The intrinsically disordered N-terminus of SUMO1 is an intramolecular inhibitor of SUMO1 interactions

  1. Zentrum für Molekulare Biologie der Universität Heidelberg (ZMBH), DKFZ - ZMBH Alliance, Heidelberg, Germany
  2. Heidelberg Institute for Theoretical Studies, Heidelberg, Germany
  3. Interdisciplinary Center for Scientific Computing, Heidelberg University, Heidelberg, Germany
  4. Department of Biochemistry, University of Washington, Seattle, WA 98195, USA
  5. Department of Molecular Life Sciences, University of Zürich, Zürich, Switzerland

Peer review process

Revised: This Reviewed Preprint has been revised by the authors in response to the previous round of peer review; the eLife assessment and the public reviews have been updated where necessary by the editors and peer reviewers.

Read more about eLife’s peer review process.

Editors

  • Reviewing Editor
    Heedeok Hong
    Michigan State University, East Lansing, United States of America
  • Senior Editor
    Qiang Cui
    Boston University, Boston, United States of America

Reviewer #2 (Public review):

Summary:

This very interesting study originated from a serendipitous observation that the deletion of the disordered N-terminal tail of human SUMO1 enhances its binding to its interaction partners. This suggested that the N terminus of SUMO1 might be an intrinsic competitive inhibitor of SUMO-interacting motif (SIM) binding to SUMO1. Subsequent experiments support this mechanism, showing that in humans it is specific to SUMO1 and does not extend to SUMO2 or SUMO3 (except, perhaps, when the N terminus of SUMO2 becomes phosphorylated, as the authors intriguingly suggest - and partially demonstrate). The auto-inhibition of SUMO1 via its N-terminal tail apparently explains lower binding of SUMO1 compared to SUMO2 to some SIMs and lower SIM-dependent SUMOylation of some substrates with SUMO1 compared to SUMO2, thus adding an important element to the puzzle of SUMO paralogue preference. In line with this explanation, N-terminally truncated SUMO1 was equally efficient to SUMO2 in the studied cases. The inhibitory role of SUMO1's N terminus appears conserved in other species including S. cerevisiae and C. elegans, both of which contain only one SUMO. The study also elucidates the molecular mechanism by which the disordered N-terminal region of SUMO1 can exert this auto-inhibitory effect. This appears to depend on the transient, very highly dynamic physical interaction between the N terminus and the surroundings of the SIM-binding groove based mostly on electrostatic interactions between acidic residues in the N terminus and basic residues around the groove.

Strengths:

A key strength of this study is the interplay of different techniques, including biochemical experiments, NMR, molecular dynamics simulations, and, at the end, in vivo experiments. The experiments performed with these different techniques inform each other in a productive way and strengthen each others' conclusions. A further strength is the detailed and clear text, which patiently introduces, describes, and discusses the study. Finally, in terms of the message, the study has a clear, mechanistic message of fundamental importance for various aspects of the SUMO field, and also more generally for protein biochemists interested in the functional importance of intrinsically disordered regions. In revision, the authors have further improved the text.

Weaknesses:

In the future, further experimental validation will be required, particularly with regards to the biological importance of the uncovered mechanism. These limitations are satisfactorily pointed out by the authors themselves in the revised manuscript.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Summary:

SUMO proteins are processed and then conjugated to other proteins via a C-terminal di-glycine motif. In contrast, the N-terminus of some SUMO proteins (SUMO2/3) contains lysine residues that are important for the formation of SUMO chains. Using NMR studies, the N-terminus of SUMO was previously reported to be flexible (Bayer et al., 1998). The authors are investigating the role of the flexible (referred to as intrinsically disordered) N-terminus of several SUMO proteins. They report their findings and modeling data that this intrinsically disordered N-terminus of SUMO1 (and the C. elegans Smo1) regulates the interaction of SUMO with SUMO interacting motifs (SIMs).

Strengths:

Among the strongest experimental data suggesting that the N-terminus plays an inhibitory function are their observations that

(1) SUMO1∆N19 binds more efficiently to SIM-containing Usp25, Tdp2, and RanBp2,
(2) SUMO1∆N19 shows improved sumoylation of Usp25,
(3) changing negatively-charged residues, ED11,12KK in the SUMO1 N-terminus increased the interaction and sumoylation with/of USP25.

The paper is very well-organized, clearly written, and the experimental data are of high quality. There is good evidence that the N-terminus of SUMO1 plays a role in regulating its binding and conjugation to SIM-containing proteins. Therefore, the authors are presenting a new twist in the ever-evolving saga of SUMO, SIMs, and sumoylation.

Weaknesses:

Much has been learned about SUMO through structure-function analyses and this study is another excellent example. I would like to suggest that the authors take some extra time to place their findings into the context of previous SUMO structure-function analyses. Furthermore, it would be fitting to place their finding of a potential role of N-terminally truncated Smo1 into the context of the many prior findings that have been made with regard to the C. elegans SUMO field. Finally, regarding their data modeling/simulation, there are questions regarding the data comparisons and whether manipulations of the N-terminus also have an effect on the 70/80 region of the core.

We thank the reviewer for insightful and constructive comments to improve our manuscript. We have now placed our findings in the context of previous structure-function analyses at several occasions, details of which can be found in our replies to the detailed comments.

We are also placing the C. elegans data into context of previously published findings on the various functions of SMO-1 in controlling development and maintaining genomic stability (lines 510ff). Finally, we addressed all questions and suggestions regarding comparison of MD simulation and NMR data, and addressed the question whether mutations in the N-terminus affected the 70/80 region. We have now clarified in the manuscript that the sum of MD and NMR data does not allow a clear-cut conclusion on the 70/80 interactions.

Reviewer #2 (Public Review):

Summary:

This very interesting study originated from a serendipitous observation that the deletion of the disordered N-terminal tail of human SUMO1 enhances its binding to its interaction partners. This suggested that the N terminus of SUMO1 might be an intrinsic competitive inhibitor of SUMO-interacting motif (SIM) binding to SUMO1. Subsequent experiments support this mechanism, showing that in humans it is specific to SUMO1 and does not extend to SUMO2 or SUMO3 (except, perhaps, when the N terminus of SUMO2 becomes phosphorylated, as the authors intriguingly suggest - and partially demonstrate). The auto-inhibition of SUMO1 via its N-terminal tail apparently explains the lower binding of SUMO1 compared to SUMO2 to some SIMs and lower SIM-dependent SUMOylation of some substrates with SUMO1 compared to SUMO2, thus adding an important element to the puzzle of SUMO paralogue preference. In line with this explanation, N-terminally truncated SUMO1 was equally efficient to SUMO2 in the studied cases. The inhibitory role of SUMO1's N terminus appears conserved in other species including S. cerevisiae and C. elegans, both of which contain only one SUMO. The study also elucidates the molecular mechanism by which the disordered N-terminal region of SUMO1 can exert this auto-inhibitory effect. This appears to depend on the transient, very highly dynamic physical interaction between the N terminus and the surroundings of the SIM-binding groove based mostly on electrostatic interactions between acidic residues in the N terminus and basic residues around the groove.

Strengths:

A key strength of this study is the interplay of different techniques, including biochemical experiments, NMR, molecular dynamics simulations, and, at the end, in vivo experiments. The experiments performed with these different techniques inform each other in a productive way and strengthen each others' conclusions. A further strength is the detailed and clear text, which patiently introduces, describes, and discusses the study. Finally, in terms of the message, the study has a clear, mechanistic message of fundamental importance for various aspects of the SUMO field, and also more generally for protein biochemists interested in the functional importance of intrinsically disordered regions.

Weaknesses:

Some of the authors' conclusions are similar to those from a recent study by Lussier-Price et al. (NAR, 2022), the two studies likely representing independent inquiries into a similar topic. I don't see it as a weakness by itself (on the contrary), but it seems like a lost opportunity not to discuss at more length the congruence between these two studies in the discussion (Lussier-Price is only very briefly cited). Another point that can be raised concerns the wording of conclusions from molecular dynamics. The use of molecular dynamics simulations in this study has been rigorous and fruitful - indeed, it can be a model for such studies. Nonetheless, parameters derived from molecular dynamics simulations, including kon and koff values, could be more clearly described as coming from simulations and not experiments. Lastly, some of the conclusions - such as enhanced binding to SIM-containing proteins upon N-terminal deletion - could be additionally addressed with a biophysical technique (e.g. ITC) that is more quantitative than gel-based pull-down assays - but I don't think it is a must.

Thank you very much for pointing towards the study of Lussier-Price. We now point out congruent findings in more detail in the discussion.

We also thank the reviewer for the advice to present and discuss the MD findings more clearly, and more explicitly specify which parameters were obtained from MD. We have made changes throughout the Results and Discussion sections.

We agree that it would be a nice addition to use ITC measurements as a more quantitative method to assess differences in binding affinities upon deletion of the SUMO N-terminus. We had tried to measure affinities between SUMO and SIM-containing binding partners by ITC but in our hand, this failed. In the study of Lussier-Price et al., the authors were able to measure differences in SIM binding upon deleting the N-terminus but only when they used phosphorylated SIM peptides. Follow-up studies, e.g., on the effect of SUMO’s N-terminal modifications should certainly include more quantitative measurement such as ITCs, however these studies will have to be picked up by others. The main PI Frauke Melchior and most contributing authors moved on to new challenges.

Reviewing Editor (Recommendations For The Authors):

Both reviewers agreed that your manuscript presents novel results and the key findings including the self-inhibitory role of the N-terminal tail of SUMO proteins in their interaction with SIM are overall well supported by the data. The reviewers also provided constructive suggestions. They pointed out that some simulation results are not clear, which could be strengthened by control analysis and by toning down the related descriptions. In addition, Reviewer 2 suggested that the conclusions from the current biochemical and simulation studies could be further reinforced by more quantitative binding measurements. We hope that these points can be addressed in the revision.

We thank both reviewers for their insightful and constructive comments and the appreciative tone. In our replies above and below we address most of the raised concerns.

We strongly recommend the change of the current title. eLife advises that the authors avoid unfamiliar abbreviations or acronyms, or spell out in full or provide a brief explanation for any acronyms in the title.

We changed the title to “The intrinsically disordered N-terminus of SUMO1 is an intramolecular inhibitor of SUMO1 interactions” to avoid acronyms in the title.

Reviewer #1 (Recommendations For The Authors):

Major:

Lines 190-262: The authors use NMR experiments and all-atom molecular dynamics (MD) simulations. They state that this approach reveals a highly dynamic interaction of the SUMO1 N-terminus with the core and that the SIM binding groove and the 70/80 region are temporarily occupied by the SUMO1 N-terminus (Fig. 3C). After comparing SUMO1, Smt3, SUMO2, and Smo1 by this approach they state that the most striking differences exist for the interaction with the SIM-binding groove, while interactions with the 70/80 region are rather comparable.

The authors then compare the average binding time data of Figure 3C, D, E, F in Figure 3G.

It is not clear which data points are included in the bar graphs of Figure 3G and how the individual data points (there are maybe 8 shown in each bar) correspond to the data shown in 3C, D, E, and F or if they are iterations (n?) of the modeled data. This should be clarified. Also, for comparison, the authors should also graph the average data of the 70/80 region.

We clarified the data shown in Figure 3G as well as 3C-F, and how It relates to each other. Indeed, Figure 3G shows 8 data points for 8 trajectories, and their average. Figure 3C-F are based on the same 8 trajectories, in this case broken down per residue of the protein. The average data of the 70/80 region does not show any significant differences across the proteins, as already pretty well visible from panels 3C-F.

Line 322: More concerning, in Figure 5, the authors model how a ED11,12KK mutations disrupt the interaction between the N-terminus and the SIM-binding groove and state that this mutation leaves interactions with the 70/80 region largely untouched. Again, it is not clear which data points are included in the bar graph 5D and 5G and how many iterations. Furthermore, data of 5B, C (SUMO1) and 5 E, F (smo1) do show clear differences between the WT and mutants affecting both the SIM binding groove and the 70/80 region. The double mutation clearly seems to affect the 70/80 region when comparing 5B, C (SUMO1) and 5 E, F (smo1), but this result is not mentioned. Indeed, the authors state that the double mutants leave the interactions with the 70/80 region largely untouched, but this is not borne out by the data presented.

We improved the clarity of the legend of Figure 5 as suggested. We also thank the reviewer for the comment on the changes in the 70/80 region, to which we point the reader explicitly now in the corresponding Results section. We, however, refrain from drawing conclusions from the MD in this case, as this change is not supported by the NMR measurements (Fig 5a). Charge-charge interactions in the charge-rich double mutants might be overstabilized in the MD simulations, a problem known for the canonical force fields used here, albeit tailoring it for IDPs. We now cite a corresponding reference. Another potential explanation for that the CMPs do not take this change up upon mutation could be a pronounced fuzziness in this region, which however, in turn, is not apparent from the simulations. We would therefore not overinterpret these differences in the 70/80 region. Our key conclusion is the loss of interactions with the SIM-binding groove – and thus of cis-inhibition – by mutations, which is supported by both, MD and NMR.

341: In their N-termini substitution experiments, the authors show that the SUMO1 core that carries the SUMO2 N-terminus (S2N-S1C) binds USP25 more efficiently than wt SUMO1. However, the SUMO1 core that carries the SUMO2 N-terminus is also reduced in its interaction with Usp25. This is concerning as the SUMO2 N-terminus was not predicted to interfere with SIM binding.

We were excited to see that the inhibitory potential could be partially transplanted by swapping the N-termini of SUMO1 and SUMO2 demonstrating that some important determinants are contained within the N-terminal tail of SUMO proteins. However, the observed effects were partial indicating that also other determinants contribute and that we do not yet understand all aspects. Obviously, the SUMO1 and SUMO2 cores are similar (also in the area comprising the SIM binding groove) but not identical, and as the inhibition arises from dynamic interactions of the N-terminus with the SIM binding area, differences in the SUMO cores and in residues flanking SUMO’s N-terminus are likely to influence the inhibitory potential as well.

Blue bars in 3G, 5D, and 6A look surprisingly similar down to the individual data points - does that mean that the same SUMO1 WT data was recycled for these different experiments? This is concerning to me.

The data displayed in the figures listed above are derived from in silico simulations and indeed display the same data set for the case of SUMO1 WT repeatedly, as we also state in the figure legends (we had done so for 5D “(identical to Fig. 3C)”, and now added the same comment to 6A, thanks for pointing this out). We show the SUMO1 WT data again to facilitate comparing the different SUMO variants in MD simulations.

Line 352 and 496: The authors used phosphomimetic mutants to assess the effect of SUMO2 N-term phosphorylation on interaction with Usp25. The data suggest a mild phenotype (6G) which is borne out by the quantization in 6H. In contrast, the effect of an array of modifications for SUMO1 (Figures 6A - C) was solely analyzed by MD simulation. If possible, this data should be confirmed, at least by using a phosphomimetic at the Ser9 position of SUMO1. Alternatively, a caveat explaining the need to confirm these predictions by actual experiments should be added to the text.

Already now we state in “Limitations of the study” that “While our MD simulations and in vitro studies with selected mutants point in this direction, we have not been able to generate quantitatively acetylated and/or phosphorylated SUMO variants to test this hypothesis.”

We agree that the hypothesis needs experimental validation. Phosphomimetic amino acids can be a useful tool in some cases but fail to mimic a phosphor group in other cases. In the past we had tested whether replacing Ser9 by a potentially phospho-mimicking amino acid (Glu) would further diminish binding of SIM-containing proteins compared to already strongly reduced binding to wt SUMO1 but the effect was too mild to yield a significant difference, at least in our assay. Whether this is due to a lack of Glu in mimicking phosphorylation of Ser9, due to limited sensitivity of our pulldown assay combined with the challenge to detect inhibition compared to an already inhibited state, or a failure in our hypothesis we were not able to clarify so far. We therefore now also added a sentence to the paragraph introducing phosphoSer9 MD simulations (now line 367) stating that this hypothesis needs to be tested experimentally.

Minor:

Line 110: the authors should include references for their summary statement that "A defining feature of SUMO proteins is the intrinsically disordered N-terminus, whose function is only partly understood." Also cite in line 119.

Thank you, we now included some references.

Line 75: Please indicate early on that the N-terminus of some SUMO proteins contains lysines for the formation of SUMO chains. Please list them.

We now list, which of the SUMO proteins used in this study contain lysine residues in their N-termini.

Line 113: Please cite studies that elucidated the sumoylation of lysines in the N-terminus of SUMO2/3 proteins.

Thank you, we now included some references.

Line 153: The authors should include additional references on Smt3 structure function analyses to provide better context. One important detail, for example, is the important finding that Yeast SUMO (Smt3) deletion can be complemented by hsSUMO1 but not hsSUMO2 and hsSUMO3. Additionally, in yeast the entire Smt3 N-terminus can be deleted without detectable effects on growth, underscoring the enigmatic role of the N-terminus (Newman et al., 2017). Caveat also applies to line 266.

Thank you, we now included some additional information and references around line 153 and below.

164: The hypothesis that the SUMO1 N-terminus interferes with SIM binding groove ignores the previous observation that deletion of the SUMO2 N-terminus does not have an effect on binding (in vitro). While this is addressed later, the authors should clarify this e.g. by stating "a unique feature of the SUMO1 N-terminus".

We now explicitly mention that this feature appears to be unique to SUMO1.

374 and 499: The authors should discuss the caveat that the deletion of the N-terminus of Smt3 does not have a phenotype in yeast in vivo (Newman et al., 2017).

We now discuss that Smt3’s N-terminus can be deleted without detectable phenotype, both in the results as well as in “Limitations of the study”.

Line 367: I feel this is overstated and I do not see any evidence that post translation modifications of the SUMO core plays a role. Therefore, I suggest: Our data and modeling are consistent with an interpretation that the N-termini of human and C. elegans SUMO1 proteins are inhibitory and that other SUMO N-termini may acquire such a function upon posttranslational modification of the N-terminus.

We agree that this is pure speculation and therefore restrict our hypothesis to modifications of the N-terminus.

Line 374 ff: Since Smo-∆N12 increases sumoylation (Fig. 2I), it is likely that the in vivo defect is due to over-sumoylation in C. elegans. The authors should discuss this possibility and quote appropriate literature e.g.: Rytinki et al., Overexpression of SUMO perturbs the growth and development of Caenorhabditis elegans. Cell Mol Life Sci. 2011 Oct;68(19):3219-32. PMID: 21253676.

In our study, we employ in vitro SUMOylation as a means to assess the SIM binding capability in an in-solution assay. For this, we use USP25 as a specific substrate known to depend on a SIM for its SUMOylation. We cannot exclude that some specific substrates depending on this same mechanism for their modification may be upregulated in modification also in the Smo-1∆N12 worms. In vivo however, the majority of SUMO substrates is not subject to SIM-dependent SUMOylation. We now added a control experiment showing that we neither observe significantly increased SUMO levels nor upregulated steady state levels of SUMOylation in these worms (Supplemental figure 8).

The phenotypes shown in the paper by Rytinki et al. do not resemble the smo-1∆N12 mutants. Rather, we observed a specific defect in the meiotic germ cells at the pachytene stage causing increased apoptosis Moreover, we show by western blot analysis that there is no global over-sumoylation occurring in smo-1∆N12 mutants (Fig. s8). Together, our data point to a germline-specific function of the SMO-1 N-terminus in maintaining genome stability (lines 510ff).

Reviewer #2 (Recommendations For The Authors):

Page2 - "Small Ubiquitin-related modifiers of the SUMO family regulate thousands of proteins in eukaryotic cells" - The authors could consider a more precise statement, e.g. that SUMO modifiers have been detected on thousands of proteins and their regulatory effect on many proteins have been demonstrated.

To be a bit more precise, the sentence now reads: “Ubiquitin-related proteins of the SUMO family are reversibly attached to thousands of proteins”. The summary has a word limit, hence we did not expand further at this place.

Page 4 - "Both events require SUMO-binding motifs (reviewed, e.g. in 7 ." - The end bracket is missing. Also, isn't it too strong a statement that paralogue specificity always requires a SIM? I don't know all the literature sufficiently well, but the authors could double-check if it is correct to say that paralogue-specific SUMOylation always depends on a SIM.

Thank you, we added the missing bracket. We agree that it would not be correct to say that paralogue-specificity always depends on a SIM. One alternative example is Dpp9, which shows a clear preference for SUMO1 without owning a SIM. Instead, Dpp9 harbors an alternative SUMO-binding motif, the E67-interacting loop, with a strong paralogue-preference (Pilla et al., 2012). We never intended to imply that a SIM is required for paralogue preference and we also rather generically wrote “SUMO binding motif” instead of “SIM”. However, in the subsequent paragraph about SUMO binding motifs we only go into details of SIMs as one of three classes of SUMO binding motifs not even mentioning the alternative classes. To make this more obvious, we now list the two other known classes of SUMO binding motifs hoping that it will shed the correct light onto our previous statement about paralogue preference.

Page 4 - In the nice discussion of different types of SIMs, the authors could consider mentioning also the special case of TDP2, which is used later by them as a model binding protein. This could provide an occasion to explain what the unusual "split SIM", mentioned on page 6, but not discussed, is, and what its relation to a normal SIM is. Also, it can perhaps be mentioned that TDP2 contacts SUMO2 not only through the two hydrophobic elements contiguous in space that mimic a SIM but also through a slightly larger interface around these regions on the surface of a folded domain.

Thank you for pointing this out. In the introduction, we extended our section on SUMO binding and now also included TDP2’s “split SIM”.

Page 11-12 - In the section "Interaction between SUMO's disordered N-termini and the SIM binding groove is highly dynamic" (and corresponding figures), it should be stated that the discussed kinetic parameters are derived from molecular dynamics simulations and not experimental measurements. It was not very clear to me. This also applies to this sentence on page 17: "First, we observed a very fast (ns) rate of the binding/unbinding process", which in its current form suggests direct observation rather than simulation.

We thank the reviewer for pointing this out, and in fact, Rev #1 made the same comment. We specified now clearly that the rates were calculated from MD simulations, in the Results and Discussion sections (on page 11-12 and 18 (previously 17)).

Page 16 - The authors could briefly mention that this relatively long disordered N-terminal tail is a specific feature of SUMO proteins that distinguishes them from ubiquitin. I guess it is obvious to people from the SUMO field, but I don't think it is explicitly stated anywhere in the text and it could be interesting for readers who are less familiar with SUMO/ubiquitin differences.

Thank you, we added a short half-sentence pointing out this difference.

Page 17 - "The N-terminal region remains fully disordered in the bound state and is thus a classic example of intrinsic disorder irrespective of the binding state." - it could be added to this sentence that this is suggested by molecular dynamics simulations and not directly observed.

We added the information that this finding is based on the MD simulations.

Page 18 - "(e.g., 41,53 or flanking the SIM binding groove24,42" - the end bracket is missing.

Thanks, we added it.

Page 19 - "Our analysis in C. elegans (Fig. 7) suggests that this N-terminal function is particularly important in DNA damage response, a pathway that is strongly dependent on the SUMO system." - this brief description of the in vivo data seems to overgeneralise them a little bit. Perhaps one can describe what was observed with slightly more nuance.

See changes on p.19, lines 510ff.

  1. Howard Hughes Medical Institute
  2. Wellcome Trust
  3. Max-Planck-Gesellschaft
  4. Knut and Alice Wallenberg Foundation