Quantitative interactome mapping of skeletal muscle insulin resistance
This article has been Reviewed by the following groups
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
- Evaluated articles (Review Commons)
Abstract
Protein-protein interactions (PPIs) are dynamic and critical to adaptive homeostasis. While there have been massive efforts to catalogue proteome-wide PPIs, global quantification of changes remains a challenge. Here, we integrate dynamic protein correlation profiling - mass spectrometry (PCP-MS) and quantitative cross linking-mass spectrometry (qXL-MS) using multiplexed stable isotope labelling to characterise global PPI remodelling following the development of chronic skeletal muscle insulin resistance (IR) with or without acute insulin stimulation. We quantify >7,000 unique PPIs amongst 5,346 proteins and show changes in the interactome network dominate the proteome response. Our data show the dysregulation of protein processing in the endoplasmic/sarcoplasmic reticulum involving changes in PPIs with protein chaperones and disulfide isomerases is a major hallmark of skeletal muscle IR. Mechanistically, we show the dysregulation of PPIs with Protein-Disulfide Isomerase 6 (PDIA6) regulates cysteine oxidation and insulin sensitivity. Taken together, we show in vivo quantitative interactome mapping is a powerful approach to understand disease mechanisms and provide new insights into protein network re-organisations with IR.
Article activity feed
-
Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.
Learn more at Review Commons
Reply to the reviewers
We provide current Revision Plan and will generate full point-by-point responses upon final resubmission.
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #2
Evidence, reproducibility and clarity
Quantitative interactome mapping of skeletal muscle insulin resistance Ng et al present a series of proteomics/interactomics studies in skeletal muscle to identity insulin regulated complexes/interactions and changes ot these in insulin resistant muscle. More mechanistically, the Authors focus on changes in interactions involving chaperones in the ER/SR, presenting interesting data on the effect of PDIA6 overexpression alters insulin sensitivity in muscle ex vivo.
Major Comments:
The section entitled "Validating the regulation of PPIs with insulin resistance in C2C12 myotubes with quantitative XL-MS". This is not really a validation …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #2
Evidence, reproducibility and clarity
Quantitative interactome mapping of skeletal muscle insulin resistance Ng et al present a series of proteomics/interactomics studies in skeletal muscle to identity insulin regulated complexes/interactions and changes ot these in insulin resistant muscle. More mechanistically, the Authors focus on changes in interactions involving chaperones in the ER/SR, presenting interesting data on the effect of PDIA6 overexpression alters insulin sensitivity in muscle ex vivo.
Major Comments:
The section entitled "Validating the regulation of PPIs with insulin resistance in C2C12 myotubes with quantitative XL-MS". This is not really a validation of th previous data as presented, but more an orthologous assay that helped pinpoint the interest in the ER. Suggest adjusting the title.
Figure 3B - the "decrease" in AS160 pS588 regulation appears to be due to increased basal, not decreased phosphorylation in after insulin. This should be commented on or clarified.
PDIA6 is down-regulated in muscle from people with T2D - so why did the authors decide to overexpress PDIA6? I note this rationale is explained in the discussion, and could be articulated better in the results.
Figure 5J and K. The TA muscles are substantially larger from PDIA6 OE mice. Are the muscle fibres also larger? Tbhis relates to the normalisation of data in K. This appears to be normalised to g tissue. If so, is the difference between control, and OE mice being driven by the increase in muscle mass - with uptake per muscle or per fibre the same?
Minor Comments:
For the PCP-MS data form C2C12 cells. The authors use an analysis of AUC to assess protein abundance, which, as they state, is important for chronic treatments if total protein is not separately quantified. However, the analysis of changes in protein distribution is less clear from the text in the results section. Intuitively, a profile that is normalised to total intensity in all fractions would provide a protein abundance-independent read-out for changes in protein distribution. Does the "local analysis" capture this same information? Could the Authors provide a little more information here?
Figure 1M - are the Authors sure that VPS41 should be in this panel. It doesn't seem to be insulin regulated, and the arrow appears to refer to movement between insulin sensitive and insulin resistant.
Figure 1N - "This includes an array of TBC1 domain-containing proteins (TBC1D15, 195 TBC1D17, TBC1D8B) that are consistently reduced with IR". Do the Authors mean the abundance was less, or that complex formation was reduced?
Optional. In general, there is a lot of text discussing the literature around proteins highlighted in the analysis. This is useful to an extent, but the Authors might consider streamlining this a little (perhaps moving some of the information ot supp tables?).
Why do the Authors think the crosslinking MS was not able to capture acute PPI changes like the PCP-MS was?
For the EDL crosslinking data. Are the Authors able to provide a comparison with C2C12 data - to highlight the differences and similarities between tissue and the cell model? This may be a challenge if the authors think most differences may be technical.
Please check - "reduces free-glycerol levels essential for fatty acid synthesis". Glycerol does not directly contribute to FA synthesis. But is needed for triglyceride synthesis.
Do the Authors think that the change in PDIA6 interactions may be a general/indirect indication of changes in ER redox and/or protein misfolding in insulin resistance?
Is PDIA6 an ER luminal protein? If so, it being phosphorylated is interesting.
Referees cross-commenting
Similarly, reviewer #1 raises important points on the description of key parts of the analysis, that will need to be addressed. I think we agree that the manuscript emcpmpasses a great deal of data, and that it is somewhat difficult to follow why PDIA6 was selected for validation. Overall, the reviews pick up on different aspects of the manuscript that could be improved.
Significance
Overall, the strength of the paper is in the underlaying proteomics workflows and analysis. The work presented of very high technical quality, and I have no doubt the data presented will be of use to the field beyond the analysis in this current publication.
However, a weakness is doubts over the relevance of the data on PDIA6 overexpression in muscle insulin resistance.
This will be of interest to those in the proteomics, interactomics and metabolism fields.
My expertise is in glucose metabolism, insulin signalling and insulin resistance.
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #1
Evidence, reproducibility and clarity
Summary
Ng et al. use a combination of quantitative (structural) proteomics tools to study the effect of insulin resistance (IR) on protein interactions in mouse muscle cells (C2C12 myotubes) and muscle tissue. First, the authors use protein correlation profiling (PCP; co-fractionation) to separate protein complexes from C2C12 cells with and without induced insulin resistance. The PCP data is then used to reconstruct protein complexes and networks based on binary interactions and to uncover changes upon IR, this is illustrated by several examples (Fig. 1). As an orthogonal method quantitative cross-linking (XL) is used to study …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #1
Evidence, reproducibility and clarity
Summary
Ng et al. use a combination of quantitative (structural) proteomics tools to study the effect of insulin resistance (IR) on protein interactions in mouse muscle cells (C2C12 myotubes) and muscle tissue. First, the authors use protein correlation profiling (PCP; co-fractionation) to separate protein complexes from C2C12 cells with and without induced insulin resistance. The PCP data is then used to reconstruct protein complexes and networks based on binary interactions and to uncover changes upon IR, this is illustrated by several examples (Fig. 1). As an orthogonal method quantitative cross-linking (XL) is used to study protein interactions in C2C12 cells. A multidimensional enrichment/fractionation scheme is designed to increase the number of cross-links/protein-protein interactions (PPIs), and again interaction networks are generated and changes between no IR vs. IR conditions assessed (Fig. 2). The XL approach is then carried forth into a mouse model of IR. Muscle tissue of mice fed on regular chow or on a high-fat diet are compared (Fig. 3). Selected XL data is validated on known structures and quantitative data from PCP and XL is compared to integrate the different interactome data (Fig. 4). Finally, the interactions of PDIA6, a key protein found to be affected by IR, are studied in more detail in the mouse model. The effect of overexpression of PDIA6 is studied using different readouts, including redox proteomics (Fig. 5). These data connect the redox imbalance / cysteine oxidation with the role of PDIA6 in insulin resistance.
Overall, the manuscript is impressive with respect to the methodological effort undertaken, generating and combining many large-scale proteomics data sets, both from a cell line and from mouse tissue. The large amount of data accumulated can be seen as a strength and as a weakness, because it is impossible to follow up on all findings (changes of interactions induced by IR observed in any sample and with any method). Nevertheless, the PDIA6 example was evaluated in more detail and with dedicated follow-up experiments. The conclusions from this experiment are plausible and presented logically. It is, however, difficult for a reviewer to quickly judge whether there would have been more promising leads for validation experiments than PDIA6.
Given the large amount of data already generated, I do not have suggestions for additional experiments. Instead, some analysis and interpretation of existing data need further clarification.
Major comments
The search strategy and statistical treatment of the cross-linking data need to be explained more clearly. The authors write that the data were searched with pLink2 and the FDR was controlled at 1%. However, it remains unclear whether the FDR was controlled at the level of peptide/cross-link-spectral-matches (PSMs/CSMs), or at the level of non-redundant site pairs or even protein-protein interactions for inter-protein links, which would be more appropriate. Controlling the FDR only at the PSM/CSM level will lead to an inflation of false positives when aggregating results at the interaction level. Moreover, it remains unclear whether all data sets were searched together or whether individual data sets or subsets of the entire data were searched individually. For example, if single files or only files belonging to a certain higher-order fraction would be searched individually, this would again lead to an underestimation of false positives. What is particularly noteworthy in this context is that in a typical dataset inter-protein links are underrepresented compared to intra-protein links for statistical reasons, while the numbers in this manuscript are much more balanced (almost equal numbers in the cell line data set, and even more inter-protein links than intra-protein links in the tissue). The authors should consider articles that discuss FDR control in XL-MS, for example by the Rappsilber group. Minimally, more details about what was searched together and at what level the FDR was controlled need to be provided.
On a similar note, it has been discussed in the literature that validating large-scale XL data on selected structures of complexes is a poor proxy for accurate FDR control, as such complexes are commonly not representative for more transient or substoichiometric PPIs, or PPIs involving low-abundant proteins.
Minor comments
Figure 1: CCT is a complex composed of eight subunits, but only seven are shown. What happened to the remaining one (CCT6)?
The authors performed a redox proteomics experiment in a PDIA6 overexpression system. However, the statement in the Conclusion section that PDIA6 overexpression promotes disulfide bond formation in interacting proteins is not directly justified because the method only quantifies cysteine oxidation, not S-S bond formation directly.
All supplementary data is not provided in an independent repository, but in a repository of the authors' institution. It is unclear whether the data could be accessed anonymously. Proteomics data need to be provided in an independent, community accepted repository such as from the members of the proteomeXchange consortium (PRIDE etc.).
A clear description of what is shown in the SI tables is missing, e.g. in the form of figure legends. In their present form, SI data are difficult to interpret. For example, I did not find information about cross-link identifications, only quantitative data on cross-link changes. However, if the identification of a cross-link is not confident in the first place (see my comments above), then the quantification will be irrelevant.
Referees cross-commenting
I trust the expertise of reviewer #2 on matters related to insulin resistance. It seems that we both agree that the PDIA6 example might require a more consistent justification throughout the manuscript.
Significance
The study is one of only a few so far that combines PCP and XL on such a large scale for a mammalian system. There are also very few studies of cross-linking on tissue. Therefore, from a methodological point of view, the study is highly innovative. The application to the muscle cell system and insulin resistance as a biological research question is furthermore very novel. As such, the study is valuable to different communities - those developing and refining experimental methods and those using them to uncover regulatory mechanisms. Another strength is that the authors made serious efforts at each step to optimize the XL method and adapt it to their sample types of interest.
The wealth of data is both a strength and a weakness of the work. Inevitably, a reviewer might argue that some aspects of the work could have been done differently. Unless someone spends a lot of time going deep into the result tables, it will be difficult to make constructive suggestions on additional targets for further investigation. Nevertheless, some statistical aspects of data analysis need to be clarified, and parts of the data analysis might need to be repeated. This, in turn, may require some reinterpretation of findings related to the XL data.
Advances: Conceptual, methodological, mechanistic
Audience: Specialized, basic research, translational
Reviewer expertise
My background is in proteomics, structural proteomics, mass spectrometry, analytical sciences, experimental methodology, and computational data analysis. I have general knowledge of biological processes, but I am not an expert on insulin resistance.
-
