TGF-β Superfamily Co-receptors in Cancer

Abstract

Transforming growth factor-β (TGF-β) superfamily signaling via their cognate receptors is frequently modified by TGF-β superfamily co-receptors. Signaling through SMAD-mediated pathways may be enhanced or depressed depending on the specific co-receptor and cell context. This dynamic effect on signaling is further modified by the release of many of the co-receptors from the membrane to generate soluble forms that are often antagonistic to the membrane-bound receptors. The co-receptors discussed here include TβRIII (betaglycan), endoglin, BAMBI, CD109, SCUBE proteins, neuropilins, Cripto-1, MuSK, and RGMs. Dysregulation of these co-receptors can lead to altered TGF-β superfamily signaling that contributes to the pathophysiology of many cancers through regulation of growth, metastatic potential, and the tumor microenvironment. Here we describe the role of several TGF-β superfamily co-receptors on TGF-β superfamily signaling and the impact on cellular and physiological functions with a particular focus on cancer, including a discussion on recent pharmacological advances and potential clinical applications targeting these co-receptors.

Introduction

The transforming growth factor-β (TGF-β) superfamily of ligands, which include TGF-βs, bone morphogenetic proteins (BMPs), growth and differentiation factors (GDFs), Activins, and Inhibins are well known for their roles in embryonic development and patterning, and also have essential homeostatic roles in adult tissue. TGF-β superfamily ligands regulate cellular function through seven type I TGF-β superfamily receptors (ALK (Activin like kinase)-1-7) and five type II TGF-β superfamily receptors (TβRII, ActRII, ActRIIB, AMHRII, and BMPRII). Type I (RI) and type II (RII) TGF-β superfamily receptors are serine/threonine kinases that form a complex together, and upon ligand binding the type II receptor phosphorylates the type I receptor to induce downstream signaling, including SMAD pathways. SMADs are key mediators of the TGF-β superfamily intracellular signaling response, which accumulate in the nucleus to facilitate transcriptional programing. TGF-β superfamily signaling may occur via a SMAD2/3 or a SMAD1/5/8 signaling pathway, both which are shuttled into the nucleus by SMAD4. The pathways activated are dependent on the receptors involved, with type I BMP receptors (ALK-1, 2, 3, 6) typically signaling through SMAD1/5/8 and type I Activin receptors (ALK-4, 7) and TGF-β receptors (ALK-5) typically signaling through SMAD2/3.¹ Additionally, as a negative feedback mechanism, SMAD6 and SMAD7 function as TGF-β superfamily signaling inhibitors. These pathways can also signal through non-SMAD pathways, including ERK1/2, JNK, p38, PI3K/Akt and Cdc42, with known crosstalk in many physiological and disease contexts.²

In addition to the vast array of possible ligand/receptor combinations, the specificity, potency, and diversity of TGF-β superfamily signaling is also mediated by cell surface co-receptors. Unlike type I and type II TGF-β superfamily receptors, co-receptors do not possess functional kinase domains and instead have short cytoplasmic domains or are GPI-anchored to the surface with no intracellular domain. These co-receptors bind to TGF-β superfamily ligands and often facilitate TGF-β superfamily signaling by promoting ligand interaction with the RII/I complex. Co-receptors are typically more abundant than type I and type II TGF-β superfamily receptors causing most cell surface ligand interactions to be initiated via co-receptors. However, depending on the co-receptor, this may enhance, alter, or antagonize that particular ligand-dependent signaling, sometimes in a cell-specific context. Co-receptors may also interact with type I and type II TGF-β superfamily receptors to influence recruitment of receptor complex components, alter their stability, and influence intracellular trafficking patterns, in addition to facilitating signaling responses. The extracellular domains of many of these co-receptors are also prone to enzymatically shedding from the cell surface, or have an alternately spliced soluble form, which allows them to travel into the extracellular space, often functioning to antagonize ligand interaction with cell surface receptors via ligand trap. However, many of the co-receptors are not exclusive to TGF-β superfamily ligands and can refine their signaling responses through other receptors via ligand-dependent and -independent signaling mechanisms, highlighting the diversity of responses that can be achieved through differential expression of these co-receptors. Notably, the dysregulation of TGF-β superfamily signaling, including through co-receptors, has been implicated in the pathophysiology of a wide variety of cancers. For an excellent review on TGF-β superfamily co-receptor structure, function, and signaling, we refer the reader to a recent review by Nickel et al.³ In this review we highlight the role TGF-β superfamily co-receptors play in the dysregulation of TGF-β superfamily signaling in cancer and discuss recent advances in targeting these co-receptors for clinical applications.

TGF-β superfamily signaling and co-receptors in cancer

TGF-β first suppresses and then promotes cancer progression through mechanisms yet to be fully defined. TGF-β initially suppresses cancer initiation through growth inhibition while promoting apoptosis, but the cancer cells that emerge have often developed resistance to these tumor suppressor effects. Mechanisms for this resistance include mutations, deletions, or epigenetic regulation of components of the TGF-β pathway, including receptors, ligands, and downstream intermediaries. Initiated cancer cells then often overexpress TGF-β to promote tumor progression and metastasis in both an autocrine manner, enhancing epithelial to mesenchymal transition (EMT), migration and invasion, and in a paracrine manner, promoting angiogenesis, matrix remodeling and inhibiting immunosurveillance in the tumor microenvironment (TME). The immunosuppressive impact of TGF-β in the TME is often a barrier to producing effective cancer immunotherapies by suppressing cytotoxic CD8+ T-cells, and by promoting the generation of CD4+ CD25+ regulatory T cells and pro-tumoral macrophages and neutrophils.⁴

TGF-β superfamily co-receptors are prone to dysregulation in cancer cells, often in an organ/tissue/cancer-specific manner. Loss of expression or overexpression of TGF-β superfamily co-receptors is a frequent occurrence in cancer that leads to increased TGF-β superfamily signaling, with the expression change being contextually dependent on whether the receptor primarily functions as an agonist or antagonist of TGF-β superfamily signaling. Additionally, non-TGF-β superfamily signaling pathways influenced by these co-receptors often become more pro-tumorigenic as a result of the co-receptor expression change. The full range of cancer phenotypes caused by the dysregulation of co-receptors is often reflected in the phenotypes exhibited by both TGF-β superfamily signaling and non-TGF-β superfamily signaling pathways. Often the release of soluble forms of these co-receptors is modulated by cancer cells and can serve as potential serum biomarkers of cancer progression. Additionally, some of TGF-β superfamily co-receptors have expression patterns during cancer that make them valuable for the detection and visualization of tumors. Targeting TGF-β superfamily signaling in cancer has been the subject of much interest over the past two decades with mixed pharmacological success hampered by the potential for on-target toxicity.⁵ However, targeting TGF-β superfamily co-receptors has garnered interest given their relative ligand/receptor specificity and contextual tissue expression and therefore potential for improved context-dependent pharmacological TGF-β superfamily signaling modulation. More recently, there has been some success in targeting TGF-β superfamily co-receptors to mitigate cancer progression, as seen with the TRC105 antibody targeting Endoglin, and as a guidance molecule for drug delivery, as seen with nanoparticle delivery targeting the cancer vasculature via neuropilin.^6,7

Here we describe several TGF-β superfamily co-receptors, defined by their ability to modulate activation of type I and type II TGF-β superfamily receptors and subsequent downstream SMAD signaling, in the absence of their own kinase activity. While extracellular components that activate TGF-β ligands from their inactivate state may also be considered TGF-β co-receptors, such as integrins and the extracellular matrix, these are not included in the current review (See Nickel, ten Dijke, and Mueller review).³

TGF-β superfamily co-receptors and their role in cancer

TβRIII (Betaglycan)

The type III TGF-β receptor (TβRIII), also known as betaglycan, is a membrane proteoglycan that is the most highly expressed TGF-β receptor in many cell types. Human TβRIII is an 851-amino acid receptor with a single-pass hydrophobic transmembrane region, a short 42-amino acid cytoplasmic domain, and a large 766-amino acid extracellular domain that is glycosaminoglycan (GAG) modified with chondroitin and heparan sulfate chains.⁸ TβRIII is able to bind with high affinity to inhibin, BMPs, and all three isoforms of TGF-β, and can also bind to basic fibroblast growth factor through heparan sulfate glycosaminoglycan chains. The extracellular domain is also composed of two subdomains that can both bind to TGF-β, but only the proximal zona pellucida domain can bind to inhibins. Recent work has identified three key residues specific to TGF-β2 and Inhibin-alpha that help facilitate binding to the proximal zona pellucida domain.⁹ However, only the distal orphan domain is sufficient to promote TGF-β2 complex formation with TβRII.¹⁰

TβRIII does not itself have a functional kinase domain. Instead, TβRIII associates with TβRII and presents TGF-β ligand isoforms to TβRI to enhance SMAD2/3 signaling (Fig. 1). This is particularly important for TGF-β2 because alone it binds 200–300 fold more weakly to TβRII than TGF-β1 and TGF-β3, whereas all three isoforms bind with similar affinity to TβRIII.¹¹ Indeed, cells lacking TβRIII require at least 2 orders of magnitude higher dose of TGF-β2 to achieve the same response.¹² However, TβRIII can also have an opposing effect on SMAD2/3 signaling at the cell surface via inhibin. TβRIII is a required co-receptor for inhibin, a known activin agonist, and while activin can initiate SMAD2/3 signaling, inhibin bound TβRIII competes with activin by sequestering the activin type II receptor, preventing the recruitment and activation the type I receptor. In the absence of TβRIII, inhibins bind far more weakly than activins to activin type II receptors.¹³ Similarly, inhibin-bound TβRIII can also antagonize BMP signaling, including BMP-induced SMAD2/3 signaling.¹⁴

Figure 1. Mechanisms regulating TβRIII-mediated signaling.

TβRIII binds TGF-β and delivers it to TβRI/II receptor complex, facilitating downstream SMAD2/3 signaling. Soluble TβRIII (sTβRIII) competes with surface TβRIII for available TGF-β, diminishing the SMAD2/3 response. TβRIII facilitates inhibin binding to ACTRII which blocks interaction ACTRI and downstream SMAD2/3 signaling. β-Arrestin 2 interaction with the TβRIII intracellular domain promotes internalization and downstream cdc42 signaling, while diminishing NF-κB signaling. GIPC interaction with the PDZ stabilizes TβRIII on the cell surface.

TGF-β1 negatively regulates TβRIII at the mRNA level via effects on the proximal promotor of the TGFBR3 gene.¹⁵ GAIP-interacting protein, C terminus (GIPC) stabilizes cell surface expression of TβRIII by binding to a PDZ binding motif in the TβRIII cytoplasmic domain, thereby enhancing TGF-β signaling.¹⁶ However, the scaffolding protein β-arrestin2 inversely impacts TβRIII surface expression by promoting endocytosis by both clathrin-mediated and clathrin-independent mechanisms via interaction with the TβRIII cytoplasmic domain, resulting in reduced TGF-β signaling.¹⁷

TβRIII knockout mice are embryonic lethal as early as embryonic day 13.5, but with the highest mortality around embryonic day 16.5, due to the essential role of TβRIII in liver and heart development.¹⁸ Additionally, TβRIII knockout mice exhibit renal hypoplasia as early as embryonic day 13.5 and testis development is impacted by embryonic day 12.5.^19,20 TβRIII is also important for T-cell development in the embryonic thymus.²¹ Due to the disproportionally high binding affinity of TGF-β2 to TβRIII compared to the TβRII, it’s not surprising that the TGF-β2 knockout mouse exhibits many of the same phenotypic characteristics observed in TβRIII knockout mice, including defects in the liver and heart.²² In adult animals, TβRIII has been identified as essential for TGF-β2 signaling for the function of a variety of cell types, including in the sertoli cells of the testis and the goblet cells of the intestines.²³ Conditional deletion of TβRIII in pituitary gonadotrope cells also revealed a role for TGFβRIII in the regulation of follicle stimulating hormone synthesis via inhibin A, but not inhibin B.²⁴

TβRIII is able to undergo ectodomain shedding, releasing a soluble form of TβRIII (sTβRIII). Shedding is initiated by proteolysis of the extracellular domain at a location proximal to the transmembrane domain, releasing sTβRIII into the extracellular matrix. Shedding can be blocked with the treatment of TAPI-2, a broad matrix protease inhibitor, but the specific sheddases involved are not well characterized.²⁵ Released sTβRIII has greatly reduced ability to present ligand to membrane-bound TβRII and instead inhibits TGF-β signaling in both a cell-autonomous and non-autonomous manner via extracellular sequestration of TGF-β ligands.^25,26 This is facilitated in large part by the bilobular structure of the ectodomain, with each lobe independently binding to TGF-β at distinct interfaces. Separation of these lobes reduces their impact on TGF-β signaling by 1–2 orders of magnitude, suggesting an essential role for their structural linkage.²⁷

Following ectodomain shedding, the remaining transmembrane-cytoplasmic fragment is stable, but gamma-secretase inhibitors are able to further increase the stability, suggesting that the transmembrane-cytoplasmic fragment is a substrate for gamma-secretase.²⁸ Interestingly, following ectodomain shedding this fragment can contribute to a blunted TGF-β2 response independent of sTβRIII.²⁸ However, the TβRIII cytoplasmic domain in coordination with the p38 MAPK pathway can also contribute to ligand-independent TGF-β signaling.²⁹ Therefore, in some situations TβRIII can function as an enhancer of TGF-β signaling, whereas in other situations it can inhibit TGF-β signaling.

Cancer

TβRIII has a role in suppressing cancer progression in multiple cancer types, including breast cancer, melanoma, prostate cancer, pancreatic cancer, colon cancer, multiple myeloma, neuroblastoma, ovarian cancer, endometrial cancer, and lung cancer.^30–34 Allelic loss of TβRIII is a frequent genetic event in many of these cancers, and loss of TβRIII is associated with cancer progression and a poor prognosis.^35,36 Additionally, there are polymorphisms of the TβRIII gene (TGFBR3) associated with increased cancer risk,³⁷ and multiple microRNAs have been implicated in promoting cancer progression and metastasis via downregulation of TβRIII.^38–41 The mechanism of action is due in large part to the TβRIII ectodomain shedding from the cell surface, releasing sTβRIII, which antagonizes the tumor-promoting and immunosuppressive effects of TGF-β signaling.⁴² Reinitiating expression of TβRIII in cancer cells and tumors inhibits cancer cell migration and invasion, metastasis, and angiogenesis, while also reestablishing immunoregulation of the tumor microenvironment (TME), permitting a proapoptotic immune response.^30,34 Also, in vivo breast and colon cancer models suggest that there is potential for treating with ectopic sTβRIII to reduce metastatic potential.⁴³ Notably, since sTβRIII can be detected in human serum and plasma, there is potential for the use of sTβRIII levels as a prognostic cancer biomarker. The evidence for this is supported thus far in pre-surgery breast cancer patients, which have approximately 5-fold lower levels of sTβRIII in their plasma compared to healthy volunteers, while HER2 positive status is associated with even lower levels of plasma sTβRIII.^36,44 Additionally, lower plasma sTβRIII levels predict disease progression and a poorer prognosis in stage III melanoma patients.³⁰

Non-TGF-β signaling via TGFβRIII interactions has also been implicated in some cancer pathologies. TβRIII negatively regulates signaling of the dimeric transcription factor NFkB, and this regulation is dependent on β-arrestin2 interaction with TβRIII.⁴⁵ NFkB regulates genes involved in many cancer related mechanisms, including apoptosis, cell migration, and immune regulation.⁴⁶ Indeed, the ability of TβRIII to suppress NFkB signaling contributes to the ability of TβRIII to inhibit breast cancer cell migration.⁴⁵ Furthermore, TβRIII interaction with β-arrestin2 promotes the activation of the small GTPase Cdc42, altering the actin cytoskeleton to inhibit migration in both normal and cancer cells.⁴⁷ The TβRIII/β-arrestin2 interaction also regulates cell adhesion by promoting co-internalization with integrin α5β1 at focal adhesion sites, which traffics to the early endosome and may recycle back to the cell surface at new sites of adhesion.⁴⁸ Interestingly, TβRIII also has influence on Wnt signaling, independent of its receptor function, influenced by the makeup of its GAG chains with heparin sulfate having an inhibitory effect.⁴⁹

Endoglin

Endoglin (CD105) has a 561-amino acid extracellular domain, containing a distal orphan domain and proximal zona pellucida domain, a single pass transmembrane domain, and a 47-amino acids cytoplasmic domain. Within the zona pellucida domain, a cysteine residue allows endoglin to form as a disulfide-linked homodimer, and there are two distinct isoforms of endoglin achieved through alternative splicing, termed the long (L)- and short (S)-endoglin. The distinction in these isoforms is the length of the cytoplasmic domain, respectively at 47 and 14 residues in length, with S-endoglin lacking a PDZ-binding motif. L-endoglin is the predominant isoform, and the presumed isoform when not specified in the literature. L-endoglin promotes proliferation and migration of endothelial cells via enhanced ALK1-Smad1/5 signaling, whereas S-endoglin enhances ALK5-Smad2/3 signaling (Fig. 2).⁵⁰ Interestingly, hyperoxia in the pulmonary endothelium is reported to enhance S-endoglin while suppressing L-endoglin.⁵¹

Figure 2. Mechanisms regulating endoglin-mediated signaling.

Long endoglin (L-Eng) facilitates delivery of BMP to the ALK1/BMPRII receptor complex and TGF-β to the ALK1/TβRII receptor complex to promote downstream SMAD1/5 signaling. Short endoglin (S-Eng) facilitates delivery of TGF-β to the ALK5/TβRII receptor complex to promote downstream SMAD2/3 signaling. Soluble endoglin (sEng) competes with surface endoglin for available ligand, diminishing the response of both the SMAD2/3 and SMAD1/5 pathways. BMP-9 binding to L-Eng activates PI3K and Akt via PDZ-bound GIPC, which is suppressed by TGF-β1. β-Arrestin 2 interaction with the L-Eng intracellular domain promotes internalization and suppresses Erk activation.

Endoglin has regions of high sequence similarity to TβRIII, which are structurally similar, suggesting that they have some overlapping function. Additionally, endoglin is reportedly able to associate with TβRIII in complex with type I and type II receptors when they are endogenously co-expressed.⁵² Despite similarities to TβRIII, endoglin has distinct ligand and type I/II receptor binding interactions. The orphan domain binds highly to BMP-9 and BMP-10 in a 2:1 complex,⁵³ whereas endoglin is unable to bind TGF-β superfamily ligands on its own. Instead endoglin requires interaction with type I and type II TGF-β superfamily receptors to bind TGF-β superfamily ligands, and the specific receptors involved determine which ligands are able to bind.⁵⁴ Specifically, endoglin can bind to TGF-β1 and TGF-β3 when interacting with TβRII, but does not under any condition bind to TGF-β2.⁵⁵ Endoglin in a complex with TβRII and both ALK1 and ALK5 enhances ALK1/Smad1/5/8 signaling in response to a two-step phosphorylation of endoglin following TGF-β1 binding.⁵⁶ Endoglin also interacts with BMP-7 and activin-A via ActRII and ActRIIB and BMP-2 via ALK3 and ALK6.⁵⁴ The endoglin zona pellucida domain also contains an integrin recognizing Arginine-Glycine-Aspartic Acid motif,⁵⁷ which is involved in endothelial integrin-mediated mural cell and leukocyte adhesion, independent of endoglin’s role with TGF-β.^58,59 Additionally, endoglin mediates non-Smad signaling function by interacting with PI3K via GIPC to activate PI3K and Akt, which is enhanced by BMP9 but suppressed by TGF-β1.⁶⁰ Endoglin also interacts with β-arrestin2, which promotes its internalization and antagonizes TGF- β-mediated ERK activation.⁶¹ New work has identified potential new binding partners to soluble endoglin, and specifically validates the binding of the E3 ubiquitin-protein ligase TRIM21 and the secreted lectin Galectin-3.⁶²

Endoglin is primarily expressed in vascular endothelial cells, in contrast to TβRIII which is more ubiquitously expressed. Though, endoglin is also expressed in macrophages, where it is essential to their immune response; in placental syncytiotrophoblasts throughout pregnancy, where it regulates trophoblast differentiation and invasion; in mesenchymal stem cells, where it promotes regulation of T-cell proliferation; in mast cells, where TGF-β regulates chemotaxis and proliferation, but a functional link to endoglin remains unclear; and in myofibroblasts, where it regulates differentiation and collagen expression.^63,64

Endoglin is one of the most studied TGF-β superfamily co-receptors because of its role in angiogenesis, and is dysregulated and targetable in the vasculature of multiple diseases, including hereditary hemorrhagic telangiectasia (HHT), preeclampsia, and cancer. Genetic deletion of endoglin is embryonic lethal at embryonic day 10.5 due to a failure to form mature yolk sac blood vessels,⁶⁵ a similar phenotype to that observed in TβRII and TGF-β1 knockout mice.⁶⁶ Endoglin promotes angiogenesis by activating pathways that promote proliferation and suppress apoptosis in the endothelium.⁶⁷ Some polymorphisms of endoglin are associated with an increased risk of cardiovascular damage,⁶⁸ and human mutations in endoglin contribute to HHT, characterized by arteriovenous malformations.^69,70 The role of endoglin in this disease is thought to be restricted to its function in veins and capillaries, but not the arteries,⁷¹ and that endothelial endoglin normally prevents aberrantly high VEGF expression that helps cause these malformations and the subsequent impact on heart function.⁷²

Similar to TβRIII, the extracellular domain of endoglin is capable of undergoing proteolytic shedding from the cell surface.⁷³ In particular, endoglin is shed by MMP-14 (MT1-MMP) and by MMP-12 from macrophages.^73,74 Also like sTβRIII, soluble endoglin is able to sequester ligand in the extracellular space, preventing it from interacting with membrane-bound endoglin and inhibiting leukocyte adhesion and transmigration.⁵⁹ However, there is some evidence that in the absence of membrane-bound endoglin, soluble endoglin may be able to present ligand to the membrane-bound TβRII/TβRI complex, albeit at a lower efficiency.⁷⁵ Indeed, recent work suggests that soluble endoglin bound to BMP9 can still complex with ALK1.⁷⁶ Thereby, it is conceivable that soluble endoglin may function to some extent as a biological rheostat to normalize endoglin signaling rates. However, increased soluble endoglin in the circulation has been associated with a number of cardiovascular pathologies including preeclampsia, heart failure, myocardial infarction, hypertension, and atherosclerosis.^77–80 Whether there is a causative role for soluble endoglin on these pathologies continues to be investigated. Recent studies suggest a role for soluble endoglin in aortic endothelial dysfunction, but not in the alteration of heart structure.^81,82 Additionally, soluble endoglin induces increased arterial pressure, which may be due to a downstream increase in BMP4 expression.⁸³ High soluble endoglin has been associated with metabolic disorders, including with cholesterol and bile acids,⁸⁴ and liver diseases, including liver cystic fibrosis, cirrhosis, and hepatocellular carcinoma.^85–87

Interestingly, soluble endoglin can travel within exosomes. Endoglin has been found in exosomes released from liver cells where N-glycosylation appears to play a role in its trafficking to exosome,⁸⁸ and in preeclampsia, heightened soluble endoglin and sFlt from the placenta may be delivered via exosomes to endothelial cells to promote endothelial dysfunction.⁸⁹

Cancer

Endoglin is overexpressed in the neovasculature of many cancers and enhances tumor angiogenesis,⁹⁰ and since non-pathological angiogenesis is rare in adults with few exceptions, endoglin is considered an effective marker of the tumor vasculature with expression levels significantly higher than VEGFRs.⁹¹ This has buoyed interested in targeting endoglin for early diagnosis and antiangiogenic therapies.^92–95 Functionally, overexpression of endoglin in the tumor vasculature can lead to more aggressive cancer with poor vascular mural coverage and enhanced metastasis in the context of colon cancer, hepatocellular carcinoma, and breast cancer models.^96–98 Further, mice haploinsufficient for endoglin have reduced tumor growth and vascular density in the Lewis Lung Carcinoma lung cancer model.⁹⁹ Endoglin shedding has also been shown to suppress tumor angiogenesis. Indeed, low soluble endoglin is associated with aggressive prostate cancer, whereas elevated soluble endoglin is associated with recurrence-free survival in patients.^100,101

While endoglin is primarily expressed in the tumor vasculature, recent studies have identified a role for endoglin in cancer cells and other tumor-associated cell types.¹⁰² For example, when expressed in cancer cells, endoglin is associated with cancer progression in endometrial cancer, ovarian cancer, hepatocellular carcinoma, Ewing sarcoma, and melanoma.^103–106 Additionally, in at least one example of a non-solid tumor, endoglin is expressed on leukemic blasts, where it enhances leukemogenic activity.¹⁰⁷ However, in other cases, endoglin expression has instead been associated with cancer suppression, including in breast cancer, esophageal squamous cell carcinoma, and prostate cancer.^108–110 Thus the role of endoglin re-expression in cancer cells is context dependent. Endoglin is also expressed in cancer-associated fibroblasts in colorectal cancer where it increases their invasive capacity and helps promote metastasis.¹¹¹ In clear cell renal carcinoma cancer stem cells, endoglin is involved in maintaining stemness and promotes an EMT phenotype, though not necessarily enhanced metastasis.¹¹² Additionally, endoglin is expressed on the surface of CD4+ T-cells and promoting the regulatory phenotype and an immunosuppressive TME.¹¹³

Amongst all TGF-β superfamily co-receptors, endoglin has thus far been the most successfully targeted for clinical applications. First, two anti-endoglin antibody-drug conjugates have demonstrated promise as tumor growth suppressors for Ewing sarcoma.¹¹⁴ Also, anti-endoglin therapy in angiosarcoma has exhibited antitumor effects via regulation of non-Smad TGF-β signaling.¹¹⁵ The monoclonal chimeric IgG1 anti-CD105 antibody (TRC105) strongly binds human endoglin. The first clinical studies with TRC105 were back in 2012 and has since had mixed results.⁶ In a Phase II clinical trial in HCC, TRC105 lacked significant single-agent activity, but has since shown promise in a preliminary Phase II study in a combination therapy with Sorafenib in HCC.^116,117 Also, a combination treatment with TRC105 and anti-VEGF (bevacizumab) antibodies improved clinical outcomes,¹¹⁸ and combined endoglin and PD-1 antibody treatment promotes tumor regression through intratumoral Treg suppression and enhanced cytotoxic T-cell response.¹¹³ There is also promise in targeting endoglin for drug delivery into the tumor neovasculature, including with liposomes bispecific for endoglin and fibroblast activation protein, and also in targeting endoglin and IP-10 to deliver drugs to the tumor vasculature.^119,120

BAMBI

BAMBI (BMP and activin membrane-bound inhibitor) was first discovered in Xenopus in an expression screen looking for factors involved in BMP4 signaling.¹²¹ BAMBI is an ortholog of human NMA (Non-Metastatic Gene A protein). BAMBI is a 260 amino acids transmembrane protein with a 130 amino acid extracellular domain and a 90 amino acid intracellular domain. The extracellular domain has similar ligand binding capacity as TGF-β and BMP type I receptors, while lacking a functional intracellular serine/threonine kinase domain. BAMBI functions as a decoy receptor that antagonizes TβRI signaling, with a structure that suggests that it may also compete for ligand. However, BAMBI instead competes for receptor complex interactions with type I receptors, and even heterodimerizes with type I receptors to impact their phosphorylation capacity. Indeed, BAMBI was shown to bind to every TGF-β and BMP type I receptor except ALK2.¹²² Furthermore, the intracellular domain of BAMBI reportedly enhances signaling inhibition by translocating to the nucleus with SMAD7.¹²³

While it is not required for normal embryonic development, BAMBI is expressed during embryogenesis in a pattern that reflects BMP.¹²⁴ In particular, it is co-expressed with BMP4, regulates BMP signaling via a negative feedback loop, and TGF-β1 can promote its expression via Smad3 activation.¹²⁵ BAMBI plays an important role in regulating TGF-β signaling in the cardiovascular system,^126,127 during adipogenesis, and inflammation.^128,129 Additionally, BAMBI is altered or implicated in a variety of conditions, including liver fibrosis, COPD, prostatic hyperplasia, and diabetic kidney disease.^130–132 BAMBI has also been identified as a target for several miRNAs which play a role in a variety of conditions including miR-HCC2 in HCC, miR-519d-3p in scar formation, miR-17-5p in nasopharyngeal carcinoma, miR-708 in melanoma, miR-20b-5p in diabetic retinal vascular dysfunction, miR-942 in liver fibrosis, and miR-338-3p in atherosclerosis.^133–138

Cancer

Due to its direct antagonism of TGF-β signaling, there were early implications that BAMBI may have a role in the regulation of tumor progression. Indeed, BAMBI is downregulated in metastatic melanoma cells which inhibits the Wnt signaling pathway while activating TGF-β signaling to suppress cell proliferation and migration.¹³⁵ BAMBI can also be epigenetically silenced in bladder cancer, which likely contributes to the aggressiveness through enhanced TGF-β and BMP signaling.¹³⁹ Also, induced expression of BAMBI was able to inhibit the formation of cancer-associated fibroblasts in the TME via TGF-β signaling inhibition in an orthotopic breast cancer xenograft model.¹⁴⁰

However, there has also been plenty of evidence to suggest that BAMBI is oncogenic. For example, BAMBI promotes metastasis in colorectal cancer and gastric cancer through crosstalk between the TGF-β and Wnt/beta-catenin pathways.¹⁴¹ Though, it was also found that the synergistic TGF-β signaling inhibitory effect of BAMBI and SMAD7 may also be contributing to increase gastric cancer metastasis.¹⁴² BAMBI inhibition also suppressed colon cancer xenograft tumor growth in vivo and colon cancer cell viability in vitro.¹⁴³ Additionally, BAMBI has been implicated in modulating the Wnt-β-catenin pathway in osteosarcomas to enhance invasion and migration.¹⁴⁴ BAMBI expression is increased in ovarian cancer.¹²³ Though, while oncogenic activity was found in vitro with ovarian cancer cells, no correlative impact on patient survival was observed.¹²³ BAMBI has also been implicated in promoting proliferation and EMT in HCC.¹³³

Reports of BAMBI expression in non-small-cell lung cancer (NSCLC) are conflicting with at least two studies showing upregulation compared to adjacent normal tissue.^145,146 However, BAMBI is downregulated in NSCLC due to epigenetic silencing, with restoration of BAMBI expression causing a reduction in tumor growth and metastasis-related phenotypes via suppression of TGF-β signaling.¹⁴⁷ This is further supported by the observation that BAMBI overexpression suppresses NSCLC xenograft tumor growth and A549 cell viability in a combination treatment with β-sitosterol.¹⁴⁸

CD109

CD109 is a GPI-anchored glycoprotein and member of the thioester-containing α2-macroglobulin/C3,C4,C5 family of proteins. CD109 consists of a 155 kDa core protein that when N-glycosylated in the endoplasmic reticulum can reach up to 205 kDa, but is then cleaved by furin to generate 180 kDa and 25 kDa fragments in the Golgi before being expressed on the cell surface, enriched on lipid rafts (Fig. 3).¹⁴⁹

Figure 3. Mechanisms regulating CD109-mediated signaling.

CD109 is enriched on lipid rafts and interacts with ALK1 and ALK5 to suppress SMAD2/3 signaling and enhance SMAD1/5 signaling. CD109 inhibits TGF-β signaling through SMAD7/Smurf2-mediated internalization and degradation of TGF-β receptors. Furin cleaves CD109 in the Golgi to generate soluble CD109 (sCD109) which gets released at the surface. Exosomal CD109 and sCD109 compete with cell surface CD109 for TGF-β ligands, suppressing TGF-β signaling.

While it was not originally identified as such, the role of CD109 as a TGF-β co-receptor was first characterized in keratinocytes with an unknown molecular identity.¹⁵⁰ Subsequent work revealed the identity of this TGF-β co-receptor to be CD109.¹⁵¹ Specifically, CD109 has been shown to reduce ALK5-mediated Smad2/3 signaling. However, CD109 was also able to enhance ALK1-mediated Smad1/5 signaling, decreasing extracellular matrix production in the epidermis.¹⁵² CD109 in association with caveolin-1 inhibits TGF-β signaling by promoting the internalization and subsequent degradation of TGF-β receptors in a SMAD7/Smurf2-mediated mechanism.¹⁵³

CD109 is normally expressed in the epidermis, testis seminiferous tubules, bone osteoblasts, prostate and bronchial epithelial basal cells, thymic epithelial cells, subpopulations of endothelial cells, activated T-cells, activated platelets, and bone marrow CD34+ cells, as well as in myoepithelial cells of mammary, salivary, and lacrimal glands.¹⁵⁴ CD109 knockout mice are born normally, but exhibit epidermal abnormalities related to its function in keratinocytes and an osteoporosis-like phenotype in the bones.^155,156

Furin cleavage can release the 180kD fragment in soluble form into the extracellular matrix, leaving behind a 25kD fragment of CD109 on the cell surface.¹⁴⁹ Soluble CD109 (sCD109) binds with all three forms of TGF-β with relatively similar affinities, and is believed to inhibit TGF-β signaling by sequestering these ligands.¹⁵⁷ CD109 can also be cleaved by mesotrypsin, leading to the formation of 120 kDa and 70 kDa fragments.¹⁵⁸ Furthermore, membrane bound CD109 can be released into the extracellular space on exosomes which likely has a biological function similar to sCD109, but could theoretically fuse with other cells or the source cell to function at the cell surface.¹⁵⁹

Cancer

In the context of cancer, CD109 expression is increased, particularly in squamous cell carcinomas (SCC), including in the uterine cervix, lung, esophagus, oral cavity, gallbladder, penis, soft tissue, and skin.¹⁶⁰ In SCC, the expression is high compared to normal squamous tissue, appearing to inversely correlate with tumor grade, and therefore may be a valuable biomarker of SCC progression.¹⁶⁰ Furthermore, serum sCD109 increases relative to tumor volume in transgenic and xenograft tumor mouse models suggesting that serum sCD109 may also serve as a potential biomarker for SCC and other tumor types overexpressing CD109.¹⁶¹ Indeed, CD109 is also highly expressed in some non-SCC cancers, including breast cancer, melanoma, lung adenocarcinoma, pancreatic cancer, bladder cancer, nasopharyngeal carcinoma, and ovarian cancer.¹⁵⁴ In many of these cancers, CD109 expression has been linked to increased invasiveness and metastasis with variable impact on patient survival. However, in contrast, CD109 expression in bladder cancer was found to be positively associated with an improved patient prognosis.¹⁶² Notably, one group has developed 2 novel antibodies to CD109 with diagnostic and prognostic value, but further work needs to be done to determine if there is therapeutic potential for the antibodies in patients.¹⁶³

In oral SCC cells CD109 overexpression enhanced proliferation, whereas reduced CD109 expression reduced the rate of growth. This was attributed to the ability of CD109 to block the anti-proliferative response to TGF-β1 via the Smad2/3 pathway.¹⁶⁴ Though, a recent in vitro study with oral SCC cells also found that CD109 suppresses EMT and inversely correlates with tumor grade, which was also attributed to suppression of TGF-β signaling, highlighting the often cited dichotomous impact of TGF-β signaling in many cancers.¹⁶⁵ However, the role of CD109 overexpression in cancer is not necessarily exclusive to its ability to regulate TGF-β signaling. In cervical SCC, CD109-mediated EGFR and STAT3 signaling are implicated in tumorigenicity and aggressiveness of the cancer,¹⁶⁶ and in lung adenocarcinoma, CD109-mediated STAT3 and EGFR signaling has been shown to promote metastasis and poor patient prognosis.¹⁶⁶ However, a recent study in lung adenocarcinoma instead found no impact on EGFR signaling and discussed the role of STAT3 as context-dependent.¹⁶⁷ They instead found that the pro-metastatic response is due to CD109 interaction with latent TGF-β binding protein-1 (LTBP1) to enhance TGF-β activation in the tumor stroma, in contrast to the TGF-β downregulating effect of CD109 in epithelial cells.^167,168 Notably though, in the context of glioblastoma cell line models, the ability of CD109 to inhibit TGF-β signaling and promote EGFR signaling may be cell-dependent and related to the level of glycosylation on the N-terminal of CD109.¹⁶⁹

While much of the established CD109 cancer research has focused on the cancer cells, a role for CD109 in other tumor-associated cells has not been as well investigated. However, it was found in HCC that reduced CD109 expression in tumor-associated endothelial cells is associated with enhanced tumor progression via the upregulation and secretion of IL-8 caused by enhanced activation of TGF-β/Akt/NF-κB pathways.¹⁷⁰ Beyond the tumor, it has also been found that bone marrow mesenchymal stem cells release sCD109, and via a paracrine effect, can mitigate malignant properties of SCC cells.¹⁷¹ Additionally, there is a rare population of circulating endothelial cells that express CD109, which are more common during cancer angiogenesis, found in the blood of breast cancer and glioblastoma patients which become less common post-treatment.¹⁷²

SCUBE proteins

SCUBE proteins are a family of secreted cell-surface glycoproteins which have three members. Structurally, SCUBE proteins have a signal peptide sequence, nine tandem repeating EGF-like domains, an N-glycosylated region, three spaced cysteine-rich sections, and a CUB (complement C1r/C1s, Uegf, Bmp1) domain.¹⁷³ SCUBE proteins are expressed during embryogenesis and in the endothelium. Investigation of SCUBE proteins as TGF-β superfamily co-receptors has been scant, but two members of the family have been described by two groups as such. Interestingly, one member appears to promote SMAD1/5/8 signaling while the other promotes SMAD2/3 signaling.

In zebrafish, SCUBE1 modulates BMP signaling and may be a BMP co-receptor.¹⁷⁴ Additionally, SCUBE1 co-immunoprecipitated with BMP2/4 and BMPRs when overexpressed in vitro.¹⁷⁴ Notably, SCUBE1 induced BMP signaling in the absence of BMP ligand, but was significantly higher when BMP2 was added.¹⁷⁴ Knockdown of SCUBE1, but not SCUBE2, significantly decreased SMAD1/5/8 phosphorylation.¹⁷⁴ More recently, this same group also found SCUBE1 to be able to bind BMP7 and confirmed SCUBE1 as a BMP co-receptor in mice.¹⁷⁵ Notably, secreted SCUBE1 has been implicated as a possible biomarker of renal and gastric cancer.^176,177 SCUBE1 can also form a heteromeric complex with SCUBE2, which has been described as a VEGF co-receptor,¹⁷⁸ suggesting there may be effects on angiogenesis as well.

SCUBE3 is expressed and secreted from cancer cells and then cleaved by MMP2 and MMP9 into two fragments, one that contains the EGF-like repeats and the other that contains the CUB domain.¹⁷⁹ In a ligand-independent manner, SCUBE3 binds directly to TβRII via the CUB domain to initiate SMAD2/3 signaling.¹⁷⁹ Because of this, it has been described as a TGF-β signaling ligand itself, rather than a co-receptor. SCUBE3 promotes lung cancer cell mobility and invasiveness in vitro and tumor progression and metastasis in vivo,¹⁷⁹ with TGF-β signaling induced by SCUBE3 contributing to these cancer phenotypes. More recently, TGF-β signaling by SCUBE3 has also been implicated in the promotion of breast cancer.¹⁸⁰

Neuropilins

Neuropilins are 120–140 kDa single pass transmembrane glycoproteins that include neuropilin-1 (Nrp1) and neuroplin-2 (Nrp2). Nrp2 has two splice variants named Nrp2A and Nrp2B. Nrp1 has higher sequence homology to Nrp2A (44%) compared to Nrp2B (15%).¹⁸¹ Neuropilins contain a short cytoplasmic region with a PDZ domain, as well as extracellular a1/a2 Cubilin homology domains, b1/b2 FV/VIII domains, and a single MAM domain.¹⁸¹ Nrp1 does not have a cytoplasmic signaling motif, but similar to TβRIII, it binds to GIPC via its PDZ binding motif.¹⁸²

Neuropilins are well known co-receptors for class 3 semaphorins, VEGF, and VEGF-C/D, but are also powerful co-receptors for TGF-β signaling.¹⁸³ Nrp1 can bind both active and latent forms of TGF-β1 via a negatively charged cleft in its b1 domain, and can then activate the latent form (Fig. 4).¹⁸⁴ Additionally, Nrp1 can interact with TβBRI, TβRII, and TβRIII with the highest affinity for TβRI.¹⁸⁴ Nrp1 with TGF-β1 can co-internalize with TβRI and colocalize in cytoplasmic vesicles, suggesting that it may regulate TβRI internalization.¹⁸⁴ Nrp1 promotes TGF-β1-mediated SMAD2/3 signaling while suppressing SMAD1/5/8 signaling in myofibroblasts and tumor cells.^184,185 However, in endothelial cells, Nrp1 downregulates SMAD2/3 activation via BMP9/ALK1 and TGF-β1/ALK5 to promote the tip-cell phenotype in sprouting angiogenesis.¹⁸⁶ Notch downregulates Nrp1 to inhibit this phenotype.¹⁸⁶ Interestingly, Nrp1 appears to be able to activate SMAD2/3 when interacting with galectin-1 in a glycosylation-dependent manner.¹⁸⁷ Initially, it was reported to do this independently of TGF-β signaling, but subsequent work by the same group suggests that TGF-β signaling is required.^187,188 The role of Nrp2 as a TGF-β co-receptor has not been as well studied, but has been shown to bind to TGF-β1 by surface plasmon resonance.¹⁸⁹

Figure 4. Mechanisms regulating neuropilin-1-mediated signaling.

Neuropilin-1 (Nrp1) activates TGF-β from its latent form via its b1 domain. Nrp1 facilitates TGF-β binding to the TβRI/II receptor complex to promote SMAD2/3 signaling and suppress SMAD1/5 signaling. In the endothelium Nrp1 instead facilitates SMAD2/3 signaling suppression with TGF-β1 and BMP9 via ALK5/RII and ALK1/RII respectively. However, endothelial Notch signaling suppress Nrp1. Nrp1 also functions as a co-receptor for Vegf receptor family to promote Akt and Erk signaling. Soluble Nrp1 (sNrp1) is transcribed in its truncated form without the MAM domain and released at the cell surface where it competes for ligand to inhibit TGF-β and Vegf signaling.

In a potential feedback mechanism, TGF-β signaling can alter the expression of neuropilins, with TGF-β1 signaling enhances expression of Nrp2, while depressing expression of Nrp1.¹⁹⁰ Also, in the vasculature, BMP4 promotes Nrp1 expression where it helps to promote arterial differentiation.¹⁹¹

Neuropilins are most widely known for their neuronal function, where they are involved in axon growth and guidance, and the vasculature, where they are involved in angiogenesis. Nrp1 knockout mice are embryonic lethal, dying by e13.5 due primarily to cardiovascular abnormalities.¹⁹² Additionally, in its capacity as a TGF-β co-receptor, Nrp1 promotes liver fibrosis in hepatic stellate cells due in part to enhanced TGF-β1 signaling and TGF-β–induced collagen secretion.¹⁹³ Neuropilins also play a role in the immune system. Nrp1 is expressed on Treg cells and requires TGF-β signaling to do so. Nrp1 can then bind TGF-β1, including the latent form, on Tregs to promote their regulatory activity.¹⁹⁴ However, neuropilins can also be expressed in epithelial cells with implication in cancer pathogenesis,¹⁹⁵ and most recently Nrp1 has been implicated in facilitating the infectivity of SARS-CoV-2.¹⁹⁶

Soluble forms of neuropilins (sNrp) contain the a1/a2 and b1/b2 domains, but lack the MAM domain.¹⁹⁷ sNrp1 is transcribed as a truncated sequence, rather than cleaved from the full-length protein. The mRNA of sNrp1 contains a short unique intron, making it identifiable from the full-length version, and was found to be expressed in liver, kidney, skin, and breast tissue.¹⁹⁷ The presumed role of sNrp1 is to sequester ligand, and has thus far been shown capable to do so with VEGFs.^197,198 Unfortunately, investigations on the role of sNrps on TGF-β signaling are lacking, and might be of interest given the ability of the membrane bound Nrp1 to bind and activate latent TGF-β.¹⁸⁴

Cancer

Neuropilins have been implicated to play a role in a variety of cancers. Overexpression of neuropilins in tumors of epithelial origin enhanced growth and invasiveness and correlates with poor prognosis.¹⁹⁹ For example, neuropilin expression correlates with tumor grade in endometrial cancer,²⁰⁰ and Nrp1 expression in breast cancer and colon cancer correlates with cancer progression.²⁰¹ Additionally, neuropilins are involved in cancer processes associated with VEGF binding in the tumor vasculature and cancer stem cells.²⁰² For example, Nrp2 promotes VEGF-mediated stemness in breast cancer stem cells.²⁰³ However, examination of the TGF-β co-receptor function of neuropilins in cancer has been limited, with most of the work focused on lung cancer, HCC, breast cancer, and colorectal cancer.

Recently, it was found in NSCLC that Nrp1 promotes TGF‑β1‑mediated EMT, migration, and invasion via interaction with TβRII, with increased expression in metastatic tissue.²⁰⁴ Additionally, Nrp1 upregulation in HCC contributes to tumor growth and vascular remodeling, where TGF-β1 binding to Nrp1 is involved in the recruitment and activation of vascular pericytes.^185,205 Nrp2 was found to promote TGF-β-mediated breast cancer cell migration and invasion.²⁰⁶ Furthermore, in colorectal cancer, TGF-β1-mediated EMT was promoted by enhanced Nrp2 expression.¹⁸⁹ Specifically, Nrp2 promoted TGF-β1-mediated SMAD2/3 activation, which could be mitigated by TβRI inhibition.¹⁸⁹

In at least two cases, Nrp2 expression is increased by TGF-β signaling to enhance cancer progression without a currently known connection to its co-receptor function. First, Nrp2 expression is increased by TGF-β signaling in HCC, resulting in increased cell migration and invasion and is associated with poor prognosis.^207,208 Secondly, expression of Nrp2B is enhanced by TGF-β signaling in NSCLC, promoting a metastatic phenotype.²⁰⁹

A role for sNrp has not been well characterized in cancer, but the evidence thus far shows that it can play an antitumor role. The initial studies on sNrp1 showed that it promotes vascular damage and cancer cell apoptosis in Rat tumors of prostate cells with VEGF sequestration as the likely cause.¹⁹⁷ Additionally, sNrp2 isoform s₉Nrp2^B was shown to prevent prostatosphere formation of prostate cancer cells by sequestering VEGF-C.¹⁹⁸

There has been some success in targeting neuropilins to inhibit cancer progression and to facilitate the targeted release of drugs. For example, Nrp1 is being used both as a target for tumors and a mechanism to endocytose nanoparticle for delivery of drugs.⁷ Additionally, in lung cancer, the anti-Nrp1 peptides named DG1 and DG2 have been shown to block Nrp1 signaling and suppress angiogenesis, tumorigenesis, and cancer cell invasion,²¹⁰ though its impact on TGF-β signaling was not investigated. Notably, the infectivity of SARS-CoV-2 can be inhibited by a monoclonal antibody targeting the extracellular b1/b2 domain of Nrp1.¹⁹⁶ While the potential for clinical use of this antibody in cancer is not clear, it has an interesting ability to block furin-mediated cleavage, preventing release of sNrp1.

Cripto-1

Cripto-1, which is also known as teratocarcinoma-derived growth factor-1 (TDGF-1), is a GPI-linked glycoprotein that was first isolated in undifferentiated human NTERA-2 teratocarcinoma cells in 1989.²¹¹ Structurally, Cripto-1 is 188 amino acids long, containing cysteine-rich EGF-like and CFC domains, a hydrophobic C-terminus, and an extracellular signal sequence. On the membrane, Cripto-1 is localized to lipid rafts and can be cleaved by GPI-phospholipase D to release a soluble form that can function as an autocrine growth factor (Fig. 5).²¹²

Figure 5. Mechanisms regulating cripto-1-mediated signaling.

Cripto-1 facilitates the binding of Nodal to ALK4/ActRIIB and ALK7/ActRIIB receptor complexes to promote SMAD2/3 signaling. Soluble Cripto-1 (sCripto-1) competes with surface Cripto-1 for nodal binding, thereby suppressing SMAD2/3 signaling. Cripto-1 in complex with GRP78 suppresses TGF-β1- and Activin-mediated SMAD2/3 signaling. Cripot-1 also induces Notch and Wnt signaling, and sCripto-1 can induce Erk/Akt signaling via interaction with Glypican-1.

Cripto-1 is a co-receptor for the TGF-β superfamily ligands Nodal, Activin, and growth and differentiation factor (GDF) 1 and 3, which promote the activation of the type I receptors, ALK4 and ALK7, and the Activin type II B receptor (ActRIIB).^213,214 The subsequent cytoplasmic signaling is carried via phosphorylated SMAD2/3, where it then translocates to the nucleus with SMAD4. Nodal does not independently bind to ALK4 and ActRII, but instead requires Cripto-1 in order to bind to these receptors via interaction with the Cripto-1 EGF-like domain.²¹⁵ Cripto-1 interacts with ALK4 via a conserved CFC motif on Cripto-1, which is essential for Nodal binding to the receptor complex and Smad2 activation.²¹⁵ Nodal can independently bind to ALK7, but is greatly potentiated by Cripto-1; in which case ALK7 will only complex with ActRIIB, but not with ActRIIA.²¹³ Independent of its co-receptor activities, Cripto-1 also inhibits TGF-β1, Activin A, and Activin B signaling when interacting with glucose-regulated protein-78 (GRP78) at the cell surface via its CFC domain,^216,217 which helps stabilize GRP78 at the surface.²¹⁸ Outside of its role in TGF-β superfamily signaling, Cripto-1 has been found to enhance Notch and Wnt signaling, and in its soluble form, Cripto-1 can activate the Src, PI3K/Akt, and MAPK pathways through its interactions with the heparan sulfate proteoglycan Glypican-1.²¹⁹ In a recent proteomics analysis, new binding partners for Cripto-1 were identified, including regulators of extracellular exosomes and the cytoskeleton.²²⁰ This analysis also revealed that subcellular localization of Cripto-1 is regulated by Myosin II activity.²²⁰

Cripto-1 knockout mice are embryonic lethal with abnormalities apparent by e7.5 with defective mesoderm formation and axial organization and subsequent failure in heart development.²²¹ Cripto-1 is considered an embryonic gene since its expression becomes greatly reduced in adulthood. However, Cripto-1 is expressed in mammary glands during pregnancy and lactation, when the soluble Cripto-1 can be found in human milk and serum.²²² Cripto-1 is considered a marker of embryonic stem cells and is transcriptionally targeted by Oct4, a key pluripotency factor.²²³ However, in adult stem cells, its aberrant expression is believed to contribute to cancer pathogenesis.²²⁴

Cancer

While not usually expressed in adult differentiated tissue, Cripto-1 is highly expressed in a broad spectrum of human tumors, including cancers of the skin, lung, oral cavity, eye, colon, gastric, bladder, pancreas, liver, breast, cervix, endometrium, and ovaries.^225–229 The mechanism of Cripto-1 re-expression in tumors is not yet well understood, but at least one study found that hypomethylation of the Cripto-1 promotor was enhancing Cripto-1 expression.²³⁰ However, in at least two tumor types, micro RNAs have been identified to target and downregulate Cripto-1 to some degree, miR-15a-16 in NSCLC and miR-15b in glioma. Mechanisms of downregulating Cripto-1 are of great interest because it is frequently involved in malignant processes of cancer, including the maintenance of cancer stem cells, promotion of EMT, and cancer cell invasion.^228,231 In particular, Cripto-1 has been found to promote invasiveness and lymph node metastasis, through its ability to drive EMT in tumors, including in NSCLC, melanoma, prostate cancer, bladder cancer, and esophageal squamous cell carcinoma.^{225,227,232,233} Cripto-1 is also implicated in promoting oncogenic transformation. For example, overexpression of a Cripto-1 transgene in mouse mammary glands causes an increased rate of mammary tumor development, and one mechanism by which Cripto-1 is shown to promote this is by inhibiting Activin signaling, which otherwise would inhibit cell growth and tumorigenesis.²³⁴ Cripto-1 is often an indicator of poor prognostic outcomes in many patients. Specifically, this has been shown in patients with HCC, NSCLC, clear cell renal carcinoma, and bladder cancer.^{225,227,235,236}

Due to the very high tumor expression of Cripto-1 compared to normal tissue, Cripto-1 has been targeted for cancer treatment with expectations of minimal impact on normal tissue. Indeed, there is a rationale for its use in combination with existing treatments as high Cripto-1 expression may be responsible for resistant to certain treatments based on pre-clinical studies with sorafenib, cytarabine, cisplatin, and taxol.^228,237 In at least one case, Cripto-1 has been attributed to this by causing resistance to apoptosis via activation of a TAK-1/NF-κB/Survivin pathway.²³⁷ However, it should be considered that Cripto-1 expression in cancer cells can be heterogeneous with high and low expressing populations, and the low expressing populations have the potential to become high expressing, suggesting a level a plasticity that may add to the complexity of treatments targeting Cripto-1 long-term.²³¹

Serum or plasma Cripto-1 has thus far been shown to be a potential biomarker for at least a subset of tumors. Evidence of this has been shown in patients with testicular germ cell tumors, oral squamous cell carcinoma, HCC, NSCLC, breast cancer, and colon cancer.^238–241 Notably, in testicular germ cell tumors, the increased soluble Cripto-1 was also observed in seminal plasma and correlated with the levels of another reported biomarker, miR-371a-3p.²³⁸ Also, hepatitis B virus-related HCC patients had an even higher soluble Cripto-1 increase compared to other HCC patients.²³⁹ In patients with NSCLC, increased serum Cripto-1 correlated with shorter overall survival rates and increased tumor burden, suggesting that in some cases serum Cripto-1 may also serve as a prognostic indicator.²⁴⁰ However, serum Cripto-1 is not necessarily implicated in causing that effect since treating with exogenous soluble Cripto-1 was shown to reduce stemness properties and self-renewal in a cancer stem cell model, and also minimized differentiation into vascular endothelial cells due to reduced ALK4/SMAD2 signaling.²⁴²

There have been many advances in the potential to clinically target Cripto-1 in cancer. In xenograft models for testicular and colon cancer, a neutralizing antibody to Cripto-1 inhibited tumor cell growth.²³⁴ Additionally, monoclonal antibodies targeting the EGF-like domain of Cripto-1 have shown some promise. One antibody, named 1B4, has thus far shown evidence of neutralizing capability in melanoma cells in vitro,²⁴³ and another antibody shows attenuated growth in leukemia and colon cancer cells, both in vitro and in vivo.²⁴⁴ Additionally, a Nodal monoclonal antibody (3D1) was found to block interaction with Cripto-1 and downstream SMAD2/3 phosphorylation.²⁴⁵ In melanoma, 3D1 was found to block anchorage-independent growth and the formation of a vasculogenic network in vitro.²⁴⁵ Also, since Cripto-1 re-expresses in many cancers, and is absent or lowly expressed in normal differentiated adult tissue, it is considered a potential target for vaccination.²⁴⁶ Using a plasmid DNA vaccination method and the 4T1 breast cancer model, vaccinated mice had reduced primary tumor growth and metastasis.²⁴⁷ In a separate study, Cripto-1 DNA vaccinated mice were protected against metastatic melanoma due in part to an improved cytotoxic CD8+ T cell response.²³²

MuSK

MuSK (Muscle-Specific Kinase) is a receptor tyrosine kinase that has been described as a BMP co-receptor. It is not the only receptor tyrosine kinase shown to modulate TGF-β and BMP signaling pathways, but does so in a ligand-dependent manner. MuSK performs its role as a BMP co-receptor independent of its receptor tyrosine kinase function. MuSK binds BMP4 at its extracellular Ig3 domain, including at low nanomolar concentrations, which then facilitates binding to the BMPRI receptors ALK3 and ALK6.²⁴⁸

MuSK is most well-known for regulating the formation of neuromuscular junctions in skeletal muscle, but is also expressed in the liver, brain, and lungs.²⁴⁹ Defects in MuSK signaling have primarily been associated with muscle disorders.²⁵⁰ However, a role for MuSK in cancer pathogenesis has not been well described in the literature with one exception; a oncogenic role for MuSK has been described in HCC, but its BMP co-receptor function is not implicated in that role.²⁵¹ Instead the secreted proteoglycan Agrin signals through MuSK to facilitate the oncogenic activity.²⁵¹

RGMa/b/c

Repulsive guidance molecules (RGM) are GPI-linked glycoproteins that were first discovered as BMP co-receptors in 2005.^252,253 There are three mammalian isoforms, RGMa, RGMb (also known as DRAGON), and RGMc (also known as hemojuvelin). These isoforms share similarity in overall structure with about 40–50% similarity in amino acid sequence. In addition to a C-terminal GPI anchor, each RGM also has an N-terminal signal peptide and a von willebrand factor (vWF) type-D domain.

RGMs are highly specific BMP co-receptors with all three membrane-bound isoforms enhancing classical BMP signaling via downstream SMAD1/5/8 activation. The role of RGMs on BMP signaling is most impactful when BMP ligand levels are too low to directly initiate signaling.^252,253 RGMa RGMb, and RGMc have all been reported to bind specifically to BMP2 and BMP4, but not BMP7, with RGMc also binding BMP6.^252–254 However, quantitation of binding kinetics via surface plasmon resonance shows that all three RGMs bind BMP2, BMP4, BMP5, BMP6, and BMP7 to some degree with BMP4 having the highest overall affinity and BMP7 the lowest.²⁵⁵ Each RGM then interacts with type I and type II BMP receptors with some variability in specific receptors. RGMa has been shown to interact with the type I receptors ALK3 and ALK6 with type II receptors ActRIIA and BMPRII, whereas RGMb interacts with ALK2, ALK3, and ALK6 with ActRIIA and ActRIIB, and RGMc interacts with ALK2, ALK3, and ALK6 with ActRIIA and BMPRII.^252–254

RGMs also function as a ligand for the receptor neogenin, signaling through which is implicated in many of the CNS functions of RGMs.²⁵⁶ However, neogenin also plays a role in promoting BMP-mediated SMAD1/5/8 signaling.²⁵⁷ Indeed, RGMs can simultaneously bind BMP-2 and neogenin, which is notable because BMP-2 causes RGM association with lipid rafts in a neogenin-dependent manner that enhances association with BMPRs.^257,258 However, the extent to which neogenin plays a role in RGMs functioning as BMP co-receptors has been controversial and remains unclear.²⁵⁹

Each RGM isoform have largely non-overlapping expression throughout the body in a variety of tissues, with RGMa and RGMb primarily expressed in the nervous system, and RGMc is primarily found the liver, heart, and skeletal muscle.²⁶⁰ RGMs play an important role in a variety of functions including neural tube closure, repulsive axonal guidance, neuronal differentiation, axonal regeneration, regulating iron homeostasis, inflammation, and immunity.²⁵⁹

As with other GPI-linked glycoproteins, RGMs can be cleaved and released from the cell surface in a soluble form with soluble RGMc being the most detectible.²⁶¹ RGMc is cleaved by furin which occurs at a polybasic site specific to RGMc, which may explain why endogenous levels of soluble RGMa and RGMb are less detectable in vivo.²⁶² Soluble RGMc plays a role in iron homeostasis; furin-mediated cleavage of RGMc is enhanced during iron deficiency and hypoxia in skeletal muscle.²⁶² Soluble RGMc antagonizes BMP signaling to downregulate hepcidin, a key regulator of iron absorption and recycling.²⁶³

Cancer

Increased BMP signaling can promote tumor progression by enhancing angiogenesis, migration, and invasion of tumors.²⁶⁴ So it might be expected that RGM-mediated regulation of BMP signaling would have implications in cancer. However, investigation of RGMs in cancer has thus far mostly been limited to a small handful of groups studying their role in breast cancer, prostate cancer, lung cancer, and colorectal cancer. Though recent studies of RGMa and RGMb in oral squamous cell carcinoma and gastric cancer have emerged.²⁶⁵ The prevailing theme is that RGMa and RGMb are involved in regulating cancer progression, and in most cases a BMP pathway-mediate mechanism is implicated or suspected. However, a role for RGMc remains unclear, and no role for a soluble RGM isoform has yet to be associated with cancer to our knowledge.

RGMa is targeted and downregulated at the transcriptional level by increased miR-210-3p in oral squamous cell carcinoma, which was associated with enhanced cancer cell proliferation and poor prognosis.²⁶⁵ RGMa is also downregulated by MicroRNA-4472 expression, which was shown to promote EMT in breast cancer cells.²⁶⁶ Lower transcript levels of RGMa was also associated with poor prognosis in breast cancer patients, while RGMb and RGMc levels were not significantly different based on prognosis.²⁶⁷ However, an earlier study by the same group found that RGMa and RGMc were barely detectable in breast cancer cells, while RGMb was. Knockdown of RGMb expression in breast cancer cells was demonstrated to enhance growth, survival, adhesion, and migration via a SMAD-dependent pathway.²⁶⁸

In prostate cancer cells, it was shown that when each RGM isoform was knocked down, loss of at least one isoform enhanced cell growth (RGMb and RGMc) and motility (RGMb) compared to control cells.²⁶⁹ However, the implications of this are unclear since no decreased expression of RGM isoforms was observed in prostate cancer cells.²⁶⁹

In lung cancer, RGMb expression is significantly lower in NSCLC tissue compared to normal tissue and correlated with advanced-stage tumors and poor patient prognosis.²⁷⁰ Similar results from another group suggests that increased lncRNA RGMB-AS1 in advanced NSCLC tumors may be targeting and downregulating RGMb, which was also associated with poor prognosis through increased growth and metastasis.²⁷¹ More recently, lncRNA RGMB-AS1 has been implicated in gastric cancer, though its impact on RGMb was not examined.^272,273

In colorectal cancer, RGMa is decreased compared to normal tissue, which was associated with tumor progression and hypermethylation of its promotor.²⁷⁴ While the consensus of much of the literature on RGMa and RGMb is that they are tumor suppressors, in at least one context RGMb has been described as a tumor promotor. RGMb expression is increased in colorectal cancer and promotes BMP pathway-mediated colorectal cancer proliferation and tumor growth,²⁷⁵ and a subsequent study by the same group suggests that RGMb may also contribute to drug resistance in colorectal cancer.²⁷⁶ Specifically, RGMb contributed to oxaliplatin resistance in colorectal cancer cells in vitro.²⁷⁶ It is not yet clear if there is a connection between the decreased RGMa and increase RGMb expression described in colorectal cancer. Lastly, RGBc expression was found to be reduced in HCC due to decreased mRNA stability, and while a role for RGBc in cancer-associated phenotypes was not examined, it has potential implication on iron metabolism in patients with HCC.²⁷⁷

Pharmacological targeting of TGF-β superfamily co-receptors

With the exception of TCR105 targeting Endoglin, clinical trials for treatments targeting TGF-β superfamily co-receptors have not yet advanced beyond Phase II, and only Endoglin, Neuropilin-1, and Cripto-1 have been targeted in cancer (Table 1). A presumable benefit of targeting TGF-β superfamily co-receptors is the potential for less toxicity compared to treatments directly targeting core components of these pathways. Indeed, both small molecule inhibitors of ALK5 and neutralizing antibodies to TGF-β ligands have exhibited pre-clinical dose limiting toxicities, resulting in cardiac valve lesions and aortic aneurysms.^5,278 Additionally, evidence of autoimmunity and atherosclerosis in the absence of TGF-β signaling in T-cells has been described.^279,280 Targeting some TGF-β superfamily co-receptors have the benefit of tissue specificity, eliminating much of the broader impact of TGF-β superfamily pathway inhibition. Though, the impact on non-SMAD pathways should be considered when targeting these co-receptors, non-SMAD signaling modulation may have parallel oncogenic benefits.

Table 1.

Targeting TGF-β superfamily co-receptors in clinical trials

Co-receptor	Drug	Mechanism	Phase	Condition Treated
TβRIII	P144	Peptide blocks TGF-β1 interaction with TβRIII	2	Skin fibrosis
Endoglin	TRC105 (Carotuximab)	Anti-Endoglin chimeric monoclonal antibody	3	Multiple cancers
Neuropilin-1	MNRP1685A	Anti-NRP1 monoclonal antibody; blocks VEGF binding	1	Solid tumors
Neuropilin-2	ATYR1923	Binds NRP2 to downregulate immune engagement	2	Pulmonary sarcoidosis
Cripto-1	BIIB015	Humanized anti-Cripto-1 monoclonal antibody	1	Solid tumors
RGMa	ABT-555 (Elezanumab)	Anti-RGMa monoclonal antibody	2	Spinal cord injury; acute ischemic stroke; multiple sclerosis
RGMc	FMX-8	RGMc fusion protein; hepcidin antagonist	2	Anemia

Non-pathological angiogenesis in adults is primarily restricted to wound healing and the female reproductive system, which makes it an ideal target for the regulation of tumor growth, particularly with the ease of drug delivery to the endothelium. Also, genetic instability of cancer cells makes the endothelium a more stable and broadly applicable target. Therefore it may be attractive to target TGF-β superfamily co-receptors expressed in the endothelium, particularly endoglin due to its specificity for the pathological neovasculature.⁹³ Surface receptors on cancer cells that are absent in normal adjacent tissue are also excellent markers to visualize or target tumors for drug delivery, blocking antibodies, or vaccination. Additionally, as an immunosuppressive pathway in cancer, TGF-β signaling may be inhibited by targeting co-receptors that facilitate immunosuppression, such as TβRIII, endoglin, and Cripto-1.²⁸¹ Targeting these receptors in combination with immunotherapies may mitigate resistance factors to these treatments and improve their overall effectiveness.

Ectopic treatment with soluble forms of these co-receptors, as a ligand trap, may have benefits over directly targeting the surface receptor. This would allow for broad mitigation TGF-β signaling within the TME, including in immune cells and CAFs. Many of these soluble co-receptors are still able to facilitate some surface receptor TGFβRI and TGFβRII activation in low ligand concentration environments, thereby, suppressing high TGF-β signaling while preventing complete inhibition that could lead to systemic toxicity.

Conclusions and Future Studies

In this review we discuss the signaling, expression, and functionality of TGF-β superfamily co-receptors with particular focus on their role in cancer pathogenesis. The co-receptors discussed here are defined by their ability to bind TGF-β superfamily ligands and to modulate TGF-β superfamily signaling. Notably, most of these co-receptors are not exclusive to the TGF-β superfamily and are also able to modulate non-TGF-β signaling pathways, for example, NFkB with TβRIII and VEGF with Nrp1. Generation of a soluble form is also a common feature of many of these co-receptors, and often have the capacity to compete with membrane-bound receptors for ligand, resulting in suppressed signaling relative to the soluble to surface receptor ratio. Additionally, while many surface co-receptors facilitate ligand interaction with type II receptors, pseudoreceptors, such as BAMBI, instead inhibit ligand interaction by binding to ligands while failing to deliver it to a type II receptor. However, these relationships are somewhat complicated by the fact that membrane-bound co-receptors can interact with receptor complexes in the absence of ligand, potentially inhibiting their activity, while soluble co-receptors are still kinetically able to facilitate ligand transfer to type II receptors to some extent. The relative occurrence of these interactions are largely dependent on ligand availability and may somewhat serve as a signaling rheostat via their cognate receptor. Indeed, co-receptors that promote ligand interaction with type II receptors are generally notable for their ability to facilitate enhanced activation in low ligand environments, which is in large part due to these co-receptors being expressed well in excess of their cognate type II receptor(s).

There are fewer commonalities between co-receptors in regards to their role in cancer due to the specific TGF-β superfamily ligand/receptor interactions involved, non-TGF-β superfamily signaling pathway interactions, and variable expression patterns. However, the variable intricacies of each of these co-receptors offer new opportunities for modulating TGF-β superfamily pathways in the treatment of cancer. Pharmacologically, co-receptors have had limited success thus far when individually targeted in cancer, Cripto-1 being a notable exception. Though some have shown considerably more potential in combinations treatments, such as with TCR105 (anti-endoglin) with anti-VEGF or anti-PD-1 treatments, and may be particularly valuable for patients that are developing resistance to current therapies. Some co-receptors also show promise as a tumor-specific target for guided drug delivery, including liposomes targeting the tumor vasculature via endoglin or neuropilins. Also, there is increasing interest in modifying the levels of the soluble forms of these co-receptors, as in sTβRIII to enhance TME immunoregulation or sCD109 to suppress SCC malignancy. Ultimately, there is still much work to be done to understand the role of TGF-β superfamily co-receptors in cancer, but recent advances continue to reveal an expanding role for these co-receptors in cancer pathogenesis and an increasing potential for developing improved cancer therapeutics by targeting these co-receptors.