A Smurf1 tale: function and regulation of an ubiquitin ligase in multiple cellular networks

Abstract

Since being discovered and intensively studied for over a decade, Smad ubiquitylation regulatory factor-1 (Smurf1) has been linked with several important biological pathways, including the bone morphogenetic protein pathway, the non-canonical Wnt pathway, and the mitogen-activated protein kinase pathway. Multiple functions of this ubiquitin ligase have been discovered in cell growth and morphogenesis, cell migration, cell polarity, and autophagy. Smurf1 is related to physiological manifestations in terms of age-dependent deficiency in bone formation and invasion of tumor cells. Smurf1-knockout mice have a significant phenotype in the skeletal system and considerable manifestations during embryonic development and neural outgrowth. In depth studying of Smurf1 will help us to understand the etiopathological mechanisms of related disorders. Here, we will summarize historical and recent studies on Smurf1, and discuss the E3 ligase-dependent and -independent functions of Smurf1. Moreover, intracellular regulations of Smurf1 and related physiological phenotypes will be described in this review.

Introduction

There is a growing interest in the significant role ubiquitylation plays in benefiting or undermining regular biological processes. Ubiquitylation is believed to regulate the physiological and metabolic homeostasis of an organism, while abnormalities and deficiencies in this process are related to a broad range of diseases, including malignancies, neurodegenerative disorders, immune disorders, and inflammation [1, 2]. During ubiquitylation, one or more ubiquitins (Ub) are sequentially delivered though ubiquitin-activating enzyme (E1), ubiquitin-conjugating enzyme (E2), and ubiquitin-protein ligase (E3), and finally tagged onto the specific lysine residue of the targeted protein [3–5] (Fig. 1a). This tagged prey is then degraded by the proteasome or undergoes other biological processes such as changing its subcellular localization (Appendix 1). More than 600 E3s have been identified and they are mainly classified into RING (really interesting new gene) and HECT (homologous to E6AP C-terminus) types [6–8] (Appendix 2). Smad ubiquitylation regulatory factor-1 (Smurf1) has been identified as a HECT type E3, and, based on its structure, it belongs to the Nedd4 family. Having been studied for more than 10 years, Smurf1 has been related to multiple biological processes such as cell growth and migration, and explored for several physiological functions in bone formation, embryonic development, and tumorigenesis.

Fig. 1.

a Ubiquitylation cascade. The conjugation of ubiquitin molecule to a substrate is processed though E1, E2, and E3. b Domain architectures of all Nedd4 E3 members in human species

Smurf1 emerged early during metazoan evolution, and is expressed in all developmental stages from embryo to mature adult, and in many organs including bone, cartilage, heart, lung, nervous system, and genital organs [9]. Smurf1 contains a catalytic HECT domain in the C terminal, two WW domains (WW1 and WW2), and a phospholipid-binding C2 domain in the N terminal region (Fig. 1b). A point mutation in the HECT domain, changing Cys699 into Ala, abolishes the catalytic activity of Smurf1. Cys699 is a crucial site in the C lobe of the HECT domain, which forms a thioester bond with the Ub molecule. Although a C699A mutant fails in ubiquitylation, it still retains the binding ability to its substrates or adaptors and thus can be used to examine Smurf1-interacting proteins. The HECT domain interacts with both E2 and Ub molecules during ubiquitylation, and in the case of Smurf1, it may also interact with the C2 domain of another Smurf1 for autoinhibition [10]. The C2 domain can bind to phospholipid-containing membranes and alter both subcellular localization and function of Smurf1. Smurf1 uses the WW domains to capture prey which contain the PY motif. Phosphorylation on specific sites of certain substrates or adaptors may be required to increase the binding affinity of the WW–PY interaction [11]. In the BMP pathway, Smad1 phosphorylation by CDK8/9 creates binding sites for the WW domains of YAP, and subsequent phosphorylation by GSK3 switches off YAP binding and adds binding sites for Smurf1 WW domains [11]. In some cases, the C2 domain and the HECT domain can also act as capturers of certain substrates [12–14]. For example, Smurf1 recognizes RhoA and hPEM-2 through its C2 domain [12, 13], whereas it targets ING2 for ubiquitylation through the HECT domain [14]. Besides direct interaction, Smurf1 also targets its prey by recruiting adaptors, which will be discussed later.

E3 activities and Smurf1 substrates in various biological processes

Historical identification of Smurf1 can be traced back to more than 10 years ago. Zhu and colleagues discovered that this new HECT ubiquitin ligase reduces cellular responses to bone morphogenetic protein (BMP) via triggering proteasomal destruction of Smad1 and Smad5, causing impaired BMP signal transduction and aberrant embryonic development in Xenopus laevis [15]. Subsequent investigations identified Smurf1 substrates not only in the BMP pathway but also in many other pathways (Fig. 2) and biological processes, including osteoblast differentiation, embryonic development, cell polarity, cell adhesion and migration, viral autophagy, and immune responses (Table 1).

Fig. 2.

Role of Smurf1 in multiple biological networks

Table 1.

Substrates of Smurf1

Substrate	Description and function	Interacting mode	References
Smad1	A receptor-regulated Smad, can be activated by BMP type1 receptor kinase and mediates BMP signal	Direct PY–WW interactiona	[15]
Smad4	The common Smad in both TGF-β and BMP pathways, mediates the signal downstream of Smad2/3 or Smad1/5, respectively	By adaptorb: Smad2, Smad6, or Smad7	[16]
Smad5	A receptor-regulated Smad, has similar function with Smad1	Direct PY–WW interaction	[15]
Smad7	A inhibitory Smad, acts negatively in both TGF-β and BMP pathways	Direct PY–WW interaction	[17]
RunX2	A transcription factor involved in osteoblastic differentiation and skeletal morphogenesis	Direct PY–WW interaction	[25, 26]
RunX3	A transcription factor acts in multiple pathways such as osteoblastic differentiation and is known as a tumor suppressor	Direct PY–WW interaction	[26]
MEKK2	A member of the serine/threonine protein kinase family. It activates other kinases in the MAP kinase signaling pathway	Direct PY–WW interaction	[28]
JunB	A member of the AP-1 family. A transcription factor downstream of the primary growth factor response	Direct PY–WW interaction	[29]
Tbx6	A member of the T-box gene family. A transcription factor in BMP pathway and acts importantly in embryonic development	By adaptor: Smad6	[27]
BMP receptor I and II	Type I and II receptors in BMP signaling. Similar to TGF-β receptor I and II	By adaptor: Smad6, Smad7	[19]
TGF-β receptor I	Type I receptor in TGF-β signaling, is activated by type II receptors upon ligand binding and mediates the signal to Smad proteins	By adaptor: Smad7	[18]
RhoA	A member of Rho family small GTPase, acts in multiple signaling pathways and regulates actin stress fibers, cell focal adhesion and cell polarity	By adaptor: Par6; or interacts with C2 domain	[32]
PEM-2	A guanine nucleotide exchange factor for cdc42, controls cell migration by regulating cdc42	Interacts with C2 domain by its PH domain	[13]
Par6	A component of the PAR polarity complex. It is important for cell polarity, epithelial tight junctions and cell transformation	Direct interactionc	[35]
Talin head	Head region of Talin protein which works essentially in integrin-mediated cell adhesion	Direct interaction	[36]
Prickle1	A component in the non-canonical Wnt signal pathway. It regulates planar cell polarity, cell convergence and extension movements in neurons	By adaptor: Par6 and Dishevelled2	[37]
TRAF1	The first member of TNF receptor associated factor (TRAF) family. It mediates anti-apoptotic signals from TNF receptors	Unknownd	[38]
TRAF2	A member of TRAF family, forms complexes with other TRAFs and plays a role in TNF-mediated MAPK/JNK and NF-κB activation	Unknown	[38]
TRAF3	A member of TRAF family, participates in TNF-CD40 signal transduction and lymphotoxin-β receptor-mediated NF-κB activation	Unknown	[38]
TRAF4	A member of TRAF family, functions in TNF receptor-mediated and Toll-like receptor-mediated signaling pathways	Direct PY–WW interaction	[38]
TRAF5	A member of TRAF family. It is required in TNF-induced signal activation and CD40 signaling	Unknown	[38]
TRAF6	A member of TRAF family, mediates multiple signals from IL-1, Toll-like, CD40, RANK and p75 NGF receptors	Unknown	[38]
MyD88	A core cytosolic adaptor protein in all Toll-like receptor signaling except TLR3	By adaptor: Smad6	[41]
STAT1	A transcription factor and a key component of Jak-STAT pathway, regulates transcription of type I interferon target genes	Direct PY–WW interaction	[42]
KLF2	A transcription factor, plays essentially in lung development, T cell differentiation and migration	Interacts with WW by both ID and ZF domains	[43]
WFS1	An endoplasmic reticulum (ER)-localized protein, plays important role in maintaining ER homeostasis	Direct interaction with Smurf1 WW domains	[44]
ING2	A member of inhibitor of growth family, participates in multiple processes such as DNA damage repair and chromatin remodeling	Interacts with HECT domain by its PHD domain	[14]

^aInteraction between Smurf1 WW domains and PY motif on the substrate

^bSmurf1 uses an adaptor protein to indirectly bind to the substrate

^cDirect interaction between Smurf1 and the substrate, but the binding regions are unclear

^dThe substrate can be ubiquitylated by Smurf1, but the interacting mode has not been determined

Cell growth and differentiation

The transforming growth factor-β (TGF-β) and BMP signaling pathways play a prominent role in a plethora of biological processes during cell development, growth, and differentiation. Two classes of cytokines, TGF-β/activin/Nodal and BMPs, bind to and activate TGF-β and BMP receptors, respectively, leading to the activation and release of receptor-regulated Smads (R-Smads): Smad1, Smad5 ,and Smad8 for BMP signaling, and Smad2 and Smad3 for TGF-β signaling. Phosphorylated R-Smads form a heteromeric transcription complex with Smad4, the common-mediator Smad (co-Smad); this complex is then transported into the nucleus and initiates downstream gene expression. This signaling cascade is terminated or blocked by the two inhibitory Smads (I-Smads), Smad6, and Smad7. Unlike its homolog Smurf2, which only functions in the TGF-β pathway, Smurf1 plays an important role in both TGF-β and BMP pathways. Smurf1 ubiquitylates and degrades Smads at all levels of TGF-β and BMP signaling cascades, including Smad1, Smad4, Smad5, and Smad7 [15–17]. Smurf1 interacts with Smad1/5 directly though the WW–PY interaction, triggering their proteasomal degradation. I-Smads can interact with TGF-β and BMP receptors, serving as scaffold proteins for Smurf1 in the ubiquitylation process. Using Smad6 as an adaptor, Smurf1 indirectly targets type I and II BMP receptors and type I TGF-β receptors for degradation, while Smad7 can help Smurf1 trigger proteasomal degradation of type I TGF-β receptors [18, 19]. Smad7 itself can also be ubiquitylated by Smurf1, a process that can be rescued by the acetylation of Smad7 [18, 20]. In addition, Smurf2 and other Nedd4 family members have been reported to have various regulatory functions in the TGF-β and BMP pathways [16, 21–24].

To regulate cell growth and differentiation, Smurf1 also has other substrates in the TGF-β and BMP pathways, including RunX2, RunX3, Tbx6, MEKK2, JunB, and TRAF4 [25–30]. RunX2 is an osteoblast-specific transcription factor, downstream of the BMP signal, which has been verified to be a target of Smurf1 [25]. Although RunX2 contains a PY motif in its C terminal region, Smurf1 interacts with RunX2 not only via the WW–PY interaction but also by recruiting Smad6 as an adaptor [26]. Similarly, Smurf1 can also target RunX3 for ubiquitin–proteasome degradation. Tbx6 (T-box 6), another downstream transcription factor of the BMP signaling, is a Smurf1 target. This ubiquitylation process requires the bridging protein, Smad6 [27]. MEKK2, a PY-motif-containing protein kinase, can be ubiquitylated by Smurf1 only when it is phosphorylated. The MEKK2–JNK cascade is involved in the MAPK (mitogen-activated protein kinase) pathway, which displays several crosstalks with the TGF-β and BMP signaling pathways [28]. JunB, a member of the AP-1 family, was recently identified as a Smurf1 substrate in mesenchymal stem cells [29]. By degrading these substrates, Smurf1 controls players at all levels of the TGF-β and BMP signal transduction. The tumor necrosis factor (TNF)-receptor-associated factor-4 (TRAF4) has been found to be targeted by Smurf1 for degradation through the WW–PY interaction [30]. TRAF4 can upregulate TGF-β signaling in Xenopus embryo and control nodal signaling, which is essential for neural crest development and morphogenesis [31].

Cell adhesion, polarity and migration

Beyond cell growth and development, broader functions of Smurf1 have been identified within cell polarity and migration. Cell adhesion, polarity, and migration are important aspects of many physiological processes, including embryonic development, neurodevelopment, and cancer metastasis. In the Cdc42-Par6-PKCζ pathway, Smurf1 triggers degradation of RhoA, a small GTPase, in cellular protrusions [32]. Smurf1 recognizes and ubiquitylates GDP-bound RhoA, suggesting that Smurf1 targets RhoA in a guanine nucleotide exchange factor (GEF)-dependent manner. To target RhoA, Smurf1 requires the plasma membrane-attaching capacity of its C2 domain, and may also need the phosphorylated Par6 adaptor, which is believed to be in charge of cell adhesion and cell polarity [33, 34]. Moreover, a recent study revealed that Smurf1 can bind to RhoA directly by the C2 domain [12]. Intriguingly, in addition to RhoA, the adaptor protein Par6 is also a substrate of Smurf1 in neurons [35]. Via the C2 domain, Smurf1 can bind to another substrate, hPEM-2, a GEF for Cdc42, for proteasomal degradation [13]. Talin, an actin and integrin binding protein, is normally cleaved into a head domain and a rod domain by calpain. The liberated Talin head is targeted by Smurf1 for degradation [36]. Phosphorylation of the Talin head by Cdk5 facilitates the interaction between Smurf1 and the Talin head. The Talin head is a protein that stabilizes cell focal adhesion by activating integrin and coupling it to actin. Thus, degradation of the Talin head leads to adhesion failure. In the non-canonical Wnt pathway, Smurf1 and Smurf2 ubiquitylate Prickle1 (Pk1) for degradation, which is one of several crucial regulators of cell convergence and extension. Smurf1 forms a polarity complex with Dishevelled (Dvl) and Par6; the latter binds to Pk1 and mediates the ubiquitylation. Degradation of Pk1 results in its asymmetric distribution in neuroectoderm cells, which is pivotal for neurulation [37].

Immunological processes and immune responses

Smurf1 may have notable functions in immunological processes. Of the seven-member TRAF family, TRAF1–TRAF6 can be ubiquitylated and degraded by Smurf1 [38]. Considering the diverse functions of TRAFs in human immune networks, we may anticipate a promising role for Smurf1 in multiple immune networks [39, 40]. In LPS-induced signal transduction though the TLR4, MyD88, and NF-κB cascade, Smurf1 (and Smurf2) ubiquitylates MyD88 and triggers its destruction using Smad6 as the adaptor. By this process, Smurf1 can inhibit MyD88-dependent TLR signaling and immune responses [41]. A recent study revealed that STAT1 is a substrate of Smurf1. STAT1 is an important transcription factor in interferon (IFN) signal amplification and is known as a component of the Jak/STAT pathway. Degradation of STAT1 by Smurf1 blocks IFN-γ signaling [42].

Other substrates and biological processes

Smurf1 also has many substrates in other pathways, for instance KLF2, ING2, and WFS1 (we are unable to list all the substrates in detail). Krüppel-like factor 2 (KLF2) is an essential transcription factor in cell development and differentiation, and inhibitor of growth 2 (ING2) has been demonstrated as a tumor suppressor that enhances the transcriptional activity of p53 [14, 43]. Smurf1 can target these two proteins for degradation. Notably, through interaction with an endoplasmic reticulum (ER)-localized protein, Wolfram syndrome protein (WFS1), Smurf1 can be recruited to the ER. Intriguingly, ER stress can trigger the degradation of Smurf1. However, how Smurf1 is degraded in this process is still unclear [44].

E3 activity-independent functions of Smurf1

Smurf1 also exhibits E3 activity-independent regulatory functions. Smurf1 can interact with the oncoprotein MDM2. The second WW domain of Smurf1 and the N-terminal region of MDM2 mediate this interaction [45]. Smurf1 stabilizes MDM2 but destabilizes the target of MDM2, p53. Smurf1 exhibits this function independently of its ubiquitin ligase activity, and it seems that it enhances the interaction between MDM2 and MDMX, which inhibits the cis-E3 activity and enhances the trans-E3 activity of MDM2 (towards p53) [45]. Like Smurf1, Smurf2 also interacts with and stabilizes the MDM2 protein. Recently, an image-based genome-wide siRNA screen showed that Smurf1 is a master regulator in viral autophagy and mitophagy. Smurf1 induces autophagosomal activity upon Sindbis or herpes simplex virus infections, and the eradication of damaged mitochondria through the membrane-targeting ability of its C2 domain in an E3 activity-independent manner [46]. However, the precise mechanisms by which Smurf1 is involved in autophagy regulation remain to be identified.

Regulation of Smurf1 expression, degradation and E3 activity

It is noteworthy that, with elaborate information on the mechanisms regulating Smurf1 at the activity and protein level, we are able to approach a comprehensive understanding of the Smurf1-associated biological networks. Currently, a wide range of intracellular regulators of Smurf1 have been found and proven to have indispensable regulatory functions in multiple pathways (Table 2). To summarize how Smurf1 is regulated intracellularly, we categorized various mechanisms according to three basic questions: (1) how is Smurf1 activated; (2) how is Smurf1 inactivated or restrained; and (3) how is Smurf1 protein expression regulated?

Table 2.

Regulators of Smurf1

Regulator	Mode of regulation	Regulatory mechanism	References
CKIP-1	Upregulates the E3 activity of Smurf1	Interacts with Smurf1 WW linker, and elevates its binding affinity to the PY motif of substrates	[47]
Cdh1	Upregulates the E3 activity of Smurf1	Interacts with Smurf1 C2 and WW1 domains, and abrogates its auto-inhibition	[10]
CCM2	Upregulates the E3 activity of Smurf1	Interacts with Smurf1 HECT domain, elevates its binding affinity to RhoA	[48]
LMP-1	Blocks the E3 activity of Smurf1	Interacts with Smurf1 WW2 domain and abolishes its interaction with PY motif of substrates	[53]
Synaptopodin	Blocks the E3 activity of Smurf1	Competitively binds to RhoA, by which inhibits Smurf1–RhoA interaction	[54]
PKA	Alters E3 preference of Smurf1	Phosphorylates Smurf1 and alters its E3 preference from Par6 to RhoA	[35]
TSC-22	Blocks Smurf1–adaptor–substrates interaction	Blocks the interaction between Smad7 and TGF-β type I receptor	[57]
Arkadia	Downregulates Smurf1–adaptor interaction	Triggers ubiquitin–proteasomal degradation of Smad7	[55]
TNF	Upregulates the mRNA level of Smurf1	Activates TNF signaling transduction and then elevates Smurf1 transcription	[62]
CKαLS	Stabilizes the mRNA level of Smurf1	Stabilizes the mRNA of Smurf1 by stabilizing one component in the mRNA splicing complex	[65]
Mir-17	Downregulates the mRNA level of Smurf1	Binds to the 3′-UTR of Smurf1’s mRNA and blocks its expression	[63]
SCFFBXL15	Downregulates the protein level of Smurf1	Act as an E3 of Smurf1, triggers ubiquitin–proteasomal degradation of Smurf1	[59]
Smurf2	Downregulates the protein level of Smurf1	Act as an E3 of Smurf1, triggers ubiquitin–proteasomal degradation of Smurf1	[58]
Trb-3	Downregulates the protein level of Smurf1	Triggers ubiquitin–proteasomal degradation of Smurf1 with unknown mechanism	[60]
WFS1	Downregulates the protein level of Smurf1	Mediates Smurf1’s ER localization. Smurf1 can be degraded there under ER stress	[44]

Corresponding descriptions of all regulators in this table can be found in the text

Several proteins interact with Smurf1 and act as adaptors or activators to enhance its E3 activity. They may contribute to Smurf1’s intracellular functions. A PH domain-containing protein, casein kinase 2 interacting protein-1 (CKIP-1), interacts with the linker region between the two WW domains of Smurf1; this interaction was observed by an augmentation of the Smurf1-mediated polyubiquitylation and degradation of Smad1/5, MEKK2, and RhoA. As the WW domains affect the substrate selection tendency, CKIP-1 may upregulate the E3 activity of Smurf1 by altering the structure of these domains to enhance their interaction with the PY motif of Smad1 and Smad5 [47]. Remarkably, CKIP-1 showed no detectable interaction with other Nedd4 family members, indicating that it is a specific Smurf1 activator. Recently, Cdh1 was reported to interact with Smurf1 and augment its E3 activity. Cdh1 was previously defined as the activator of anaphase-promoting complex (APC) during the cell cycle progression. Cdh1 has been shown to bind to both the C2 and WW1 domains to disrupt the homodimer-mediated autoinhibition of Smurf1, resulting in an increased ubiquitylation and proteasomal destruction of its substrates [10]. Cerebral cavernous malformation 2 (CCM2) protein binds to the Smurf1 HECT domain through the PTB-binding motif, resulting in augmentation of the Smurf1-mediated degradation of RhoA. This effect is believed to be achieved by the specific localization of Smurf1 via both the C2 domain and the interaction with CCM2 on the endothelial cell membrane [48]. As discussed previously, phosphorylation of Par6 by TGF-β type II receptor complex can promote Smurf1-dependent ubiquitylation of RhoA [32, 34, 49]. In addition to Par6, modifications of other Smurf1 adaptors may also alter its E3 activity: for example, acetylation of Smad7 by p300 prevents Smad7 from proteasomal degradation [20].

To address the manner by which Smurf1 is inactivated, we should start from its autoinhibition. Autoinhibition is not an exceptional event among Nedd4 family members. For instance, Smurf2 autoinhibits itself by head–tail docking of its C2 and HECT domains, while the WW domain of Itch interacts with the HECT domain for autoinhibition [50, 51]. Smurf1 does not exhibit intramolecular autoinhibition among its domains [33, 52]. However, Smurf1 shows homodimer-mediated intermolecular autoinhibition [10]. The C2 domain of one Smurf1 interacts with the HECT domain of another Smurf1 protein, and its HECT domain binds to another one’s C2 domain. However, the conditions or requirements under which the autoinhibition of Smurf1 occurs are still unclear. Once liberated from the silent mode, Smurf1 activity can be restrained or affected by various players, including LMP-1, Synaptopodin, and PKA. LIM mineralization protein-1 (LMP-1) is a LIM domain-containing protein that interacts with the WW2 domain of Smurf1, and as a result abolishes the binding of Smurf1 to the PY motif of Smad1 and Smad5 [53]. Synaptopodin is also considered as a suppressor of Smurf1 as it competes with Smurf1 on binding to RhoA [54]. Protein kinase A (PKA) is not strictly seen as an inhibitor of Smurf1. While Smurf1 targets Par6 for proteasomal degradation, phosphorylation on Thr306 of Smurf1 by PKA can stabilize Par6 but increase Smurf1-mediated ubiquitylation and destruction of RhoA. This E3 affinity switch leads to axon initiation [35]. The adaptor-dependent E3 activity of Smurf1 can be altered by regulating certain adaptors. For example, the RING E3, Arkadia, was identified to process proteasomal degradation of Smad7, thereby avoiding Smad7-mediated down-regulation of TGF-β signaling [55, 56]. Moreover, TGF-β-stimulated clone 22 (TSC-22) can block the interaction between Smad7 and TGF-β type I receptor, consequently restraining ubiquitylation of the receptor by Smurf1 [57].

Although it is an E3 ligase, Smurf1 is unavoidably destroyed. Smurf1 can be ubiquitylated by a handful of protein ligases such as itself, Smurf2 and SCF^FBXL15, and some Smurf1-interacting proteins can trigger its degradation. Autoubiquitylation of Smurf1 was observed under the upregulation of its E3 activity by CKIP-1 and Cdh1 [10, 47]. Being triggered by either CKIP-1 or Cdh1, this increased Smurf1 degradation also leads to a Smurf1 gain-of-function phenotype because of the increased Smurf1 E3 activity. In contrast, ubiquitylation of Smurf1 by other E3s will result in a Smurf1 loss-of-function phenotype. In human breast cancer cells, Smurf2 ubiquitylates Smurf1 and induces its ubiquitin–proteasomal degradation, by which Smurf2 may rescue important substrates of Smurf1 to prevent malignant migration of tumor cells [58]. A cullin E3 ligase complex named SCF^FBXL15, which contains FBXL15 (F-box and LRR domain-containing protein 15), can target Smurf1 for ubiquitylation by interacting with the N lobe within the HECT domain and transferring Ub onto Lys residues in the WW-HECT linker region [59]. In addition to these well-identified E3 ligases, Smurf1 may undergo destruction by unknown E3s or proteolytic mechanisms. It has been previously shown that Smurf1 ubiquitylates type I and II BMP receptors (BMPRI and BMPRII) for destruction, yet a novel mechanism has been reported to defend these processes. A BMPRII-interacting protein called Tribbles-like protein 3 (Trb3) is released from the tail of BMPRII under BMP stimulation and then triggers degradation of Smurf1 and protects the BMP signal transduction [60]. However, the E3 ligase involved in the Trb3-mediated Smurf1 ubiquitylation has not yet been identified. In addition, as described above, targeting WFS1 may induce Smurf1 itself to locate to the ER, and under ER stress Smurf1 can be degraded by unknown mechanisms [44].

At the transcriptional level, tumor necrosis factor (TNF) and interleukin 1β can upregulate the expression of Smurf1 in osteoblasts, resulting in osteogenesis inhibition [61–63]. In the pulmonary arterial hypertension (PAH) mode, hypoxia induces the expression of Smurf1. This may be an etiologic process of the blockage of the BMP pathway in PAH patients [64]. At the posttranscriptional level, the nuclear protein kinase CKαLS stabilizes hnRNP-C, a component of an mRNA splicing complex, and thus maintains the expression of Smurf1 [65]. At the posttranscriptional level in mesenchymal stem cells (MSC), SMURF1 mRNA levels can be reduced by miR-17 microRNA. miR-17 binds to the 3′-UTR of SMURF1 mRNA and blocks its expression, resulting in the induction of osteogenesis [63]. So far, no transcription factor or epigenetic modification of the SMURF1 gene has been identified.

Roles of Smurf1 in physiological phenotypes and diseases

Apparently, the processes that involve Smurf1 indicate that aberrant expression and dysfunction of this protein may lead to severe defects and diseases. Many biological processes involving Smurf1—including cell growth, differentiation, polarity, adhesion, and migration, as well as viral autophagy and immune responses—are implicated in diverse phenotypes in bone, embryo, heart, neuron, and the immune system, and in related disorders and pathological changes.

Osteogenesis and bone disorders

Bone formation and resorption are two dynamic processes in maintaining skeleton homeostasis. The osteoblast responds to bone increase, whereas the osteoclast, a macrophage in bone tissues, regulates bone loss. Osteoblasts originally derive from mesenchymal stem cells (MSCs) in bone marrow through the MSC-osteoblast precursor-osteoblast lineage differentiation, followed by differentiation into osteocytes after stimulation. Components of the BMP pathway such as Smads, MEKK2, and related transcription factors such as RunX2 and JunB are master players in the differentiation and growth of osteoblasts. Smurf1 ^−/− mice do not show distinct phenotypes, but age-dependent increase in bone mineral density (BMD) and osteoblast differentiation along with high levels of JunB and MEKK2 have been observed [28, 29], suggesting a negative regulatory function of Smurf1 in bone formation. In accordance with this phenotype, CKIP-1 ^−/− mice exert the same phenotypes of induced BMD and osteoblast markers [47]. In contrast, after administrating rats with siRNA against SCF^FBXL15, which degrades Smurf1, decreased BMD was observed [59]. At the cellular level, altering wild-type Smurf1 into a catalytic mutant in osteoblasts greatly rescued the bone loss phenotype [25]. Overexpression of Smurf1 or downregulation of Trb3 in C2C12 cells inhibited their osteogenic activity and differentiation [60, 61, 66]. When examining chondrocyte development, transgenic mice with overexpressed Smurf1 in chondrocytes significantly reduced endochondral ossification, and overexpression of both Smad6 and Smurf1 in these transgenic mice caused postnatal dwarfism [66]. The balance between bone formation and resorption changes to increasing bone loss in an age-dependent manner, and osteoporosis occurs during natural aging. Interrupting the Smurf1 E3 activity in the BMP signaling pathway is a candidate strategy in future osteoporosis therapies.

Embryonic development

As a master regulator in TGF-β and BMP signaling, Smurf1 may play a vital role in embryonic development. The ortholog of Smurf1, XSmurf1, is expressed from the egg stage to the swimming tadpole during Xenopus laevis embryonic development. The mRNA level peaks during the egg to gastrula stages. Overexpression of XSmurf1 triggers dorsalization of mesodermal tissues, differentiation of the ectodermal germ layer, and enhanced responses to TGF-β and activin signaling in the animal cap [15]. Moreover, XSmurf1 is required for neural rather than mesodermal development. Blocking the expression or function of XSmurf1 in Xenopus embryo by specific morpholino oligos or the C699A mutant greatly enhances the BMP pathway activity, and leads to several dorsal neuroectodermal developmental defects in terms of reduced neural tissue differentiation, defective anterior neural tube closure, and loss of f-actin bundles at neural fold hinge points [67]. In Drosophila, DSmurf is a negative regulator of the DPP (the ortholog of BMP in Drosophila) pathway. DSmurf plays a vital role in regulating Drosophila embryonic development by targeting MAD (the ortholog of Smad1) for degradation. A DSmurf mutant showed defects in hindgut morphogenesis and loss of dorsal hindgut integrity [68]. Also, in germline stem cells of Drosophila, DSmurf downregulates DPP signal transduction via targeting the DPP receptor thickveins [69]. Although Smurf1 ^−/− mice were viable, fertile, and born at the expected Mendelian ratio, and showed a normal embryonic phenotype, Smurf1 and Smurf2 double-knockout mice manifested embryonic lethality. This double-knockout model displayed disorganized neuroectoderm and lateral expansion of the floor plate as well as shortened A–P axis during early embryonic development, indicating that a lack of Smurf1 and Smurf2 leads to defects in convergence and extension movements. In addition, the double-knockout mice also displayed severe neural tube closure defects. Moreover, in Smurf1 ^−/− Smurf2 ^+/− mice and Smurf1 ^+/− Smurf2 ^−/− mice, misorientation and disorganization of sensory hair cells in the cochlea were observed [28, 37]. Also, overexpression of Smurf1 in cultured mice embryonic lung was found to inhibit lung epithelial branching [70].

Epithelial-mesenchymal transition and cancer

Most Nedd4 family members are highly related to oncogenesis, as a number of their substrates, such as PTEN, p53, p63, p73, KLF2, Smads, and RhoA, are players in apoptosis, growth arrest, DNA damage repair, and cell morphogenesis [71–74]. Smurf1 may also play an important role in multiple cancers including breast cancer, gastric adenocarcinoma, and pancreatic cancer. As previously summarized, by triggering degradation of RhoA, Talin head, and hPEM-2, Smurf1 increases membrane protrusive behavior, impacts cell adhesion, and promotes abnormal cell migration. Downregulation of RhoA by the Par6–Smurf1 complex induces tight-junction dissolution and initiates epithelial–mesenchymal transition (EMT), which may be important for tumor invasion [34]. Smurf1 induces EMT and tight-junction dissolution in epithelial cells, and then promotes cell invasion [75]. In tumor cells, Smurf1 promotes lamellipodia formation and cell migration by downregulating RhoA and related downstream pathway components [76]. In pancreatic cancer, highly expressed Smurf1 was seen by fluorescence in situ hybridization, and knocking down Smurf1 by RNA interference led to reduced cell invasion [77]. These findings are in line with previous results of a genomic hybridization assay, by which Smurf1 was speculated to be a potential oncogenic factor in pancreatic cancer [78, 79]. Another genomic hybridization assay, performed with gene expression array and protein analysis, revealed that Smurf1 may also be a candidate regulatory factor in gastric adenocarcinomas [80]. Moreover, Smurf1 targets and ubiquitylates KLF2 and ING2 for degradation, and stabilizes the MDM2–MDMX complex to enhance proteolysis of p53 [14, 43, 45]. These findings suggest a potential but unknown role of Smurf1 in oncogenesis. In contrast, Smurf2 plays the opposite role in oncogenesis and cancer invasion. A recent study showed cancerization in a wide spectrum of organs ranging from liver to lung in aged Smurf2 ^−/− mice, and proposed that Smurf2 may be a guardian that prevents tumorigenesis by controlling RNF20 turnover [81]. In breast cancer, ubiquitylation and degradation of Smurf1 by Smurf2 was found along with suppression of cell migration ability [58]. This indirect evidence indicates that Smurf1 is a harmful regulator, promoting cell migration in breast cancer.

Other biological phenotypes

Besides bone formation, embryonic development and tumor invasion, Smurf1 plays vital roles in neurite outgrowth, hypertension, and immune responses. By reducing RhoA in neural cells, Smurf1 was demonstrated to promote neurite outgrowth. Similarly, further investigation revealed that Smurf1 is required for neurite formation and elongation in enteric nervous system precursors [82, 83]. Together with the finding that Smurf1 controls planar cell polarity, we could imagine the considerable role of Smurf1 in nervous system development and neurocyte differentiation. In heart development, EMT of endocardial cells in the atrioventricular cushion is vitally important for heart valve formation. Amplified Smurf1 expression in mice ventricular endocardial cells was found to induce EMT and cell invasion, and loss of epithelial character [49, 75]. Upon TNF stimulation, the Smurf1 level elevates in the hypoxia-induced pulmonary arterial hypertension (PAH) rat model [64]. By targeting MyD88 and STAT1 for proteasomal degradation, Smurf1 may contribute to a diminished antiviral response and enhanced viral replication in macrophages [41, 42]. In contrast, Smurf1 mediates normal viral autophagy when murine embryonic fibroblasts are infected with Sindbis or herpes simplex viruses, and Smurf1 ^−/− mice show accumulated damaged mitochondria as the cellular mitophagic function is impaired [46]. All these findings indicate a potential role of Smurf1 in the immune system.

Perspectives

The accumulating scientific discoveries on Smurf1 have shown the notable role this ubiquitin ligase plays in multiple biological networks, and future investigations would focus on both translational and basic research. The function of Smurf1 in viral autophagy as well as innate immunity has just emerged, but the regulatory mechanisms within this function are still unclear. The fact that the slight phenotype of Smurf1-knockout mice is asymmetric to the multiple functions of this ubiquitin ligase is puzzling. Many Smurf1 substrates play important roles in certain biological processes, but their regulation by Smurf1 does not show dominant phenotypes in the animal model. Therefore, there is a clear gap between the expected functions of Smurf1 according to its many reported substrates and the true phenotypes of Smurf1 using genetically modified animal models. This suggests that the Smurf1/substrates correlation could very much be context dependent. Furthermore, there are spatial and timing restraints that control the specific interaction between Smurf1 and its selected pool of substrates for certain physiological activities. Future studies should be performed to establish tissue-specific knockout mice models of Smurf1 and to explore in more detail the functions of Smurf1 in physiological and pathological conditions.

A long-standing question in understanding E3s is why closely related E3 members have distinct substrate specificity and regulatory mechanisms. Compared with the close homolog Smurf2, Smurf1 exerts a plethora of distinct functions, such as the exclusive role of Smurf1 in regulation of viral autophagy and the unique accelerated bone formation phenotype of Smurf1-deficient mice. Overlapping but distinct substrate specificity has been observed between Smurf1 and Smurf2 despite the high homology between the WW domains. We propose that at least three mechanisms might explain these differences. First, the specific activators of Smurf1, including CKIP-1 and Cdh1, interact with Smurf1 but not Smurf2. These auxiliary factors help Smurf1 exert its ligase activity and target the substrate for degradation to control bone formation. The small differences in amino acid sequence might account for the specificity. For example, CKIP-1 recognizes the LNxVxCxEL motif in the WW domain’s linker of Smurf1 [47]. This motif is not found in Smurf2. Interestingly, when this motif was introduced into Smurf2, CKIP-1 could interact with the mutant Smurf2 [47]. Second, Smurf1 is predominantly localized in the cytoplasm, near the plasma membrane, although it can also shuttle into the nucleus. In contrast, Smurf2 is predominantly localized in the nucleus, though it can shuttle out to the cytoplasm. Furthermore, Smurf2 can target the nuclear RNF20 for degradation and regulates the chromatin conformation, but Smurf1 cannot [81]. The differences in their subcellular localization and tissue expression profile may also contribute to their functional differences. Third, the specific regulatory mechanisms are very important in guiding Smurf1 into various functions. At the basal level, Smurf1 exerts higher ligase activity than Smurf2. This is mainly caused by the intramolecular autoinhibition of Smurf2. It is notable that Smurf1 contains two WW domains, whereas Smurf2 contains three (Fig. 1b). The WW2 and WW3 domains of Smurf2 are homologous to those of Smurf1. However, the function of the Smurf2 WW1 domain remains unknown. An interesting explanation is that it provides the C2 domain with enough conformational space to interact with the HECT domain. In the case of Smurf1, the C2 domain cannot interact with the HECT domain due to insufficient distance. Thus, the HECT domain can exert the catalytic function. This hypothesis requires further investigation to support it with structural evidence.

Smurf2 has been recently shown as a tumor suppressor [81]. Genomic ablation of Smurf2 in mice resulted in increased susceptibility to various types of cancers in aged mice. A similar phenotype has not been observed in Smurf1-knockout mice. So far, the role of Smurf1 in both carcinogenesis and tumor invasion has not been determined. Further basic research is required to fill the blanks in our basic understanding of Smurf1. Besides cancer, other chronic disorders are becoming a more severe problem, which may not be fatal but are less susceptible to current therapies. Osteoporosis is certainly one of them. The negative role Smurf1 plays in osteogenesis has been well established. This inspires us to perform translational studies aiming at finding new therapeutic solutions to the related bone formation disorders.

Appendix 1 Ubiquitylation cascade and chain linkages

Ubiquitylation. A highly conserved 76 amino acid polypeptide, ubiquitin, can be initially activated by an E1 in an ATP-dependent process; with the active-site cysteine residue in E1 forms a high-energy thioester linkage with the carboxyl terminus of an Ub molecule. The activated Ub is then trans-thiolated from E1 to an active-site cysteine residue of one E2. Finally, Ub molecule is donated to a specific lysine residue (abbreviated as Lys or K) of a substrate through an E3-dependent manner (Fig. 1a). Ubiquitylation may occur once or multiple times, resulting in mono-ubiquitylation (attaching a single Ub molecule at a lysine residue), multi-ubiquitylation (attaching a single Ub at multiple lysine residues), or poly-ubiquitylation, in which the poly-ubiquitin chain is elongated on certain Lys residue on the ubiquitin by sequential cycles of ubiquitin assembling [84, 85].

Ubiquitin chain linkages. There are seven lysine residues in the ubiquitin, among which K48 and K63 are two major sites for poly-ubiquitin chain elongation (Fig. 1a). The fate of an Ub chain-tagged protein may be determined by the type of the Ub linkage. K48-linked poly-ubiquitin chain-attached proteins are usually detained and destructed by 26S proteasome, whereas K63-linked poly-ubiquitylation or mono-ubiquitylation and multi-ubiquitylation generally function in other biological processes.

Appendix 2 RING and HECT E3s

RING type E3s. The RING E3 ligases facilitate E2-dependent ubiquitylation that interact and bring target proteins close enough to let E2s transfer ubiquitin directly to specific internal Lys residues of the substrates. Those E3s can work by monomers, dimmers or complexes containing multiple subunits. RING E3 complexes include cullin RING ligase (CRL) superfamily (includes SCF, BTB, and SOCS/BC type of E3 complexes) and anaphase-promoting complex (APC) [86, 87].

HECT type E3s. HECT E3s receipt Ub molecule, by form an Ub-thioester intermediate, from their E2 donors and then transfer those ubiquitins onto their specific interacting preys. The characteristic HECT domain is a bilobal domain with an N lobe and a C lobe. Based on the N terminus architecture, HECT E3s can be generally divided into Nedd4 family (contains successive C2 and WW domains from N terminus) (Fig. 1b), HERC family (contains the RLD domain in N terminus), and other HECTs.