used assay for in vitro studies (Figure 2), pTreg cell differentiation may not majorly contribute to tumour-infiltrating Treg cells. Furthermore, Treg cell maintenance and function in the tumour does not depend on TGFβ signaling in a transgenic model of breast cancer³². In a transplantable squamous cell carcinoma model, neutralization of TGFβ did not affect Treg cell frequency in tumours, but it diminished the anti-PD-1-induced expansion of Treg cell populations and synergized with anti-PD-1 to suppress tumour development⁵⁸. Whether TGFβ promotes pTreg cell differentiation under the condition of anti-PD-1 treatment remains to be determined.

[H2] Cytotoxic T lymphocytes.

TGFβ promotes IL-7Rα expression and thymic CD8⁺ T cell lineage commitment¹⁷, whereas it suppresses antigen-driven proliferation of peripheral CD8⁺ T cells⁵⁹, partly via DGKζ- and PTPN22-dependent mechanisms^60,61. In addition, FOXP1 [Au:OK?] interacts with TGFβ-activated SMAD2 and SMAD3 to repress the MYC and JUN transcription factors that drive proliferation of CD8⁺ T cells⁶². TGFβ also inhibits expression of T-bet and Eomes, which support CD8⁺ T cell differentiation into cytotoxic T lymphocytes (CTLs)⁶³. Furthermore, SMAD2 and SMAD3 partner with ATF1 to inhibit expression of CTL effector molecules, including IFNγ and granzyme B⁶⁴. Chronic antigen stimulation of CTLs results in a dysfunctional state of T cell ‘exhaustion’ characterized by high expression of inhibitory receptors including PD-1. TGFβ enhances PD-1 expression in a SMAD3-dependent manner in association with SMAD3 recruitment to the Pdcd1 promoter region⁶⁵. Thus, TGFβ engages a number of gene expression programmes to attenuate CTL differentiation and function (Figure 2).

TGFβ has been shown to inhibit cancer immunosurveillance by CTLs in transplantable tumour models. Blocking TGFβ signalling in CD8⁺ T cells inhibits the growth of transplanted thymoma cells and this is associated with expansion of antigen-specific CTLs⁶⁶. Mice with CD8⁺ T cell-specific deletion of TGFBR1 or treated with a TGFBR1 kinase inhibitor display increased rejection of an implanted colon cancer cell line, and this is associated with high expression of CXC-chemokine receptor 3 (CXCR3), which supports CTL trafficking to the tumour⁶⁷. High numbers of Treg cells are frequently observed in transplanted tumours, and their depletion results in CTL-mediated tumour inhibition⁶⁸. Of note, Treg cells express high levels of integrin αvβ8, which activates TGFβ, and T cell- or Treg cell-specific deletion of Itgb8 (which encodes integrin β8) or treatment with an integrin β8-blocking antibody suppresses tumour growth in a CD8⁺ T cell-dependent manner^69,70. Treg cells also express high levels of the TGFβ-tethering molecule GARP, and Treg cell-specific deletion of GARP or treatment with a GARP–TGFβ1 blocking antibody enhances the anti-tumour activity of anti-PD-1 in association with increased CTL effector function⁷¹. These findings suggest that Treg cells may suppress CTL-mediated cancer immunity by presenting and activating TGFβ1 in the tumour microenvironment, although αvβ8 and TGFβ1 expressed by cancer cells may also promote tumor immune evasion^70,72,73.

However, the impact of CD8⁺ T cell-targeted TGFβ signalling blockade on cancer progression is not robust in genetic models of solid tumours. In a transgenic model of prostate cancer, transfer of tumour-antigen-reactive CD8⁺ T cells that express a dominant-negative mutant of TGFBR2 (TGFDNR) delays, but does not prevent, tumour progression⁷⁴, although in another study with CD8⁺ T cells recognizing a different oncoprotein antigen, the anti-tumour effect of the transferred cells can be substantially enhanced if other lymphocytes are depleted[Au:OK? The meaning of the second part of this sentence isn’t clear to me]⁷⁵. In a transgenic model of breast cancer, deletion of TGFBR2 specifically in CD8⁺ T cells results in enhanced CD8⁺ T cell activation in tumour-draining lymph nodes with tumour-infiltrating CD8⁺ T cells displaying high levels of granzyme B and low levels of PD-1 expression. However, tumour development is not inhibited³². The discrepancy between transplantable and autochthonous cancer models could be because the commonly used epithelial tumour-derived cancer cell lines exhibit a more mesenchymal phenotype⁷⁶, and are more likely to be suppressed by CTLs with a blood-circulating effector–memory-like state, which manifests predominantly under conditions of TGFβ inhibition (Figure 2).

Of note, recent studies have revealed a critical function for TGFβ in promoting the establishment of an epithelial tissue residency programme in CTLs. In infection models, TGFβ signalling is required at multiple stages of resident memory CTL development: from naive T cell priming in secondary lymphoid organs by αvβ8-expressing dendritic cells (DCs)⁷⁷ to T cell retention in non-lymphoid tissues. The latter [Au: Edit OK?] can be mediated by keratinocyte integrin activation of CD8⁺ T cell-produced TGFβ1 in the epidermis⁷⁸, and by T-bet-expressing Treg cells that are recruited to the vicinity of CD8⁺ T cells to activate TGFβ1 via αvβ8⁷⁹. An important target of TGFβ in promoting epithelial tissue residency is the αE integrin (also known as CD103), which is likely induced in CD8⁺ T cells through SMAD2–SMAD3-mediated reversal of transcriptional repression by SMAD4–SKI–SKIL^80,81. Indeed, SMAD4 suppresses TGFβ target genes including CD103, both transcriptionally and epigenetically, in CD8⁺ T cells prior to TGFβ exposure, and SMAD4 deletion leads to microbiota-mediated accumulation and epithelial cell-induced activation of CD8⁺ T cells, resulting in severe intestinal inflammation⁸². Paired with β7 integrin, CD103 binds to E-cadherin on epithelial cells, and facilitates CTL-mediated cancer immunosurveillance by enhancing target cell cytotoxicity⁸³. Importantly, tumour-infiltrating CTLs that resemble CD103⁺ resident memory T cells have been associated with improved patient prognosis in a number of epithelial cancer types⁸⁴. In a lung cancer model, CD103 is required for CTL recruitment within epithelial tumour islets, and TGFβ signalling enhances CD103-dependent T cell adhesion and integrin-linked kinase (ILK)-mediated signalling to promote cancer cell killing⁸⁵. These findings reveal that TGFβ promotes resident memory-like CTL-mediated cancer immunosurveillance within the epithelial cancer microenvironment (Figure 2), and findings from transplantable tumour models should be interpreted with caution. [H2] Regulatory CD8⁺ T cells. Unprimed mice harbour populations of highly self-reactive MP CD8⁺ T cells that express Eomes and the IL-2/IL-15 receptor β chain CD122⁸⁶. A subset of these cells express inhibitory receptors of the LY49 family and CXC-chemokine receptor 5 (CXCR5), and are preferentially localized near or within B cell follicles to inhibit Tfh cell-mediated humoral immune responses⁸⁷. Similarly to tTreg cells, thymic selection of CD122⁺LY49⁺ regulatory CD8⁺ T cells is promoted by TGFβ signalling at a young age⁸⁸, which might be mediated by TGFβ suppression of clonal deletion of autoreactive T cell progenitors. In addition, ablation of TGFBR2 in mature T cells results in diminished expression of the transcription factor Helios in regulatory CD8⁺ T cells, which synergizes with Eomes deletion to deplete regulatory CD8⁺ T cells and trigger autoimmunity⁸⁹. Notably, follicular lymphoma patients have a population of tumour-associated cytotoxic CXCR5⁺CD8⁺ T cells that inhibit Tfh cell-mediated B cell differentiation⁹⁰. These cells can be induced and expanded by TGFβ and IL-23, and exhibit a gene expression profile similar to that of regulatory CD8⁺ T cells⁹⁰. Furthermore, this gene signature is positively associated with survival of follicular lymphoma patients⁹⁰, supporting a lymphoma surveillance function of this unique TGFβ-induced regulatory CD8⁺ T cell subset (Figure 2).

[H2] TGFβ in innate-like T cell responses in cancer.

A fraction of αβ T cells that react strongly to self-peptides presented by MHC class I or MHC class II molecules are agonistically selected and differentiate into intraepithelial T lymphocytes (IELs)⁹¹. In addition, lipids presented by the MHC class I-related molecule CD1d drive differentiation of invariant natural killer T (iNKT) cells and subsets of γδ T cells⁹². γδ T cell populations can also be selected by non-polymorphic proteins such as the butyrophilin family⁹³. These unconventional lineages of αβ T cells and γδ T cells can directly traffic to and reside in the tumour without experiencing priming in secondary lymphoid organs.

[H2] Cytotoxic innate-like T cells.

In transgenic models of epithelial cancers, a population of tissue-resident αβTCR⁺ lineage killer innate-like T cells (ILTCKs) that express the innate lymphocyte activation receptor NK1.1 and the integrin CD103 expand in tumours⁹⁴. αβTCR⁺ ILTCKs exhibit potent lytic granule-mediated cytolytic activities against cancer cells, and are phenotypically distinct from CD8⁺PD-1⁺ T cells that do not kill cancer cells⁹⁴. Following early encounter with cognate antigens, αβTCR⁺ ILTCKs arise from FcεRIγ-expressing IEL-like thymic progenitors, and seed healthy and tumour tissues independently of priming in secondary lymphoid organs⁹⁵. Expansion and effector differentiation of αβTCR⁺ ILTCKs in the tumour are driven by the cancer cell-expression of IL-15⁹⁵ and TGFβ signalling (B.G.N, unpublished observations). Thymic selection of IELs has also been shown to require TGF-β signaling as a likely consequence of attenuated clonal deletion^18,96. In addition, FcεRIγ-expressing TCRαβ ILTCKs are induced in human colon cancer⁹⁵. These findings suggest that TGF-β promotes thymic development and peripheral differentiation of TCRαβ ILTCKs to support their prewired anti-tumor function (Figure 3).

Figure 3. TGFβ control of prewired innate lymphocyte and innate-like T cell responses in cancer.

a, In the thymus, TGFβ is required for development of CD1d/lipid agonist antigen-reactive invariant Natural Killer T (iNKT) precursor (iNKTp) and MHC/peptide agonist antigen-reactive intraepithelial lymphocyte (IEL) precursor (IELp), likely through attenuated clonal deletion. In addition, TGFβ is required for the differentiation of IL-17-producing iNKT cells (iNKT17) and IL-17-producing ɣδ lineage T cells.

b, In the bone marrow, the innate lymphoid cell progenitor (ILCp) gives rise to ILC1, ILC2, and ILC3, while the natural killer (NK) progenitor (NKp) gives rise to NK cells. TGFβ promotes differentiation of ILC2 in part via ST2 expression.

c-d, In the tumor, TGFβ inhibits NK cell activation and effector function against mesenchymal phenotype cancer cells (c), while promoting cytotoxic ILC1s as well as killer innate-like T cells (ILTCKs) of both αβ and γδ T cell lineages that co-localize with epithelial cancer cells in part through CD103 interaction with E-cadherin (d). The cancer surveillance functions of NK cells, ILC1s and ILTCKs are additionally regulated by IL-15 and activation receptor signaling pathways.

In addition to TCRαβ ILTCKs, a population of tumor-resident NK1.1⁺CD103⁺ TCRγδ ILTCKs expand in epithelial cancers⁹⁴, and they are also dependent on cancer cell-expressed IL-15 as well as TGF-β signaling for terminal differentiation (B. G. N., E.R.K, unpublished observations). In mice, development of γδ T cells commences in the fetal thymus and occurs in overlapping waves, with each wave expressing specific pairs of TCRγ and TCRδ chains⁹⁷; yet the TCR usage and thymic origin of γδTCR⁺ ILTCKs remain to be determined. A subset of γδ T cells that expresses high levels of cytotoxic molecules and innate lymphocyte-associated markers expands in human breast tumours, and its gene expression signature predicts better patient survival^98,99. Notably, these cells can express the tissue residency marker CD103, and are functionally skewed towards cytolysis and IFNγ production⁹⁹. In vitro studies have revealed that TGFβ enhances cell contact-dependent cytotoxicity of γδ T cells through upregulation of CD103¹⁰⁰, supporting a positive role of TGFβ in γδTCR⁺ ILTCK-mediated cancer immunosurveillance (Figure 3).

[H2] Helper innate-like T cells.

iNKT cells are cytokine-producing T cells whose development is supported by TGFβ signalling in part through enhanced survival of iNKT precursors^20,21,101 (Figure 3). Three subsets of mature iNKT cells — iNKT1, iNKT2 and iNKT17 cells — are characterized by their expression of IFNγ, IL-4 and IL-17, respectively, and their differentiation initiates in the thymus. Thymic development of iNKT17 cells is dependent on TGFβ signalling¹⁰² (Figure 3), and SMAD4 is required for IL-17 production by these cells [Au:OK?] in peripheral tissues¹⁰². There is evidence for both pro- and anti-tumour functions of iNKT cells. Mice deficient in CD1d show protection against tumour development [Au:OK?] in multiple cancer models^103,104, but administering α-galactosylceramide, which induces iNKT activation and cytokine production, also shows a protective effect in models of cancer [Au:OK?] ^105,106. Whether and how TGFβ regulates iNKT cell responses in cancer remains to be determined.

Thymic development of IL-17-producing γδ T cells is dependent on TGFβ1¹⁰⁷ (Figure 3), and these cells promote cancer progression in part through recruitment of neutrophils¹⁰⁸. In contrast, the development of skin-resident dendritic epidermal T cells (DETCs), another population of γδ T cells, is not affected in TGFβ1-deficient mice¹⁰⁹. DETCs produce high levels of IL-13, which act on epithelial cells to suppress cutaneous carcinogenesis¹¹⁰. DETCs also promote antibody class-switching to IgE by directing an IL-4-dependent Th cell response, which protects against skin-tumour formation by inhibiting epithelium damage¹¹¹. Whether TGFβ regulates DETC-mediated type 2 cancer immunity is open to future investigation.

[H2] TGFβ in B cell responses in cancer.

B cells mediate humoral immune responses by secreting antibodies, and this involves the activation of conventional B2 cells in secondary lymphoid organs. Blockade of TGFβ signalling in B cells causes enhanced activation and proliferation, whereas class switching to IgA is attenuated¹¹². Peyer’s patches are major sites of IgA production and IgA responses are dependent on B cell expression of GARP and follicular dendritic cell (FDC) expression of αvβ8^113,114. However, blockade of TGFβ signalling does not affect IgA production in lymph nodes, and delayed inhibition of TGFβ signalling in germinal center (GC) B cells does not impair IgA class switching in Peyer’s patches¹¹⁵. Iterative cycling of GC B cells between the light zone, where they are stimulated by Tfh cells and FDCs, and the dark zone, where they undergo somatic hypermutation, is crucial for antibody affinity maturation. GC B cells show high levels of TGFβ-dependent SMAD2 phosphorylation, and TGFβ signalling in GC B cells promotes their transition from the light zone to the dark zone of the GC as well as antibody affinity maturation¹¹⁵. Thus, TGFβ has pleiotropic functions in control of B cell responses (Figure 2).

In mouse models of breast cancer with high mutational burdens, B cells are crucial for efficacious immune checkpoint blockade (ICB) therapies that promote production of cancer cell-reactive IgG¹¹⁶. Increased numbers of B cells and B cell-associated tertiary lymphoid structures have also been associated with better ICB outcomes in a number of cancer types^117–119. Whether and how TGFβ regulates the anti-cancer B cell response will be an interesting topic for future study. In contrast, depletion of B cells enhances CTL-dependent inhibition of prostate tumours in mice treated with oxaliplatin, a platinum-based chemotherapy[Au:OK?]¹²⁰. Oxaliplatin induces tumour infiltration by a population of IgA⁺PDL1⁺IL-10⁺ plasma cells, and blockade of TGFβ signalling in B cells attenuates their differentiation into plasma cells [Au:OK?], potentiating the oxaliplatin-induced inhibition of tumour growth¹²⁰. Considering that TGFβ signalling is selectively required for IgA class switching at an early stage of B cell activation in Peyer’s patches¹¹⁵, TGFβ may be playing a similar role in the prostate tumour-associated IgA⁺ plasma cell response, but where this effect occurs and whether it occurs in other tumour contexts remain unknown.

[H1] TGFβ in innate immunity in cancer

The innate immune system includes lymphoid and myeloid cell lineages that form the first line of defense against immune challenges and regulate adaptive immune responses. Similar to its influence over T and B cells, TGFβ exhibits pleiotropic functions in control of innate immune cell functions in cancer.

[H2] TGFβ in innate lymphocyte responses in cancer.

Innate lymphocytes manifest effector functions that mirror those of T cells and innate-like T cells. Natural killer (NK) cells recirculate and can induce target cell cytotoxicity, while innate lymphoid cells (ILCs) reside in peripheral tissues, produce an array of inflammatory cytokines, and can also trigger lytic granule-mediated cytotoxicity.

[H2] Cytotoxic innate lymphocytes.

Depletion of TGFBR2 from NKp46-expressing innate lymphocytes, which include NK cells, [Au:OK?] inhibits cancer cell metastasis¹²¹, while expression of an active form of TGFBR1 in these cells enhances cancer cell metastasis¹²¹ and accelerates growth of methylcholanthrene-induced fibrosarcoma¹²². In addition, deletion of SMAD4 causes enhanced cancer cell metastasis, which is associated with a phenotypical change in NK cells similar to what occurs in cells expressing an active form of TGFBR1¹²³. This suggests that TGFβ suppresses NK cell-mediated surveillance of cancer metastasis via release of SMAD4 transcriptional repressive activity, such as that mediated through SMAD4–SKI–SKIL. However, it remains possible this effect is due to signalling by other TGFβ superfamily members that also use SMAD4. TGFβ inhibits IL-15-induced activation of the metabolic regulator mTORC1 in NK cells, which might account for their reduced proliferation, diminished expression of activation receptors, and impaired cytotoxic activities¹²¹. In addition, TGFβ impedes IL-2-induced glycolysis and oxidative phosphorylation in human NK cells, which may limit their anti-tumour activity [Au:OK?] ¹²⁴. These findings collectively indicate that TGFβ can suppress NK cell-mediated cancer surveillance (Figure 3).

While NK cells recirculate, tumour-resident cytotoxic innate lymphocytes are induced in transgenic models of epithelial cancers, and these cells also suppress tumour development^94,125. These cells do not develop from mature NK cells, but differentiate from ILC progenitors¹²⁵. Maintenance of CD103⁺ group 1 ILCs (ILC1s) is dependent on cancer cell-expressed IL-15 as well as TGFβ signalling, and ablation of TGFBR2 in ILC1s results in accelerated tumour growth¹²⁵. Cytotoxic ILC1s are also induced in human epithelial cancers and these cells are associated with better survival rates in patients with chromophobe renal cell carcinoma¹²⁶. In co-cultures with head and neck cancer cells, the differentiation of CD94⁺NKp80⁺CD16⁻ ILC progenitors into cytotoxic ILC1s is dependent on TGFβ signalling¹²⁷. Thus, similar to what is seen with resident memory-like CTLs and ILTCKs, tissue-resident [Au:OK?] cytotoxic ILC1s require TGFβ signalling for surveillance of epithelial cancers (Figure 3).

[H2] Helper innate lymphocytes.

ILC2s produce inflammatory cytokines in response to ‘alarmins’ such as IL-33. Blockade of TGFβ signalling reduces ILC2 progenitor numbers and attenuates ILC2 development in association with low expression of the IL-33 receptor ST2¹²⁸ (Figure 3). Administration of IL-33 expands ILC2s and suppresses tumour development in a mouse pancreatic cancer model via recruitment of DCs that activate CD8⁺ T cells in tumours¹²⁹. In a mouse melanoma model, ILC2-derived GM-CSF promotes the expansion and effector function of eosinophils to suppress tumour development¹³⁰. In line with these observations, high levels of ILC2 infiltration track with good clinical prognosis in patients with pancreatic cancer or melanoma^129,130. In contrast to these positive effects of TGFβ in promoting anti-tumour ILC2 responses, in the AOM–DSS-induced colorectal cancer model, TGFβ signalling promotes conversion of ILC3s into a population of IL-10-producing regulatory-type ILCs, and co-transfer of these regulatory ILCs with cancer cells results in accelerated tumour growth¹³¹.

[H2] TGFβ in mononuclear phagocyte responses in cancer.

DCs, monocytes, and macrophages constitute the mononuclear phagocytes of the immune system. DCs present antigens to T and B cells in secondary lymphoid organs, and are critical regulators of priming-dependent adaptive immunity. Monocytes circulate in blood and differentiate to tissue macrophages which, together with macrophages seeded early during development, maintain tissue homeostasis and control immune responses.

[H2] Dendritic cells.

As sentinels of antigenic challenge, DCs differentiate from the common DC progenitor distinct from monocytes, and include conventional DC subset 1 (cDC1) and cDC2¹³². Conditional ablation of Tgfbr2 with CD11c-Cre transgenic mice shows no overt phenotypic changes in DCs in secondary lymphoid organs and yet results in lethal autoimmunity¹³³. This discrepancy may be due to poor DC specificity of CD11c-Cre, which targets other leukocyte populations including activated T cells¹³⁴. Notably, blockade of TGF-β signaling selectively impairs differentiation of intestinal CD103⁺CD11b⁺ cDC2s, but not CD103⁺CD11b⁻ cDC1s, in association with defective pTreg and Th17 cell differentiation¹³⁴. How this distinct TGF-β DC regulation pathway affects T cell responses in intestinal malignancy has not been explored.

[H2] Monocytes and macrophages.

Tissue macrophages can differentiate from embryonic progenitors, including epidermal Langerhans cells and alveolar macrophages, which both require autocrine TGFβ1 for their maintenance^135,136. In a model of cutaneous squamous cell carcinoma, mice with constitutive depletion of Langerhans cells are protected from DMBA-induced tumour development as a consequence of defective conversion of DMBA to DMBA-trans-3,4-diol¹³⁷. In contrast, inducible depletion of Langerhans cells causes accelerated tumour growth due to failed recruitment of NK1.1⁺ innate lymphocytes into the epidermis¹³⁸. While TGFβ plays key roles in development of these cells, whether and how it further influences their function in cancer remains to be explored. [Au: Can you explain what this is telling us about TGFb in the context of cancer here?]

In addition to tissue-resident macrophages, monocytes are frequently recruited to tumours and differentiate into tumour-associated macrophages¹³⁹. TGFβ can induce chemotaxis of monocytes¹⁴⁰ and can influence the monocyte-derived macrophage phenotype, such as in the mammary gland¹⁴¹ and in the intestine¹⁴². In the context of cancer, blockade of TGFβ signalling in myeloid cells, including macrophages, results in diminished tumour growth and reduced cancer cell metastasis^143–146. Likewise, in a DMBA-induced breast cancer model, inducible expression of a dominant-negative mutant of TGFBR2 in macrophages leads to reduced tumour incidence¹⁴⁷. Administration of a STING agonist induces type 1 interferon production and suppresses growth of mammary tumours orthotopically transplanted into the fat pad, but not the growth of mammary tumours that spontaneously arise in transgenic mice[Au:OK?]¹⁴⁸. The differential response is due to excessive macrophage TGFβ signalling in the genetic model, which inhibits STING-induced IRF3 phosphorylation and the ensuing type 1 interferon production¹⁴⁸. Conversely, activated IRF3 suppresses TGFβ-induced SMAD3 phosphorylation¹⁴⁹. These findings reveal cross-inhibitory functions of TGFβ and the IRF3 pathway, which may account for the immunosuppressive function of TGFβ in macrophages in the setting of cancer [Au:OK?].

[H2] TGFβ in granulocyte responses in cancer.

Neutrophils are the most abundant granulocytes in the circulation and are frequently recruited to tumours. They predominantly have pro-tumour functions, driving cancer cell genome instability, cancer cell proliferation and tumour angiogenesis¹⁵⁰. In a transgenic model of colon cancer, neutrophil-specific deletion of Tgfbr1 inhibits metastasis in association with reduced numbers of tumour-infiltrating neutrophils¹⁵¹ — this is in line with the chemotactic activities of TGFβ on neutrophils^152,153. Thus, TGFβ may induce neutrophil recruitment to promote tumour development.

[H1] Targeting TGFβ for cancer immunotherapy

The pleiotropic roles of TGFβ in the context of cancer make it a prominent, but complex, target for therapy. Although TGFβ can promote cancer cell invasion and dissemination, its tumour suppressor functions can be observed [Au:OK?] in advanced cancers¹⁵⁴. Furthermore, loss-of-function mutations affecting TGFβ receptors in cancer cells can lead to a spillover effect, with TGFβ instead acting on non-cancer cell populations in the tumour microenvironment^155,156. In addition, while mesenchymal stromal cell TGFβ signalling inhibits T cell infiltration into tumours^157–160, blockade of TGFβ signaling in fibroblasts induces forestomach carcinoma as a consequence of excessive inflammation-induced DNA damage^161,162. Systemic blockade of TGFβ signalling can also trigger cardiac toxicity¹⁶³, limiting safety profiles of some drug programmes. Cancer immunotherapies that block inhibitory pathways such as PD-1 can show great efficacy but are often associated with immune-related adverse events (IRAEs)¹⁶⁴. Given the role of TGFβ in inhibiting peripheral autoreactive lymphocytes¹⁹, such immune-related toxicities remain potential side effects of TGFβ-blocking drugs. Despite these constraints, several TGFβ-targeting programmes are being developed for cancer immunotherapy.

[H2] Systemic blockade of the TGFβ pathway.

Interest in the pro-tumour functions of TGFβ has led to the development of antagonists that intervene with almost all steps of TGFβ signalling (Figure 4). In preclinical studies, these drug programmes show immune system-dependent therapeutic efficacy particularly in settings of combination therapies.

Figure 4. Strategies to target the TGFβ pathway for cancer therapy.

Pharmacological interventions of the TGFβ pathway are grouped into two categories: first, systemic blockade that acts on 1) the latency-associated peptide (LAP) and TGFβ complex, 2) the LAP/TGFβ tethering molecule GARP, 3) the LAP/ TGFβ activating integrin, 4) the active form of TGFβ, or 5) the TGFβ receptor with antibodies and the ectodomain of TGFBR2 (TGFβ Trap)-based biologics as well as small-molecule kinase inhibitors; and second, targeted blockade with bispecific molecules to deliver TGFβ Trap to targeted cell populations such as CD4⁺ T cells with 4T-Trap or PD-L1-expressing cancer cells with PD-L1-Trap, or overexpressing a dominant-negative mutant of TGFBR2 (TGFDNR) in tumor antigen-specific T cell receptor (TCR) T cells or cancer cell-reactive chimeric antigen receptor (CAR) T cells for cell therapy.

[H2] Blocking antibodies.

SRK-181 is a blocking antibody that selectively binds to latent TGFβ1 and inhibits its activation¹⁶⁵. In transplantable tumour models, SRK-181 synergized with anti-PD-1 to suppress tumour development in association with the expansion of CD8⁺ T cell and Treg cell populations in the tumours¹⁶⁵. Of note, unlike 12.7, which is a high-affinity antibody that neutralizes all TGFβ family members, SRK-181 exhibits low cardiovascular toxicities¹⁶⁵. Robust CD8⁺ T cell-dependent anti-tumour immunity has also been reported following treatment with TW7–28G11 and TW7–16B4, which are mouse [Au:OK?] antibodies that bind to the TGFβ1–LAP complex or LAP, respectively, and inhibit TGFβ1 activation^166,167. Of note, unlike SRK-181, these antibodies deplete LAP⁺ T cells, including Treg cells, which likely contributes to their single-agent effects.

Docking of the TGFβ1–LAP complex on the T cell surface is mediated by GARP. 58A2 is an antibody that binds to the GARP–TGFβ1–LAP complex and inhibits TGFβ1 activation⁷¹. A 58A2 variant that does not bind Fc receptors synergizes with anti-PD-1 to suppress growth of CT26 tumours in association with enhanced effector functions of tumour-infiltrating CD8⁺ T cells⁷¹. Treg cell-specific deletion of GARP nullifies the anti-tumour effects of 58A2[Au:OK?]⁷¹, suggesting that the GARP–TGFβ1–LAP complex expressed on Treg cells is the functional target, although endothelial cells may also be a target in an MC38 tumour model¹⁶⁸.

As discussed above, latent TGFβ1 and TGFβ3 complexes are activated by the integrins αvβ6 and αvβ8. ADWA-11 is an αvβ8 blocking antibody, and a non-Fc-binding form of ADWA-11 can synergize with anti-PD-1, anti-CTLA4, an agonist antibody against 4–1BB and radiation therapy to suppress the growth of transplanted tumours[Au: Edit OK?]⁶⁹. Itgb8 mRNA is enriched in tumour-associated Treg cells, and T cell-specific deletion of Itgb8 recapitulates the tumour suppression phenotype of ADWA-11⁶⁹. In contrast, experiments using an αvβ8 blocking antibody, C6D4, showed that cancer cell-expressed αvβ8 activates TGFβ1–LAP produced by T cells^72,73. Activation of TGFβ can also be mediated by cancer cell-expressed αvβ6 in pancreatic cancers, and its blockade either enhances or suppresses cancer progression in tumour-bearing mice treated with the chemotherapy gemcitabine^169,170. It is currently unknown to what extent the opposing outcomes are due to pleiotropic functions of TGFβ on cancer cells and/or stromal cells.

Blocking antibodies against active forms of TGFβ are among the first antagonists tested in preclinical and clinical studies. Administration of a pan-TGFβ blocking [Au:OK?] antibody 1D11 revives T cell-dependent cancer immunity in combination with radiation, anti-CTLA4 or anti-PD-1^33,171,172. A humanized version of 1D11, fresolimumab, has entered a phase 2 trial for metastatic breast cancers, and prolongs patient median survival when combined with radiation therapy¹⁷³. Several other pan-TGFβ neutralizing antibodies as well as an antibody that blocks TGFβ1 and TGFβ2, but not TGFβ3, have also shown efficacy in preclinical studies^58,174,175. However, systemic blockade of TGFβ can suppress anti-tumour immune responses as well. In a model of head and neck cancer, vaccination-triggered cancer immunity is attenuated by a pan-TGFβ antibody, as it inhibits the generation of tissue-resident CD8⁺ T cells¹⁷⁶.

[H2] TGFβ traps.

AVID200 is a fusion protein of the TGFBR2 ectodomain to human Fc, and selectively neutralizes TGFβ1 and TGFβ3¹⁷⁷. A phase 1 dose escalation study showed that AVID200 is well tolerated as a monotherapy, with some cancer patients experiencing stable disease after treatment. The ectodomain of TGFBR2 has also been fused with anti-PD-L1 as a bifunctional anti-PD-L1–TGFβ-Trap molecule, M7824, that better suppresses tumour development than anti-PD-L1 alone in transplantable tumour models¹⁷⁸. M7824 (Bintrafusp alfa) shows a good safety profile and clinical efficacy in phase 1 trials¹⁷⁹, but does not produce better outcomes as a monotherapy in lung and biliary tract cancers than anti-PD-1 or anti-PDL1. Additional anti-PD-L1 or anti-CTLA4 and TGFBR2 extracellular domain fusion proteins have been generated, and shown better efficacies in preclinical models than anti-PD-L1 or anti-CTLA4 monotherapy¹⁸⁰. Anti-CTLA4–TGFβ-Trap effectively depletes Treg cells¹⁸⁰, but its safety profile has not been examined. [Au: Edit OK?]

[H2] Small-molecule inhibitors.

Galunisertib (LY2157299) is an orally available TGFBR1 kinase inhibitor, and has shown anti-tumour activity as a single-agent or in combination with anti-PD-L1 in preclinical models^159,181,182. An intermittent dosing of galunisertib appeared safe, and prolonged survival in a subset of patients with hepatocellular carcinoma in combination with the tyrosine kinase inhibitor sorafenib in phase 1b/2 clinical trials^183–185. In addition, galunisertib increased median survival time of patients with unresectable pancreatic cancer in combination with gemcitabine¹⁸⁶, but had limited clinical activity in combination with the anti-PD-L1 antibody durvalumab¹⁸⁷. Several other TGFBR1 kinase inhibitors, including vactosertib (TEW-7197), LY3200882 and PF06952229, are also being tested in clinical trials.

[H2] Targeted blockade of the TGFβ pathway.

The pleiotropic functions of TGFβ engender complex outcomes when using a systemic TGFβ blockade approach. Non-selective drug delivery also limits the maximum-tolerated dose of TGFβ antagonists that can be achieved in therapeutically relevant targets. To overcome these limitations, targeted blockade of TGFβ signalling in selected leukocyte populations have emerged as a novel immunotherapy strategy (Figure 4).

[H2] Antibody-based targeting.

Following the finding that TGFβ primarily targets CD4⁺ T cells to induce tumour immune tolerance³², a bispecific receptor decoy 4T-Trap was developed with fusion of the extracellular domain of TGFBR2 to ibalizumab, a non-immunosuppressive CD4-specific antibody¹⁸⁸. 4T-Trap effectively suppresses TGFβ signalling in CD4⁺ T cells, and triggers vasculature remodelling and starvation-associated cancer cell death, which is further enhanced by VEGF neutralization¹⁸⁸. The bi-functional molecule anti-CTLA4–TGFβ-Trap (M7824) and the PD-L1–TGFβ bispecific antibody YM101 can also be classified as targeted TGFβ blockers^178,180,189. Notably, anti-CTLA4–TGFβ–Trap exhibits better anti-tumour activity than the combination of anti-CTLA4 and a systemic TGFβ antagonist¹⁸⁰. M7824 and YM101 are delivered to PD-L1-expressing cells, including APCs and cancer cells, but this targeted delivery does not appear to exert better cancer immunity compared to combinations of anti-PD-L1 and a nontargeted TGFβ antagonist^178,189, and may induce undesirable outcomes by blocking TGFβ signaling in cancer cells.

[H2] Cell engineering-based targeting.

Expression of TGFDNR in tumour-reactive CD4⁺ or CD8⁺ T cells effectively suppresses cancer progression in a transplantable melanoma model¹⁹⁰. In a transgenic model of prostate cancer, TGFDNR also enhances CD8⁺ T cell-mediated cancer immunity with varying efficacies^74,75. This therapeutic strategy has shown efficacy in a small group of Epstein-Barr virus (EBV)⁺ Hodgkin lymphoma patients transfused with autologous EBV antigen-specific CD8⁺ T cells¹⁹¹. In addition to antigen-specific T cells, chimeric antigen receptor (CAR)-based T cell therapy can be enhanced through expression of TGFDNR, deletion of the Tgfbr2 gene, or expression of a secreted form of an anti-PD-1–TGFBR2 extracellular domain fusion protein in preclinical models^192–194. A recent phase 1 trial of TGFDNR-expressing CAR T cells targeting the prostate-specific membrane antigen (PSMA) appears feasible and generally safe with one patient experienced a marked clonal CAR cell expansion¹⁹⁵. Of note, high and persistent expression of TGFDNR in CD8⁺ T cells can trigger lymphoproliferative disorders in mouse models^196,197, cautioning potential safety liabilities of this approach.

TGFβ signalling can be repurposed to enhance cancer T cell immunity through the expression of a TGFBR2–41BB chimeric molecule to ectopically induce signalling of 41BB, a co-stimulatory T cell receptor¹⁹⁸. Targeting TGFβ with an engineered anti-TGFβ single-chain variable fragment-based CAR has also been shown to effectively activate T cells in response to TGFβ¹⁹⁹. In addition to T cells, NK cells exhibit enhanced anti-cancer immunity in preclinical models when engineered to express TGFDNR, a SMAD3 shRNA or a chimeric molecule combining TGFBR2 with DAP12 (a transmembrane signalling adapter protein in NK cells)^200,201.

[H1] Concluding remarks

As an evolutionarily ancient regulatory cytokine, TGFβ has pleiotropic functions within and beyond the immune system. Its pro- or anti-tumour immune activities depend on its source, dose, context and its leukocyte targets, as well as on the cancer type and disease stage. TGFβ regulates priming-dependent immune responses during lymphocyte activation and differentiation in secondary lymphoid organs and in the tumour itself. Similarly, prewired innate immune cell and innate-like adaptive lymphocyte responses can be influenced by TGFβ, both during their development and in the tumour.

Although TGFβ-mediated inhibition of Th1 and Th2 cell differentiation and induction of Th17 cell differentiation generally fosters tumour progression, TGFβ has been shown to promote both pro-tumour and anti-tumour functions of cytotoxic lymphocyte lineages. TGFβ-mediated suppression of effector memory-like CTL and NK cell responses promotes immune evasion of mesenchymal-type cancer cells, but it bolsters immunosurveillance of epithelial cancers by resident memory-like CTLs as well as by tumour-resident ILC1s and ILTCKs.

A deeper understanding of how TGFβ influences tissue-associated immune responses in homeostasis will further inform the role TGFβ plays when cancer disrupts this balance. Such disruption can lead to pathological TGFβ signaling, impacting both immune and non-immune cells and leading to unexpected outcomes. A spillover effect is observed when cancer cells lose responsiveness to TGFβ through somatic mutations in TGFβ receptors, thus leading to an increase in TGFβ signaling in non-cancer cells and modulation of the tumour microenvironment.

Given the multifaceted physiological roles of TGFβ, both in the immune system and beyond, therapeutic approaches to target this cytokines for cancer therapy need to be stringently vetted for safety, as treatments targeting other immune inhibitory pathways can trigger IRAEs. Potential therapies also need to be rigorously tested for efficacy in tumour models, particularly in autochthonous tumor models that more closely recapitulate human malignancies. The pipeline of TGFβ antagonists has expanded to cover almost all steps of the molecular signalling pathway, and cell-specific targeting of TGFβ function has also been leveraged for targeted immunotherapy. It is hoped that our growing appreciation of the intricacies of TGFβ signalling in various cancers will reinvigorate attempts to target this pathway for cancer treatment in the future.

Glossary

Innate immunity

immunity mediated by leukocytes that express germline-encoded receptors that recognize common, broad patterns associated with pathogens and cell stress

Adaptive immunity

immunity mediated by leukocytes that express an antigen-specific receptor (T cell receptor or B cell receptor) that drives both their development in primary lymphoid organs and their function in the periphery

Prewired immunity

immune responses involving innate myeloid cells as well as innate lymphocytes and innate-like adaptive lymphocytes that reside in premalignant tissues or traffic to tumours directly after development in primary lymphoid organs

Priming-dependent immunity

immune responses involving adaptive lymphocytes that are primed in secondary lymphoid organs before homing to tumours

Autochthonous tumor model

tumour model where tumour initiation is by cells of an endogenous organ in an intact animal via processes such as oncogene expression or carcinogen exposure