Smurf1 is required for lysosome damage responses by autophagic recognition
How can lysosomal damage be transduced to the autophagosomal system? It was reported that recognition of endomembrane injury by Galectins could inhibit mTOR and promote lysosomal biogenesis23,24. Both Galectins and Smurf1 have been established in protection of Mycobacterium tuberculosis, which can cause endomembrane damage25, by recruiting proteasomes and components of the autophagy machinery4,26. However, to our knowledge, there is no literature about a possible connection between Galectins and Smurf1 in response to endomembrane damage. To evaluate whether Smurf1 is involved in lysosome injury, we used the lysosomotropic compound LLOMe (L-Leucyl-L-leucine methyl ester) to damage lysosomal membranes because autophagy receptors are among the best markers of autophagosome in degrading damaged lysosomes27 as they can form discernible puncta to recognize ubiquitinated cargo for selective autophagy. LLOMe-induced endogenous aggregation of LC3 and Ub puncta significantly reduced Smurf1 knockdown compared with parental cells (Fig. 1a-d). In addition, Smurf1 was indispensable for the colocalization of LC3-Ub puncta under LLOMe treatment (Fig. 1c-d). These results demonstrated that Smurf1 was required for recruiting selective autophagy machinery components to damaged lysosomes. Next, we screened cargo receptors, such as p62, NBR1, NDP52, OPTN and TAX1BP1, which are known to play a role in “eat-me” signals in response to endomembrane recognition. We found that Smurf1 was more or less co-localized with these receptors (Extended Fig. S1a). Importantly, an indirect interaction between Smurf1 with p62, NBR1, NDP52, OPTN, TAX1BP1 and LC3 was observed in both co-immunoprecipitation (Co-IP) and pull-down assays (Fig. 1e; Extended Fig. S1b-d). In addition, enhanced ubiquitination levels of p62 and OPTN, but not DNP52 and TAX1BP1, were observed in the presence of Smurf1 (Extended Fig. S1f-h). The recognized lysosomes were degraded by autophagy. Consistently, we confirmed the colocalization between Smurf1 and ATG proteins, as evidenced by the colocalization of Smurf1 with ubiquitin, ULK1, FIP200, Beclin-1 and ATG16 (Extended Fig. S2a). We further observed that Smurf1 indirectly bonded with Beclin-1, but not ULK1, ATG16 and FIP200, in Co-IP and pull-down assays (Extended Fig. S2b-e). However, the ubiquitination modification of Beclin-1 was not observed upon overexpressing Smurf1 (Extended Fig. S2f).
We found that endogenous Smurf1 dots were significantly increased upon LLOMe-induced damage for 2 h, indicating a link between Smurf1 and lysosome autophagic response (Extended Fig. S3a-b). By overexpression of Smurf1, we found an increase in Smurf1 colocalization with the lysosomal-specific marker LAMP1 (Lysosomal Associated Membrane Protein 1) in response to LLOMe, suggesting that Smurf1 was recruited to autophagosome after lysosomal damage in a time-dependent manner (Fig. 1f-g). Considering that Smurf1 is an E3 ubiquitin ligase, we tested whether Smurf1 has a role in the ubiquitination of injury lysosomes. Immunofluorescence assays showed that Smurf1 was recruited to Ub-positive lysosomes, marked by LAMP1, upon LLOMe treatment (Fig. 1g; Extended Fig. S3c-d), and knocking down Smurf1 abolished the ubiquitination of lysosomes marked both by LAMP1 and LAMP2 (Lysosomal Associated Membrane Protein 2; another lysosome specific marker like LAMP1) (Extended Fig. S3d). The diminished ubiquitinated lysosome was rescued by overexpressing wild-type Smurf1, but not the ligase-deficient Smurf1 mutant (C699A), which lost the ability to ubiquitinate lysosomes (Extended Fig. S3e-f). Of note, we also found that the recruitment of p62 and LC3 to LAMP1+ lysosome was at least partially dependent on Smurf1 E3 ubiquitylation activity (Fig. 1g; Extended Fig. S3e-f), indicating that the E3 ligase activity of Smurf1 was required for lysosomal ubiquitination. Next, we evaluated the type of ubiquitin chain of lysosomes (we focused on LAMP2 as the marker for lysosomes) affected by Smurf1. The results showed that LLOMe induced both K48-linked poly-ubiquitination (K48-Ub) and K63-linked poly-ubiquitination (K63-Ub) of LAMP2, which were significantly decreased in Smurf1 suppression compared with parental controls (Fig. 1h-i; Extended Fig. S3g).
Smurf1 promotes damaged lysosomes clearance
Given that the crucially role of Smurf1 on autophagic elimination of mycobacteria in host defense4 and involved in mitophagy28, it is of great interest to explore whether Smurf1 mediates the clearance of damaged lysosomes. Galectins especially Gal3 have been widely established for damaged lysosome markers that they interact with glycans, which only present in the luminal side of the damaged lysosomal membrane, and form the puncta upon lysosomal membrane damage24,26. We firstly monitored the recruitment capability of endogenous Gal3 upon lysosomal injury. There was an increase of Gal3 puncta, normally diffused throughout the cells, at 1 hour of LLOMe in both Smurf1 knockdown and parental cells (Fig. 2a-b). During washout of LLOMe, Gal3 profiles recovered, but less efficiently in Smurf1 knockdown cells, evidenced by Gal3 dots began to decrease and disappeared 12 h upon LLOMe washout in parental cells, but not in Smurf1 knockdown cells (Fig. 2a-c). These results raise the possibility that clearance of LLOMe induced-Gal3 dots was blocked in Smurf1 knockdown cells. To further confirm Gal3 dots were recruited to lysosome, consistently, we found the Gal3 dots were co-localized with lysosomal-specific marker LAMP1 both in Smurf1 knockdown and parental cells upon LLOMe treatment (Fig. 2a). Of note, small amount of Gal3 puncta have also noticed in Smurf1 knockdown cells before LLOMe treatment (Fig. 2a), this may indicate that Smurf1 is required for maintaining lysosome integrity.
The autophagic process is strictly regulated by different ATGs (autophagy-related genes). These genes, which about 38 ATGs have been found and is highly conserved in yeast and mammals, are essential molecules for autophagy and participate in different stages of autophagy29. To identify whether Smurf1 is involved in the autophagic degradation of damaged lysosomes, we tested whether Gal3-positive lysosomes could co-localizate with 13 ATGs which are essential for varies autophagy stages (Phagophore formation: ULK1, FIP200, ATG13, ATG14, Beclin-1 and VPS34; LC3 conjugation system: ATG4B; Elongation and autophagosome formation: Atg9A and WIPI1; ATG12 conjugation system: ATG5, ATG7 and ATG16) and one autophagosome-lysosome fusion protein (UVRAG). As expected, LLOMe induces Gal3+LAMP2+ puncta co-localization with the above ATGs in parental cells, but not in Smurf1 knockdown cells (Extended Fig. S4a). This indicates blockage of Smurf1 suppresses the recruitment of these ATGs, thus, inhibits the lysophagy. Indeed, the co-localization of Gal3 and/or LAMP2 with Ub-, LC3- and p62-positive dots were also significantly blocked by knocking down of Smurf1 in response to LLOMe (Fig. 2d; Extended Fig. S5a-b).
To confirm Gal3 is prohibited for autophagic degradation by inhibition of Smurf1, we constructed tfGal3 (GFP- and RFP-tandem fluorescence tagged Gal3), which GFP signals is rapidly quenched, but the RFP signals maintain relatively stable exposed to acidic lysosome. As shown in Extended Fig. 5c, RFP- and GFP- signals were detected in Smurf1 depleted and parental cells in the treatment of LLOMe at 1 hour (Extended Fig. S5c). The GFP signals have attenuated at 6 h and abolished at 24 h after LLOMe washout in parental cells, but not in Smurf1 knocking down cells (Extended Fig. S5c-d). By comparison, the decreased GFP signals was suppressed by autophagic inhibitor, Bafilomycin A1, by which blocks the fusion of autophagosome and lysosome (Extended Fig. S5d). In addition, to gain insight into the effect of Smurf1 on the degradation of endogenous Gal3, we employed tfLC3 as a reporter of autophagic flux. Both indicate autophagosomes, and only RFP signals represent autolysosomes. The ratio of autophagosomes (RFP+GFP+ puncta) with autolysosomes (RFP+ puncta), which is also colocalized with endogenous Gal3 dots, decreased after LLOMe washout in parental cells. By contrast, clearance of Gal3+ autophagosomes were strongly hindered in Smurf1 knocking down cells (Fig. 2f-g). Altogether, these observations revealed that Smurf1 is required for the clearance of damaged lysosome.
To confirm the Smurf1 mediates the clearance of only damaged lysosome but not the lysosome as a whole, we also stained the specific lysosomal fluorescent probe Lysotracker, which monitors the acidic intact lysosome, during washout. Fluorescent staining indicated knocking down of Smurf1 failed to clear the damaged lysosome (Gal3+ LAMP2+) and to promote lysosomal biogenesis during the washout, evidenced by the intensity of Lysotracker+LAMP2+ was markedly decreased accompanied with augmented Gal3+ dots in Smurf1 depleted cells following LLOM exposure and/or during the washout (Fig. 2j-k). Of note, Lysotracker+LAMP2+ recovered, and Gal3+ LAMP2+ damaged lysosome attenuated in Control cells upon washing out LLOMe (Fig. 2j-k). In sum, Smurf1 plays a vital role in lysosomal homeostasis by promoting both the clearance of the damaged lysosome and also the lysosomal biogenesis.
Smurf1 promotes lysosomal biogenesis by activating TFEB
Given that TFEB is an established transcription factor links autophagy for lysosomal biogenesis 7, we wonder whether Smurf1 regulates TFEB nuclear translocation for regeneration of lysosomal machinery. First, to identify whether Smurf1 induced the lysosomal homeostasis is dependent on TFEB, we knocked down TFEB in overexpressed Smurf1 and parental cells. We found that TFEB is required for the lysosomal clearance and biogenesis (Extended Fig. S6a-b). Second, Smurf1 remarkedly stimulates the nuclear translocation of TFEB evidenced by knocking down smurf1 suppressed, overexpression of smurf1 increased, the nuclear portion of TFEB at 2 hours LLOMe (Fig. 3a-d). Additionally, subcellular fraction assay confirms Smurf1 robustly induced TFEB accumulation in the nuclear (Fig. 4e-f). To examine whether TFEB translocation is dependent on Smurf1 E3 activity, we knocked down endogenous Smurf1 and overexpressed both the catalytic mutant of Smurf1-C699A and wildtype Smurf1. It showed that Smurf1-C699A, which lost intrinsic E3 activity, decreased the nuclear portion of TFEB compared to control at 2 h LLOMe (Extended Fig. S6c-d). Consistent with this notion, a set of lysosomal biogenesis (LAMP1, MCOLN1 and Cathepsin B; CTSB) and ATG (ULK1, p62, Beclin1 and ATG16) genes, which are the direct targets of TFEB, were notably elevated upon Smurf1 overexpression in TFEB-dependent manner (Fig. S6e-g).
Previous studies have shown that the subcellular localization and activity of TFEB are highly controlled by its phosphorylation status. Dephosphorylated TFEB is mediated by MCOLN1-Calcineurin pathway. Calcineurin phosphatase dephosphorylates TFEB and its release from 14-3-3 to drive its nuclear translocation14. To identify whether TFEB nuclear import is dependent on Calcineurin phosphorytase activity, we knocked down endogenous Smurf1 and overexpressed Calcineurin. It showed that overexpression of Calcineurin failed to induce TFEB unclear import by knocking down Smurf1 (Fig. 3g), suggesting Smurf1 is required for Calcineurin activation. Indeed, Cyclosporine A (CsA), a Calcineurin inhibitor, blocked Smurf1 induced TFEB nuclear translocation and the mRNA level of TFEB targets (Fig. S6h-j). Calcineurin is composed of a catalytic calcieurin A (CnA) subunit (includes a calmodulin-binding, CBD; CNB-binding domain, BBD; and an autoinhibitory, AID domain), which encoded by PPP3CA, PPP3CB and PPP3CC, and a Ca2+ binding regulatory cineurin B (CNB) subunit encoded by PPP3R1 and PPP3R2 genes. PPP3CB (includes a catalytic CD domain, 67-351 amino acid; calmodulin-binding and CaM-binding domain, 351-474; and an autoinhibitory AID domain, 475-525) is the most significant hit identified by the primary screening was the calcineurin catalytic subunit isoform beta30 (Gene ID:5532). To identify which component of calcineurin pathway plays a required role in the TFEB nuclear import, we suppressed MCOLN1-Calcineurin pathway by knocked key components of including MCOLN1, PPP3CB and PPP3R1, and checked the subcellular localization of TFEB. It showed that TFEB not able to localized in nuclear by overexpression of MCOLN1, PPP3CB and PPP3R1, but PPP3CB-1-474 (removed the AID domain) can, in Smurf1 knocked down cells (Fig. 3g-h; Extended Fig. S6k), suggesting that MCOLN1-Calcineurin pathway can override for the nuclear import of TFEB mediated by Smurf1. We also found that the overexpression of Smurf1 had a similar effect of promoting TFEB nuclear translocation during 2 hours chase as the overexpression of constitutive activation of PPP3CB-1-474 (Fig. 3g-h; Extended Fig. S6k). This data strongly suggest that Smurf1 may help PPP3CB to remove its AID domain from CD domain. Additionally, Smurf1 induced lysosomal hemostasis is dependent on MCOLN1-Calcineurin pathway evidenced by knocking down PPP3CB or MCOLN1 significantly blocked the lysosomal biogenesis and ATG genes expression (Extended Fig. S6i-m). Importantly, subcellular cytoplasmic and nuclear assay verify that knocking down PPP3CB or MCOLN1 blocked TFEB nuclear import (Fig. 6i-k).
Smurf1 activates TFEB independent of mTOR
In normal sedentary conditions mTORC1 phosphorylates TFEB on the lysosomal surface31. During starvation and lysosomal damage mTORC1 dissociates from the lysosomal surface and its activity is inhibited23. However, immunofluorescence assay identified overexpression of Smurf1 attracted mTOR to damaged lysosomal at 2 hours LLOMe chase (Fig. 4a-b). Given that mTOR dissociates from the lysosome is dependent on the p6232, we also found overexpression of Smurf1 depletes p62 in the cytoplasm to lysosome (Fig. 4a-c), thus may blocks the mTOR dissociation from lysosome. Next, we tested the function of mTOR by detection of the phosphorylation of its downstream targets. Interestingly, LLOMe prohibits the function of mTOR evidenced by decreased phosphorylation of P70S6K, 4EBP1 compared to control, with or without overexpression of Smurf1 (Fig. 4d; Extended Fig. S7a), suggesting mTOR is not involved in Smurf1 mediated LLOMe response.
To examine whether Smurf1 induced TFEB nuclear translocation is dependent on mTOR, we found TFEB was blocked in the cytoplasm in the absence of Smurf1 in the treatment of mTORC1 inhibitors EBSS or Torin1 (Extended Fig. S7b-e). Similarly, subcellular fractionation assays confirmed the fraction of unclear portion of TFEB is suppressed by knocking down of Smurf1 in the treatment of EBSS or Torin1 (Extended Fig. S7b-i), suggesting Smurf1 is required for mTORC1 inhibitors induced nuclear TFEB. To test the hypothesis that Smurf1 regulates TFEB nuclear translocation in an mTORC1-independent manner, we immune-stained the lysosomal marker by suppression of calcineurin pathway. It showed that knocking down PPP3CB depleted LAMP1 expression with or without Smurf1 (Fig. 4e-g), further strengthened calcineurin pathway plays a key role in the lysosomal biogenesis.
Galectin-3 interacts with Smurf1 and PPP3CB
Previously, it reported that lysosomal integrity monitored by galectin-8 (Gal8), which by recruiting NDP52 to activates antibacterial autophagy33. Further study also established that Gal8 also inhibits mTOR in response to lysosomal injury23. Additionally, endomembrane Gal3 exposition leads to recruits endosomal sorting complexes required for transport (ESCRT) components to damaged lysosomes to repair24. To sought to whether Smurf1 was also recruited to lysosome directly dependent on endomembrane damage modulation galectins, immunofluorescence assays were conducted to identify whether Smurf1 associates with galectins exposed to LLOMe. It showed that Gal3 is required for recruitment of Smurf1 to damaged lysosome evidenced by loss of Gal3, but not Gal8, significantly reduced the colocalization of Smurf1 and LAMP1 puncta in response to LLOMe (Fig. S8a). Importantly, Smurf1 were co-localized with Gal3 and LAMP2 (Fig. 5a), suggesting Smurf1 can be recruited by Gal3, but not Gal8. Indeed, both Co-IP and Pull-down assay revealed that Smurf1 indirectly bound to Gal3, but not Gal8, in LLOMe enhanced and E3 ligase activity independent manner (Fig. 5b; Fig. S8b-d). Our data also showed that Smurf1 is not able to ubiquitylate Gal3 (Fig. S8e). Next, we mapped the specific Gal3 (containing non-lectin N-terminal domain, NT; and carbohydrate recognition domain, CRD) and Smurf1 (containing C2, WW, and HECT domains) interaction domain by constructing truncated Gal3 and Smurf1 as indicated in Fig. d-e. We found Gal3 NT domain and Smurf1 WW domain are required for their indirect interaction ex vivo (Fig. 5c-e). To examine whether Gal3 could also recruits PPP3CB to form the complex, we overexpressed Smurf1 and PPP3CB with or without knockdown of Gal3. It showed that Gal3 is required for PPP3CB recruitment (Fig. 5a). By mapping the binding domain of Gal3 and PPP3CB (containing N-term domain, catalytic domain, CD; calcineurin B binding domain, CNB; calmodulin binding domain, CaM; and autoinhibitory domain, AID). Our data revealed that their interaction was dependent on the N-term domain (1-21 amino acid) of PPP3CB and the CRD domain of Gal3 (Fig. 5f-h; Fig. S8f-i). Previously studies have shown that the proline-rich sequence of N-term domain is involved in substrate recognition as identified by PPP3CB-P14G/P18G point mutation and PPP3CB-22-524 truncated mutation34. Indeed, we verify that PPP3CB-P14G/P18G mutation no longer interacts with Gal3 (Fig. 5i; Fig. S8j). The LxVP and PxIxIT motifs recognizes and binds to catalytic domain of calcineurin35. Therefore, we analyzed and found LIVP motif in the sequence of Gal3, which is located at the amino acids 114-117 of CRD domain. We then deleted LIVP amino acids, and found no interactions were observed between PPP3CB and Gal3-∆LIVP (Fig. 5g), suggesting LIVP Gal3 motif is required for its interaction with PPP3CB.
Given that the indirect binding between Smurf1 and Gal3, we next to explore the inter mediator for the interaction between Smurf1 and Gal3. It showed that PPP3CB is required for the indirect association of Smurf1 and Gal3 evidenced by the binding affinity between Smurf1 and Gal3 was abolished by suppression of PPP3CB (Fig. S9a). Consistently, overexpression PPP3CB promoted the interaction of Smurf1 and Gal3 (Fig. S9b). Ca2+ dependent calmodulin induces the conformational change in AID, thereby releasing its inhibitory role on the catalytic effect of CNA36. However, how AID is release from the catalytic domain and/or whether the release is sufficient for calcineurin activation remains poorly understood. Immunoprecipitation assay evidenced that PPP3CB inhibitor (CsA) significantly suppressed, but constitutively activated PPP3CB (PPP3CB-1-474 amino acid) increased, the binding affinity between Smurf1 and Gal3 (Fig. S9c). Previously studies have shown that calcineurin can regulate a wide range of phosphorylation of serine, threonine and tyrosine from its substrates37,38. We wonder if PPP3CB affect the phosphorylation status of Smurf1 and/or Gal3. Immunoprecipitation assay evidenced that CsA significantly increased, but PPP3CB-1-474 decreased both the phosphorylated threonine and tyrosine of Gal3, and threonine of Smurf1 (Fig. S9d-e), indicating that Smurf1 and Gal3 can be phosphorylation modified by PPP3CB. Additionally, we also checked knocking down Smurf1 is not required for the interactions between PPP3CB and Gal3 (Fig. S9f-g). To identify the role of regulatory subunit PPP3R1 in the recruitment by Gal-3, interestingly, we found knocking down PPP3R1 decreased, and overexpression of PPP3R1 increased binding affinity of Smurf1 and Gal3 (Fig. S9h-i), suggesting PPP3R1 also plays a role in the formation of the complex. Two alternative possibilities can explain the cause for PPP3CB promoting the interaction of Smurf1 and Gal3. Firstly, de-phosphorylated Gal3 may change its autoinhibitory conformation evidenced by overexpression or constitutive activation of PPP3CB promotes the disassociation of NT-CRD (Fig. S9j-l). Another one is that PPP3CB serves as a ligand of Gal3 and directly interacts with CRD domain of Gal3 and the interaction was strengthened in the treatment of LLOMe (Fig. S8f-g). Thus, these will facilitate the open conformation of Gal3 and contributes its recruitment activity.
Smurf1 promotes PPP3CB CD and AID domain disassociation
How Smurf1 plays a role in the Gal3-Smurf1-Calcineurin complex? To begin with, we tested whether Smurf1 could directly binds to PPP3CB. As expected, our results identified that Smurf1 directly binds PPP3CB independent of E3 ligase activity, and their binding affinity was significantly enhanced in response to LLOMe treatment (Fig. 6a-b; Fig. S10a-c). Interestingly, Gal3 is also required for the interaction of Smurf1 and PPP3CB evidenced by knocking down Gal3 diminished the Co-IP of Smurf1 and PPP3CB (Fig. 6c), suggesting may help to stabilize the interaction between Smurf1 and PPP3CB. Again, overexpression of Gal3 significantly increased their binding affinity in response to LLOMe (Fig. 6d). Next, we mapped the specific interaction domain of Smurf1 and PPP3CB, and identified Smurf1 WW domain directly interact with CD domain of PPP3CB (Fig. 6e-g), further strengthened the previous finding that the dephosphorylation of Smurf1 by PPP3CB. It is reported that intracellular calcium controlled protein calcineurin phosphatase activity by inducing the release of autoinhibitory AID from the CD domain39. Given that Smurf1 interacts with the CD domain of PPP3CB, we next explore whether Smurf1 affects the association between CD and AID domain of PPP3CB. Strikingly, our data showed that deletion of Smurf1 promoted the binding, and Gal3 overexpression rescued the association of CD and AID of PPP3CB (Fig. 6h). Overexpression of Smurf1 or Gal3 or the treatment LLOMe could completely release the inhibitory effect of AID from CD domain (Fig. 6i; Fig. S10d), and Gal3 deletion could reverse the disassociation of AID from CD in Smurf1 overexpressed cells (Fig. 6j). Altogether, LIVP Gal3 motif is required for its interaction with PPP3CB. Smurf1 directly interacts with PPP3CB for the disassociation of AID from CD domain (Fig. 6k).
Smurf1 promotes TFEB unclear translocation via the ubiquitylation of PPP3CB
Considering that Smurf1 is a E3 ubiquitin ligase and interacted with PPP3CB identified above. We determined whether Smurf1 contributes to the TFEB nuclear translocation through ubiquitylating PPP3CB. Our data showed that Smurf1 suppression markedly inhibited the ubiquitination of PPP3CB (Fig. 7a). The ubiquitination level of PPP3CB was restored upon re-expression with Smurf1-CS (codon switch), but not upon catalytically inactive form Smurf1-CS-C699A, in Smurf1-deficient cells (Fig. 7a). It indicates a role of E3 ligase activity of Smurf1 in the regulation of PPP3CB ubiquitination. To further verify which domain of Smurf1 is responsible for the PPP3CB ubiquitination modification, we first generated a series of truncated mutants of Smurf1 to dissect functional domains of the Smurf1 on PPP3CB ubiquitination. It revealed that the ubiquitination of PPP3CB by Smurf1 was dependent on the C2 recognition and HECT domain E3 activity of Smurf1 (Extended Fig. 11a). We next explore which types of ubiquitin chains were modified to PPP3CB mediated by Smurf1. Interestingly, we found that K63-linked polyubiquitin was selectively conjugated to PPP3CB mediated by Smurf1 (Extended Fig. 11b-f). We further confirmed K63-linked polyubiquitin is conjugated to PPP3CB by using K63R-linked or K48R-linked ubiquitin variants. Our data revealed that PPP3CB was markedly modified by K48R-linked ubiquitin but not K63R-linked ubiquitin (Fig. 7b). These results support our findings that PPP3CB was mainly modified with K63-linked polyubiquitin by Smurf1.
Next, to identify which domain of PPP3CB was ubiquitinated by Smurf1, we first found that the polyubiquitination levels of PPP3CB truncated forms of both 352-525 and 1-66 were significantly abolished even in the presence of Smurf1, but the 1-158 truncated form was significantly polyubiquitinated by Smurf1 (Fig. 7c; Extended Fig. 7g-i). The data indicates that the ubiquitination sites of PPP3CB by Smurf1 are probably located at residues 67-158. Studies have shown that ubiquitin is usually attached to lysine residues (K) of substrates or ubiquitin itself, and there are only three lysine residues K85, K109 and K146 within residues 67-158. In order to know which lysine residues is ubiquitinated by Smurf1, we generated point mutants of PPP3CB by substituting the lysine residues with arginine residues (R). Our data showed that K146R point mutant, but not K85R and K109R failed to be polyubiquitinated by Smurf1 (Fig. 7d; Extended Fig. 11j). We next study the functional importance of ubiquitylation modification at K146. It showed that K146R mutation had a weaker effect on promoting TFEB nuclear import compared to wild-type PPP3CB (Fig. 7d; Extended Fig. S11k). Collectively, these results showed that ubiquitination of PPP3CB at K146 is necessary for TFEB nuclear localization.
Next, we sought to determine the role of Smurf1 in TFEB nuclear import. Previously, we identified Smurf1-mediated the release of AID from the CD domain is essential for PPP3CB activation. In this case to exclude the role of ubiquitination of PPP3CB mediated by Smurf1.We observed the effect of Smurf1 on TFEB nuclear localization in the cells transfected with wild-type PPP3CB or PPP3CB-K146R mutation respectively. Nuclear and cytoplasmic separation assay showed that Smurf1 further strengthen the TFEB nuclear import caused both in overexpression of PPP3CB and PPP3CB-K146R groups compared to control (Fig. 7e). The observation decreased TFEB nuclear import caused by PPP3CB-K146R mutation compared to wild-type PPP3CB group with or without Smurf1 overexpression may due to ubiquitination modification of PPP3CB (Fig. 7e). Overall, the data suggest a pivotal role of Smurf1 on PPP3CB activation and TFEB nuclear import both by dissociation autoinhibitory domain from the catalytic domain, and ubiquitination of PPP3CB.
Smurf1 regulates Gal3 recruitment of PPP3R1
Immunofluorescence assay showed that PPP3R1 also recruited to lysosomes upon LLOMe treatment in Gal3 dependent manner (Extended Fig. 12a). To identify the role of PPP3R1 in the formation of complex, as expected, we first found that PPP3CB directly binds with PPP3R1 in LLOMe enhanced manner (Fig. 8a-b). Considering that PPP3CB directly associates with Gal3, we also checked the interaction between PPP3R1 and Gal3. Indeed, the results showed that PPP3R1 indirectly binds with Gal3 (Fig. 8c). Similarly, LLOMe treatment also promoted the association of PPP3R1 and Gal3 (Fig. 8d). We next mapped the key interaction domain of Gal3 with PPP3R1, and showed the NT domain of Gal3 is essential for the association with PPP3R1 (Fig. 8e-f). Furthermore, PPP3CB overexpression increased, suppression of PPP3CB abolished, the interactions of PPP3R1 and Gal3 (Fig. 8g-h), suggesting PPP3CB is required for the interaction between Gal3 and PPP3R1. Interestingly, we detected that Smurf1, but not MCOLN1, was also binds with Gal3 in LLOMe enhanced manner (Extended Fig. 12b-e). Given that both Smurf1 and PPP3R1 are indirectly bind with the NT domain of Gal3, we asked whether Smurf1 affected the interactions between PPP3R1 and Gal3. Our data indicated that suppression of Smurf1 decreased, overexpression of Smurf1 increased, the interactions of PPP3R1 and Gal3 (Fig. 8i-j), suggesting Smurf1 promotes the recruitment of PPP3R1 by Gal3 (Fig. 8k).
Smurf1 interacts with and ubiquitylates TFEB
Surprisingly, we found that Smurf1 also directly interacts with TFEB and their binding affinity was not dependent on LLOMe or Smurf1-C699A (Fig. 9a-c; Extended Fig. S13a-b). TFEB contains the glycine-rich (Gln rich) domain, the basic helixloop-helix (bHLH) domain, the leucine zipper (Zip) domain, and the proline-rich (Pro rich) domain. We then identified that the HECT domain of Smurf1 interacts with the Pro rich domain (366-443 amino acids) of TFEB (Fig. 9d-f; Extended Fig. S13c-d). Additionally, suppression of Smurf1 significantly block the ubiquitination of TFEB compared to control in the treatment of MG132 (Fig. 9g). Overexpression of Smurf1, but not Smurf1-C699A, could rescue the ubiquitination of TFEB in Smurf1 deleted cells (Fig. 9g; Extended Fig. S13e). To determine which type of lysine ubiquitin chain can be added in TFEB by Smurf1, we carried out immunoprecipitation assays in the presence of different variants of ubiquitin proteins. Interestingly, our data showed that Smurf1 increased K11-, but not K6-, K27-, K29-, K33-, K48- or K63- linked, ubiquitination of TFEB (Fig. 9i; Extended Fig. S13f-i).
We also found HECT domain of Smurf1 could ubiquitinates of TFEB in vitro and in vivo (Fig. 9j; Extended Fig. S13j-k). Additionally, the potential TFEB ubiquitination sites mediated by Smurf1 are resided between 366-443 amino acids evidenced by the ubiquitination level was strongly decreased in of TFEB deleted with 1-365 amino acids (Extended Fig. S13l). We then generated point mutants of K430R and K431R between 366-443 amino acids and showed that K430R, but not K431R and K460R, failed to be ubiquitination by Smurf1 (Fig. 9k; Extended Fig. S13m). Furthermore, we found K430R mutation resulted in a decreased TFEB nuclear import (Fig. 9l), suggesting that Smurf1 mediated ubiquitination of TFEB at K430 is critical for TFEB nuclear translocation. Altogether, these results showed the Smurf1 interacts with and ubiquitylates TFEB (Fig. 9n)
Gal3-Smurf1-PPP3CB complex stabilizes TFEB
Given the data that Gal3-PPP3CB working together to recruits their partner Smurf1 and PPP3R1, we next asked if which factors influence the stability between the complex and TFEB. Firstly, we screened the other component of the complex and showed that the interactions or binding affinity of Smurf1 and TFEB is not dependent on single overexpression or knockdown Gal3, PPP3CB or PPP3R1 (Fig. S14a-h), suggesting that the de-phosphorylation of both Smurf1 and TFEB regulated by PPP3CB did not disturb for their association. To further confirm the status of phosphorylation of TFEB plays a role in its association with Smurf1, we introduced S211 mutation of TFEB, which can be phosphorylated by mTORC1 kinase, and conducted the pull-down assay. It showed that the interaction between Smurf1 and TFEB independently on non-phosphorylatable mutant S211A and the phosphomimic mutant S211D and S211E of TFEB compared to wild-type TFEB (Fig. S14i), suggesting S211 phosphorylated TFEB can also be recruited by the complex. Of note, Smurf1 is connected to mTORC1 inhibitory autophagy re-activation by facilitating TFEB recruitment to PPP3CB for its de-phosphorylation. Interestingly, overexpression or knockdown Gal3 and PPP3CB them together promoted or decreased Smurf1 and TFEB interaction affinity, emphasizing the importance of Gal3-Smurf1-PPP3CB complex in stabilizing the its interaction with TFEB (Fig. 10a-b).
Next, we try to figure out how Gal3-Smurf1-PPP3CB complex regulates the association between PPP3CB and TFEB. Previous studies demonstrated that PPP3CB interacts with and de-phosphorylates TFEB in vivo. However, whether PPP3CB is directly interacts TFEB and the mapped binding domain have remained elusive. We firstly screened the possible interacted partner of PPP3CB and indeed found that LLOMe promotes the binding affinity of PPP3CB with Gal3, Smurf1, and PPP3R1, but not TFEB (Fig. 10c), suggesting lysosome damage promotes the formation of Gal3-Smurf1-PPP3CB complex. We also showed that PPP3CB directly interacts with TFEB (Fig. 10d-e; Fig. S15a). Additionally, it revealed that AID domain of PPP3CB interacted with 444-476 amino acids of TFEB (Fig. 10g-i; Fig. S15b-d). Secondly, to identify whether Smurf1 or Gal3 plays a role in stabilizing the TFEB, it interests to note that suppression of Smurf1 decreased, while overexpression of both wild-type Smurf1 and Smurf1-C699A increased, the binding affinity of PPP3CB and TFEB (Fig. 10j-k). Consistently, Gal3 is required for dephosphorylation of TFEB evidenced by knocking down Gal3 decreased, while overexpression of Gal3 increased, the interaction of PPP3CB and TFEB (Fig. 10i-m). Of note, we found that the enhanced PPP3CB and TFEB interaction mediated by Smurf1 overexpression was abolished by Gal3 deletion (Fig. 10n). Overall, we hypothesis that Gal3 and Smurf1 contributed PPP3CB from “close” to “open” form, which facilitate TFEB docking to the AID domain.
PPP3R1 corrects TFEB conformation for its dephosphorylation
To identify how PPP3R1 plays the molecular role in connection of TFEB activation, we initially checked if PPP3R1 could directly binding with TFEB. Indeed, PPP3R1 also directly binds with TFEB in LLOMe independent manner (Extended Fig. S16a-b), suggesting PPP3R1 may work together with PPP3CB to help TFEB re-position to fit into the CD domain of PPP3CB. In addition, we mapped the 166-319 amino acids of TFEB is responsible for the binding of PPP3R1 (Extended Fig. S16c-e). Additionally, suppression of PPP3CB, but not Smurf1, abolished the association of PPP3R1 and TFEB (Extended Fig. S16f-g), suggesting PPP3CB is required for the interaction of PPP3R1 and TFEB. Conversely, PPP3CB also promotes the interaction of PPP3R1 and TFEB both in vivo and in vitro (Extended Fig. S16h-i), further indicating PPP3CB and PPP3R1 may work together to activate TFEB. Of note, overexpression of PPP3CB enhanced this interaction but this role was remarkedly inhibited by knocking down of Gal3 (Extended Fig. S16i), suggesting Gal3 is required for the activation of TFEB. Taken together, the results suggest a pivotal role of PPP3R1 in stabilization of TFEB (Extended Fig. S16k).
TFEB feedback stabilizes the Gal3-Smurf1-PPP3CB-PPP3R1 complex
Considering that TFEB interacts the AID domain of PPP3CB, we asked whether TFEB could also disassociation AID from CD domain. Our data showed that TFEB overexpression has no effect on the disassociation of CD and AID domain (Fig. S17a), suggesting TFEB is passively recruited to PPP3CB. In addition, Smurf1-PPP3CB, Gal3-PPP3R1 and Gal3-Smurf1 did not interrupted, but Gal3-PPP3CB interaction was increased upon TFEB overexpression (Fig. S17b-e), suggesting TFEB may also maintain the stability of the complex. In contrast, Smurf1-PPP3CB, Gal3-PPP3R1, Gal3-Smurf1 and Gal3-PPP3CB interactions were decreased by knocking down TFEB (Fig. S17f). Alltogether, these data suggested that TFEB also plays a crucial role in the stability of Gal3-Smurf1-PPP3CB-PPP3R1 complex as a positive feedback mechanism.