Requirement of Smurf-mediated endocytosis of Patched1 in sonic hedgehog signal reception

Reviewer #1:

In this manuscript, Yue et al investigated the role of Smurf-mediated Ptch1 ubiquitination in the regulation of Shh signaling. […] However, the authors should address the following concerns either by discussion or by additional experiments before publication is recommended.

1) Whereas the evidence for Smurf/PY-mediated Ptch1 endocytosis and degradation is strong, it is not so clear how this process is promoted by Shh. Although Shh-induced upregulation of Smurf could contribute, other mechanisms may exist. Have the authors examined whether Shh promotes the binding of Smurf to Ptch1? For example, does Shh treatment increase the FRET between GFP-Smurf2CG and Ptch1-RFP shown in Figure 9A-C? On the other hand, Smurf-mediated degradation of Ptch1 could be Shh independent, as suggested by Casali (Science Signaling, 2010). For example, Ptch1Δ2, which has the Shh binding domain deleted (Briscoe et al., Mol Cell 2001), might still be regulated by Smurf/PY. Furthermore, Huang et al suggested that Smurf prefers degrading ligand-unbound Ptc (Huang et al., PLOS Bio 2013). How could the authors reconcile their finding that Shh promotes Ptch1 degradation? Could they examine whether Ptch1Δ2 is degraded more or less effectively by Smurf than Ptch1 in the presence of Shh?

We thank this reviewer for raising these very important issues. Our previous and new data indicate that Shh promotes the Smurf-mediated endocytosis of Ptch1 in several ways. First, Smurfs are preferentially localized in the nucleus in normal cells (Kavsak et al., Mol Cell 6:1365-75, 2000) and play important roles in maintaining the genomic stability (Blank et al, Nature Medicine 18:227-34, 2012). In the revised manuscript, we show that Shh promotes a re-pivoting of Smurf2 from the nucleus to the cytoplasm (Figure 7B, and Figure 7–figure supplement 1). Second, our data also show that Shh induces Smurfs expression (Figure 5E, 5G). So, these two events should lead to an increase of the effective cytoplasmic concentration of Smurfs. Third, as requested, we conducted a new FRET experiment and found that Shh indeed promotes the colocalization of Ptch1 and Smurf2 (Figure 7E). Fourth, we further add new data showing that ShhN treatment enhances the ubiquitin modification of Ptch1 (Figure 9D), consistent with our data showing that Shh promotes Ptch1 turnover (Figure 8).

In Huang et al, the authors ectopically expressed activated Smo mutants, Smo^SD, in the entire A-compartment, which drastically increased the level of Ptc (Huang et al, Figure 4D). They argue that DSmurf prefers the ligand-unbound Ptc as a substrate because ectopic expression of DSmurf reduced Ptc staining selectively in the A compartment. However, comparing their Figure 4D and 4E, one could find that the intensity of Ptc staining at the A/P boundary was also reduced by DSmurf, notwithstanding the fact that Ptc is normally high at the boundary. On the other hand, since the authors did not examine the distribution of Hh in the disc that received the ectopically expressed Smo^SD, it would be an unsupported assumption that the elevated Ptc in the A compartment was still in the unbound form. After all, the Hh ligand is normally restricted to the compartmental border by the high level of Ptc there. If the border stripe of high level Ptc was made to expand, Hh zone should expand with it. Furthermore, it is well known in the field that Ptc and Smo, when over-expressed, tend to form a nonphysiological complex (Stone et al, Nature 384:129, 1996, and Taipale et al, Nature 418:892, 2002).This raises a possibility that the nonphysiological Ptc-Smo complex could trigger an “unfolded protein response” of some sort that leads to the DSmurf-mediated destruction. This type of degradation is very different from the one that we describe in our manuscript, although both could be mediated by the Smurf E3 ubiquitin ligases, even specifically.

Notwithstanding the above analysis, assuming DSmurf does prefer the ligand unbound form of Ptc for degradation, this would put the site of DSmurf action in the A compartment, where Ptc level is low and Smo is in an inactive state. However, their data indicated that Smo has to be activated in order to promote Ptc degradation. In Huang et al, there is no data that either indicate or imply the source of the activated Smo for activating the Smurf-mediated Ptc turnover or to explain this conspicuous conflict.

We measured the turnover rate of the loop2 mutant of Ptch1 in wt MEFs, and found that the effect of Shh ligand induction was abolished (Figure 8E, 8F). We further quantified the turnover rate of Ptch1 in Smo^null cells, and found that Shh still promotes Ptch1 turnover there (Fig.8G, 8H). Moreover, we did not detect interaction between Smo and Smurfs by co-IP experiments, even though Smurf was shown to bind Ptch1 readily (Figure7–figure supplement 2). So, these results demonstrate that Smurfs likely promote Ptch1 endocytic turnover through direct binding, rather than using Smo as an intermediate, as suggested by Huang et al. However, Smo probably still has a long term feedback role through enhancing downstream Smurf gene expression.

2) Huang et al argued that Smurf-mediated ubiquitination and degradation of Ptc are promoted by activated forms of Smo (Smo^SD) in Drosophila (Huang et al., PLOS Bio 2013). Have the authors examined whether Shh promotes Ptch1 degradation through Smo? For example, does overexpression of mammalian Smo^SD promote Smurf-mediated ubiquitination/degradation of Ptch1?

As stated above, we examined Ptch1 turnover in Smo^null cells, and found that Shh still promotes Ptch1 turnover. We also found by Co-IP experiment that Ptch1 binds Smurf but Smo does not (Figure 7-figure supplement 2). These data strongly argue that Shh-induced, Smurfs-mediated Ptch1 endocytic turnover is independent of Smo.

3) The authors showed that mutating the PY motifs or Smurf1/2 affected both Ptch1 ciliary exit and endocytosis. Is the failure of Ptch1 ciliary exit the result of defective endocytosis? Or could Shh induce Ptch1 ubiquitination in the primary cilium, which may directly regulate ciliary exit of Ptch1? Is there any evidence that Smurf1/2 can be found in the primary cilium with or without Shh treatment? In Figure 6C, can Shh trigger ciliary exit of Ptch1 in the absence of Smurf1/2? Does pharmacological blockage of Ptch1 endocytosis/degradation affect Ptch1 ciliary exit?

It is our interpretation that Ptch1Δ2PY fails to exit cilia because of defective endocytosis. Despite an initial hypothesis, we found neither endogenous nor transfected Smurfs in the cilia with or without Shh treatment. Our data also show that the Shh-induced ciliary export of Ptch1 was compromised when Smurf1 and Smurf2 were knocked down with siRNAs (Figure 5–figure supplement 1). We further show that blocking Ptch1 endocytosis with Leupeptin also blocked its ciliary exit (this result were not included in the previous submission, but is now added as Figure 3–figure supplement 1 in the revised manuscript).

4) The effect of Δ2PY-GFP on Smo ciliary localization presented in Figure 4 does not match the quantification well, especially at 4 hours after Shh treatment where there is almost no difference in the ciliary Smo levels between Δ2PY-GFP and Ptch1-GFP (Figure 4C) while there is a 2-told difference in the quantification (Figure 4D). The authors need to provide a better image reflecting the quantification. Of note, it has been shown that Shh/Ptch1 regulates both the ciliary localization and conformation of Smo (Zhao et al., nature 2007). Have the authors examined whether Δ2PY-GFP affect mSmo conformation using FRET analysis?

We replaced the images in the old Figure 4C with better ones in the revision (new Figure 3C). By using a sophisticated FRET imaging approach, Zhao et al elegantly demonstrated that Hh induces phosphorylation and a conformational change of Smo c-tail that result in Smo dimerization and activation of downstream signaling. Their work also extended this observation to mammalian Smo. However, this regulation, albeit a likely key event in the Shh pathway activation, lies downstream to Ptch1 functions. Since we have demonstrated that Shh-induced Ptch1 endocytic turnover is independent of Smo, and analyzed extensively the ciliary trafficking of Smo, another well recognized key event of the Shh pathway activation, we felt that examining Δ2PY-GFP on mSmo conformation would be a repetition of an already well-addressed issue. In addition, setting up the FRET experiment on Smo conformation would not be a trivial endeavor, if one needs to do it properly. If this reviewer and the editors deem this FRET experiment absolutely essential, which we would respectfully disagree, we will perform as demanded, provided that we are granted additional time.

Reviewer #2:

In this manuscript Yue et al. uncover a role for the Hect E3 ligases Smurf1 and Smurf2 in promoting Hedgehog-dependent changes in the subcellular localization of the Patched receptor that leads to their increased turnover. […]

The strongest aspects of this manuscript are the loss of function experiments conducted with the Sf1-/-,Sf2fl/fl MEFs and GPCs. Indeed, the complete absence of Sf1 and Sf2 leads to a remarkable inhibition of Smo and Gli3 ciliary localization and blunting of Shh-promoted induction of Gli1 levels in MEFs. These results are strongly supported by the experiments in Figure 11 showing a reduction of neural progenitors in cerebellar slice cultures knockout for Sf1 and Sf2 and an inability of Shh to promote the in vitro proliferation of granule cell progenitors when Sf1 and Sf2 are knocked out. These experiments strongly support an important functional requirement of Smurf proteins for Hedgehog signal transduction.

In terms of mechanisms describing the function of Smurf proteins, the evidence presented in the manuscript are however disappointing in that they are too often not convincing, confusing or incomplete. For example, according to their model, Hedgehog ligands are shown to promote the localization of Patched in caveolae, a transitory localization that promote the Smurf dependent ubiquitination of Patched and its endosomal routing to the lysosomes where it is degraded. First of all, although scattered evidence suggests that caveosomes and endosomes may physically interact in specific contexts, the authors present their evidence supporting a role of Rab proteins and endosomal trafficking in promoting Patched exit from caveolae as a well defined and accepted mechanism. However, caveolae-mediated endocytosis is most often described to be separate from endosomal sorting. Although this could represent a novel sorting mechanism for cell surface receptors, the characterization of this process needs to be strengthened and better discussed.

We agree with this reviewer that caveolae was a recently recognized alternative route for internalization of membrane-bound ligand-receptor complexes, but this phenomenon was actually noted more than two decades ago. At that time, a term of “potocytosis” was coined to distinguish it from the Clathrin-mediated endocytosis (Anderson RG, Science 255:410-1, 1992; Gleizes PE, Eur. J. Cell Biology 71:144-53, 1996), because the cargo of potocytosis was thought to be emptied directly into the cytosol. Later studies demonstrated that caveolae-mediated internalization actually feeds into the conventional endocytic pathway, and “caveosomes”, which were previously regarded as independent organelles distinct from endosomes, were actually late endosomes modified by the accumulated Caveolin-1 therein (Hayer et al, J Cell Biol 191:615-29, 2010; Sandvig et al, Curr Opin Cell Biol 23:413-420, 2011). To clarify this issue, we made modifications in the Introduction and cited several key references.

Moreover, all of the evidence supporting the localization of Ptch in different subcellular fraction relies on overexpression experiments and on colocalization with overexpressed markers tagged with fluorescent proteins (especially important for Rab7). These experiments should be repeated using endogenous proteins and images obtained at higher resolution to more precisely follow the fate of Ptch trafficking and more convincingly support the implication of caveolae and/or endosomal trafficking.

Antibodies again mouse Ptch1 are not commercially available, precluding a direct visualization of the endogenous Ptch1, which is present at extremely low level in cells (Rohatgi et al Science). Fluorescence labeled Rab5, Rab7, and Lamp1 are widely used for marking early endosomes, late endosomes, and lysosomes, and the data in question were generated through confocal imaging on a newly acquired Zeiss LSM710 microscope. We have repeated the experiments in question using Ptch1GFP and antibodies against endogenous Rab5, Rab7, and Lamp1, respectively. The data are displayed in new Figure 2 and Figure 2–figure supplements 1 and 3. Signals from antibody staining of endogenous proteins were quite low, probably reflecting the low abundance of the interacting species or the low avidity of this commercial antibody, nevertheless, colocalization between Ptch1GFP and Rab7 poitive late endosomes was confirmed. We also showed colocalization between Ptch1GFP and Lamp1 positive lysosomes using leupeptin to block proteolysis. However, we were unable to detect colocalization between Ptch1GFP and early endosomes (Rab5) without or with ShhN treatment, confirming our previous finding that Ptch1 traverses from lipid rafts directly to late endosomes, bypassing early endosomes. Finally, the d2PY mutant was not colocalized with any of these vesicles. We want to emphasize that these confocal images presented were taken in z-stack using a 63x oil lens. Some images may appear fuzzy, particularly in colocalizing areas/vesicles. This is likely because only a very small fraction of cytoplasmic Ptch1 is channeled to the endocytic pathway; the bulk of forced expressed Ptch1 still turns over via proteasomes (Figure 1F).

In addition in my opinion the biggest question that is left unanswered is how ubiquitination of Patched by Smurf proteins contributes to its function. Do Smurfs lead to Patched mono-ubiquitination or to K63 or K48 ubiquitin chain conjugations? Is ubiquitination involved in Patched endocytosis per se or in its sorting from endosomes to lysosomes? Does Hedgehog ligand promote the interaction of Patched with Smurfs? Do Hedgehog ligands promote Patched ubiquitination?

In our humble opinion, elucidation of the type of Smurfs-mediated ubiquitin modification of Ptch1is certainly informative, but is nevertheless a mechanistic detail in our investigation. It is also extremely difficult to visualize monoubiquitination of Ptch1 under natural settings, given the size of this protein. We did however use mutant forms of ubiquitin and found that Smurf2 promotes Ptch1 to undergo both K63 and K48 ubiquitin chain-mediated ubiquitination (new Figure 9C). We further show that Shh-N promotes interaction of Ptch1 with Smurfs (new Figure 7E) and Ptch1 polyubiquitination (new Figure 9D). Because Ptch1 ΔPY is accumulated in Caveolin-positive lipid raft but not in late endosome (Figs.1A, 1B, and 2A, 2B), we believe that Smurf-mediated Ptch1 ubiquitination is involved in sorting of Ptch1 from lipid raft to late endosomes.

There also seems to be a disconnection between the results obtained using the Ptch-d2PY mutant (when rescuing the Ptch1-/- MEFs) and the results obtained in the Sf1, Sf2 double KO cells. Indeed, whereas the Shh-promoted accumulation of Smo and Gli1 activation are blunted in the dKO cells, Smo accumulation is only reduced when the d2PY mutant is expressed (4C,D). Since the interaction between the d2PY mutant and Smurf proteins seems to be completely abolished (9E) how is this explained? If there is more Ptch1-d2PY in cilia, why do Smo enters at all?

We replaced the d2PY images in old Figure 4C as well as those in old Figure 8B with new ones that better reflect the corresponding statistic graphs. We apologize for those images that may have exaggerated the difference. Judging from the data graphs, it is clear that the reduction in Smo ciliary localization and Gli1 activation caused by d2PY deletion is clearly in line with that by Smurfs knockdown (compare Figure 3D, time point 1-4 hours vs. Figure 6B, 6C).

With regard to the last question, the current paradigm of Ptch1 inhibiting Smo by preventing the latter entry into cilia is based on the observation that Smo moves in whereas Ptch1 moves out of cilia under the influence of Shh (Rohatgi et al, Science 317:372-8, 2007). However, there is no evidence to indicate that the presence of these two membrane receptors in the cilium is mutually exclusive. To the contrary, there are published studies reporting cyclopamine actually promotes Smo entry into the cilium, suggesting that Smo and Ptch1 can co-exist in cilia.

Reviewer #3:

In the manuscript entitled “Requirement of Smurf-Mediated Endocytosis of Patched 1 in Sonic Hedgehog Signal Reception”, Yu et al. present evidence that Smurf1 and Smurf2 promote ubiquitination of PTCH1 resulting in endocytic turnover that is required for HH pathway activation. In particular, the authors provide significant experimental data examining the subcellular localization of PTCH1 and the role of two PPXY motifs in regulating PTCH1 localization turnover, and downstream effects on HH pathway function. While, overall the results appear to be of high quality, there are some issues with both interpretation of the data and proper acknowledgement of previous work that the authors must address.

Major comments:

1) There is an unfortunate lack of proper citation of previous work by other labs in this field. Two essential examples include the recent publication of work identifying a role for Smurfs in regulating Drosophila Ptc turnover (Huang et al., PLOS Biology, 2013), and work from Tom Kornberg that defined a role for the PPXY motif in regulating the turnover of vertebrate PTCH1 (Kawamura et al., JBC, 2008). These two papers directly impact the current study by Yue et al., and this work should be considered in the context of these previous studies.

We have cited these two papers and discussed extensively the Huang’s recent publication.

2) In Figure 5, the authors utilize Ptch1-/- MEFs to address differences in the ability of PTCH1 and PTCH1Δ2PY to promote ligand-dependent signaling. However, the authors miss an opportunity to distinguish between the ligand-dependent and ligand-independent effects of PTCH1 in the HH pathway. They should use these cells and constructs to examine the ability of PTCH1 or PTCH1Δ2PY to antagonize SMO in the absence of ligand. That is, Ptch1-/- MEFs display constitutive HH pathway activation; however, re-expressing PTCH1 rescues this pathway activity. The question is whether PTCH1Δ2PY is equally effective? Do the authors observe equivalent antagonism of SMO in these cells? Or is PTCH1Δ2PY a more effective antagonist of SMO than wt PTCH1? These are straightforward questions to address since the authors have all the necessary tools and reagents in hand.

We did the experiment as requested and the results indicate that Δ2PY is equally effective as the wt Ptch1 in antagonizing Smo in Ptch1-/- MEFs (Figure 4C). This is different from the results obtained from Shh-induced signaling events.