Cellular and viral peptides bind multiple sites on the N‐terminal domain of clathrin

Short peptide motifs in unstructured regions of clathrin‐adaptor proteins recruit clathrin to membranes to facilitate post‐Golgi membrane transport. Three consensus clathrin‐binding peptide sequences have been identified and structural studies show that each binds distinct sites on the clathrin heavy chain N‐terminal domain (NTD). A fourth binding site for adaptors on NTD has been functionally identified but not structurally characterised. We have solved high resolution structures of NTD bound to peptide motifs from the cellular clathrin adaptors β2 adaptin and amphiphysin plus a putative viral clathrin adaptor, hepatitis D virus large antigen (HDAg‐L). Surprisingly, with each peptide we observe simultaneous peptide binding at multiple sites on NTD and viral peptides binding to the same sites as cellular peptides. Peptides containing clathrin‐box motifs (CBMs) with the consensus sequence LΦxΦ[DE] bind at the ‘arrestin box’ on NTD, between β‐propeller blades 4 and 5, which had previously been thought to bind a distinct consensus sequence. Further, we structurally define the fourth peptide binding site on NTD, which we term the Royle box. In vitro binding assays show that clathrin is more readily captured by cellular CBMs than by HDAg‐L, and site‐directed mutagenesis confirms that multiple binding sites on NTD contribute to efficient capture by CBM peptides.

likely the AmphyiphysinCBM peptide is just not long enough on its own to allow it docking in Royle box and simply needs some flanking residues. This could be tested.
3) In the linker of beta2 adaptin (621 IPSQGDLLGD LLNLDLGPPV 640), the sequence contains both a CBM motif and an arrestin box motif ([LI][LI]GxL.). The authors might want to compare which motif favors binding to the arrestin box.
4) The Amphiphysin sequence also contains two signals: a clathrin box motif and a W box motif (351 LLDLDFDPFK PEVTPAGSAG VTHSPMSQTL PWDLWTTSTD 390). The linker in between is about 20 aa. Miele, et al 2004 showed that the longer peptide in amphiphysin containing these two motifs enhances the binding of clathrin NTD by 6 folds (from 28uM to <5uM) (ref). The authors may wish to test the longer peptide to check the binding preference.
Referee: 2 Comments to the Author Main points.
This detailed structural/biochemistry paper analyses the binding of short, linear 'clathrin box motif' (CBM) peptides to the betapropeller of the clathrin N-terminal domain (NTD). Perhaps the most significant finding is that NTD binding peptides of beta-2 adaptin, amphiphysin and HDAg-L2 bind to both the clathrin and arrestin box sites (in contrast with previous findings). The authors also defined a fourth peptide binding site on the NTD which they called the 'Royle box'. This is a clear, straightforward structural paper and I have little to add other than some minor suggestions.
First, it would be helpful to have a broader context (perhaps in the discussion) to frame the detailed information presented. Presumably these different proteins can't all bind the NTD at the same time. How do the authors envisage that the different cellular binding partners 'share time' on the NTD? Is it an ordered, sequential series of binding interactions or some kind of competitive freefor-all?
Second, how confident are the authors in the 'Royle box' as a 'consensus' binding site given that similar peptides bind in different orientations and that binding does not involve the consensus CBM motif? This is especially true since given the high concentrations of peptides used in the study (~3mM(+)).

Minor points:
Abstract -Sentence starting 'Surprisingly, with each peptide …'. The sentence doesn't scan, presumably ',' should be replaced by 'and' to read: 'Surprisingly, with each peptide we observe simultaneous peptide binding at multiple sites on the NTD and viral peptides binding to the same sites as cellular peptides.

Author Rebuttal
Dear Prof Marsh, Please find attached a revised version of our manuscript entitled "Cellular and viral peptides bind multiple sites on the Nterminal domain of clathrin" (Manuscript ID TRA-16-0533).
We thank the reviewers for their supportive comments and suggestions for informative further experiments. We have addressed their concerns by performing significant further experiments, including the elucidation of a new crystal structure between clathrin NTD and an extended peptide containing the human amphiphysin I clathrin-box motif. A point-by-point response to the reviewers' comments is attached.
We hope that you will now find our revised article suitable for publication in Traffic.

Yours sincerely, Stephen Graham and Bernie Kelly
Editor's comments Regarding your manuscript "Cellular and viral peptides bind multiple sites on the N-terminal domain of clathrin" submitted for consideration for publication in Traffic, I asked two colleagues who are experts in the field to review the paper and their verbatim comments are appended below. Although both reviewers point out the lack of any physiological link, you will see that both are enthusiastic about the work and recommend publication in Traffic. They do make some suggestions on additions that will clarify the conclusions, in my view you should go at least some way to addressing these points as well as incorporating the corrections indicated by reviewer 2.
Although I cannot accept your manuscript for publication at this point, I believe that you will be able to address the referees' concerns and I look forward to receiving your revised manuscript. To expedite handling when you resubmit please include a response outlining how you have addressed each of the referees' concerns.
Referee 1: In this study, the authors have solved the high resolution structures of clathrin TD in complex with various CBM peptides. Surprisingly, in addition to binding to clathrin-box sites, all CBM peptides also showed density at the arrestin box binding region. The new interaction provides a structural basis to explain the redundancy of clathrin recruitment. Moreover, several CBM peptides from amphiphysin or Hepatitis D virus large antigen were found to bind to the Royle box site. This is the first report to confirm the "Royle" binding site at the structural level. They further carry out GST pull down assays in combination with mutagenesis work to evaluate these binding surfaces.
Overall, the manuscript is well-written and the structures support the conclusion that CBM peptide can bind to multiple sites in clathrin NTD. The main limitation of the study is that it does not explore the physiological consequences. Nevertheless, the in vitro and high resolution structural studies more than cross the bar for publication in Traffic.
The authors should consider the following questions. I would encourage but not insist on carrying out further controls experiments to clarify these points.

1)
In Figure 2: The unbiased density maps of HDAg-L1, HDAg-L2 in the arrestin box and Royle box are disconnected and do not resolve side chain positions unambiguously. Perhaps this would be clearer at a lower contour level? If the assignments cannot be better supported the authors should be clear about the ambiguity (Figure 4).
The reviewer is correct that the unbiased difference electron density for HDAg-L1 and HDAg-L2 peptides at the Royle box was less extensive than either the density for Amph4T1 at this site or for HDAg-L1 and HDAg-L2 at the clathrin box ( Figure 2). However, extensive attempts to model the HDAg-L1 and -L2 peptides in different conformations or orientations failed to yield models with acceptable fit to density and good stereochemistry. We believe that the poor-quality difference density arises from the viral peptides being less well-ordered than Amph4T1 when bound at the Royle box. Comparing the B values of peptides bound at the clathrin, arrestin and Royle boxes with the B values of surrounding residues (Table S1, added to manuscript) confirms that the Amph4T1 peptide bound at the Royle box is better ordered than the HDAg-L peptides, both in absolute terms and when compared to surrounding residues. The same is generally true when comparing the binding of HDAg-L peptides at the Royle box versus the clathrin box. A note to this effect has been added to the Results section.
In terms of illustrating the quality of our peptide models in the vicinity of the Royle box, we fear that reducing the difference density contour level of HDAg-L peptides at the Royle box in Figure 2 may mislead readers as to the strength of our evidence for peptide binding. We have therefore included an additional panel in Figure 4 showing the fit of each refined peptide to a feature-enhanced map [Afonine et al (2015) Acta Cryst. D71, 646-666]. This will allow the reader to get an impression of the model's fit to density that isn't confounded by differences in local map strength (and thus scale of peak heights) between regions of the peptides, with the added benefit that feature-enhanced maps reduce model bias. Lastly, we note that our final refined models and structure factors have been deposited with the PDB and are thus freely available to the community for inspection and re-analysis.

2)
In Figure 2: there is clear evidence of CBM peptide in Royle box for Amphiphysin4T1. Since AmphiphysinCBM wasn't predicted to dock in this region, it may indicate that the artificial residues (LERPHRD) may have driven peptide binding to the Royle box. More likely the AmphyiphysinCBM peptide is just not long enough on its own to allow it docking in Royle box and simply needs some flanking residues. This could be tested.
The peptides Amph4T1 and AmphCBM used for crystallisation were the same length, with sequences ETLLDLDFLE and ETLLDLDFDP, respectively. We realise that using the same name for the longer GST-peptide fusion proteins and the peptides used for crystallisation was confusing, for which we apologise. We have modified the manuscript to specifically identify when we're referring to the peptides used for crystallisation (by adding pep to the name of the peptide, e.g. Amph4T1pep).
This point notwithstanding, we agree with the reviewer that premature truncation of the peptide used for crystallisation might have prevented binding of a peptide derived from the naturally-occurring amphiphysin sequence at the Royle box. We have therefore crystallised NTD in complex with a longer peptide encompassing the amphiphysin CBM (AmphCBMlongpep, sequence ETLLDLDFDPFK). This structure reveals clear binding of the AmphCBMlongpep peptide at the clathrin and arrestin boxes, but not at the Royle box. We thus conclude that is was indeed the non-native residues in Amph4T1pep that facilitated binding of this peptide at the Royle box. We have included details of this new structure solution and refinement, along with figures illustrating AmphCBMlongpep bound at the arrestin and clathrin boxes, but not the Royle box, as supplementary material (Table S2 and Figure  S2). Together with the structures of AP2CBMpep and AmphCBMpep, the AmphCBMlongpep structure supports our conclusion that the CBM and (as yet undefined) Royle box consensus sequences aren't identical.

3)
In the linker of beta2 adaptin (621 IPSQGDLLGD LLNLDLGPPV 640), the sequence contains both a CBM motif and an arrestin box motif ([LI][LI]GxL.). The authors might want to compare which motif favors binding to the arrestin box. This is an interesting question we hadn't considered -we thank the reviewer for raising it. Given that the peptide motifs overlap (the final leucine of the arrestin box motif being the first of the CBM), a single β2 adaptin molecule could obviously not bind at both sites using both motifs simultaneously.
To address this question we have generated two GST-tagged peptide constructs containing the arrestin box motif of β2 adaptin (AP2arrL and AP2arrS) and compared their ability to capture His-NTD-NEMO to that of the β2 adaptin CBM (AP2CBM). Two GSTarrestin motif constructs were used to avoid issues that might arise from the peptide's carboxylate group immediately following the final leucine of the LLGDL arrestin box motif: AP2arrS ends with the subsequent residue of β2 adaptin (Leu) while AP2arrL ends with the sequence 'AlaSerSer', corresponding to the residues that follow the LLGDL arrestin box motif of arrestin2L [Kang et al (2009)]. To probe relative binding of these peptides to the arrestin box we used the GST fusions as baits to precipitate both wild-type His-NTD-NEMO and a mutant (Q98A+F91K+F9W) where the clathrin and Royle boxes, but not the arrestin box, had been disrupted. We found that AP2CBM and AP2arrS captured wild-type or Q98A+F91K+F9W His-NTD-NEMO more efficiently than did AP2arrL, suggesting that the arrestin consensus motif (LLGDL) alone binds the arrestin box more weakly than does the CBM or an extended arrestin motif (LLGDLL). However, we also noticed that GST-AP2arrL and GST-AP2arrS capture wild-type His-NTD-NEMO much more efficiently than they do the Q98A+F91K+F9W mutant. This strongly suggests that the arrestin consensus motif (LLGDL) is able to bind the clathrin or Royle boxes in addition to binding the arrestin box. This unexpected finding shows that the degeneracy of clathrin binding extends beyond the CBM to the arrestin-box motif. We have included this data as an additional figure in the main text ( Figure 6) and a short sections describing this experiment and its implications to the Results and Discussion, respectively.

4)
The Amphiphysin sequence also contains two signals: a clathrin box motif and a W box motif (351 LLDLDFDPFK PEVTPAGSAG VTHSPMSQTL PWDLWTTSTD 390). The linker in between is about 20 aa. Miele, et al 2004 showed that the longer peptide in amphiphysin containing these two motifs enhances the binding of clathrin NTD by 6 folds (from 28uM to <5uM) (ref).