Signaling mechanisms of myofibroblastic activation: outside-in and inside-out

Abstract

Myofibroblasts are central mediators of fibrosis. Typically derived from resident fibroblasts, myofibroblasts represent a heterogeneous population of cells that are principally defined by acquired contractile function and high synthetic ability to produce extracellular matrix (ECM). Current literature sheds new light on the critical role of ECM signaling coupled with mechanotransduction in driving myofibroblastic activation. In particular, transforming growth factor β1 (TGF-β) and extra domain A containing fibronectin (EDA-FN) are thought to be the primary ECM signaling mediators that form and also induce positive feedback loops. The outsidein and inside-out signaling circuits are transmitted and integrated by TGF-β receptors and integrins at the cell membrane, ultimately perpetuating the abundance and activities of TGF-β and EDA-FN in the ECM. In this review, we highlight these conceptual advances in understanding myofibroblastic activation, in hope of revealing its therapeutic anti-fibrotic implications.

1. Introduction

Principally characterized by accumulation of disorganized extracellular matrix (ECM), fibrosis is the endstage hallmark of a broad range of diseases. The central cellular mediator of fibrosis is myofibroblast with a prominent contractile/synthetic phenotype. While myofibroblasts are heterogeneous [1, 2], they commonly express smooth muscle proteins such as a smooth muscle actin (α-SMA) that forms contractile stress fibers. Moreover, myofibroblasts produce large amounts of ECM proteins, including collagen I, extra-domain A containing fibronectin (EDA-FN), and matrix metalloproteinases (MMPs). In recent lineage tracing studies, de novo expressed periostin was identified as a robust marker of myofibroblasts [3]. Many terms have been used for the process that generates myofibroblasts; e.g. myofibroblast (trans)differentiation or activation or transformation, fibroblast activation, fibroblast-to-myofibroblast transition or conversion, etc. Current literature reveals that rather than terminally differentiated, myofibroblasts represent a range of highly plastic cell states. Thus, in this review we opt to use the term “myofibroblastic activation” [4–6] attempting to reconcile the deviatory terminologies with updated understanding of myofibroblasts.

A classical view is that myofibroblasts derive from fibroblasts for wound repair, where they produce and contract granulation tissue to fill and close a lesion. Recent studies have revealed that myofibroblasts derive from a wide variety of cell types [7, 8], maybe triggered by signaling cues beyond injury [9]. Early-stage myofibroblastic activation is deemed protective. However, in fibrotic diseases it fails to terminate, resulting in progressive and chronic damage through continuous synthesis of fibrotic scar tissue [10]. The biological basis for the occurrence of fibrosis instead of resolution remains unclear. Myofibroblast-mediated fibrotic tissue remodeling occurs in most organs, especially at the end stage of major diseases ranging from scleroderma, atherosclerosis, heart failure, liver cirrhosis, renal fibrosis, and many cancers [11]. As such, a better understanding of signaling mechanisms underlying myofibroblastic activation is imperative for producing knowledge that could be broadly applied in anti-fibrotic therapeutic developments.

Extensive research has been done to uncover many signaling pathways responsible for myofibroblastic activation and fibrosis, for which a large number of reviews have been published, mostly in specific disease contexts. With new investigative tools available, ECM remodeling and mechanotransduction have recently received refreshed enthusiasm [12, 13], likely for their reciprocal actions that profoundly influence myofibroblastic activation [14, 15]. Of particular note, the duet of transforming growth factor β1 (TGF-β) and the alternatively spliced EDA-FN constitute a potent myofibroblastic activator [16]. They form a positive feedback loop by coupling with TGF-β receptors and mechanotranducer integrins on the cell surface (see Figure 1). This circuit transmits mechanical and biochemical signals from the ECM into the cell, triggering transcriptional reprograming (outsidein). Consequently, elevated production of α-SMA stress fibers and EDA-FN in turn mediates cell contraction and ECM alteration and further TGF-β activation (inside-out). To highlight this integrated concept, the current review focuses on TGF-β1/EDA-FN and associated integrin and intracellular signaling that orchestrate myofibroblastic activation.

Figure 1. Outside-in and inside-out signal transduction in myofibroblastic activation.

Outside-in: Activated TGF-β1 (freed from LAP) binds to the TGF-βR complex stimulating intracellular signaling that promotes α-SMA production. In parallel, ECM-to-cell mechanical transduction through the integrin/focal adhesion (FA) pathway activates RhoA, leading to assembly of α-SMA stress fibers.

Inside-out: TGF-β1-activated EDA-FN splicing increases the EDA-FN protein in the ECM; cell contraction executed by α-SMA stress fibers alters the LAP (gray) conformation via the integrin/LAP interaction releasing TGF-β1 (blue) from LAP.

Positive feedback loops: 1. TGF-β1 stimulates the production of EDA-FN which in turn facilitates the ECM incorporation and activation of TGF-β1 via the LTBP-1/EDA-FN/integrin interaction network. 2. ECM-integrin mechano-transduction activates RhoA which enhances the assembly of α-SMA stress fibers that execute cell contraction and integrin mechano-transduction, further augmenting RhoA activation. 3. TGF-β1-stimulated stress fiber production ultimately enhances TGF-β1 release from the ECM via mechano-activation through the integrin/LAP interaction. 4. The assembly of G-actin (monomer α-SMA) into F-actin (stress fiber) allows for the escape of MRTF from G-actin sequestration in the cytosol and its nuclear translocation and SRF activation, which in turn propagates α-SMA and stress fiber production.

Notes:

1. Latent TGF-β forms SLC with the LAP “straight jacket”, where it is trapped; SLC and LTBP forms LLC.

2. Green arrows indicate TGF-βR and integrin/FA mutual modulations.

3. Activated RhoA promotes MRTF nuclear translocation by enhancing F-actin assembly while reducing G-actin in the cytosol.

2. Precursors of Myofibroblasts

Origins of myofibroblasts, particularly in vivo, have been hotly debated [17]. Recent lineage tracing studies overall support the assertion that the main sources of myofibroblasts are resident mesenchymal cells such as fibroblasts and pericytes [18, 19]. It has been shown that periostin-expressing myofibroblasts in the heart are derived from resident fibroblasts of the TCF21 lineage, but not from endothelial, immune/myeloid or smooth muscle cells [3]. Fibroblasts are systemically ubiquitous ECM-producing cells that assume a spindle-shaped morphology in cell culture [20–22]. Myofibroblasts are distinguished from fibroblasts by highly upregulated α-SMA stress fibers and contractility, de novo expression of periostin and EDA-FN, increase of smooth muscle myosin heavy chain, vimentin, focal adhesion proteins, and elevated collagen secretion [23]. Lineage tracing experiments also identified that resident pericytes in the kidney [24] and those in the liver (hepatic stellate cells, HSCs) [18, 25] account for the major sources of fibrogenic myofibroblasts. Moreover, a new report indicates that Gli1⁺ mesenchymal stem cell (MSC)-like cells, which reside in the perivascular niche of many organs, substantially contribute to myofibroblasts and injury-induced organ fibrosis [26]. Interestingly, a novel population of Lin^-integrin-α7^-Sca1⁺PDGFRα⁺ multipotent mesenchymal progenitor cells, termed fibro/ adipogenic progenitors (FAPs), were recently identified [27]. These are the main source of myofibroblasts in injured skeletal muscle [28, 29], but are also found in multiple other tissues [30, 31].

Other cell types have also been proposed as precursors of myofibroblasts. Earlier reports suggested that epithelial cells undergoing EMT [21] or endothelial cells following EndoMT [32, 33] generate major populations of myofibroblasts. However, recent lineage tracing analyses refuted those assessments [1, 34]. While the origin of myofibroblasts from vascular SMCs has not been extensively studied, this paradigm has been suggested as an explanation for the pathogenesis of artery stiffness [35]. Fibrocytes, the circulating blood-borne cells similar to fibroblasts but expressing α-SMA and CD45, were reported to give rise to myofibroblasts [36]. However, using Collagen1a1-GFP as a robust cardiac fibroblast marker, a new report demonstrated that the majority of infarct fibroblasts are of epicardial origin but not derived from bone marrow lineages, EndoMT, or blood [37].

Despite many publications, the relative contributions of different cell sources to myofibroblasts and fibrosis have yet to be clearly defined in various organs or tissues [2]. Application of unambiguous lineage tracing technologies to distinguish the origins of myofibroblasts is an active direction of current research, and may successfully address this question [8, 38]. The answers will likely be contextually dependent on tissue and pathology [25], and precisely targeting a defined cell population for effective anti-fibrotic therapy may motivate extensive studies to delineate myofibroblastic mechanisms.

3. Myofibroblast-activating ECM Signaling

Defining the signal transduction that leads to myofibroblastic activation is inherently difficult due to the involvement of nearly all the major signaling pathways. Recent literature highlights a previously underappreciated role of ECM signaling as a primary mediator of myofibroblastic activation. While many signaling molecules are involved as myofibroblast activators, e.g. platelet derived growth factor and connective tissue growth factor [39, 40], TGF-β1 has been identified as mediating predominant impacts in most of the studies [41–43]. Therefore, we will focus on the TGF-β1 interactions with the ECM and its critical influence on myofibroblastic activation. Immune-mediated myofibroblastic activation involving TGF-β1 has been extensively investigated [44] but will not be included in the scope of this review.

3.1. Latent TGF-β1 “caged” in the ECM

Before maturation and activation, TGF-β is stored as a latent protein in the ECM, where it is bound to a latency associated peptide (LAP) that is conjugated to a latent TGF-β binding protein (LTBP) [45]. TGF-β1 and its pro-protein LAP-1 are translated from the same transcript. This pro-TGF-β1 dimerizes via disulfide bonding [46]. Subsequently, the linkage between LAP-1 and TGF-β1 is cleaved by a furin-like convertase and the LAP-1 dimer folds into a “straight jacket” that sterically trap the TGF-β1 dimer (Figure 1), resulting in the heterotetrameric small latent complex (SLC) [46]. The SLC is typically disulfide bonded with LTBP-1 at both LAP-1 monomers, forming the large latent complex (LLC), which is secreted and incorporated into the ECM. Four LTBPs exist in various tissues but with consensus high expression in elastic tissues such as heart, lung, and skeletal muscle [47]. They have different affinities to each of the three TGF-β proteins, and LTBP-1 is thought to sequester the most TGF-β1 in the ECM [48]. LTBP-2 is not known to conjugate to the TGF-β1 SLC; LTBP-3 and LTBP-4 have not been extensively studied for this function [49]. Aside from tethering the SLC to the ECM to mechanically secure TGF-β1’s latency [50], LTBP-1 may also serve a chaperone-like function to assist the folding and secretion of pro-TGF-β1 [51]. The LLC is conjugated to the ECM at LTBP-1 by transglutaminase [52]. The N-terminus of LTBP-1 is known to bind to FN [47, 49]. A recent study suggests that Fibrillin-1 binds to the C-terminal LTBP-1 at a site adjacent to the LAP-binding motif [53]. The Fibrillin-1 assembly into the ECM is FN dependent [48, 54]. The functional significance of TGF-β1 sequestration in the ECM is still not clear [45]. Studies provide interesting clues from Marfan syndrome, where Fibrillin-1 mutation leads to excessive TGF-β1 activation. This may be explained by insufficient Fibrillin-1 to keep TGF-β1 latent in the LLC, but the mechanisms can be far more complex [55].

3.2. Activation of TGF-β1

Compelling evidence indicates that the activation or “uncaging” of TGF-β1 from the ECM is a critical early step in driving myofibroblastic activation [42]. The release of mature TGF-β1 from the SLC can be triggered by a plethora of biological, physical, and physico-chemical factors (e.g. pH) [56], with integrins and proteases being well-established mediators [45]. The LAP protein contains an Arg-Gly-Asp (RGD) sequence which binds a number of integrins. All av integrins are known to bind RGD, with αvβ6 and αvβ8 as the best studied TGF-β1 activators [45]. A recent report demonstrates αvβ1 as another important TGF-β1 activator that binds LAP-1 [57]. The critical role of αv in myofibroblastic activation has been confirmed in several mouse models of organ fibrosis [57–59]. The literature to date on TGF-β1 activation involving αvβ6 is mostly focused on a traction-based mechanism [60]. As LAP-1 is anchored to the ECM via LTBP-1 and also bound to the cell surface through the binding of its RGD motif to αvβ6, a traction force between ECM and cells can deform the LAP-1 “straight jacket” thus liberating TGF-β1 [61] (Figure 1). Atomic structural information for this model has lately been updated [62]. This mechanism is also extrapolated to the “pulling” force generated by myofibroblast stress fiber contraction [63], or from ECM alone in a cell-free system [50]. There is evidence for an in vivo scenario where mechanical straining of the ECM primes the SLC (or LAP) for TGF-β1 activation triggered by cell contraction [64]. How αvβ8 activates ECM-caged TGF-β1 is less understood. Unlike αvβ6, the mechanism seems to involve recruitment of MMPs and protease activities independent of traction [65, 66].

Several proteinases are known to activate TGF-β1 without necessarily involving traction. MMP14 (aka: MT1-MMP) has been shown to degrade LAP-1 but is dependent on simultaneous cell-surface binding of the αvβ8 integrin to the RGD motif of LAP [67]. MMP-2 has been shown to directly cleave LAP-1 and LTBP-1 [68–70]. The role of MMP-9 is more complex, because it cleaves the soluble form but not ECM-bound form of LTBP-1 [71]. BMP1 has been shown to cleave LTBP-1 at two specific sites, liberating the LLC from ECM and resulting in activation of TGFβ1 after cleavage of LAP by non-BMP proteinases [69]. Serine proteinases, specifically plasmin, thrombin, elastase, chymase, trypsin, and kallikrein have been known for decades to cleave LTBP-1 and stimulate TGF-β1 activation [68, 72, 73]. However, the TGF-β1-activating mechanisms of these proteases are still not well understood in vivo in the context of myofibroblastic fibrosis. Thrombospondin-1, a non-protease protein, is well-documented in vitro and in vivo for its critical role in activating TGF-β1 and promoting fibrosis [74]. Its action likely involves an interaction with the N-terminal region of LAP-1 that alters the SLC conformation [75]. The other factors that turn TGF-β1 from latency to an active form have been updated by informative reviews [23, 45, 60].

3.3. The FN splice variant containing EDA (EDA-FN)

Based on studies using different precursor cells, three basic conditions are considered necessary for myofibroblastic activation [16], including strained ECM [76] and activated TGF-β1 [42, 77] as discussed above, and EDA-FN. Cellular but not plasma FN contains splice variants A (EDA or EIIIA) and B (EDB or EIIIB). EDA-FN is not expressed under normal conditions but is highly upregulated in myofibroblasts. While EDA-FN is recognized as necessary for myofibroblastic activation in various different precursors and fibrotic conditions in vitro and in vivo [78–82], EDB-FN is known as a hallmark of tumor angiogenesis [83]. It is worth noting that EDA-FN is reported to be dispensable for transformation of hepatic stellate cells and portal fibroblasts into myofibroblasts in culture and in mouse models of liver fibrosis [84], suggesting that the role of EDA could be compensated for by other factors.

It has been suggested that the insertion of EDA-FN into bulk fibronectin ECM may at a micro-level change overall material stress-strain properties [85]. The EDGIHEL motif at the EDA C’-C region binds to fibroblast-expressed integrins α4β1 and α9β1 [80, 86, 87]; the α4 and α9 subunits are considered the essential EDA-binding components [80, 88]. Another EDA-binding integrin, α4β7, has been found to be important in mouse lung myofibroblastic activation [89, 90], but the α4β7-binding EDA sequence has yet to be identified [89]. Of note, a recent report demonstrates that EDA-FN forms a signaling axis with Toll like receptor 4 (TLR4) driving fibrosis in scleroderma [81].

In addition to EDA and EDB, FN type III domains also contain RGD and PHSRN motifs. They have been shown to bind the canonical FN-recognition integrin α5β1 [91], as well as allbpi and α3β1 [92]. Interestingly, while the binding of LAP-1’s RGD with integrins facilitates TGF-β1 release from latency, as discussed above, the binding of FN’s RGD and its synergy site (the PHSRN motif) with integrins is thought to activate the integrin signaling that promotes FN filament assembly [92, 93]. However, whether this synergistic integrin-binding property of the RGD and PHSRN motifs plays an important role in TGF-β1 activation is not clear.

3.4. The TGF-β1/EDA-FN positive feedback

Activated TGF-β1 and EDA-FN positively affect and are affected by the cellular-mediated changes they induce in the ECM, essentially linking their functionality together in an ECM-cell feedback loop. It has been suggested that EDA-FN plays a positive role in activating TGF-β1signaling, possibly in multiple ways. First, there is direct evidence indicating that EDA plays a supporting role for the interaction of LTBP-1 with FN and its incorporation into the ECM [16, 64, 94]; blocking the EDA-FN/LTBP-1 interaction impairs TGF-β1 activation [16, 95]. The detailed molecular mechanism has yet to be revealed. Second, the EDA/integrin interaction has been reported to enhance the interaction of the LAP-1 RGD with integrins [96], implicating a role in facilitating the release of TGF-β1 from the ECM. In addition, in view of a network formed by protein-protein interactions in the LAP-1/LTBP-1/EDA-FN/integrin complex, it is an intriguing question as to whether perturbation of ECM may modify the LAP-1 conformation and hence the TGF-β1 activation status via the EDA-FN/integrin binding. Given the complexity of this ECM network and technical challenges, it will take careful future studies to delineate this potential mechanism. Moreover, further studies are also required to elucidate whether the EDA-FN binding with integrins assists with the incorporation of latent TGF-β1 into the ECM.

Interestingly, it has been shown that TGF-β1 directly regulates EDA-FN expression during myofibroblastic activation. For example, TGF-β1 has been identified as a modulating factor in EDA-FN splicing. Exposure of lung fibroblasts to TGF-β1 modulates serine-arginine-rich splicing factors (SRps), which influence the splicing of EDA into the cellular FN transcript. One study showed that active SRp40, SRp55, and SRp75 increased threefold and SRp20 increased eightfold after 24-hour TGF-β1 exposure. Additionally, several U5 proteins and helicases were modified. These changes modified EDA inclusion into the FN transcript [97]. This effect was further explored in a recent publication, where it was shown that TGF-β1 stimulates SRp40 via the PI3K-Akt pathway, which will be discussed later in this review. Akt was shown to directly bind SRp40, which bound the EDA-FN exon splicing enhancer on pre-mRNA and significantly increased the proportion of total cFN containing an EDA domain [98]. The observation in both of these studies that TGF-β1 mediates cellular redistribution of SRp proteins is intriguing and potentially worth further study, along with the signaling pathways governing this TGF-β1 guided induction of EDA-FN.

Due to the extracellular availability and high expression in fibrosis, EDA-FN has been used clinically as an indicator of fibrotic disease and tissue remodeling in humans [99]. Moreover, EDA function-blocking antibodies and competitive peptides inhibit activation of TGF-β1 and myofibroblasts [16, 100]. As such, the relationship of TGF-β1 and EDA-FN signaling pathways will continue to attract research interest directed toward the identification of new mechanisms for therapeutic targeting.

4. Signal Transmission and Integration at the Cell Membrane

At the plasma membrane, the TGF-β receptor complex is activated by TGF-β1 that is released from the ECM; Integrins interact with EDA-FN and the ECM in a mechanoreceptive manner. These two major events integrate to transmit ECM signals into the cell resulting in RhoA-stimulated stress fiber formation and transcriptional/ posttranscriptional reprograming that orchestrate myofibroblastic activation.

4.1. TGF-β1 signaling via TGF-β receptors

In the canonical pathway, TGF-β1 activates the TGF-β receptor complex which in turn phosphorylates Smad2 and Smad3. These two signaling proteins then complex with Smad4 and translocate to the nucleus to initiate transcription of many genes [46, 101]. The Smad2/3 signaling is known to elevate the expression of ECM-remodeling proteins such as collagen I and tissue inhibitor of metalloproteinases 2 (TIMP2)-an important regulator of MMP activity [102]. However, the role of this canonical TGF-β pathway in myofibroblastic activation was not confirmed in vivo until last year [103]. This study demonstrated that cardiac fibroblast-specific Smad2/3 deletion in mice inhibits the fibrogenic gene program and ECM remodeling [103]. Myofibroblastic transcription can also be activated through non-Smad pathways such as ERK and p38 [104, 105]. Revealing a crosstalk between canonical and non-canonical TGF-β pathways, a most recent report showed that TGF-β1 induced human lung fibroblast activation through the Raf1/ERK/Smad pathway [106]. Given the intricate TGF-β1 downstream signaling networks, the pathway(s) that mediates TGF-β1-induced myofibroblastic activation is likely tissue/signaling context dependent [107].

Since TGF-β1 activates intracellular signaling via TGF-β receptors, regulation of these receptors has profound influence on myofibroblastic activation [4]. TGF-βR endocytosis and degradation via clatherin-dependent or caveolin-1 dependent processes have been well established [108]. A relatively new thread of knowledge is that the TGF-βRI and TGF-βRII activities are regulated by their dissociation and association. In a recent report whereas TGF-βRI was found to be selectively enriched at focal adhesions, TGF-βRII was selectively excluded [109]. Reduction in Rho/ROCK signaling, as mediated by decrease in cellular tension, diminished this segregation and allows the formation of complete TGF-βRI/TGF-βRII complexes. Further, the av integrin subunit was shown to form a tension-mediated complex with TGF-βRI, TGF-βRII, and cofilin, although the significance of this regulation remains elusive. Taken together, these results suggest that the TGF-βR complex is modulated in part by mechanotransduction and that specific integrins, such αv -that are known to activate ECM-tethered TGF-β-may have multifaceted roles in the activation of TGF-β1 signaling.

4.2. Integrin signaling through focal adhesions

The ability for diverse isoforms of integrins to cluster and to mechanotransduce ECM mechanical, topographical, or motif-directed cues is primarily mediated by the formation of focal adhesions (FAs) [13, 110]. These structures were previously described as maturing by gaining increasing mechanical strength and α-SMA incorporation throughout the morphological development of myofibroblasts. Focal adhesions are dynamic structures known to consist of over 150 proteins including focal adhesion kinase (FAK) and Src, another tyrosine kinase. FAK activation is essential for mechanotransduction of integrin signaling during myofibroblastic activation. Activated integrins allow for autophosphorylation of FAK and its binding with Src, which results in mutual activation of the kinase domains of both molecules [111, 112]. They link clusters of integrins at the plasma membrane to the cytoskeleton and mechanoreceptive cytosolic signaling factors.

Focal adhesion biology is extremely complex. Focal adhesion mechanotransduction of myofibroblast pathways has been extensively investigated yet the inner workings of their subcomponents are not well understood [113]. The canonical integrin-FAK-Src signaling and downstream RhoA-activating pathways have been frequently reviewed [114–116]. Therefore, we will only highlight Hic-5, a paxillin family member that has recently garnered new attention as an essential component of myofibroblastic activation [117, 118]. Otherwise known as TGF-β1-induced transcript 1 (TGFB1H) [119], Hic-5 is induced by TGF-β1 and also induces TGF-β1 transcription [120] and regulates the coupling between focal adhesion activities and ECM degradation [121]. The Hic-5 protein shuttles between focal adhesions and the nucleus [122] in a process at least partially dependent on cellular mechanical stress [123, 124]. While knockdown of Hic-5 inhibits myofibroblastic activation, its precise regulatory role is only beginning to be unveiled. Recent studies show that Hic-5 transcription is directly regulated by and enhances myocardin-related transcription factor A (MRTF-A) nuclear accumulation [125], a downstream factor induced by non-canonical TGF-β1 signaling and mechanical stress (to be discussed later). Further, it appears that both MRTF-A and Hic-5 are required for the induction of α-SMA and formation of super-mature focal adhesions. MRTF-dependent Hic-5 expression is also regulated in a Smad3-dependent manner, although the exact mechanism is still unclear [117]. Signaling pathways connecting Hic-5 to α-SMA expression have yet to be elucidated, but are likely dominated by Hic-5 synergizing with extracellular stress to promote RhoA activation. Hic-5 may also affect myofibroblastic activation indirectly through modulating steroid receptor transcription [122, 126].

4.3. Crosstalk between focal adhesions and TGF-α signaling

The activities of integrins/focal adhesions have been shown to engage in crosstalk with TGF-β1 signaling. It is known that FAK-activated Src facilitates the activation of phosphoinositide 3-kinase (PI3K), a lipid inositol-modifying heterodimer that phosphorylates PIP2 into PIP3 [127] [128]. This further activates the downstream Akt1/mTOR pathway [129–131]. The phosphatase PTEN abolishes PI3K signaling by catalyzing the PIP3 conversion back to PIP2. TGF-β1-dependent alternative splicing of EDA-FN has been shown as dependent on the PI3K activation of Akt, which binds to the EDA-FN-inducing enzyme SRp40 [98]. As activated Akt1 potently induces EDA-FN production in addition to total FN [132], it appears possible that EDA-FN induces its own expression via FAK-PI3K-Akt activation mediated by the EDA/integrin interaction, thus resulting in a matrix-mediated positive feedback loop. While TGF-β1 modulates the PI3K-Akt pathway via FAK activation [133], its inhibition of PTEN has also been shown as necessary for both TGF-β1-induced α-SMA expression and EDA-FN production [134]. However, this interaction itself inhibits Smad-based TGF-β1 signaling, as decreased PTEN results in increased Smad7 which promotes TGF-βR degradation [22, 135].

The FAK/Src axis has also been linked to TGF-β1 signaling in stress fiber formation of myofibroblasts via TGF-βRs. Sequestering of Src by FAK decreases the activation of dynamin, a protein essential for endocytosis [136], and may hence reduce TGF-βR degradation [137]. Additionally, it has been shown that TGF-β1 signaling results in TGF-βR/Src-dependent association of RhoA and calveolin-1, activating RhoA, a necessary signaling step for TGF-β1-induced fibronectin synthesis [138]. Another recent study has shown that Src activation is an essential component for facilitating the profibrotic effects of TGF-β1, including the expression of α-SMA and the formation of stress fibers [139].

Interestingly, TGF-β1stimulation can robustly upregulate Pyk2 expression that leads to EMT [140]. Since Pyk2 is an FAK homolog that enhances RhoA activation [141], it is possible that TGF-β1 signaling may activate RhoA in an FAK-independent manner. Independent and compensatory regulations of downstream RhoA effectors by TGF-β1 and FAK signaling may provide an explanation for the ability of either TGF-β1 or EDA-FN to fully and independently induce myofibroblastic activation in vivo.

5. Intracellular Signaling

A cascade of signaling events downstream of integrin/FAK/Src involving ERK and MAPK pathways are known to induce myofibroblastic activation [142, 143]. Moreover, FAK and Src facilitate the RhoA GTPase activation by regulating guanine nucleotide exchange factors (GEFs) and GTPases-activating proteins (GAPs) in response to mechano-stimulation [144]. Updates on these pathways related to myofibroblasts have been covered by excellent recent reviews [143, 145, 146]. Therefore, we will focus on activated RhoA and downstream pathways responsible for α-SMA stress fiber production and associated nuclear regulations.

5.1. RhoA signaling in actin stress fiber formation

Rho GTPases have been shown to control the assembly of F-actin in all eukaryotic species examined to date. The best-characterized Rho-like proteins are RhoA, Rac1, and Cdc42. We will focus on RhoA because its activation is primarily mediated by mechanotransduction factors and is a specific requirement in the induction of α-SMA stress fiber formation in myofibroblastic activation [147]. RhoA activates the signaling molecules ROCK and mDia. ROCK phosphorylates and inactivates the myosin light chain phosphatase (MLCP) and can directly phosphorylate and activate the myosin light chain. Both of these ROCK actions facilitate the activation of non-muscle myosin II at focal adhesions [148, 149]. Additionally, ROCK further activates LIM kinases, which phosphorylate and inhibit the actin-cleaving enzyme cofilin, thereby decreasing actin degradation and while increasing the torsional rigidity of existing actin filaments [150]. While ROCK induces stability and myosin accumulation, mDia acts in a VASP-dependent manner to polymerize actin and allows the formation of lengthy filaments [151, 152]. These molecules balance the activity of one another as mDia induces a Src-dependent Cas/Crk/DOCK180 cascade resulting in decreased ROCK activity [153, 154]. Together, ROCK and mDia pathways interact to induce the formation of stress fibers via actin polymerization.

Several reports have shown that TGF-βRI phosphorylates and activates Par6. Subsequently, Par6 can dimerize with Smurf1, and the resulting complex can target RhoA at tight junctions for its ubiquitination and degradation. This process facilitates the dissolution of tight junctions and has been determined to be necessary for TGF-β1 induced EMT [155, 156]. As this process contradicts the necessity for RhoA-induced α-SMA stress fiber formation, the regulations of RhoA in EMT and in myofibroblastic activation are likely spatially and/or temporally separate processes.

5.2. MRTF-directed nuclear signaling

It has been well documented that the activation of serum response factor (SRF), a transcription factor, is essential for the transcription program that drives myofibroblastic activation [157]. RhoA activation facilitates the nuclear translocation of SRF co-activators, with MRTFs (A and B) being prominently important for instigating myofibroblastic phenotypes [158].

Factors known to directly interact with SRF include Myocardin (MyoCD), MRTFs, Elkl, β-catenin, YAP (Yes-activating protein)/TAZ (transcriptional co-activator with PDZ binding motif), and Smad3. MyoCD is an SMC and cardiac muscle specific SRF cofactor required for significant activation of SRF-directed transcription and is constitutively localized to the nucleus. Acting in a similar fashion as MyoCD, tissue ubiquitous MRTF is a cofactor for SRF activation. However, unlike MyoCD, MRTF is typically localized in the cytoplasm and bound to G-actin in an inactive state. RhoA stimulates the assembly of actin stress fibers thereby reducing the G-actin/F-actin ratio. Consequently, MRTF is freed from G-actin sequestration and is allowed to translocate to the nucleus, where it activates SRF-mediated transcription of myofibroblastic factors such as α-SMA (Figure 1). Elk1 acts as an inhibitor of MRTF, through direct antagonistic SRF binding or through operator binding adjacent to the SRF-binding CArG motif.

Recent research reveals an interplay between canonical TGF-β1 signaling and MRTF in mechanotransduction [147], e.g. nuclear Smad3 recruits GSK-3β to phosphorylate MRTF leading to subsequent ubiquitination and degradation of MRTF [159]. Regulations of MRTF stability also link the Wnt/β-catenin and YAP/TAZ pathways to TGF-β1 signaling in the nucleus, as discussed below.

5.3. MRTF-regulatory nuclear factors

Wnt cytokines act canonically through the Frizzled receptor complex and activate the scaffolding protein Disheveled, which degrades Axin thereby inhibiting constitutive destruction of β-catenin, which can then translocate to the nucleus [160]. Nuclear translocation of β-catenin has been linked to increased expression of TGF-β1 in addition to α-SMA [157, 161–164].

The YAP/TAZ pathway has been recognized as sensitive to mechanotransduction. Inhibition of YAP/TAZ signaling blocked fibroblast-to-myofibroblast transformation and renal fibrosis in mice [165]. Similar to β-catenin, inhibited YAP/TAZ degradation promotes their translocation to the nucleus where they complex with the transcription factor TEAD to co-activate transcription [166–168]. There is evidence that cellular tension increases YAP/TAZ nuclear concentrations [5]. Interestingly, while RhoA and the actin cytoskeleton are thought to be necessary to maintain YAP/TAZ activation, this complex does not seem to be directly regulated by G-actin/F-actin levels, implicating either upstream RhoA interaction or stress fiber interaction with an unknown mediator [169, 170].

Another pathway recently implicated in myofibroblastic activation involves zinc finger E-box-binding homeobox 2 (ZEB2) protein [171], which is a transcription factor and constitutive repressor of Meox½. Reduced expression of Meox½ is associated with increased expression of EDA-FN and α-SMA during myofibroblastic activation [172].

Studies suggest that these major nuclear factors may all crosstalk with the TGF-β1/Smad3 pathway at the resolution of nuclear MRTF stability, which is highly regulated. While nuclear Smad3 has been shown to promote MRTF-A degradation via GSK-3β-mediated phosphorylation followed by ubiquitination, this action is inhibited by nuclear β-catenin [159, 160, 173]. Moreover, nuclear-localized YAP/TAZ bind MRTF-A and enhance its activity while down-regulating Smad3 [163, 174]. The mechanisms by which these functions occur have yet to be clearly addressed. Despite Smad3 nuclear inhibition of MRTF, TGF-β1 has been shown to increase nuclear MRTF concentration as mentioned above. Whether this TGF-β1 action is executed via Smad or non-Smad pathways is unclear. The context-dependent role of TGF-β1 signaling in MRTF regulation warrants further investigation. In addition, the TGF-β1/Smad3 pathway inhibitor, Ski, has been found to bind and repress ZEB2 thereby derepressing Meox2 and suppressing myofibroblastic activation [175]. Though contradicting a preconceived inhibitory effect on myofibroblasts, this observation implicates an emerging significance of Ski in stimulating myofibroblasts. Whether Ski and ZEB2 directly regulate MRTF in the myofibroblastic context remains an open question.

5.4. Epigenetic regulation

Beyond aforementioned classical pathways, interest in epigenetic regulators, including DNA and histone modifiers and readers of the modifications [176, 177], microRNAs [178], and long noncoding RNAs [179], is rapidly growing. These chromatin-associated factors exert profound influence on transcriptomic dynamics and cell state transition without involving DNA sequence changes [14]. It has been demonstrated that both genetic and pharmacological inhibition of DNMT (DNA methyltransferase) mitigates mouse renal myofibroblasts and fibrosis [180, 181]. Inhibitors of HDACs have shown anti-fibrotic effects in several mouse models [182]. Whereas miR29a mitigates [183], miR21 aggravates [184] TGF-β1-induced myofibroblastic programs, both expected as potential interventional targets [182, 185]. Interestingly, there is also evidence suggesting that miR29a is a possible mediator of the EDA-FN-driven fibrosis in scleroderma [81]. More recently, the bromo and extraterminal domain (BET) family of epigenetic readers (binding to acetyl marks on histones) have attracted significant attention due to their potentially crucial role in regulating myofibroblastic activation [186–188]. It has been observed that pharmacological blockage of the BET family effectively inhibits fibrosis in the liver, kidney, and heart [187, 189–191].

How the epigenetic regulators cooperate with the aforementioned nuclear factors in fibrogenic pathways is an intriguing question that is soliciting growing research. As opposed to traditional perception that epigenetic factors regulate transcription globally, accumulating evidence suggests significant specificity of those regulations depending on cell type, extracellular cue and signaling context [192]. This feature may open a way for future epigenetic interventions to mitigate fibrotic progression with careful assessments of possible side effects [182, 190].

6. Other Myofibroblastic Regulators

Aside from SMA and EDA-FN, many other products resulting from myofibroblastic nuclear reprogramming in turn modulate the ECM architecture and signaling. They participate in the inside-out signal transmission that further accentuate myofibroblastic activation. Given a large number of those products, we will only highlight a few that have recently re-energized myofibroblast research.

6.1. Hyaluronic Acid

Evidence suggests that the ECM glycosaminoglycan hyaluronic acid (HA) and its receptor CD-44 are involved in myofibroblastic activation [193]. It has been shown that high molecular weight HA inhibits fibrotic pathways but low molecular weight HA appears to induce them [194]. Interestingly, naked mole rats express unique HA molecules that are five times the length of those in humans or mice, and they have seven-fold longer domestic lifespan than mice [195]. While the large HA confers cancer resistance in naked mole rats, the relationship between unique HA molecules and lifespan is not clear [196]. High molecular weight HA has been shown to induce CD-44-assisted translocation of activated TGF-βR to caveolin-1 lipid rafts, facilitating the previously discussed Smurf1/Smad7-mediated TGF-βR degradation [197]. However, inhibition of HA synthase (HAS) also antagonizes TGF-β1-induced expression of α-SMA [198]. Moreover, TGF-β1-mediated HAS2 activation results in some aspects of myofibroblastic activation [193]. These results seem to contradict the previously observed inhibition of myofibroblasts by the HA-CD-44 signaling [197]. Alternatively, as a viscous molecule, HA may also regulate myofibroblastic activation through its interactions with the ECM [199].

6.2. Matricellular proteins

Matricellular proteins reside in the ECM but do not serve a primary structural role [200]. They modulate the activities of proteases, cytokines/growth factors, and cell-surface receptors [201]. Regarded as a new myofibroblast marker in lineage-tracing studies [3], periostin prevents ventricular rupture but also contributes to myofibroblast activation and cardiac fibrosis [202, 203]. Secreted modular calcium-binding protein 2 (SMOC2) is a member of the secreted protein acidic and rich in cysteine (SPARC) family, and was recently found to promote myofibroblast-mediated fibrosis in the mouse retina [204]. Syndecans comprise a family of transmembrane proteins, their extracellular domains interacting with the ECM. Whereas beneficial effects of over-expressing Syndecan-4 have been found in preventing mouse lung [205] and cardiac [206] fibroblast-to-myofibroblast transition and subsequent fibrosis, opposite effects in mouse renal fibrosis have also been reported [207], likely involving its ectodomain shedding [208]. While more detailed mechanistic studies are required, these and other new findings exhibit an attractive potential of matricellular proteins as new interventional targets in the battle against fibrosis.

7. Conclusion

TGF-β1 signaling is generally regarded as the most important activator of myofibroblasts [6, 42, 43]. Latent TGF-β1 is stored in the ECM, and its release and binding to TGF-βRs transmit ECM signaling cues into the cell. This incites transcriptional and posttranscriptional reprograming (e.g. EDA-FN splicing) and cytosolic remodeling (e.g. RhoA activation and stress fiber assembly). These orchestrated events in turn lead to increased ECM accumulation and mechano-responsive activities, and ultimately, further TGF-β1 activation via the integrin/ECM interaction network. Therefore, TGF-β1-activated pathways in conjunction with mechanotransduction constitute outside-in and inside-out signaling loops that potently propagate myofibroblastic activation. As myofibroblasts mediate fibrosis in a broad range of pathologies, methods to break these vicious cycles may result in viable therapeutic options. We have therefore reviewed recent literature on myofibroblastic signaling mechanisms with a primary focus on TGF-β1 and associated pathways in the context of ECM/cell signaling integration. While studies have revealed several ECM-cell-ECM and intracellular positive feedback loops, antagonizing mechanisms have also been noted. In the myofibroblastic intracellular signaling network, RhoA is a cytosolic hub protein that perpetuates α-SMA stress fibers. This RhoA effect is converted into transcriptional reprogramming via nuclear translocation of MRTF-A, another hub protein that integrates the modulations from other nuclear factors including Smad3, β-catenin, and YAP/TAZ. Encouraging therapeutic progress has recently been achieved in preclinical models by targeting the major myofibroblast pathways to ameliorate fibrosis [10, 13, 17, 41, 57, 209]. More in-depth investigations into the myofibroblastogenic mechanisms in vitro and in vivo are required. New knowledge should help create effective anti-fibrotic therapies by precisely targeting the signaling vicious cycles that are propagated by defined myofibroblast or precursor populations.

Abbreviations

α-SMA

α-smooth muscle actin

BET

bromo and extraterminal domain containing epigenetic readers

ECM

extracellular matrix

EDA-FN

extra domain A-containing fibronectin

EMT

epithelial-to-mesenchymal transition

FA

focal adhesion

FAK

FA Kinase

FN

fibronectin

GEF

guanine nucleotide exchange factor

HA

hyaluronic acid

HAS

hyaluronic acid synthase

LAP

latency associated peptide

LLC

large latent complex

LTBP

latent TGF-β binding protein

MRTF

myocardin-related transcription factor

MSC

mesenchymal stem cell

MyoCD

myocardin

RGD

the tripeptide sequence of Arg-Gly-Asp

SLC

small latent complex

SMC

smooth muscle cells

SRF

serum response factor

TGF-β

transforming growth factor

TGF-βR

TGF-β receptor

TAZ

transcriptional coactivator with PDZ binding motif

YAP

Yes-activating protein

ZEB2

zinc finger E-box-binding homeobox 2.