Mapping protein-metabolite interactions in E. coli by integrating chromatographic techniques and co-fractionation mass spectrometry

  1. Mateusz Wagner
  2. Jieun Kang
  3. Catherine Mercado
  4. Venkatesh P. Thirumalaikumar
  5. Michal Gorka
  6. Hanne Zillmer
  7. Jingzhe Guo
  8. Romina I. Minen
  9. Caroline F. Plecki
  10. Katayoon Dehesh
  11. Frank C. Schroeder
  12. Dirk Walther
  13. Aleksandra Skirycz

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

In our pursuit of understanding the protein-metabolite interactome, we introduced PROMIS, a co-fractionation mass spectrometry (CF-MS) technique focusing on biosynthetic and regulatory processes. However, the challenge lies in distinguishing true interactors from coincidental co-elution when a metabolite co-fractionates with numerous proteins. To address this, we integrated two chromatographic techniques— size exclusion and ion exchange—to enhance the mapping of protein-metabolite interactions (PMIs) in Escherichia coli . This integration aims to refine the PMI network by considering size and charge characteristics, resulting in 994 interactions involving 51 metabolites and 465 proteins. The PMI network is enriched for known and predicted interactions validating our approach’s efficacy. Furthermore, the analysis of protein targets for different metabolites revealed novel functional insights, such as the connection between proteinogenic dipeptides and fatty acid biosynthesis. Notably, we uncovered an inhibitory interaction between the riboflavin degradation product lumichrome and orotate phosphoribosyltransferase (PyrE), a key enzyme in de novo pyrimidine synthesis. Lumichrome supplementation mimicked the biofilm formation inhibition observed in a ΔpyrE mutant strain, suggesting lumichrome role in integrating pyrimidine and riboflavin metabolism with quorum sensing and biofilm formation. In summary, our integrated chromatographic approach significantly advances PMI mapping, offering novel insights into functional associations and potential regulatory mechanisms in E. coli .

Article activity feed

  1. Review Commons

    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

    'The authors do not wish to provide a response at this time.'

  2. Review Commons

    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 #3

    Evidence, reproducibility and clarity

    The authors expand upon a previous method (PROMIS) that was developed to detect metabolite-protein interactions from cell extracts through co-elution of proteins and metabolites in size exclusion chromatography. Here, the authors use ion exchange chromatography. By comparing PROMIS and IEX datasets, the authors increase the confidence of detected interactions being true hits. The addition of IEX data significantly improves the power of the method to identify previously known/predicted protein-metabolite interactions. The authors also purify two proteins identified by the method and experimentally validates their novel interactions with dipeptide metabolites.

    In general the paper is well written and provides a clear assessment of the two methods. The validation of two interactions is welcomed.

    There are some issues that could be addressed with better clarification:

    • While the two datasets for omniTICC appear to show good overlap, the two PROMIS replicates appear to have very little. Due to this low overlap, one of the metabolite interactions chosen for validation (dipeptide) was chosen by manually lowering the correlation cutoff of PROMIS as well as relying on two previous PROMIS datasets. Authors could comment on possible reasons for this low overlap.
    • Experimentally validating the effect of Val-Leu on the enzyme activity of FabF would be a very good addition, but is not crucial to the paper.

    Minor:

    Line 173: In STITCH the authors find 1012 known or predicted interactions, while their method finds 92 of these interactions. Authors could comment on limitation of the unbiased method which may lead to missing these interactions (for example, sensitivity of MS detection of low abundance metabolites).

    Minor:

    • Authors could comment on what are some drawbacks to using docking to estimate binding sites. Computational screening could be a powerful way to prioritize hits for validation, so could be worth a more detailed discussion here.
    • In figure 2 the color labelling of the experiments could be presented more clearly.
    • Authors could add in the text or methods how they calculate the Pearson correlation coefficient they use for determining significance. Overall the methods are well presented and the technical approach is very well described and impressive. In general the authors present the statistics such as likelihood over chance detection, in a way that helps to evaluate the accuracy of the separate and combined CF approaches.

    Some comments:

    • The PROMIS method and the omniTICC method are performed only in duplicate, where the two PROMIS experiments also differ in proteomics workflow. Could that be why the overlap between PROMIS experiments is so poor? Why not perform the experiments in biological triplicates?
    • Authors correct for poor replicate overlap by lowering the requirements of one subset of metabolites. What's the argument for not doing this with all metabolite subsets? Would this interaction have been discovered just by using PROMIS in three different experiments (i.e. more biological replicates)? The benefit of adding the IEX analysis is clouded by the poor overlap of PROMIS data
    • Line 353 - It would also be good to elaborate as to why other methods would be likely to miss this interaction. Is it because the authors method does not require the users to choose specific metabolites or proteins to focus on beforehand?
    • Authors could comment on whether any of the high confidence interactions were also been observed in different IP experiments with E. coli (such the Lip-SMAP method reported by Piazza et al., 2018).
    • Authors could comment on other ways to identify the potential binding site. For example, limited proteolysis of the protein in presence of metabolite has been used previously
    • There is massive interconnectivity between NMPs and ribosomal proteins. It would be good to comment on this. Do the authors believe these are true interactions? Could there be another reason for this cluster? Do the STITCH interactions also show that these proteins interact with so many NMPs?
    • Line 157 - The authors state that "isopropylmalate co-eluate with tens of proteins". This feels like a bit of an understatement if the actual numbers are: 92, 119, 303, 287

    Minor:

    • How did the number 1479 proteins come about? Are those detected in all 4 datasets? What about those detected in only some datasets? This information could be more clear

    Significance

    General assessment

    The paper is a welcome comparison of using two co-elution MS methods for identifying protein-metabolite interactions. It is clear that these types of interactions are important for modulating protein activity but are not well studied. The paper provides a clear workflow for both the experimental and data analysis portions of interaction proteomics. The relatively low number of validated hits could limit its significance outside the specialised field.

    Advance.

    The paper makes a conceptual advance in interaction proteomics. It is perhaps not unexpected that combining two different interaction proteomics methods gives more accurate target identification than a single method alone. The paper also strengthens the evidence that dipeptides play a role in regulation that may be conserved in bacteria and plants.

    Audience.

    The audience for this paper is likely researchers that are actively involved in the field of interaction proteomics. The link between dipeptides and feedback regulation has been developed by the group over a few papers, and this report provides further evidence. The paper also shows that a weak kinetic effect in vitro can actually lead to a significant effect in vivo, which is a very interesting finding that is a good point of reference for future in vitro validation efforts. Often the effect of metabolites in vitro is weak, which could lead to the (possibly false) conclusion that it is not relevant in vivo (see e.g. weak effects of metabolites on enzyme activity in vitro in a recent LiP-SMap paper (https://doi.org/10.1038/s42003-023-05318-8)

    Reviewer background:

    Reviewers are active in the field of interaction proteomics and have previously used the Lip-SMAP method as well as thermal proteome shift method.

  3. Review Commons

    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

    In this work, Wagner et al aimed to comprehensively map protein-metabolite interaction in E. coli. They do so by extending a previously developed SEC-based co-elution method with an IEX-based co-elution method. The results is a more extensive and robust protein-metabolite network. As a validation, selected predicted protein-metabolite interactions are confirmed at the level of binding, as well as enzyme kinetics and even phenotype.

    The key conclusions of the paper are mostly convincing, even more so because the authors provide a critical evaluation of the method. My only concern is that the results from the PROMIS replicates seem to be quite different (e.g. Fig 1h and Suppl Fig 1a). One wonders whether, given this apparent variation, two replicates are sufficient to define the protein-metabolite interactions with great confidence.

    The author took care to make all data (proteomics & metabolomics) publicly available and the methods are clearly described.

    The manuscript is well-written and the figures are clear, although Fig 1 would benefit from and explanation of the color coding in the legend. Personally, I feel that less emphasis on lumichrome would help focus the conclusion section.

    Significance

    This work is of high significance because it describes a method to comprehensively map protein-metabolite interaction that could relatively easily be applied to any organism. The work is of high quality and similar to work by for example Link et al (https://doi.org/10.1038/nbt.2489) or Piazza et al (https://doi.org/10.1016/j.cell.2017.12.006), but technically less challenging which increases its potential impact.

    Beyond these technical advances, global studies like these are a great resource for anyone working on the functional characterization of proteins. Often, a (predicted) protein-metabolite interaction can be a crucial lead to find the function of a protein.

  4. Review Commons

    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

    In this manuscript, the authors demonstrate a strategy for uncovering metabolite-protein interactions, rooted in the concept of identifying co-fractionating metabolites and proteins through independent analyses of each fraction using untargeted proteomics and metabolomics. This approach is applied to E. coli protein lysates.

    The same research group has previously advocated for utilizing the simultaneous elution of metabolites and proteins as an indication of interactions, as seen in Luzarowski et al. (2019) and Veyel et al. (2017), where metabolite-protein interactions in yeast and plants were investigated, respectively. Here, a variation of the original proposed method (PROMIS) is introduced, involving an additional chromatography separation (ion exchange) alongside the original size exclusion separation proposed in PROMIS. By incorporating an extra dimension to reduce sample complexity, both experimental and data analysis efforts are effectively doubled, entailing the processing of 48 additional fractions for proteomic and metabolomic analyses, in addition to the 40 already included in the original PROMIS protocol.

    While previous studies have demonstrated the isolation of small molecule-protein complexes following ion-exchange based separation (Chan et al., 2012), the method described here does not introduce any fundamentally new concepts. However, it remains to be determined whether the addition of an extra separation dimension genuinely aids in accurately classifying new interactions. Although coincidental co-elution effects may be mitigated, there is a risk of disrupting genuine complexes through excessive handling of lysates.

    In my view, there are several points that require clarification:

    1. Line 138: There's an assumption that the metabolite profile will alter in IEX upon binding to proteins. While this might occur, it's not definite. Additionally, it's unclear why peaks associated with protein binding could also increase in intensity, and how this signifies a binding event.
    2. Lines 147-154: The most crucial dataset in this paper isn't adequately delineated here. We learn that by combining SEC and IEX, 1479 proteins and 58 metabolites are identified, but what about when only one separation method is employed? What advantages does using IEX provide?
    3. Line 170: Similar to point 2 regarding PMI.
    4. Line 174: To gauge whether the proposed strategy aids in accurately classifying new interactions, the authors examined if their predicted interactions also appear in the STITCH database. Out of the 994 PMI in the network, 92 were found in STITCH. I'm uncertain if STITCH is the most suitable metric for this assessment, given it likely hasn't been updated since 2016. How does this PROMIS-IEX protocol mirror known interactions in E. coli's central carbon metabolism, for instance, such as those detailed in this publication: doi: 10.15252/msb.20199008?
    5. Certain details of Figure 1 are challenging to grasp and inadequately explained in the figure legend. What do the colours of the heatmap in 1d represent?
    6. In Figure 1h, the proteins co-eluting with 2-isopropylmalic acid in the four separations decrease to 5 from tens of proteins, with only one known interactor of 2-isopropylmalic acid (LeuA) among them. Is this outcome favourable or unfavourable? Are the other four proteins false positives?
    7. Figure 2b: Why is this expected to work particularly well for NMPs? Is there a specific biological rationale behind it?

    Furthermore, among the classified interactions, two were corroborated through microscale thermophoresis and protein docking. One involved the enzyme FabF and the dipeptide Val-Leu. The decision to delve deeper into this pair likely stemmed from the prior focus on dipeptide-protein interactions, extensively discussed in Luzarowski et al.'s 2019 manuscript. The second interaction pertained to lumicrome and PyrE.

    Significance

    To summarise, this paper advocates for heightening the complexity of experimental and computational analyses in studying metabolite-protein interactions through co-fractionation techniques. While it's anticipated that increased separation would enhance results, I remain unconvinced that the data presented conclusively demonstrates this. Overall, I believe the proposed method possesses only a modest level of originality and novelty, as outlined at the beginning of my review. Nonetheless, the substantial experimental effort and data generation warrant publication following additional meticulous quality control evaluation.

  5. Version published to 10.1101/2024.02.14.580258v1 on bioRxiv