A Comprehensive Review on the Role of Co-signaling Receptors and Treg Homeostasis in Autoimmunity and Tumor Immunity

1. Introduction

Immune surveillance is an essential component of the defense mechanism against infections and tumor progression. However, aberrant immune responses against self-antigens can result in autoimmune disease[1, 2] and failure to recognize and/or mount an effective immune response against tumor antigens can allow cancer to grow[3, 4]. Immune system comprises of many lymphoid and myeloid cell types which coordinately function to maintain immune homeostasis and ensure mounting of protective responses against infections while attenuating deleterious responses against self-antigens. T-cells are the most important effector cell types that mediate various immune responses and therefore, have been preferred targets for immunomodulation. T cells can be broadly classified as T-effector (Teff) cells and T-regulatory (Treg) cells based on the paradoxical nature of their function. Foxp3 is the lineage-specific transcription factor exclusively expressed in Tregs and not in Teff cells at least in mice[5]. The major discriminating factor between Treg and Teff cells is their affinity for self-antigens[6]. During thymic selection, T-cell clones expressing high-affinity T-cell receptors (TCRs) for self-antigens are either deleted by negative selection or rendered anergic. However, thymic negative selection is imperfect in that self-reactive T-cell clones often escape negative selection, migrate to the periphery and contribute to autoimmunity. However, T-cells expressing TCRs with an intermediate affinity for self-antigens gain Foxp3 expression and become Tregs are positively selected and migrate to the periphery where they help maintain peripheral self-tolerance[6]. In addition, CD4⁺CD25⁻Foxp3- Tconv cells can differentiate into induced Treg (piTregs) cells in the periphery and suppress excessive inflammatory response directed against microbial antigens[7]. Physiologically, the antigen-specific Teff cells can eliminate infectious agents[8] and kill tumor cells[9]. However, uncontrolled or excessive Teff response can result in tissue destruction as seen in inflammatory diseases.[10] Similarly, inappropriate Teff response against self-antigens can result in autoimmune diseases[11]. In contrast, Treg cells can prevent or attenuate pro-inflammatory and autoimmune responses of Teff cells and help maintain immune homeostasis [12, 13]. In addition, Tregs can also suppress tumor-specific immune responses and thus facilitate immune evasion by tumor-cells resulting in tumor growth[14]. Therefore, understanding the molecular underpinnings regulating T-cell responses is not only of immense interest but also of clinical importance.

T-cell activation and expansion are regulated through a bi-signaling mechanism [15, 16]. Binding of MHC bound antigenic peptides displayed on antigen presenting cells (APCs) to TCRs of antigen-specific T cells activates the primary (first) TCR signal. This primary TCR signal when combined with secondary co-signals resulting from interactions between co-stimulatory/co-inhibitory ligands expressed on APCs and their cognate receptors expressed on T-cells determine the fate of T-cell response [17, 18]. Therefore, the co-signaling molecules play a critical role in regulating the immune response. Co-signaling molecules can be divided into two superfamilies based on their structure; 1) Immunoglobulin superfamily (Ig SF), 2) Tumor Necrosis Factor Superfamily (TNFSF)[19]. Ig SF co-signaling receptors are grouped based on their extracellular domain architecture[19] and consist of CD28, PVR-like, T-cell Immunoglobulin and Mucin (TIM) and CD2/Signaling Lymphocytic Activation Molecule (SLAM) families. TNF receptors share a conserved cysteine-rich signature containing domain[20] and include Glucocorticoid-Induced TNFR-related protein (GITR), OX40, 4-1BB, CD27, TNF-RI & II, Herpes Virus Entry Mediator (HVEM), etc.

Based on their function, co-signaling molecules can be divided into co-stimulatory and co-inhibitory molecules. Major co-stimulatory molecules include CD28, ICOS, TNFR-II, GITR, OX40, and 4-1BB, etc, and these co-stimulatory interactions play an essential role in many facets of the immune response such as T cell-antigen priming, expansion, survival, and differentiation and effector functions[19]. On the other hand, co-inhibitory molecules counteract costimulatory signaling and antagonize afore-mentioned facets of an immune response. Major co-inhibitory molecules include CTLA4, PD1, LAG3, TIM3, and TIGIT[19]. In general, co-stimulatory interactions are involved in the activation and expansion of naïve T-cells while co-inhibitory interactions provide feedback inhibitory signals to activated T cells to prevent excessive immune response [21]. Thus, the outcome of an immune response can be modulated through sequential co-signaling in a context-dependent manner to maintain immune homeostasis.

In general, co-signaling receptors, other than CD28 which is constitutively expressed, are predominantly expressed on Teff cells largely upon TCR-activation, but not on naïve T-cells. In contrast, Tregs constitutively express several co-signaling receptors like CTLA4, GITR, and OX40. In addition, many co-signaling receptors are preferentially over-expressed on Tregs which include PD1, ICOS, TIGIT, LAG3, TIM3, TNF-RII, and 4-1BB[22, 23]. This preferential overexpression is due to relatively higher TCR-signal strength experienced by Tregs compared to Teff cells during thymic selection. Such preferential overexpression of co-signaling receptors on Tregs might allow selective targeting for therapeutic utility. Overexpression of co-inhibitory receptors by Tregs is one of the prominent mechanisms by which they subvert APC-T-cell communication and suppress Teff cell response in autoimmunity[24]. Agonistic antibodies to co-inhibitory receptors and antagonistic antibodies to co-stimulatory receptors have shown promising results in treating autoimmune diseases[21]. However, tumor cells and tumor-infiltrating myeloid cells such as Myeloid-Derived Suppressor Cells (MDSCs)[25] and Tumor-Associated Macrophages (TAMs)[26] hijack this mechanism by expressing co-inhibitory receptors to maintain an immune suppressive microenvironment to evade immune surveillance[27]. In addition, tumor-infiltrating Tregs also significantly contribute to immune evasion by tumors through co-inhibitory receptor signaling[28] (Fig-1).

Activation of co-stimulatory receptors such as CD28, OX40, and GITR promote Teff cell proliferation and effector functions. Agonistic antibodies to TNFRSFs such as OX40, GITR, 4-1BB, CD27, and CD30 have been used for inducing antitumor Teff cell responses in many cancers[29] (Fig-1). Therefore, agonistic signaling through these receptors is generally thought to regulate Teff cell response and contribute to immune enhancement. However, that notion was called into question upon discovering reduced Tregs in mice deficient in co-stimulatory receptor CD28 and accelerated autoimmune diabetes in CD28^−/− NOD mice.[30] The second breakthrough occurred when it was found that co-stimulation through TNF-RII, GITR, and OX40 coupled to TCR signal strength facilitated thymic Treg selection and differentiation, and combined inhibition of these signaling pathways significantly impaired Treg generation in the thymus[31]. Thus, it has become evident that these co-stimulatory receptors play a key role in Treg differentiation and functions as well.

Immune Checkpoint Blockade (ICB) using anti-CTLA4 and anti-PD-1 antibodies have been successfully used to overcome immune evasion by tumor cells and facilitate an effective anti-tumor immune response[28]. Although modulating tumor micro-environment with ICB has resulted in better therapeutic efficacy than other interventions, still many patients are refractory to this approach, attributed to either compensatory mechanisms mediated through other co-inhibitory receptors or due to yet unknown inhibitory mechanisms. Moreover, autoimmune adverse events observed in patients who received ICB[32, 33] have revealed the challenges in targeting co-signaling receptors for selective immunomodulation.

Here, we critically review the role of important co-stimulatory and co-inhibitory molecules such as CD28, CTLA4, PD1, TIGIT, TIM3, LAG3, TNF-RII, GITR, OX40 and 4-1BB in Treg homeostasis and autoimmunity, and their involvement in tumor immune evasion and autoimmune complications arising upon targeting those molecules in cancer immunotherapy.

2. Immunoglobulin superfamily co-signaling receptors

3. TNF Superfamily co-signaling molecules

Co-signaling receptors and ligands of the TNF superfamily are summarized in Table-2. TNFSF ligands predominantly assemble as homotrimers, with the exception of mouse GITRL (dimeric) and LTaB2 (heterotrimer). In addition to their membrane-bound form, TNFSF ligands can also be produced in soluble form upon proteolytic cleavage of their ectodomains[20]. TNFRSF co-signaling receptors share a cysteine-rich domain with three disulfide bonds adjoining a core CXXCXXC motif. There are significant differences in the number of cysteine-rich domains between family members, varying from only one in BAFFR to six in CD30[113]. TNFSF receptors can be classified into three subtypes based on their cytosolic signaling domains.1) Death receptors containing death domains -DR3, DR6, TNFRI, TRAILR1, and TRAILR2. 2) TRAF-interacting receptors - TNFRII, LTbR, GITR, OX40, 41BB, CD30, CD40, HVEM, RANK, CD27, and BAFFR. 3) Decoy receptors lacking the cytosolic domain- TRAILR3, TRAILR4, DcR3[114]. High-affinity TNFSF ligand binding to their cognate receptors induces receptor clustering in target cells leading to the recruitment of adaptor proteins and resulting in the initiation of intracellular signal transduction[115]. Depending upon the signaling context, TNFRSF signaling can induce cell death, differentiation, survival or proliferation[20]. However, factors determining the outcome of TNFRSF signaling are not fully known and thus targeting TNFRSF signaling for the desired outcome remains challenging.

Unlike IgSF which constitute both co-inhibitory and co-stimulatory molecules, TNFRSF predominantly consists of co-stimulatory molecules. TNF signaling has been implicated in the development and functions of many cell types. TNFRI and II bind to TNF-α and lymphotoxin-α (LTα)[116]. Binding of TNF-α to TNFRI can promote cell death through activation of the death domain-containing adaptor proteins such as Fas-associated death domain (FADD) and TNFR1-associated death domain (TRADD). In contrast, binding of TNF-α to TNFR-II can lead to the activation of TRAF-mediated NF-kB activation resulting in cell survival and proliferation[117].

The TNFR-II expression is mainly restricted to immune cell types, neural and endothelial cells, whereas TNFR-I is widely expressed on many other cell types. TNFRII is expressed on activated, but not naive T-cells and delivers a costimulatory signal for their expansion and differentiation. It is highly expressed on a subset of naive Tregs and can promote their survival and proliferation [119, 127]. GITR (TNFSRF18) is expressed at low levels on naïve T-cells, B cells, and NK-cells, whereas it is constitutively expressed at higher levels on Tregs[119]. Activated T-cells also express higher levels of GITR[128]. Its cognate ligand GITRL is expressed on APCs and medullary thymic epithelial cells [31]. GITR has potent costimulatory functions and can induce a robust proliferation of Teff cells[128] and it is essential for CD4+ and CD8+ T-cell differentiation/expansion during infections[129]. 4-1BB (TNFRSF9) is expressed at low levels on naive Tregs, higher levels on activated T-cells and Tregs, and on several endothelial and epithelial cells. Its ligand 4-1BBL is expressed on professional APCs [119]. 4-1BBL/4-1BB ligation provides a co-stimulatory signal for T-cell expansion, survival, effector functions, and differentiation into memory phenotype [130, 131]. OX40 (CD134), similar to GITR, is expressed on activated T-cells and constitutively expressed on Tregs[121]. OX40 ligand is (OX40L) transiently expressed on activated APCs and highly expressed on Innate Lymphoid Cells (ILC)-2 and ILC-3[132]. OX40 signaling can promote Treg proliferation in the absence of canonical antigen presentation through MHC class-II in an IL-2 dependent manner [133]. In the context of TCR signaling, OX40L/OX40 ligation increases the proliferation and effector functions of TCR-activated T-cells and can also contribute to memory T cell differentiation[121]. CD30 (TNFRSF8), is another co-stimulatory molecule whose expression is induced on activated CD4/CD8 T-cells and B-cells and its cognate ligand CD30L is expressed on activated CD4/CD8 T-cells, macrophages, and B-cells[134]. CD30L/CD30 interaction can modulate immune responses by regulating T-cell survival/apoptosis and proliferation[135]. CD30 signaling has been implicated in the thymic negative selection and CD30 deficient mice show impaired negative selection and suffer from autoimmunity[135]. CD27 is expressed on naïve CD4+ and CD8+ Tconv cells, a subset of Tregs, B-cells and NK cells under resting state and its expression is transiently increased upon TCR activation[136], whereas its cognate ligand CD70 is expressed on thymic epithelial cells[137] and activated DCs, T and B cells, NK cells and Tregs[136]. CD27-CD70 interaction is a strong co-stimuli inducing CD4+ and CD8+ T-cell proliferation and Teff cell differentiation[138]. HVEM is a unique TNFRSF member that can interact with TNF ligand LIGHT (Lymphotoxin-related Inducible ligand that competes with HSV Glycoprotein-D for Herpesvirus entry mediator on T-cells) to induce co-stimulatory signaling for T-cell activation [139]. It is expressed on both resting and activated T-cells, B-cells, NK cells, DCs [140]. LIGHT acts as a counter-regulator for HVEM-BTLA mediated co-inhibitory pathway[141, 142] and in addition to HVEM, LIGHT can signal through LT-bR[143] and DcR3[144] which are expressed on various non-hematopoietic cell types.

4. Regulation of Treg homeostasis by co-signaling receptors

Tregs exhibit their inhibitory functions through multiple mechanisms such as, by modulating APC functions, producing suppressive cytokines, nutrient deprivation, IL-2 exhaustion and cytolysis. Tregs constitutively express co-inhibitory receptor CTLA4 which compete with co-stimulatory receptor CD28 for binding to co-stimulatory ligands CD80/CD86 expressed on APCs. Ligation of CTLA4 to CD80/CD86 on APCs leads to trans-endocytosis of these co-stimulatory ligands resulting in the ablation of T-cell co-stimulation [46]. Tregs produce immunosuppressive cytokines like IL-10, TGF-b and IL-35 which can suppress Teff cell proliferation and pro-inflammatory cytokine production [145, 146] [79] [147]. IL-10 and IL-35 play critical roles in the prevention of experimental colitis, and Tregs deficient for IL-10 [145, 146] and IL-35 had impaired suppressive function and failed to protect against colitis and IBD [79]. In addition, interaction of co-inhibitory receptor TIGIT expressed on Tregs with dendritic cells induces the production of immunosuppressive cytokines such IL-10 and TGF-b which suppress Teff cells. Tregs express higher levels of two proteins namely, CD39 and CD73, which catalyze ATP degradation to AMP, and AMP to adenosine sequentially. Adenosine in turn can down-regulate NF-KB signaling upon binding to A2A receptors expressed on Teff cells and APCs, resulting in the inhibition of effector cytokine and chemokine production [148]. In addition, constitutive expression of high affinity CD25 (IL-2Ra) by Tregs allows for higher consumption of IL-2 relative to Teff cells and results in IL-2 exhaustion and reduced proliferation and survival of Teff cells [149]. Besides, Tregs can also induce apoptosis of target cells such as APCs, CD4+/CD8+T-cells and NK cells via granzyme-A/B and perforin-mediated cytolysis [150, 151] [152].

CD4⁺ Tregs can be broadly classified into two major types CD4⁺Foxp3⁺ Tregs and CD4⁺Foxp3⁻IL-10⁺ Tr1 cells. In addition, CD8+Foxp3+ Tregs have been reported to constitute a minor proportion of Treg pool[153]. CD4⁺Foxp3⁺ Tregs can be further classified as natural Tregs (nTregs) and induced Tregs (iTregs) or adaptive Tregs, based on their site of differentiation. Regulation of Treg homeostasis by co-signaling receptors is depicted in Fig-2. The nTregs are generated in the thymus (tTregs) through a bi-phasic process. Thymocytes expressing TCRs with intermediate affinity for self-antigens differentiate into CD4⁺CD25⁺Foxp3⁻ and CD4⁺CD25⁻Foxp3^lowTreg precursors in a TCR-dependent phase [31, 154, 155]. Next, in an initial TCR-independent phase these Treg precursors gain Foxp3 expression through IL-2-dependent STAT5 activation[156]. In general, thymus-derived nTregs represent phenotypically stable lineage-specific Tregs which prevent aberrant autoimmune response against self-tissues due to their higher TCR affinity for self-antigens. Studies involving neonatal thymectomy have revealed the critical importance of tTregs in preventing autoimmunity[157]. Furthermore, lack of cognate antigen expression in the thymus leads to impaired differentiation of antigen-specific Tregs and loss of peripheral tolerance to the corresponding antigen [158]. In addition, extra-thymic de novo differentiation of iTregs occurs in the periphery from CD4⁺CD25⁻Foxp3⁻ Tconv cells through signals emanating from TCR, IL-2R and TGF-βR activation. The iTregs function mainly to tame excessive inflammatory response elicited against non-self-antigens such as food and microbial antigens[5]. Studies on iTregs have shown their importance for graft and gut tolerance [159, 160]. However, there is no convincing marker yet identified to differentiate between nTregs and iTregs although Helios/Nrp1 expression has been used to distinguish them in naïve mice[161].

The B7-CD28/CTLA-4 co-signaling pathway plays a vital role in thymic[31] and peripheral[162] Treg/Teff cell development and functions[163]. During T-cell development in the thymus, CD28 is highly expressed on CD4+CD8+ DP thymocytes and expressed at relatively low levels in CD4+ and CD8+ SP T-cells. B7.1 and B7.2 ligands are expressed at low levels in the thymic cortex and higher levels in medulla[164]. CD28^−/− mice on both C57BL6[162] and NOD background[30] had significantly reduced CD4⁺CD25⁺Foxp3⁺ Tregs. Blockade of CD28 signaling followed by adoptive transfer of Tregs led to rapid loss of transferred Tregs indicating the critical role of CD28 signaling for the survival of Tregs in the periphery [165]. However, Treg specific CD28^−/− mice had only a 25-30% reduction in their thymic Tregs and no significant reduction in the periphery, indicating a Treg extrinsic role for CD28 signaling to maintain Treg homeostasis. More importantly, CD28^−/− Tregs were functionally compromised and Treg specific CD28^−/− mice developed spontaneous skin and lung autoimmunity [166]. Collectively, these studies indicate that CD28 signaling is required for thymic Treg development and survival, and for the expansion of Tregs in the periphery, and is indispensable for their suppressive functions.

CTLA4 is constitutively expressed by Tregs and is one of the target genes of Foxp3 [167]. The autoimmune symptoms arising in both Foxp3 and CTLA4 deficient mice are similar in nature indicating the convergence of CTLA4 signaling with Treg functions[168]. Ctla4^−/− mice developed severe lymphoproliferative disease and inflammation in multiple organs including severe myocarditis and pancreatitis and died by 3-4 weeks of age [169]. Though ligation of CTLA4 on Teff cells delivers co-inhibitory signal independent of CTLA4+ Tregs, Ctla4^−/− Tregs showed impaired functional ability in suppressing wild-type Teff cells, indicating an inherent role for Ctla4 signaling in Treg functions [5, 170, 171]. In contrast to these germline CTLA4 deletion studies, recent studies using conditional deletion of CTLA4 during adulthood showed increased peripheral expansion of Tregs with intact functions. In a recent study, T-cell specificCTLA4 expression was deleted during adulthood by crossing Ctla4^fl/fl mice to Lck^Cr+e mice and the resultant Ctla4fl^/fl.Lck^cre mice were protected against experimental autoimmune encephalomyelitis (EAE) and did not develop autoimmune symptoms [172]. The increase in Treg expansion seen in CTLA4 deficient adult mice could be either due to enhanced CD28 signaling in Tregs or cell-intrinsic inhibitory effect of CTLA4 on cell cycle progression[173]. In another study, Ctla4^fl/fl mice was crossed to tamoxifen-inducible Cre recombinase Rosa26^Cre/ERT2+ mice and CTLA4 ablation was induced by tamoxifen treatment. Ablation of CTLA4 expression during adulthood increased Tregs and these mice were protected against peptide-induced EAE, but developed accelerated collagen-induced arthritis (CIA). In addition, these mice developed severe, but not fatal, autoimmunity despite increased Tregs [174]. The discrepancy between these two studies can be explained by the experimental approach used wherein the earlier study lead to complete deletion of Ctla4 gene and the later might cause a partial ablation of Ctla4 due to tamoxifen treatment. In addition, differences in MHC class II haplotype might also contribute to the differences in autoimmune susceptibility observed in the later study which had H-2^q haplotype instead H-2^b haplotype in the earlier study. Nevertheless, attenuated Treg functions and appearance of autoimmune symptoms in cancer patients treated with anti-CTLA4 blocking antibodies suggest a key role of CTLA4 signaling in human Treg functions [175]. Thus, it is pertinent that CTAL4 can contribute to Treg functions in a cell-extrinsic manner by modulating APCs while inhibiting their expansion in a cell-intrinsic manner.

PD-1 co-inhibitory signaling can regulate T-cell repertoire and peripheral tolerance mechanisms to control autoimmunity. The PD1 expression has been noted in CD4⁻CD8⁻ DN thymocytes, whereas PD-L1 and PD-L2 are widely expressed in the thymic cortex and medulla respectively [176]. Tregs express both PD1 and PD-L1 under resting as well as activated conditions[58]. Unlike CTLA4 which inhibits Treg expansion, PD-L1-Ig augments iTreg generation from naïve T-cells and enhances Foxp3 expression and suppressive functions [177]. Moreover, it has been demonstrated that PD-1 expression in Tregs is indispensable for their suppressive functions, and loss of PD1 expression in Tregs accelerates the generation of exFoxp3 Tregs (plastic phenotype of Tregs which lose Foxp3 expression and produce pro-inflammatory cytokines) and thereby flare autoimmunity[178]. In contrast, to previous reports of an obligatory role for PD-1/PD-L1 signaling on iTreg generation, sustained Foxp3 expression and Treg lineage stability[177], Wong et al. reported an inverse correlation between PD1 expression and Treg functions, and blockade of PD1 signaling in NZB/NZW.F1 mice resulted in enhanced CD4+PD1^lowFoxp3^hi and CD8+PD1^lowFoxp3^hi Treg survival/functions[179, 180]. This apparent discrepancy between the two studies may be attributable to either difference in mouse strains used; especially NZB/NZB.F1 mice which are prone to autoimmune disease.

LAG3 is preferentially over-expressed on natural Tregs and induced Tregs, and its expression is induced on activated CD4/CD8 cells and NK cells. LAG3 is highly expressed in type-1 regulatory (Tr1) cells and co-expression of CD49b along with LAG3 is used as markers to characterize Tr1 cells [108]. Moreover, CD4⁺CD25⁺ Tregs from LAG3^−/− mice had a reduced suppressive capacity indicating its involvement in Treg functions [181]. TIM3 is expressed on a subset of Tregs and TIM3⁺Foxp3⁺ Tregs were superior in IL-10 production than TIM3⁻Foxp3⁺ Tregs and were potent suppressors of IFN-γ and TNF-α production [182]. Similarly, TIGIT expression is confined to a subset of Tregs and TIGIT⁺Foxp3⁺ Tregs show increased demethylation of the Treg-specific demethylated region in Foxp3 gene loci compared to TIGIT⁻Foxp3⁺Tregs, and express higher levels of nTreg signature genes such as Ctla4, Cd25, and Foxp3, and possess superior suppressive capacity compared to TIGIT⁻ Tregs[183]. This could be explained by the fact that TIGIT is a target gene of Foxp3 [184] and perhaps, TIGIT signaling enhances Foxp3 expression and Foxp3-dependent Treg signature genes through a positive feedback mechanism.

Unlike CD28 co-stimulation which positively regulates both nTreg and iTreg differentiation, TNFRSF co-stimulatory molecules play a dual role in Treg differentiation (Fig-2). TNFRSF members like TNFRII, GITR and OX40 play a secondary role to CD28 signaling in facilitating thymic Treg differentiation. Mahmud et al. found that CD28^−/− mice had reduced TNFR-II⁺/GITR⁺/OX40⁺ Tregs in their thymus. Additionally, co-stimulation through TNF-RII, GITR, and OX40, when coupled to TCR signal, increased thymic Treg precursor differentiation and also synergized with IL-2-induced STAT5 signaling to increase thymic Treg maturation. Furthermore, combined inhibition of these signaling pathways significantly impaired Treg generation in the thymus [31]. OX40^−/− mice had reduced Tregs in the thymus, but not in the periphery. More recently, we have shown that soluble OX40L-Fc can induce TCR-independent proliferation of murine and human thymic Treg precursors and matured Tregs in an IL-2 dependent manner [185, 186]. Like OX40 other TNFRSF members like GITR [187], TNFRII [187, 188] and 4-1BB [189] can also drive the proliferation of Tregs.

In contrast to its critical role in thymic Treg differentiation, OX40 signaling has been shown to inhibit iTreg differentiation in the periphery by antagonizing TGF-β signaling[190]. We have observed a similar negative effect of other TNFSF ligands such as 4-1BBL, GITRL, and TNF-α (Unpublished observations BP) on iTreg differentiation. Given the significant overlap in signaling cascades activated by these TNFRSF members and their functional outcomes, it is not surprising to observe similar effects on iTreg induction. Though TNFRSF co-stimulation can inhibit Foxp3 induction in iTregs, there is no clear evidence on the molecular mechanism of direct inhibition of Foxp3 transcription or RNA/protein stability in nTregs. It is plausible that the TNFRSF co-stimulation of Tconv cells with TCR stimulation might produce proinflammatory cytokines like IL-6 which can antagonize iTreg differentiation; instead promote Th17 cell differentiation in combination with TGF-β [191, 192]. In addition, OX40, GITR, 4-1BB, and TNFRII are capable of activating PI3K-AKT signaling[193] which can suppress Foxp3 induction by down-modulating transcription factors driving Foxp3 expression such as Foxo1 and Foxo3a [194]. However, once iTreg differentiation occurs, TNFRSF signaling might promote their proliferation in synergy with IL-2 signaling [195, 196]. Another possible mechanism by which OX40 signaling inhibits Foxp3 expression could be the induction of IL-2 exhaustion by proliferating Tregs expanded by OX40 signaling. Such IL-2 deprivation might attenuate STAT5 activation required for Foxp3 transcription in Tregs. In line with this notion, a recent study demonstrated that OX40 signaling-induced negative effects on Foxp3 expression were due to exhaustion of IL-2 and addition of exogenous IL-2 drove Treg proliferation with sustained Foxp3 expression [197]. Several studies have shown that Tregs expanded by TNFRSF signaling have compromised suppressive functions[187, 198-200] and it is claimed that activation of these co-signaling receptors in Teff cells rendered them resistant to Treg mediated suppression[201, 202]. Conversely, other studies have shown that Tregs expanded using 4-1BBL, TNF-α, OX40L, and GITRL were functionally competent [185, 187-189]. However, there is no mechanistic proof available as to how TNFR signaling causes loss of suppressive functions in expanded Tregs or impart Treg unresponsiveness to Teff cells. Since Foxp3 is a transcriptional repressor for TCR-induced expression of inflammation-associated genes like Ifn-g, Il-2, and Zap70 in Tregs[167], it can be postulated that reduced Foxp3 expression due to IL-2 exhaustion by expanded Tregs may compromise its ability to repress inflammatory genes leading to a labile phenotype and attenuated suppressive functions. Altogether, the effect of TNFR signaling on Treg differentiation, expansion and functions is clearly dependent on the activation state of Tregs and the local cytokine milieu. In contrast, CD28 signaling is essential for both thymic and peripheral Treg differentiation and proliferation, and CTAL4/PD-1 signaling are indispensable for Treg functions, but not expansion.

5. Role of IgSF co-signaling pathways in autoimmunity

6. Role of TNFSF signaling in autoimmune diseases

7. Role of Tregs and co-inhibitory receptors in tumor immune evasion, and checkpoint blockade in tumor therapy

A growing body of evidence suggests that immune suppressive functions of Tregs favor tumor progression by dampening anti-tumor immunity[311]. Increased frequencies of Tregs were observed in the peripheral blood and tumor tissues in lung[312], head and neck[313], pancreatic[314], gastric and esophageal[315], liver[316], breast[317] cancers and melanoma[318]. Though few studies have indicated that infiltration of Tregs is associated with better prognosis in colorectal carcinoma (CRC)[319], recent reports have clarified the issue by noting accumulation of FOXP3^hi suppressive Tregs in case of poor prognosis and FOXP3^low non-suppressive Tregs in cases with better prognosis[320]. Expression of co-inhibitory molecules such as CTLA4, PD-1, TIGIT, LAG3 and TIM3 by cancer tissues, tumor infiltrating T-cells and Tregs was identified as a key mechanism promoting tumor immune evasion[321]. Since Tregs preferentially overexpress these co-inhibitory receptors, they represent a putative target for immune checkpoint inhibitors that can reverse immune suppression. The effect of different checkpoint inhibitors on Teff/Treg balance and tumor progression is summarized in Table-5. Iplimumab (anti-CTLA4 Ab) was approved by Food and Drug Administration (FDA) in 2011 for the treatment of unresectable metastatic melanoma[322]. Pembrolizumab (anti-PD-1 Ab) was approved for treating advanced or unresectable melanoma in 2014 and is currently used for the treatment of metastatic Non-Small Cell Lung Cancer (NSCLC) with PDL1 expression and recurrent Squamous Cell Carcinoma of the Head and Neck (SCCHN). Nivolumab, another PD1 blocking antibody was approved in 2014 for treating unresectable and metastatic melanoma with progression even after ipilimumab treatment, NSCLC and metastatic renal Cell Carcinoma (RCC). Atezolizumab (anti-PD-L1) antibody was approved in 2015 for treating NSCLC, and in 2016 for treating Urothelial carcinoma [323]. In addition, combination of ipilimumab and nivolumab was approved in 2015 for treating unresectable metastatic melanoma[322]. Studies using IgG2a isotype anti-CTLA4 antibodies found that tumor-infiltrating Tregs were selectively depleted through an Fc-dependent mechanism involving Fcγ-receptor expressing macrophages within TME, leading to increased CD8+ Teff/Treg ratio at tumor sites with a concomitant increase in peripheral blood Tregs [324, 325]. PD1/PD-L1 pathway has been found to be critical for Treg generation[326] and thus contribute significantly to tumor immune evasion[327]. Stage III/IV melanoma patients who were vaccinated with a cocktail of peptides including the glycoprotein 100 (gp100) 209-217 (210M) (gp100-2M) and MART-1 26-35 (27L) (MART 27L) heteroclitic peptide analogs showed increased PD-L1+ Tregs and PD-1+ CD8+ T-cells in their circulation. Interestingly, blockade of PD1 signaling in these patients reduced Foxp3 expression in Tregs and helped overcome Treg mediated inhibition of CD8+T-cell proliferation and effector cytokine production [328]. Moreover, combined blockade of CTLA4 and PD1/PD-L1 along with Flt3 vaccine increased Teff/Treg ratio in B16 melanoma tumors and doubled the tumor rejection efficiency of monotherapies[329], [330].

Though CTLA4 and PD-1 blockade has improved therapeutic efficacy in several types of solid tumors and melanoma, still many patients are refractory to this approach. This lead to the search for other co-inhibitory molecules which might compensate for the loss of CTLA4/PD1 signaling and thus might favor tumor immune evasion. Upregulation of another co-inhibitory molecule TIM3 was noted in murine lung adenocarcinoma treated with anti-PD1 antibody[343]. Lung cancer patients had increased TIM3 expression in tumor-infiltrating CD4⁺ and CD8⁺ T-cells, but not in peripheral blood. The frequency of IFN-γ+ T-cells was reduced within CD8⁺TIM3⁺ cells compared to CD8⁺TIM3⁻ cells. More importantly, 70% of tumor- infiltrating CD4⁺TIM3⁺ cells were found to be FOXP3⁺ Tregs and correlated with poor prognosis[344]. In preclinical tumor models, TIM3 was expressed at low levels on peripheral blood Tregs but highly expressed on tumor-infiltrating Tregs[345] and were predominantly PD1⁺ as well. TIM3⁺ Tregs accumulated in the tumor before exhaustion of CD8⁺ T-cells and co-blockade of PD1 and TIM3 yielded improved efficacy than PD1 blockade alone[99]. Currently, humanized anti-TIM3 antibody (TSR-022) is being tested in patients with advanced solid tumors[321].

LAG3 is expressed on tumor infiltrating lymphocytes (TILs), especially on Tregs. CD4⁺CD25^hiFoxp3⁺LAG3⁺ Tregs from PBMCs, tumor invaded lymph nodes and visceral metastatic tissues of melanoma and CRC patients showed preferential proliferation and produced higher amounts of immunosuppressive cytokines such as IL-10 and TGF-β, and also exerted potent suppressive activities in contact-dependent assays[346]. More interestingly, Scurr et al. have characterized CD4⁺CD25⁺Foxp3⁻ LAG3⁺LAP⁺ intra-tumor T-cell population from CRC patients which can produce IL-10 and TGF-β and show 50% more suppressive activity than Foxp3⁺Tregs[347]. Combined inhibition of LAG3 and PD-1 using blocking antibodies resulted in potent antitumor response accompanied by reduced tumor infiltration of CD4⁺CD25⁺Foxp3⁺ Tregs and increased CD8⁺ Teff cell functions[335]. Humanized LAG-3 (BMS-986016) antibody is under phase I and phase II clinical trials in patients with colorectal, cervical, ovarian, renal cell carcinoma, head and neck, and squamous small cell carcinoma, gastric and hepatocellular cancers. In addition, anti-LAG3 antibody (MK-4280) is currently being tested as either monotherapy or in combination with an anti-PD1 antibody against advanced solid tumors[321].

TIGIT is co-expressed with PD1 on tumor infiltrating CD8⁺ T-cells and Tregs[336]. Emerging preclinical studies shed light on the inhibitory role of TIGIT on anti-tumor immunity [336, 348]. TIGIT signaling has been shown to predominantly regulate anti-tumor immune responses via Treg mediated suppression. Intra-tumoral TIGIT⁺ Tregs were found to be highly suppressive and they expressed other co-inhibitory molecules such as Pdcd-1, Lag3, Ctla4, and TIM-3, and suppressor cytokines IL-10 and TGF-β compared to TIGIT⁻ Tregs. Adoptive transfer of CD8⁺ T cells from WT or Tigit^−/− mice mixed with WT CD4⁺Foxp3⁺ and Foxp3⁻ T cells into Rag2^−/− mice implanted with B16F10 tumor cells showed a dominant role of TIGIT⁺Foxp3⁺ Tregs than TIGIT⁺CD8⁺ T-cells in tumor evasion. It was also noted that TIGIT synergized with TIM-3 in restraining anti-tumor responses [348]. In contrast, another study observed no significant difference in the frequencies of tumor-infiltrating Tregs upon TIGIT and PD-L1 blockade despite increased CD8⁺ Teff functions and tumor regression. TIGIT expression correlated with CD8⁺ T-cell exhaustion and TIGIT blockade increased effector cytokine (IFN-γ and TNF-α) production from CD8⁺ T-cells in a Treg-independent manner [336]. This discrepancy could be attributed to the differences in the mouse models and experimental approaches used in these two independent studies. However, combined inhibition of TIGIT and PD1 improved tumor antigen-specific CD8⁺ T-cell response in melanoma patients [349]. Thus, TIGIT blockade may promote anti-tumor immunity through both Treg dependent and independent mechanisms. Currently, anti-TIGIT antibodies BMS-986207 and MTIG7192A are in clinical trials against advanced solid tumors and advanced metastatic tumors either alone in combination with anti-PD1 and anti-PD-L1 antibodies[321].

8. TNFSF members in tumor immunotherapy

In addition to overcoming T-cell suppression by targeting co-inhibitory receptors of IgSF members, activation of Teff cell through by co-stimulation through TNFRs, with concomitant suppression of Tregs, has also been beneficial in tumor immunotherapy[29]. Many TNFRSF agonists were tested in the past for their ability to induce anti-tumor immunity owing to the crucial role in T-cell activation and proliferation. Here, we restrict our focus to TNFRII, GITR, OX40, and 4-1BB which could target both Teff and Treg cells. TNFRII was found to be preferentially overexpressed on several tumor tissues and tumor-infiltrating Tregs which were highly suppressive[350]. In contrast to TNF-α signaling through TNFRI which causes cell death, alternative signaling through TNFRII can induce cell proliferation [119]. Therefore, TNFR-II agonism was shown to be associated with the proliferation of tumor cells and tumor-infiltrating Tregs, which synergistically promoted tumor progression[350]. Therefore, the selective targeting of TNFRII using TNFRII antagonist is proposed to be beneficial in cancer immunotherapy[350]. In a recent study, the anti-TNFR2 antagonistic antibody has been shown to induce cell death in OVCAR3 cancer cells which overexpress TNFR2. Moreover, TNFR2 antagonists also promoted cell death of Tregs isolated from ovarian cancer ascites more effectively than Tregs from healthy donor samples [351]. In addition, TNFR2 blocking antibody in combination with TLR9 agonist CpG ODN reduced tumor-infiltrating Tregs and increased IFN-γ+CD8+T-cell infiltration leading to tumor regression and tumor-free survival[352].

4-1BB co-stimulation can promote cytotoxic CD8+ T-cell survival/activity and protect tumor infiltrating T-cells from activation-induced cell death [353]. Consistent with this, 4-1BB agonistic antibody increased cytotoxic activity of TILs and showed anti-tumor activity in preclinical models (Table-5). However, in human clinical trials, 4-1BB agonists produced significant liver toxicity and the trial was halted although it showed some anti-tumor effects[354]. Currently, lower doses of 4-1BB agonists are being tested in combination with other agents such as anti-LAG3 antibodies[355]. On the other hand, 4-1BB intracellular domain is used in engineering CAR-T-cells with B-cell antigen CD19 and CD3-domains to enhance T-cell survival and effector functions, and CART19 cells have shown disease remission for more than 10 months in a chronic lymphocytic leukemia patient[356].

OX40 is highly expressed on TILs, especially on Tregs, in many cancers including melanoma, colon cancer, head and neck cancer, breast cancer and B-cell lymphoma[357] and used as a marker for tumor antigen-specific Tregs[358]. The agonistic anti-OX40 antibody has been shown to induce an anti-tumor immune response in several preclinical tumor models including B16 melanoma[359], fibrosarcoma[360], CT26 colon cancer[361] and GL261 glioma[362]. Consistent with the effect of OX40 signaling in memory T-cell formation, OX40 mediated antitumor response was driven through CD4⁺ and CD8⁺ memory T-cell responses, and OX40 agonist treated mice were found resistant to tumor re-challenge[363]. In addition, the agonistic OX40 antibody has also been shown to deplete intra-tumoral Tregs which express higher levels of OX40 through FcγR mediated ADCC caused by myeloid and NK cells present within the TME [340] which resulted in altered Teff/Treg ratio in tumors. However, in clinical trials murine anti-human OX40 agonistic antibody showed increased tumor infiltration of Tregs despite increased Teff cell proliferation [364] which could be due to Treg proliferation induced by constitutively expressed OX40 upon OX40L binding. Currently, humanized anti-OX40 agonistic antibody is in a phase II clinical trial in combination with stereotactic radiation and/or cyclophosphamide for many cancer types[29].

Similar to an OX40 agonistic antibody, GITR agonist DTA-1 has been shown to promote anti-tumor responses in several syngeneic murine tumor models including B16 melanoma [365], CT26 colon cancer [366], and fibrosarcoma [367]. Moreover, combination therapy with GITR+OX40 agonists further enhanced anti-tumor response [366]. GITR agonist increased infiltration of large numbers of IFN-γ secreting CD4⁺ and CD8⁺ T-cells into regressing tumors. Similar to OX40 agonist, GITR agonist also caused increased infiltration of Foxp3+ Tregs into tumors [367] and induced Treg depletion through FcγR mediated ADCC [368]. However, in contrast to co-stimulatory functions of murine GITRL on T-cells, human GITRL expression undermines NK cell immune-surveillance of acute myeloid leukemia. Several human AML cell lines and AML patients were found to express GITRL and GITRL signaling induced the production of immunosuppressive IL-10[369]. Thus, OX40 and GITR agonists are dependent on FcγR expressing cell types in the TME for the ADCC. However, in an immunosuppressive tumor micro-environment lacking these FcγR expressing cell types, OX40 and GITR agonists may induce Treg proliferation as evidenced by increased Tregs seen in the tumor sites in some clinical trials [364, 367].

9. Immune-related adverse events (IRAE) resulting from checkpoint blockade

Although immune checkpoint blockade using anti-CTLA4/anti-PD1/anti-PD-L1 can produce long-lasting anti-tumor responses, it can also lead to many IRAEs and most of them are autoimmune disorders due to the critical role played by these co-inhibitory signaling molecules on immune tolerance[370]. These autoimmune adverse events span a wide spectrum of immune dysfunctions including vitiligo, mucosal inflammation, colitis, endocrine disorders such as hypothyroidism, hypophysitis (reduced levels of pituitary hormones), diabetes mellitus, liver and kidney dysfunctions, polyarthritis, pancreatitis and hematologic syndromes [371]. Almost 60% of patients receiving anti-CTLA4 therapy develop autoimmune adverse events with colitis being the most common [372]. Thyroiditis and pneumonitis[373] are more common in anti-PD1/PDL-1 treated patients[374]. There are reports suggesting that patients who are more responsive to checkpoint blockade suffer from severe autoimmune adverse events compared to unresponsive patients, indicating the efficacy-toxicity relationship [375, 376]. However, other studies have failed to show a correlation between IRAE and efficacy[377]. Non-specific immunomodulation through checkpoint blockade can shift the balance of Teff and Treg cells, and perturb immune homeostasis resulting in autoimmune adverse events. Anti-CTLA4 antibodies can promote antitumor immunity by overcoming the intrinsic suppressive effects of CTLA4 signaling on Teff cell activation and expansion, and its extrinsic effect of CD80/CD86 trans-endocytosis which is mainly mediated by Tregs. In addition, constitutive expression of CTLA4 by Tregs allows targeting these suppressor cells for ADCC mediated by Fcγ receptor expressing macrophages present in the TME [324]. Thus, anti-CTLA4 blockade can abolish Treg numbers/functions. By adoptively transferring OVA- specific CTLA4^−/− Tregs and OVA-TCR transgenic T-cells into β-cell-specific OVA-peptide expressing transgenic Rag2^−/− mice, it was shown that CTLA4 expression in Tregs is necessary to confer antigen-specific tolerance [378]. Similarly, CTLA4 blockade used in cancer treatment can lead to inhibition of Tregs including antigen-specific Tregs which are essential for maintaining peripheral tolerance, resulting in autoimmune adverse events. Unlike CTLA4 which is constitutively expressed on most Tregs, the PD1 expression is confined to a subset of Tregs. Interestingly, in breast cancer patients, PD-L1 expression in tumor cells and FOXP3 expression in tumor-infiltrating Tregs have been shown to mutually upregulate each other’s expression[379]. Thus, in addition to the direct co-inhibitory effect on Teff cell activation, PD1/PD-L1 signaling can also promote tumor evasion by inducing Treg mediated immune suppression. In line with these findings, PD-1 blockade has been shown to attenuate FOXP3 expression, Treg suppressive function and induce antigen-specific CD8+ T-cell proliferation in melanoma patients [328]. Earlier, PD-1 expression in Tregs was found to be correlated with MOG-antigen specific EAE development and PD-1^−/− mice developed severe EAE even in the absence of adjuvant [237]. Besides, PD-L1 signaling has been shown to augment adaptive Treg differentiation and regulate Treg lineage stability [177]. Therefore, in addition to promoting anti-tumor immunity, PD1/PD-L1 blockade can also attenuate Treg function and cause loss of self-antigen specific tolerance resulting in autoimmune adverse events.

IRAEs are more prevalent in CTLA4 blockade than PD-1[371] which correlates with constitutive expression of CTLA4 on all Tregs compared to the PD1 expression on a subset of Tregs. Moreover, CTLA4 expression in Tregs is one of the key determinants of Treg function and their blockade might severely affect their functions than PD1 blockade could. This is consistent with severe autoimmunity and lymphoproliferative disorder observed in CTLA4^−/− mice [169] compared to moderate autoimmunity seen in PD1^−/− mice [380]. Supporting this notion, a report on thoracic melanoma patients treated with anti-CTLA4 developed severe pan-colitis with significantly reduced peripheral blood Tregs, increased IFN-γ⁺/IL-17⁺ CD4⁺ T-cells and granzyme-B⁺ CD8⁺ T-cells[381]. Preclinical mouse models also showed a correlation between IRAEs and altered Treg/Teff ratio upon checkpoint blockade[382]. Moreover, prolonged or transient depletion of Tregs in tumorbearing Foxp3-DTR mice with diphtheria toxin (DT) administration showed IRAEs similar to that of patients treated with anti-CTLA4/anti-PD1 antibodies [383]. In addition, transient depletion of Tregs in tumor-bearing mice, followed by treatment with anti-CTLA4/anti-PD1/anti-4-1BB exacerbated IRAEs and anti-4-1BB agonist was the most toxic [383]. On the other hand, targeting TIM-3, LAG3 and TIGIT pathways in combination with PD1 which are predominantly expressed on intra-tumoral Tregs, but not constitutively on Tregs, may have reduced autoimmune adverse events compared to CTLA-4 blockade [371, 383, 384]. However, their efficacy and safety are yet to be determined in larger clinical trials. Though 4-1BB, OX40, and GITR agonists promote effective anti-tumor responses, they have been implicated in the exacerbation of autoimmune diseases in mice that are pre-disposed to spontaneous autoimmunity and experimentally induced autoimmune disease models. FcγR-mediated depletion of Tregs using anti-OX40 /GITR antibodies and activation Teff cells not only promotes anti-tumor response, but can also lead to loss of peripheral tolerance to tissue antigens. The effect of checkpoint inhibitors and anti-OX40/GITR agonists on Teff/Treg compartments to modulate antitumor immunity and consequential autoimmune adverse events are summarized in Fig-3. Earlier, the 4-1BB clinical trial was halted for safety concern and checkpoint blockade with anti-CTLA4 and anti-PD1/PDL1 has already been demonstrated to induce fatal autoimmune diseases in some patients [33]. Therefore, further investigations are needed to fully understand the potential benefits and undesirable side effects associated with the combination of anti-CTLA4/anti-PD-1/anti-PD-L1 with anti-GITR, anti-OX40 and anti-4-1BB antibodies for tumor immunotherapy. Moreover, relevant clinical studies using a larger cohort are required to draw definite conclusions on how Tregs are related to autoimmune complications arising from checkpoint blockade. Development of relevant animal models of IRAE might help to study the possible consequences of checkpoint blockade on immune homeostasis and identify the molecular leads that can be targeted to fine tune the balance to restore immune homeostasis while minimizing autoimmune adverse events.

10. Conclusion

CTLA4 is constitutively expressed on Tregs and activated Teff cells and is indispensable for maintaining immune tolerance. Thus, blockade of this pathway for cancer treatment might not only overcome tumor-specific immune suppression but could also break physiological self-tolerance to cause autoimmunity. On the other hand, PD-1//PD-L1 are not constitutively expressed on Tregs, however, they play an essential role in tissue tolerance as noted by their expression on many non-immune cell types and the development of spontaneous autoimmunity in PD-1^−/− mice[58]. Second line co-inhibitory receptors such as LAG3, TIM3, and TIGIT seem to offer an advantage as they are preferentially overexpressed on tumor-infiltrating Tregs and mice deficient in these co-inhibitory molecules do not develop spontaneous autoimmunity. However, more extensive studies are needed to determine their safety and efficacy. Though TNFSF members are potent stimulators of Teff cell response, their context-dependent dual role on Tregs should also be considered for effective targeting. Despite playing a secondary role to CD28 and CD40 signaling in the induction of autoimmunity, activation of 4-1BB, GITR, and OX40 signaling can aggravate ongoing autoimmune response. Therefore, pre-existence of IRAEs should be tested before using combinations of OX40, GITR and 4-1BB agonists with anti-CTLA4/PD1 blockade which might worsen existing autoimmune reactions. Moreover, murine and human immune systems exhibit several striking differences [186, 385] and relevant humanized mouse models must be developed to study human tumors and to ensure successful clinical translation. Currently, more suitable animal models are being developed to study the mechanism of pathogenesis of autoimmune adverse events arising from checkpoint blockade[382]. These animal models may help understand the molecular mechanisms affecting immune homeostasis upon ICB that can be potentially corrected to prevent IRAEs.