Targeting TGF-β Mediated SMAD Signaling for the Prevention of Fibrosis

Fibrosis occurs when there is an imbalance in extracellular matrix (ECM) deposition and degradation. Excessive ECM deposition results in scarring and thickening of the affected tissue, and interferes with tissue and organ homeostasis – mimicking an exaggerated “wound healing” response. Many transforming growth factor-β (TGF-β) ligands are potent drivers of ECM deposition, and additionally, have a natural affinity for the ECM, creating a concentrated pool of pro-fibrotic factors at the site of injury. Consequently, TGF-β ligands are upregulated in many human fibrotic conditions and, as such, are attractive targets for fibrosis therapy. Here, we will discuss the contribution of TGF-β proteins in the pathogenesis of fibrosis, and promising anti-fibrotic approaches that target TGF-β ligands.


INTRODUCTION
In fibrotic disease, increased deposition of extracellular matrix (ECM) proteins compromises tissue architecture, and interferes with normal organ function. Fibrosis is frequently observed in the heart, liver, lungs, and kidneys, but can arise in any tissue that has suffered chronic insult. Triggers for fibrosis can be either biological (e.g., persistent bacterial/viral infections, genetic disorders or tissue injury) and/or environmental (e.g., pollutant/chemical exposure or allergens) (Macneal and Schwartz, 2012). Fibrosis is primarily driven by inflammatory cytokines including the interleukins (Nikolic-Paterson et al., 1996;O'Reilly et al., 2012), and members of the transforming growth factor-β (TGF-β) superfamily (Roberts et al., 1986;Meng et al., 2016), such as TGF-β1, activin A and activin B. Many of these ligands are expressed by infiltrating inflammatory cells, which migrate toward damage tissues. Of interest, these TGF-β ligands not only promote ECM deposition, but they also become concentrated within the accumulating matrix, thereby accelerating the pro-fibrotic response.
Mechanistically, TGF-β1 promotes fibrosis by driving the differentiation of quiescent fibroblasts into matrix secreting myofibroblasts (Vaughan et al., 2000;Lijnen et al., 2003). Under duress fibroblasts differentiate into a proto-myofibroblast lineage, which in the presence of TGF-β1 (and ED-A fibronectin), become fully differentiated myofibroblasts (Gabbiani, 2003; Figure 1). Notably, TGF-β1 drives the production of α-smooth muscle (α-SM) actin, which gives the myofibroblasts their contractility. These cells represent activated fibroblasts, with a high synthetic capacity for ECM proteins (Ignotz and Massague, 1986). In response to heightened TGF-β1 signals, dominating myofibroblasts deposit excessive ECM, which compromises the local tissue architecture. TGF-β1 can further exacerbate this response by acting as a chemoattractant for some inflammatory cells (Wahl, 1992;Ludviksson and Gunnlaugsdottir, 2003), and is abundantly expressed by infiltrating macrophages in fibrotic tissue (Denholm and Rollins, 1993).
Although TGF-β1 is the best characterized pro-fibrotic factor within the family, TGF-β2 also displays potent fibrotic activity. TGF-β2 accumulates in the bile ducts in human fibrotic liver disease (Milani et al., 1991), and has been implicated in the fibrotic response associated with glaucoma (Wordinger et al., 2014). Remarkably, TGF-β3 appears to have anti-fibrotic activity in some tissues. TGF-β3 plays a key role in regulating epidermal and dermal cell motility during wound repair, a TGFβ-isoformspecific effect (Occleston et al., 2008). TGF-β3 is expressed at high concentrations during wound repair, and unlike TGF-β1 and -β2 isoforms, can promote wound healing without fibrotic scarring (Ferguson et al., 2009).
Activins -Similar to TGF-β1, activins can trigger a pro-fibrotic response in several tissues via activation of the SMAD-2/3 cascade. Activins promote the proliferation of fibroblasts, their differentiation into myofibroblasts (Ohga et al., 1996;Ota et al., 2003;Yamashita et al., 2004), and the accumulation of ECM (Yamashita et al., 2004;Murakami et al., 2006). Serum concentrations of activin are elevated in patients suffering cystic fibrosis (Hardy et al., 2015), acute respiratory failure (de Kretser et al., 2013), chronic kidney disease (Agapova et al., 2016), and heart failure (Yndestad et al., 2004). Indeed, increased serum activin is a hallmark of many human chronic conditions and can have catastrophic consequences for affected patients. Our studies have shown that activins can drive the multi-organ wasting syndrome, cachexia (Chen et al., 2014), and that high levels of activin can induce a marked fibrotic response in skeletal muscle and liver, characterized  (Presser et al., 2013). TGF-β2 accumulates in the bile ducts in human fibrotic liver disease, encouraging collagen deposition (Milani et al., 1991).
TGF-β1 drives differentiation of renal epithelial cells into α-SMA positive myofibroblasts, which also secrete collagen (Fan et al., 1999). In human glomerular disease, TGF-β1 protein expression is positively correlated with interstitial fibrosis severity and ECM production (Goumenos et al., 2001).  (Azuma, 2012) has been approved for the treatment of IPF. TGF-β1 targeting mAbs have also been tested in human IPF patients (Voelker et al., 2017), and TGF-β2 mAbs have anti-scaring activity in human glaucoma patients (Mead et al., 2003).
to improve wound healing during clinical trials (Ferguson et al., 2009).
Activin A ActRIIA/ActRIIB, ALK4/7, induces SMAD 2/3 signal PRO Activin A is upregulated in rat models of liver fibrosis, and drives collagen production from hepatocytes (Sugiyama et al., 1998). Activin A is produced by activated HSCs in vitro and in vivo, and its activity appears to be unopposed by follistatin (Patella et al., 2006).
Activin A promotes cell proliferation, induces differentiation into myofibroblasts, and promotes expression of collagen in primary cultured renal interstitial fibroblasts in rats (Yamashita et al., 2004). Activin A expression is upregulated in multiple mouse models of chronic kidney disease (Agapova et al., 2016).
Heart failure patients have elevated activin A serum levels, which correlate with disease severity.
Cardiomyocytes are the primary source of activin A production in the heart as it fails (Yndestad et al., 2004). Follistatin can improve muscle function and reduce fibrosis in two models of muscle disease (Mendell et al., 2015(Mendell et al., , 2017. Other ligands traps such as sActRII and propeptides work effectively in mouse models of activin-induced fibrosis (Chen et al., 2015;Agapova et al., 2016).
Myostatin is a negative regulator of muscle mass and overexpression results in a pro-fibrotic response, with increased myofibroblasts and ECM production .
Follistatin and soluble ActRII blocked myostatin signaling during clinical trials, leading to a reduction in fibrosis (Attie et al., 2013;Mendell et al., 2015;Campbell et al., 2016 BMP-9 derivatives have been examined as bone regenerative agents (Bergeron et al., 2009, Bergeron et al., 2012. No specific BMP-9 intervention has been trialed to date for fibrosis therapy. Frontiers in Pharmacology | www.frontiersin.org by an influx of differentiated myofibroblasts and accompanied ECM deposition. Significantly, we have shown that the fibrotic pathology induced by activins can be fully reversed (Chen et al., 2014(Chen et al., , 2015, highlighting the potential of anti-activin therapy to treat muscular dystrophies, in which patients suffer severe muscular fibrosis. Activin is one of four major TGF-β ligands that signal through the activin type II receptors, ActRIIA/ActRIIB. Activin A, activin B, myostatin (GDF-8), and GDF-11 can all complex with ActRIIA/ActRIIB and initiate SMAD-2/3 intracellular activity. Despite this, all four ligands have non-overlapping bioactivities in vivo owing to their cell/tissue specific expression and distinct preferences for type I receptors (ALK4, ALK5 or ALK7). Myostatin is expressed almost exclusively in skeletal muscle, and can mimic the pro-fibrotic response observed under high activin conditions . Myostatin promotes the proliferation of fibroblasts in muscle, and induces the expression of ECM proteins including collagen and fibronectin both in vitro and in vivo . Given the structural homology of GDF-11 with myostatin (Walker et al., 2017) and shared receptor contacts, it is predicted that at high local concentrations GDF-11 will also exhibit pro-fibrotic activity.
Unlike the TGF-β isoforms, activins are secreted in an 'active' form. To constrain their activity, activins (and myostatin) are regulated extracellularly by follistatin. Follistatin binds directly to activin-related ligands, shielding their receptor contact sites so as to limit their signaling potential. Follistatin has opposing activity to activins/myostatin in fibrosis; attenuating early liver fibrosis (Patella et al., 2006), and lung fibrosis (Aoki et al., 2005) in murine models.

ACTIVATORS OF THE SMAD-1/5 AXIS AND FIBROSIS
Whilst hyper-activation of the SMAD-2/3 pathway is reported to be pro-fibrotic, signaling through the alternate SMAD-1/5 pathway is anti-fibrotic. Of all the SMAD-1/5 activators, BMP-7 has the most well documented anti-fibrotic activity. BMP-7 is expressed only in select adult tissues, including the kidney, and its expression declines in rodent models of renal fibrosis. BMP-7 counteracts TGF-β1 induced induction of myofibroblasts and ECM in multiple models of organ fibrosis (Zeisberg et al., 2003;Weiskirchen and Meurer, 2013). Addition of exogenous BMP-7 can both prevent and even reverse fibrosis in models of kidney disease (reviewed in Weiskirchen and Meurer, 2013;Li et al., 2015). BMP-7 mediates its anti-fibrotic activity through the type I receptor, ALK-3 [also termed BMPRIA, (Sugimoto et al., 2012)]. Expression of ALK-3 increases in the early stages of kidney disease, and loss of ALK-3 exacerbates TGF-β1 mediated fibrosis, suggesting that ALK-3 is protective for fibrosis. ALK-3 is also a docking receptor for related BMP ligands, BMP-2, BMP-4 and BMP-6. Both BMP-2 and BMP-6 can similarly attenuate kidney fibrosis in rodent models of kidney disease (Yang et al., 2009;Dendooven et al., 2011), and BMP-6 expression is increased in humans and mice suffering non-alcoholic fatty liver disease (NAFLD) and is protective for hepatic fibrosis (Arndt et al., 2015). Interestingly, loss of BMP-4 is speculated to drive opposing pro-fibrotic response in cardiac tissue (Sun et al., 2013), and increased BMP-9 in the liver induces a fibrotic-like response (Breitkopf-Heinlein et al., 2017).
Like activins, BMPs are secreted from cells in an activated state. Once secreted, many of the BMPs are sequestered to the ECM by propeptide-mediated binding to fibrillins. Fibrillin binding creates a local concentration of BMPs and is thought to facilitate release of the active BMP proteins (reviewed in Wohl et al., 2016). Notably, perturbed fibrillin/microfibril assembly is a feature of human fibrosis (Wohl et al., 2016), and likely alters the bioavailability of associated BMPs. BMP activity is also restricted by extracellular antagonists Gremlin, DAN, Chordin and Noggin (reviewed in Weiskirchen et al., 2009).

TARGETING TGF-β SIGNALING FOR THE TREATMENT OF FIBROSIS
TGF-β targeted therapies in fibrosis are designed to reduce activin/TGF-β signaling via SMAD-2/3, or alternatively, promote a BMP-mediated SMAD-1/5 signal. Many approaches are currently being explored, and some have reached the clinic (e.g., TGF-β3, follistatin, BMP-7). Leading approaches and emerging new strategies targeted to TGF-β mediated SMAD signals are described here.
TGF-β targeted therapies -A plethora of TGF-β targeted therapies have been explored (reviewed in Akhurst and Hata, 2012), some of which have been designated specifically for fibrosis. These include, TGF-β1 antibodies for kidney fibrosis (Akhurst and Hata, 2012;Voelker et al., 2017), peptides (Llopiz et al., 2009), and receptor decoys for lung fibrosis (Yamada et al., 2007). Interventions for TGF-β2 mediated fibrosis include peptides for cardiac and skin fibrosis (Santiago et al., 2005;Hermida et al., 2009), and antibodies for glaucoma-related scarring (Mead et al., 2003). TGF-β3 in the form of Juvista TM (Renova) was found to improve wound healing in phase I/II human clinical trials, but failed in phase III (Ferguson et al., 2009). However, few of these approaches have resulted in positive patient outcomes. Indeed, in a recent study testing a TGF-β1 specific antibody in a diabetic model of kidney fibrosis, disease progression was not improved (Voelker et al., 2017). Several factors likely impede the clinical effectiveness of these approaches, including the presence of multiple pro-fibrotic factors in advanced disease.
Despite their non-selectively, the small molecule inhibitors have advanced furthest owing to their economical production and ease of administration. Many of these approaches target the kinase activity of the TGF-β type I (TβRI, or ALK-5) and II (TGFβRII) receptors, including compounds GW788388 (Petersen et al., 2008) and SB-525334 (Grygielko et al., 2005), and have proven anti-fibrotic activities in vivo. Significantly, the small molecule inhibitor Pirfenidone TM (Azuma, 2012) has been approved for the treatment of lung fibrosis in humans. Though encouraging, the promiscuous nature of these small molecule inhibitors renders them more prone to side effects (Noble et al., 2011).
Interventions that target the intracellular phosphorylation of SMAD-2/3 proteins, can also reduce TGF-β triggered fibrosis (reviewed in Munoz-Felix et al., 2015). SMAD-7, which sequesters SMAD-2/3, protects against renal fibrosis upon viral gene-delivery in a mouse model of kidney fibrosis (Terada et al., 2002). SIS3, a selective compound that targets only SMAD-3 proved to reduce the expression of ECM proteins and delayed the progression of diabetic nephropathy in a mouse model .
Activin and myostatin targeted therapies -Broad-spectrum TGF-β signaling inhibitors including SMAD-7, SIS3, and many of the small molecule kinase inhibitors can also block activin and myostatin induced signaling (Inman et al., 2002;Rodgers et al., 2014). The pleiotropic nature of these inhibitors is unfavorable for systemic use, and there is a pressing need for more tailored ligand therapies. 'Ligand traps' such as follistatin, soluble forms of the activin type II receptors (sActRII), and propeptides offer improved ligand selectivity, and their increased size favors serum retention (Wakefield et al., 1990).
Follistatin therapy is the leading approach for SMAD-2/3 blockade in activin/myostatin triggered fibrosis. Gene therapy approaches deploying follistatin have reached phase I/II clinical trials in humans (Mendell et al., 2015(Mendell et al., , 2017. Local expression of follistatin is sufficient to improve muscle function and reduce fibrosis in inflammatory myopathy (Mendell et al., 2017), and muscular dystrophy (Mendell et al., 2015). Similarly, the clinically relevant forms of sActRII (RAP-011 or ACE-011/sotatercept forms) can effectively reverse fibrosis in murine models of kidney disease (Agapova et al., 2016). However, clinical advancement of sActRII has been impeded by its apparent off-target vascular effects when used for human therapy (Attie et al., 2013;Campbell et al., 2016).
Propeptides, which are natural by-products of TGF-β assembly, and have an affinity for the ECM (Harrison et al., 2011), are attractive anti-fibrotic agents. We have shown that modified forms of the activin propeptides (propeptide-Fc fusion proteins) can both potently and specifically attenuate activin-mediated pathologies in mouse skeletal muscle (Chen et al., 2015(Chen et al., , 2017. Similarly, we and others have found that the myostatin propeptide can revert muscle damage and attenuate fibrosis in mouse models (Qiao et al., 2008;Hamrick et al., 2010;Chen et al., 2017). This approach has also been demonstrated for the TGF-β propeptide, termed 'latency associated propeptide' or LAP, which has been shown to attenuate TGF-induced pathologies in vivo (Bottinger et al., 1996). However, as the TGF-β1 LAP has comparable affinity for all three TGF-β isoforms, and also GDF-8/-11, further modifications are required to improve its selectivity.
BMP-targeted therapies for fibrosis -The anti-fibrotic activity of BMP-7 encouraged its application as a human therapy. However, like many TGF-β proteins, BMP-7 is poorly made and processed in mammalian cells (Swencki-Underwood et al., 2008), and upon delivery, is likely rapidly cleared from the blood (Coffey et al., 1987). To address these limitations, Sugimoto et al. (2012), developed a small molecule BMP-7 mimetic (AA123), which has the same anti-fibrotic activity as recombinant BMP-7 in a mouse model of kidney disease. A similar BMP-7 mimetic is currently in phase II clinical trials for acute kidney injury (Thrasos Therapeutics, Canada).

CONCLUSION
Extracellular matrices provide a structural framework for cells and additionally serve as a scaffold for growth factors. Excessive ECM deposition, as observed in fibrosis, compromises tissue and organ structure and function and can lead to organ malfunction and failure. TGF-β proteins are major regulators of fibrosis, and the balanced activities of the pro-fibrotic and anti-fibrotic ligands ensure tissue homeostasis. TGF-β targeted therapies in fibrosis are designed to suppress the pro-fibrotic activity of TGF-β isoforms/activins/myostatin, or heighten the activity of anti-fibrotic BMPs. Promising anti-fibrotic TGF-β targeted therapies involve the use of ligand traps (follistatin, soluble receptors, propeptides) which sequester and deter activation of pro-fibrotic signals. Though encouraging, a major hurdle for clinical transition of these ligand traps is their in vivo stability, tissue-specificity, and minimisation of side effects. Indeed, one of the only drugs approved for TGF-β inhibition in lung fibrosis triggered gastrointestinal upset and the appearance of rashes (Noble et al., 2011) in treated patients, and activin blockade using the sActRII ligand trap caused vasculature complications (Attie et al., 2013;Campbell et al., 2016). Additionally, the importance of TGF-β ligands in non-fibrotic processes requires titrated treatments, so as not to ablate homeostatic functions. The ultimate goal for future TGF-β targeted fibrotic therapies is to identify ligand specific inhibitors with an extended signaling range that can act precisely within fibrotic tissue.

AUTHOR CONTRIBUTIONS
KW collected supporting evidence and wrote manuscript. KJ assisted with reference collection and collating. CH edited and provided guidance for the manuscript.