Smad ubiquitination regulatory factor 1 (Smurf1), an homologous to E6AP C‐terminus (HECT)‐type E3 ubiquitin ligase, performs a crucial role in the regulation of the bone morphogenetic protein (BMP) signalling pathway in both embryonic development and bone remodelling. How the stability and activity of Smurf1 are negatively regulated remains largely unclear. Here, we report that F‐box and LRR domain‐containing protein 15 (FBXL15), an F‐box protein of the FBXL family, forms an Skp1‐Cullin1‐F‐box protein‐Roc1 (SCF)FBXL15 ubiquitin ligase complex and targets Smurf1 for ubiquitination and proteasomal degradation. FBXL15, through its leucine‐rich repeat domain, specifically recognizes the large subdomain within the N‐lobe of the Smurf1 HECT domain and promotes the ubiquitination of Smurf1 on K355 and K357 within the WW‐HECT linker region. In this way, FBXL15 positively regulates BMP signalling in mammalian cells. Knockdown of fbxl15 expression in zebrafish embryos by specific antisense morpholinos causes embryonic dorsalization phenocoping BMP‐deficient mutants. Injection of FBXL15 siRNAs into rat bone tissues leads to a significant loss of bone mass and decrease in bone mineral density. Collectively, our results demonstrate that Smurf1 stability is suppressed by SCFFBXL15‐mediated ubiquitination and that FBXL15 is a key regulator of BMP signalling during embryonic development and adult bone formation.
Protein ubiquitination is a highly ordered multi‐step enzymatic cascade catalysed by ubiquitin‐activating enzymes (E1s), ubiquitin‐conjugating enzymes (E2s), and ubiquitin ligases (E3s) (Hershko and Ciechanover, 1998). E3 ubiquitin ligases are classified into the really interesting new gene (RING) finger and homologous to E6AP carboxyl terminus (HECT) domain categories (Rotin and Kumar, 2009). Smad ubiquitination regulatory factor 1 (Smurf1) and Smurf2 are close members of the HECT‐type neural precursor cell‐expressed developmentally downregulated gene 4 (Nedd4) family ligases with a C2‐WW‐HECT architecture, and they have critical roles in the regulation of the transforming growth factor β (TGF‐β) and bone morphogenetic protein (BMP) signalling pathways (Zhu et al, 1999; Yamashita et al, 2005), as well as the Wnt (Narimatsu et al, 2009) and RhoA pathways (Wang et al, 2003). Smurf1 was originally identified as the E3 ligase of Smad1/5, and it dorsalizes ventral mesoderm and neuralizes ectoderm by modulating BMP signalling in Xenopus embryos (Zhu et al, 1999). Smurf1 also has a specific suppressing role in adult bone formation, but not in embryonic bone development or adult bone resorption (Zhao et al, 2003, 2010; Yamashita et al, 2005). Smurf1−/− mice exhibit an age‐dependent increase in bone mass due to augmented BMP responses (Yamashita et al, 2005). Given the significant role of Smurf1 during embryonic development and adult bone homoeostasis maintenance, the stability and activity of Smurf1 should be tightly controlled.
We previously demonstrated that the PH domain‐containing protein casein kinase 2‐interacting protein 1 (CKIP‐1) functions as an auxiliary factor to enhance Smurf1 activation through binding to the WW domains linker (Lu et al, 2008). CKIP‐1−/− mice display an age‐dependent bone mass increase due to decrease of Smurf1 activity (Lu et al, 2008). Regarding the negative regulation of Smurf1, a recent study showed that Smurf2 interacts with Smurf1 to induce the degradation of Smurf1, whereas Smurf1 failed to induce the degradation of Smurf2. Knockdown of Smurf2 in human breast cancer cells resulted in the enhancement of cell migration in vitro and bone metastasis in vivo (Fukunaga et al, 2008). Because the study was performed with a breast cancer cell line MDA‐MB‐231, the physiological role of Smurf1 regulation by Smurf2 under developmental conditions remains uncharacterized. To date, how Smurf1 is negatively regulated still remains unclear.
To search for more regulators of Smurf1, a yeast two‐hybrid screening using Smurf1 as bait was performed (Li et al, 2010). One of the candidate interactors is the F‐box protein: F‐box and leucine‐rich repeat (LRR) domain‐containing protein 15 (FBXL15). F‐box proteins are the variable substrate adaptors for the Skp1‐Cullin1‐F‐box protein‐Roc1 (SCF) ligase complexes, and they dictate the substrate specificity (Skaar et al, 2009a, 2009b). The SCF ligases mediate the timely proteolysis of important eukaryotic cellular regulators and are involved in diverse cellular processes, including cell‐cycle progression and tumourigenesis (Cardozo and Pagano, 2004). Cullin1 acts as the scaffold to recruit the Skp1‐F‐box protein pair to its N‐terminus and Roc1 plus an E2 enzyme to its C‐terminus (Skaar et al, 2009a). The F‐box protein usually contains an N‐terminal F‐box, which mediates Skp1 binding and then is linked to the rest of the SCF complex, although not all F‐box proteins form E3 complexes (e.g. FBXL10 and FBXL11). In addition, the F‐box protein contains certain substrate‐recognition modules, and based on this homology, they fall into three groups: FBXW subfamily with WD40 repeats, FBXL with LRR, and FBXO with other domains (Jin et al, 2004). There are a total of 69 human F‐box proteins, but nearly 70% of them have not yet been matched with any substrates, a category that included the here identified FBXL15.
Here, we show that FBXL15 forms a functionally active SCF complex and targets Smurf1 for ubiquitination and proteasomal degradation. Importantly, we identify a critical role of FBXL15 in both the determination of dorsalization/ventralization during embryonic development and bone homoeostasis maintenance in the adult stage. So far as we know, this is the first evidence to show the physiological role of FBXL15 in BMP signalling and to indicate the negative regulatory mechanism of Smurf1 stability control.
Identification of F‐box protein FBXL15 as an Smurf1‐interacting protein
A yeast two‐hybrid screen was performed using the WW‐HECT of human Smurf1 as bait, as the WW domains are responsible for substrate recognition, whereas the HECT is the catalytic domain (Rotin and Kumar, 2009). One of the recovered interactors encoded the full length of the F‐box protein FBXL15 (Figure 1A). This prey attracted our interest for further investigation because most F‐box proteins serve as substrate‐recognition subunit of SCF ubiquitin ligases (Cardozo and Pagano, 2004). FBXL15 is highly conserved across species, with nearly 95% identity among mammals (Supplementary Figure S1A). Human FBXL15 contains an F‐box domain (aa 22–64) in the N‐terminus and six LRR repeats (aa 113–269) distributed in the resting regions (Figure 1A).
To ascertain whether the observed interaction was direct, the FBXL15‐Smurf1 association was tested in an in vitro binding assay. GST‐fused FBXL15 protein, but not GST alone, interacted with Smurf1 (Figure 1B). We then tested whether the interaction between the two proteins occurs in a physiological context. For this purpose, we first generated a specific antibody, which recognized both endogenous and ectopic FBXL15 protein and designed FBXL15‐targeting siRNAs (Supplementary Figure S1B and C). The FBXL15 siRNA had no off‐target effects on other F‐box proteins (Supplementary Figure S1D). Using this antibody, we observed that FBXL15 was widely expressed in mouse heart, liver, spleen, bone, muscle, brain, and kidney, as well as in various human cell lines (Supplementary Figure S1E and F). Importantly, an immunoprecipitate of endogenous Smurf1 contained significant amounts of FBXL15, whereas a pre‐immune IgG failed to immunoprecipitate any FBXL15 (Figure 1C). FBXL15 interacted with wild‐type Smurf1 and the ligase‐inactive C699A mutant to similar extents (Figure 1D). FBXL15 and Smurf1 were colocalized in the cytoplasm (Figure 1E; Supplementary Figure S1G). Together, these data demonstrate that FBXL15 interacts with Smurf1 both in vitro and endogenously in cultured cells.
Deletion analysis revealed that the LRR domains of FBXL15 and the HECT domain of Smurf1 were both sufficient and necessary for the interaction (Figure 1F and G). The HECT domain is the catalytic region of Smurf1 and consists of two lobes: an N‐terminal lobe and a C‐terminal lobe. The N‐terminal lobe consists of a large and a small subdomain. The small subdomain of the N‐lobe contains the E2‐binding site, whereas the C‐lobe contains the catalytic cysteine (C699), which forms a thioester intermediate with ubiquitin (Rotin and Kumar, 2009). In contrast, the role of the large subdomain remains less well known. Interestingly, the large subdomain of Smurf1 was both required and sufficient for interaction with FBXL15 (Figure 1H and I). Thus, the LRR domains of FBXL15 and the HECT large subdomain of Smurf1 mediate their interaction.
SCFFBXL15 targets Smurf1 for proteasomal degradation
LRR domains of FBXL family are usually responsible for substrate recognition (Skaar et al, 2009a). The fact that the LRR of FBXL15 mediates the interaction with Smurf1 led us to investigate whether Smurf1 is the substrate of SCFFBXL15. Considering that not all F‐box proteins form E3 ligase complexes with Cullin1 and Roc1, we first examined the interaction of FBXL15 with common SCF subunits in mammalian cells. As shown in Figure 2A, FBXL15 could be co‐immunoprecipitated with endogenous Cullin1, Skp1, and Roc1, but the ΔF mutant could not be, indicating that it does indeed form a genuine and typical SCF complex. Coexpression of FBXL15 with Smurf1 resulted in a significant decrease in the Smurf1 protein level, whose effect was blocked by treatment with the proteasome inhibitor MG132 (Figure 2B; Supplementary Figure S2A). Remarkably, FBXL15 downregulated the stability of both the Smurf1 WT and C699A, ruling out the possibility that FBXL15 promotes Smurf1 degradation by augmenting the Smurf1 cis‐E3 activity. Additionally, neither the F‐box nor the ΔF truncates of FBXL15 had any significant effects on Smurf1 levels (Figure 2C), indicating that both the direct interaction between FBXL15 and Smurf1 mediated by the LRR and the intact SCF complex mediated by the F‐box are required for FBXL15 to efficiently regulate Smurf1. In contrast, overexpression of Smurf1 had no significant effect on the protein level of FBXL15 (Supplementary Figure S2B).
Next, we asked whether endogenous SCFFBXL15 affects the Smurf1 protein. To this end, FBXL15, Cullin1, or Roc1 was each depleted by two independent siRNA duplexes in HEK293T cells. We observed a 1.7–2.5‐fold increase in the Smurf1 protein level in cells depleted of FBXL15, Cullin1, or Roc1 relative to control cells (Figure 2D). Furthermore, the half‐life of endogenous Smurf1 was approximately 6 h in control cells and it was greatly increased to >9 h after depletion of FBXL15 (Figure 2E, top panels; Supplementary Figure S2C). Introduction of the siRNA‐resistant FBXL15 mutant reversed the effects of FBXL15 siRNA and resulted in accelerated degradation of Smurf1 (Figure 2E, middle and bottom panels). Importantly, the regulation of Smurf1 by FBXL15 was specific. Indeed, overexpression of six other F‐box proteins did not affect Smurf1 levels (Figure 2F). Thus, endogenous FBXL15 causes the destabilization of endogenous Smurf1 protein.
The SCF complex has been shown to have a role in controlling cell‐cycle progress. We next tested whether the stability of Smurf1 has any cell‐cycle dependence. The results indicated that Smurf1 (and Smurf2) were upregulated when the cells entered into S and G2 phases (2–8 h after thymidine‐aphidicolin release) (Figure 2G, lanes 2–5). At the later stage of G2 phase (10 h) and the M phase (14 h), FBXL15 were upregulated, companied by a significant decrease of Smurf1/2 levels (lanes 6–7). In contrast, the expression of Cullin1 and Roc1 remains constant during this process. Importantly, in the FBXL15‐depleted cells, the levels of Smurf1 (and Smurf2) were constant within the 18 h after thymidine‐aphidicolin release (lanes 8–15). These results suggested that SCFFBXL15 might have a role in the G2/M progression at least partially through targeting Smurf1 (and Smurf2) for degradation.
SCFFBXL15 promotes the ubiquitination of Smurf1
As most F‐box proteins involve the substrate's turnover by ubiquitination, we sought to determine whether FBXL15‐mediated Smurf1 degradation is a consequence of ubiquitination. As shown in Figure 3A, in the absence of ectopic FBXL15, the ubiquitination of Smurf1 WT (lane 2) and CA (lane 4) can be both detected. The latter was less than the former and should be mediated by endogenous FBXL15. FBXL15 overexpression strongly enhanced the ubiquitination of Smurf1 (lanes 3 and 5). Knockdown of endogenous FBXL15 or Cullin1 or Roc1 resulted in decrease of Smurf1 ubiquitination (Figure 3B), indicating SCFFBXL15 indeed functions as E3 ligase of Smurf1 in cultured cells.
The E2 used by FBXL15 being unknown, we tested two proteins that are frequently used by mammalian RING finger E3s: UbcH5c and UbcH7. Incubation with FBXL15 purified from cultured cells, in the presence of either UbcH5c or UbcH7, led to the ubiquitination of His‐Smurf1 (Figure 3C). To further evaluate the capacity of FBXL15 to directly ubiquitinate Smurf1, we carried out the ubiquitination of Smurf1 in vitro using a semi‐purified system. Bacteria‐expressed and purified Smurf1 ligase could catalyse the auto‐ubiquitination in the presence of HA‐Ub and E1/E2/ATP (Figure 3D, lane 2). Incubation of this mixture with purified GST‐FBXL15 protein alone or GST/Cul1/Roc1/Skp1 elution (the Cul1/Roc1/Skp1 were expressed as Myc‐tagged proteins in HEK293T cells and the cell lysate was pull down by GSH beads) had no significant effects on the Smurf1 ubiquitination (lanes 3 and 4). Strikingly, incubation with GST‐FBXL15/Cul1/Roc1/Skp1 (lane 5), but not the GST‐FBXL15‐ΔF/Cul1/Roc1/Skp1 complex (lane 6), resulted in a dramatic increase of Smurf1 ubiquitination, indicating that an intact SCF complex was both sufficient and required for promoting the Smurf1 ubiquitination. Again, this ubiquitination was independent of Smurf1 ligase activity since Smurf1‐C699A was effectively ubiquitinated by GST‐FBXL15/Cul1/Roc1/Skp1 (lane 9), whose effect also required the entire SCF complex (lanes 8 and 10). Notably, FBXL15 seems not to function as an adaptor to change the structure of Smurf1 for its auto‐ubiquitination since incubation with GST alone or GST‐FBXL15 protein had no significant effects on Smurf1 auto‐ubiquitination under the in vitro ubiquitination conditions (Figure 3E, lanes 3 and 4). As a positive control, incubation with the Smurf1 cofactor CKIP‐1 (Lu et al, 2008) resulted in a dramatic increase of Smurf1 auto‐ubiquitination in vitro (lane 5). In addition, Smurf2 depletion in HEK293T cells had no significant effect on the FBXL15 activity of regulating Smurf1 ubiquitination (Supplementary Figure S3), implicating that FBXL15 might regulate Smurf1 ubiquitination in an Smurf2‐independent manner. Collectively, these data demonstrate that SCFFBXL15 complex is a bona fide E3 ligase for Smurf1.
Lysines 355 and 357 of Smurf1 are major ubiquitination sites for FBXL15
As the HECT domain is essential for Smurf1 association with FBXL15, we asked whether HECT domain alone is sufficient for degradation by FBXL15. We observed that in contrast to the significant degradation of Smurf1 WT by FBXL15 (Figure 4A, lanes 1–3), the level of HECT alone was not decreased by FBXL15 (lanes 7–9). In contrast, WW plus HECT was markedly reduced by FBXL15 (lanes 4–6). Examination of the WW plus WW‐HECT linker found total three lysines, K324, K355, and K357 (Figure 4A), all of which are located within the WH linker. Deletion of this linker from full‐length Smurf1 blocked the FBXL15‐promoted Smurf1 degradation (lanes 10–12), indicating that the degradation signal is located within this linker.
Mutations of all three lysines to arginines completely blocked the degradation of Smurf1 by FBXL15 (Figure 4B). Further individual mutation of each lysine showed that K357 was the primary and K355 the secondary site for degradation (Figure 4C). K355+357R double mutation resulted in resistance against the degradation (Figure 4D, lanes 13–16 versus 5–8) and abrogated the FBXL15‐mediated ubiquitination of Smurf1 (Figure 4E). Note that the steady‐state level of K355+357R in the absence of ectopic FBXL15 gradually decreased upon cycloheximide treatment, as did that of wild‐type Smurf1 (Figure 4D, lanes 9–12 versus 1–4), indicating that Smurf1 was also degraded by ligases other than FBXL15, such as the auto‐degradation by itself. These data strongly suggest that Smurf1 K357 is the primary ubiquitination site for FBXL15‐induced Smurf1 degradation, and K355 is a secondary site.
FBXL15 antagonizes the effect of Smurf1 on BMP signalling
We then sought to better characterize the functional relevance of FBXL15 and Smurf1 for BMP signalling. Ectopic expression of Smurf1 resulted in a dramatic decrease in Smad1 and Smad5, but not Smad3 levels (Figure 5A). Further coexpression of FBXL15 antagonized this effect of Smurf1 (Figure 5A). Upon BMP stimulation, Smurf1 significantly targeted the phosphorylated Smad1/5 for ubiquitination and degradation (Sapkota et al, 2007). Through this manner, Smurf1 activation contributes to shut off the BMP signalling. Depletion of FBXL15 resulted in an increased Smurf1 level and attenuated Smad1/5 phosphorylation, even at the early response stage (Figure 5B).
We further delineated the function of FBXL15 in BMP signalling with a luciferase reporter assay. BMP‐2 treatment remarkably stimulated the BMP‐responsive BRE‐luc activity and Smurf1 overexpression inhibited it. When Smurf1 was coexpressed with FBXL15, FBXL15 substantially antagonized the inhibitory effect of Smurf1 (Figure 5C, column 5 versus column 4). In contrast, neither FBXL15 ΔF nor the F‐box alone had any significant effect on Smurf1 (columns 6 and 7). Additionally, other F‐box proteins we examined, including FBXL1/Skp2, FBXL3, FBXL5, and FBXL21, showed no evident antagonistic effects on Smurf1 (Supplementary Figure S4).
To explore whether FBXL15 regulates BMP response through targeting Smurf1, Smurf1 WT and its ubiquitination‐resistant mutants were tested for their ability to regulate BRE activity in the absence or presence of ectopic FBXL15 (Figure 5D). As shown, all of the examined Smurf1 forms displayed strong inhibitory effects on BRE activity in the absence of ectopic FBXL15. Coexpression of FBXL15 completely removed the inhibitory effects of Smurf1 WT, K324R and K355R, and partially the K357R mutant. However, FBXL15 failed to reverse that of ubiquitination‐resistant K355+357R mutant, strongly indicating that FBXL15 regulates the activity of Smurf1 by targeting Smurf1 for ubiquitination and degradation.
Furthermore, we showed that knockdown of endogenous FBXL15 or Cullin1 or Roc1 resulted in reduced stimulation of BRE activity upon BMP‐2 treatment (Figure 5E). Depletion of FBXL15 also reduced the upregulation of target genes downstream of Smad1/5 (Figure 5F). Collectively, these data suggest that FBXL15 is a positive regulator of BMP signalling in cultured mammalian cells, at least partially through directing the degradation of Smurf1.
Smurf2 is also ubiquitinated and degraded by FBXL15
Due to the conserved protein sequences and the common domain architecture, we next investigated whether other members of the Nedd4 family could also be regulated by FBXL15. We first examined the association of Nedd4 family members with FBXL15. All nine members, when expressed in ectopic forms, could be co‐immunoprecipitated with FBXL15, although they displayed differential binding affinities (Figure 6A). This result is not surprising, as the HECT large subdomains of each E3 of Nedd4 family exhibit relatively high homologies. Next, we examined whether FBXL15 could control the steady‐state levels of these ligases. We observed that Smurf1, Smurf2, and WWP2, but not the other six ligases, were targeted for degradation by FBXL15 (Figure 6B). In vivo ubiquitination assays showed that the ubiquitination of Smurf1, Smurf2, and WWP2 could also be significantly enhanced by FBXL15, whereas other E3s could be not (Figure 6C). Among the family, only Smurf2 possesses a lysine identical to the critical ubiquitination site K357 of Smurf1 (Supplementary Figure S5A). WWP2 contains K498 and H500 in the Smurf1 K355‐ and K357‐corresponding positions. Neither K498R nor H500K mutations had any significant effects on the WWP2 ubiquitination promoted by FBXL15 (Supplementary Figure S5B), suggesting that those residues other than K498 and H500 might be responsible for this ubiquitination, which needs further investigations.
The regulatory effects of FBXL15 on Smurf2 stability were further verified. GST pull down and endogenous co‐immunoprecipitation assays (Supplementary Figure S6A and B) indicated that Smurf2 could interact with FBXL15 both in vitro and endogenously in cultured cells, similar to Smurf1. The degradation–promoting effect of FBXL15 on Smurf2 was dependent on the proteasome activity and the integrity of the SCF complex formed by FBXL15 (Figure 6D). Furthermore, the possible effect of FBXL15 on Smurf2's regulation of TGF‐β signalling was measured. Smurf2 overexpression inhibited ALK4‐induced activation of TGF‐β signalling, and this inhibition was significantly reversed by FBXL15, depending on both the F‐box and the LRR domains (Figure 6E). Knockdown of endogenous FBXL15 resulted in a significant decrease in ALK4‐induced 3TP‐luc activity (Figure 6F). Taken together, these results demonstrate that Smurf2, like Smurf1, is ubiquitinated, degraded, and inhibited by SCFFBXL15.
Zebrafish fbxl15 is required for dorsoventral patterning in early development of embryos
Smurf1 has been demonstrated to antagonize BMP signalling and have a critical role in dorsoventral patterning of Xenopus embryos (Zhu et al, 1999; Murakami et al, 2003; Xia et al, 2010). We hypothesized that FBXL15 is involved in embryonic dorsoventral patterning as well. To test this, we used zebrafish as an animal model. We identified a zebrafish cDNA, zgc:85882 (NM_212942), from the GenBank database, which encodes a putative peptide sharing an identity of 65.0% to human FBXL15 (Supplementary Figure S1A) and should be zebrafish fbxl15. We first examined the spatiotemporal expression pattern of fbxl15 during embryonic development by whole‐mount in situ hybridization. Compared with the sense control (Figure 7A′–E′), fbxl15 is maternally expressed and its transcriptions are evenly distributed until the 6‐somite stage (Figure 7A–D). At 24 h postfertilization (hpf), fbxl15 transcripts are mainly present in head region (Figure 7E). The existence of fbxl15 transcripts in early embryos suggests a potential function in early development.
Next, we investigated fbxl15 function in the zebrafish embryos by knocking down fbxl15 expression with two different morpholinos of fbxl15. Both morpholinos efficiently blocked the expression of fbxl15‐5′UTR‐GFP fusion plasmid in zebrafish embryos, while the control morpholino fbxl15‐cMO1 with mismatched nucleotides for fbxl15‐MO1 had no effect (Supplementary Figure S7). Injection of 7.5 ng fbxl15‐MO1 into one‐ to two‐cell wild‐type embryos resulted in a kind of classic dorsalized phenotype at 24 hpf, manifesting a shorter and curved posterior trunk, together with extensive cell death in the neural system (Figure 7F–H). Since morpholinos often cause certain degrees of apoptosis due to off‐target activation of p53 (Robu et al, 2007), we co‐injected 7.5 ng fbxl15‐MO1 and 7.5 ng p53‐MO to block the non‐specific effect. Such a co‐injection still led to the dorsalized phenotype, but avoided widespread cell death (Figure 7I), which suggests a bona fide requirement of fbxl15 for early development.
To further confirm the implication of fbxl15 in dorsoventral patterning, we examined the expression of some dorsal and ventral markers at the shield stage. In fbxl15‐MO1‐injected embryos, the expression of the dorsal marker gsc was expanded (100%, n=92), while the expression of the ventral markers eve1 (87.5%, n=96) and gata2 (100%, n=87) were almost eliminated. Compared with fbxl15‐MO1 injection alone, co‐injection of 50 pg fbxl15 mRNA had a partial rescue on the marker genes expression. For example, only 60.3% of embryos showed gsc expansion (n=58), 40.9 and 44.9% showed eve1 (n=66) and gata2 (n=107) reduction (Figure 7J–M). Similar phenotype was observed with fbxl15‐MO2 (Supplementary Figure S8). In addition, the rescue effect was confirmed by qRT–PCR experiments as shown in Figure 7N. Taken together, these results suggest that fbxl15 is required for the specification of the dorsoventral fate during embryonic development.
Fbxl15 contains an N‐terminal F‐box domain, which mediates binding to Skp1 and then is linked to the rest of the SCF complex. Injection of 100 pg fbxl15‐ΔF mRNA caused gsc expansion and eve1 and gata2 reduction (Supplementary Figure S8), which mimicked the effect of fbxl15‐MOs. These results suggested that Fbxl15‐ΔF functions as a dominant‐negative form and F‐box domain is essential for the roles of FBXL15 in dorsoventral patterning in zebrafish embryos. Additionally, overexpression of 50 pg skp1 mRNA or 50 pg fbxl15 mRNA, respectively, almost had no effect on gsc expression at shield stage, while co‐injection of these two mRNA was able to cause obvious inhibition of gsc at the same stage (Supplementary Figure S9). It is indicated that Fbxl15 can work together with Skp1 to regulate the dorsoventral patterning in zebrafish embryos. In the light of the results mentioned above, we speculate that Fbxl15 functions dependently on the SCF complex.
Furthermore, we tested the ability of Fbxl15 to degrade Smurf1 in zebrafish embryos. Compared with gfp mRNA injection (Supplementary Figure S10A), injection of smurf1‐gfp fusion mRNA produced weaker GFP fluorescence (Figure 7O). Co‐injection of fbxl15 mRNA with gfp mRNA did not change GFP fluorescence levels (Supplementary Figure S10B), but fbxl15 and smurf1‐gfp mRNA co‐injection produced extremely low levels of GFP fluorescence (Figure 7P). Correspondingly, co‐injection of fbxl15‐MO1 or fbxl15‐MO2, respectively, both enhanced the Smurf1‐GFP fluorescence in zebrafish embryos at the same stage (Figure 7Q and R). These results imply that Fbxl15 is able to destabilize Smurf1‐GFP. As demonstrated above, FBXL15‐promoted Smurf1 degradation in vitro requires K355 and K357 within the WW‐HECT linker region. We found that fluorescence produced in zebrafish embryos by Smurf1Mt‐GFP that had K355R and K357R mutations in Smurf1 was not affected by fbxl15 mRNA injection (Figure 7S and T). Similarly, fbxl15 overexpression promoted degradation of Smurf2‐GFP in zebrafish embryos (Figure 7U and V). We conclude that Fbxl15 has the ability to facilitate Smurf1 and Smurf2 degradation in zebrafish embryos.
FBXL15 is involved in the regulation of bone homoeostasis in adult rats
We next asked whether FBXL15 also has a role in adult development, especially in the maintenance of bone homoeostasis. Smurf1 has been determined to specifically suppress bone formation in mammals (Zhao et al, 2003, 2010; Yamashita et al, 2005). Consistent with this notion, the deficiency of its auxiliary factor CKIP‐1 in mice resulted in augmented bone mass and enhanced osteoblast activity as well as reduced Smurf1 activity (Lu et al, 2008). We then hypothesized that FBXL15, as a negative regulator of Smurf1, has a positive role in bone formation. If true, FBXL15 depletion should lead to a decrease in bone mass.
To verify this hypothesis, we evaluated the effect of FBXL15 depletion by RNA interference (RNAi) in normal adult rats on bone mass. First, we designed three independent siRNAs specifically targeting the rat FBXL15 gene and determined the knockdown efficiency in the rat osteoblast‐like cell line UMR106. Quantitative real‐time PCR analysis showed that all of the examined three siRNAs displayed high knockdown efficiency of >90% (Figure 8A). Second, the siRNA #1 with the highest knockdown efficiency (96.4%) were integrated with a liposome‐based bone‐targeting delivery system (BTDS) (Jeffs et al, 2005) to be delivered to and enriched at the bone formation surface. Third, we examined the apparent bone mineral density (apparent BMD), three‐dimensional architecture parameters, and bone histomorphometric parameters in trabecular bone after administration of the FBXL15 siRNA in 6‐month‐old female rats. The microCT analysis showed that FBXL15 siRNA treatment for 6 weeks significantly decreased the apparent BMD, relative bone volume, trabecular thickness (Tb.Th), and trabecular number (Tb.N) compared with the age‐matched (AM) group, non‐sense siRNA group (NC), or BTDS only group (without siRNA) (VC) (P<0.05 for all) (Figure 8B–E). No significant difference was found in trabecular separation among all the groups (P>0.05) (Figure 8F). Importantly, less organized three‐dimensional architecture and lower bone mass in trabecular bone were clearly found in the rats treated with the FBXL15 siRNA compared with the above control groups (Figure 8G). Overexpression of FBXL15 in rat osteoblast cell line UMR106 was sufficient to induce an increase of ALP activity, but the FBXL15‐ΔF was not (Supplementary Figure S11). Collectively, these data strongly indicate that FBXL15 is a positive regulator of bone mass maintenance in adult mammals.
Compared with RING‐type E3s, there still remain many more unanswered questions regarding HECT‐type E3s, among which the question how HECT domain E3 activity is regulated constitutes a particularly significant gap. Nedd4 family members are the best characterized subgroup of HECT‐type E3s, and several mechanisms in multiple layers of regulation have been determined, such as E3‐substrate recognition, E2–E3‐binding affinity, and relief of intra‐molecular autoinhibition (Ogunjimi et al, 2005; Gallagher et al, 2006; Oliver et al, 2006; Oberst et al, 2007; Wiesner et al, 2007; Bruce et al, 2008; Lu et al, 2008; Mund and Pelham, 2009). Among this family, Smurf1 and Smurf2 are attractive because they have an important role in the regulation of adult bone remodelling, embryonic development and tumourigenesis through controlling the BMP/TGF‐β, Wnt‐PCP, and RhoA pathways (Zhu et al, 1999; Wang et al, 2003; Zhao et al, 2003; Izzi and Attisano, 2004; Yamashita et al, 2005; Chen and Matesic, 2007; Bernassola et al, 2008; Inoue and Imamura, 2008; Suzuki et al, 2008; Jin et al, 2009; Narimatsu et al, 2009). Although recent studies have identified the activation mechanisms of Smurf1 (Lu et al, 2008) and Smurf2 (Ogunjimi et al, 2005; Wiesner et al, 2007) by CKIP‐1 and Smad7, respectively, how their activity is negatively regulated still remains unclear.
In this study, we identified an F‐box protein FBXL15 as a substrate‐recognizing subunit of the SCF ligase complex to target Smurf1, and also Smurf2, for ubiquitination and proteasomal degradation (Figure 8H). FBXL15 interacts with Skp1 and Smurf1 through its F‐box and LRR domains, respectively; thereby promotes the ubiquitination of Smurf1 on two lysines within the WW‐HECT linker. The degradation of Smurf1 by FBXL15 is dependent on the integrity of a functional SCF complex. To our knowledge, this is the first case to show a typical HECT E3 (Smurf1) acts as a substrate to be degraded in trans by a RING E3, especially by a multi‐subunit SCF complex. Under the in vitro ubiquitination assay system, FBXL15 alone seems not to change the auto‐ubiquitination ability of Smurf1 itself, whereas the previous identified Smurf1 co‐factor, CKIP‐1 (Lu et al, 2008), can do it (Figure 3E). Also, depletion of Smurf2 in human embryonic kidney HEK293T cells also had no significant effects of the FBXL15 on Smurf1 ubiquitination (Supplementary Figure S3). However, we cannot rule out the possibility that the SCFFBXL15 ligase complex coordinates with Smurf1 to control the ubiquitination and stability of Smurf1 itself, since the purified Cullin1/Roc1/Skp1 complex in the ubiquitination assays of this study (Figure 3D) might also co‐purify the endogenous FBXL15 and Smurf1 proteins. In addition, we cannot rule out the possibility that SCFFBXL15 complex might coordinate with Smurf2 to control Smurf1 stability in certain cell types or pathophysiological conditions. In this regard, a recent study showed that Smurf2 interacts with Smurf1 to induce the degradation of Smurf1 to prevent migration of breast cancer cells (Fukunaga et al, 2008). Previous studies have shown that RING E3s can be targeted by certain HECT E3s; for example, Cbl is degraded by Nedd4‐1 and Itch/AIP4 (Magnifico et al, 2003), whereas RNF11 is a target of Smurf2 (Subramaniam et al, 2003). A recent study demonstrated that the N‐end rule pathway is mediated by a complex of RING‐type Ubr1 and HECT‐type Ufd4 ligases (Hwang et al, 2010). Our current data, combined with these reports, suggest that the crosstalk between RING E3 and HECT E3 may be more general than once thought.
WW domains of Smurf1 function as the substrate‐recognition module, whereas the HECT C‐lobe and the small subdomain of the N‐lobe are responsible for ubiquitin thioester intermediate formation and E2 binding, respectively (Huang et al, 1999; Rotin and Kumar, 2009). The WW domains linker is also involved in the substrate binding and functions as an adaptor‐targeting site for Smurf activation (Lu et al, 2008; Chong et al, 2010). In contrast, the function of HECT N‐lobe large subdomain and the role of WW‐HECT linker are still unknown. This study provides new insight into these issues: the large subdomain is the recognition site of FBXL15, and the WH linker contains the ubiquitination sites K355 and K357 (Figure 8H). Consistent with the former notion, all nine members of Nedd4 family were co‐immunoprecipitated with FBXL15, implying that the HECT large subdomain contains a certain motif or sequence that is commonly recognized by the LRR domain. Therefore, the in trans ubiquitination represents a novel regulatory mechanism to coordinate the homoeostasis of Smurf1/2 and the linker regions may serve as regulatory sites for E3 activity.
One of the long‐standing central issues in SCF ligase studies is to identify the physiological substrates, as the substrates of nearly two‐thirds of F‐box proteins have not been characterized (Skaar et al, 2009a, 2009b). In this regard, this study identified Smurf1 and Smurf2 as endogenous substrates of FBXL15 (but not other F‐box proteins) and highlighted a positive role of FBXL15 in BMP signalling both in zebrafish and in mammals. Depletion of fbxl15 in zebrafish embryos resulted in embryonic dorsalization with altered dorsal and ventral marker expression (Figure 7), mimicking bmp2b‐deficient mutant swirl (Kishimoto et al, 1997) and bmp7‐deficient mutant snailhouse (Dick et al, 2000). This evidence is the first to directly show a role for an F‐box protein in the determination of the dorsalization–ventralization axis. Furthermore, FBXL15 depletion in adult rats resulted in a decrease of bone mass, bone mineral density, Tb.N and Tb.Th (Figure 8). Although we cannot rule out the possibility that FBXL15 may target more substrates than Smurf1/2 in different tissues or developmental stages, our present evidence at least partially reflects the in vivo physiological relevance between FBXL15 and Smurf1. Deeper investigations surely need to be performed to verify the significance of FBXL15 in vivo and the correlation between FBXL15 and Smurfs in certain spatial‐temporal conditions. Generation of an FBXL15‐deficient mice model should represent one of such approaches and is now in preparation.
Materials and methods
Immunoprecipitation and immunoblotting
For general cell lysis, transfected cells were harvested and lysed in HEPES lysis buffer (20 mM HEPES pH 7.2, 50 mM NaCl, 0.5% Triton X‐100, 1 mM NaF, 1 mM dithiothreitol) for 30 min at 4°C and boiled with 2 × SDS/PAGE loading buffer for 10 min. For immunoprecipitation, cell lyates were prepared in 500 μl HEPES buffer supplemented with protease inhibitor cocktail (Roche) and phosphatise inhibitors (10 mM NaF and 1 mM Na3VO4). Immunoprecipitations were performed using mouse anti‐Myc (2 μg, MBL), anti‐Flag (2.5 μg, Sigma), or Smurf1 (2 μg, Abcam) monoclonal antibodies for 4 h at 4°C followed by incubation with protein A/G‐agarose beads (Santa Cruz) overnight at 4°C. Beads were then washed three times in HEPES lysis buffer and examined by immunoblotting with the indicated primary antibodies and appropriate secondary antibody, followed by detection with SuperSignal chemiluminescence kit (Pierce). Quantification of the visualized bands by densitometry was analysed using Scion Image software.
In vivo ubiquitination assays
To determine in vivo ubiquitination, HEK293T cells were transiently transfected with HA‐ubiquitin, FBXL15, and either Smurf1 (WT/CA/K355+357R) or other members of Nedd4 family, as indicated. Thirty‐six hours after transfection, cells were treated with MG132 for 12 h and then cells were harvested. To rule out non‐specific ubiquitinaton detection, immunoprecipitations were prepared in 500 μl denatured modified RIPA buffer (10 mM Tris–HCl, pH 7.5/5 mM EDTA, 150 mM NaCl, 1% Nonidet P‐40, 1% sodium deoxycholate, 0.025% SDS, protease inhibitors) supplemented with protease inhibitor cocktail (Roche) and phosphatase inhibitors (10 mM NaF and 1 mM Na3VO4) and then incubated with anti‐Myc, anti‐Smurf1, or anti‐Flag antibody for 3 h and protein A/G‐agarose beads for another 8 h at 4°C. After washing for three times, ubiquitinated conjugates were detected by immunoblotting with indicated antibodies.
In vitro ubiquitination assays
His‐Smurf1 (wild type and C699A mutant) and GST‐FBXL15 were expressed in Escherichia coli BL21 and purified according to the manufacturer's instructions. SCFGST‐FBXL15 and SCFGST‐FBXL15‐ΔF complexes were purified by GST‐affinity column from HEK293T cell lysates expressing GST‐FBXL15 or GST‐FBXL15‐ΔF, together with Myc‐Cul1, Myc‐Skp1, and Flag‐Roc1. Cul1, Skp1, and Roc1 proteins were purified by immunoprecipitation with Myc‐antibody or Flag‐antibody plus protein A/G agarose. The assays to assess the in vitro ubiquitination of Smurf1 were carried out in 30 μl ubiquitination assay buffer (50 mM Tris pH 8.0, 50 mM NaCl, 1 mM DTT, 5 mM MgCl2, and 3 mM ATP), with 0.7 μg of E1, 1 μg of UbcH5c (E2), 15 μg HA‐ubiquitin (all from Boston Biochem), 0.7 μg of His‐Smurf1 (wild type or mutant as indicated), and 1.5 μg GST‐FBXL15 or SCFGST‐FBXL15 or SCFGST‐FBXL15‐ΔF complexes (SCF complex contains Myc‐Cul1, Myc‐Skp1, Flag‐Roc1, and GST‐FBXL15 or GST‐FBXL15‐ΔF) were incubated at 30°C for 1.5 h and terminated with sample buffer before western blotting with anti‐Smurf1 monoclonal antibody.
The FBXL15 siRNA‐targeting human FBXL15 (1#: 5′‐CACCCUGGAGCUUCAAAUATT‐3′, 2#: 5′‐GGAACUGCCCAGAACUCCATT‐3′, 3#: 5′‐GCCUGAGCCGCUUGCGGAATT‐3′), the FBXL15 siRNA‐targeting rat FBXL15 (1#: 5′‐CACCCUUUGUCAACCUACATT‐3′, 2#: 5′‐CACCCUGGAAUUUCAAAUATT‐3′, 3#: 5′‐GGCCAAAGCUAGUAAAGCUTT‐3′), the Cullin1 siRNA (1#: 5′‐GGUUAUAUCAGUUGUCUAA‐3′, 2#: 5′‐CAACGAAGAGUUCAGGUUU‐3′), the Roc1 siRNA (1#: 5′‐GAAGCGCUUUGAAGUGAAA‐3′, 2#: 5′‐GCAUAGAAUGUCAAGCUAA‐3′) (Salahudeen et al, 2009), the Smurf2 siRNA (5′‐CCUUCUGUGUUGAACAUAA‐3′), and the non‐targeting siRNA (5′‐UUCUCCGAACGUGUCACGUTT‐3′) were synthesized by Shanghai GenePharm. All siRNA transfections were performed using Lipofectamine 2000 (Invitrogen) according to the commercialized protocol, and the RNAi efficiency was assessed by western blot analysis.
Zebrafish Tuebigen strain was used. Embryos were raised in Holtfreter's solution at 28.5°C and staged as previously described (Kimmel et al, 1995).
mRNAs, morpholinos, and microinjection in zebrafish embryos
mRNA was in vitro synthesized using mMESSAGE mMACHINE® kit (Ambion), purified with RNeasy Mini kit (Qiagen), and dissolved in RNase‐free water. Morpholinos were synthesized by Gene Tools, LLC, and dissolved in water. The related morpholinos sequences were as follows: fbxl15‐MO1, 5′‐CAGGTTTTTGAT CCATGCTGACACT‐3′, blocking translation of fbxl15; fbxl15‐cMO1, 5′‐CAGCTTT TTCATCGATGGTCACAGT‐3′, a control morpholino with six mismatched nucleotides (underlined) according to fbxl15‐MO1; fbxl15‐MO2, 5′‐ CCTGTGGTCCGTGTACATTATTATT‐3′, blocking translation of fbxl15; p53‐MO, 5′‐GACCTCCTCTCCACTAAACTA CGAT‐3′, which was used to suppress non‐specific activation of morpholinos (Robu et al, 2007). The 5′UTR of Zebrafish fbxl15 was amplified by PCR and inserted into XhoI and BamHI sites of pEGFP‐N3 vector for examining the efficiency of fbxl15‐MO. About 1–1.5 nl mRNAs or morpholinos solution was injected into embryos at one‐ to two‐cell stages. When co‐injection of mRNA and morpholino was performed, the mRNA was first injected, followed by the morpholino injection. For mRNA rescue experiment, the fbxl15 sequence region complementary to the fbxl15‐MO1 sequence was mutated by PCR with the modified forward primer 5′‐GGAATTCCACCATGGACCAGAAGCCGGACGAGCGCATGCAAAGCC‐3′, so that fbxl15‐MO1 could not bind to the modified fbxl15 mRNA. The mutant fbxl15‐ΔF is constructed by deleted F‐box domain (1–66 amino acids) at the NH2 terminus of fbxl15.
Whole‐mount in situ hybridization
Digoxigenin‐UTP‐labelled antisense RNA probes were generated by in vitro transcription with Roche kit. Whole‐mount in situ hybridization followed the standard procedure. The stained embryos after in situ hybridization and GFP‐expressing embryos were photographed using an SPOT Insight camera under the Leica MZ16 microscope.
Statistical evaluation was carried out using a Student's t‐test.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank Drs Dahua Chen, Di Chen, Ye‐Guang Chen, Hugues Dardente, Kohei Miyazono, Yongfeng Shang, Wesley I Sundquist, James A Wohlschlegel and Yue Xiong for providing materials. This research was supported by the Chinese National Basic Research Programs (2011CB910602, 2007CB914601), the Chinese National Natural Science Foundation Projects (31071233, 30830029, 30970601, 30830068, 30921004) and the National Key Technologies R&D Program for New Drugs (2009ZX09503‐002, 2009ZX09301‐002).
Author contributions: LZ, FH, and SJ co‐organized and performed project planning, data analysis, discussion, and writing; YC and SH performed the main experimental work of Smurf1‐FBXL15 interaction, Smurf1 ubiquitination by FBXL15, and molecular mechanism investigations; CX performed the main experimental work of FBXL15 function analysis in zebrafish embryos; KL and GX participated in the bone homoeostasis analysis; JW performed the yeast two‐hybrid screening; AM was involved in the zebrafish phenotype analysis.
- Copyright © 2011 European Molecular Biology Organization