Structures of Multiple Peptide Resistance Factor from Pseudomonas aeruginosa

  1. Shaileshanand Jha
  2. Kutti R. Vinothkumar

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

The aminoacylation of lipid head group in many bacteria is carried out by bi-functional enzymes called MprF, which encode for a soluble synthase domain that typically transfers lysine or alanine from a tRNA to lipid head groups, and the modified lipid is translocated across the leaflets by a transmembrane domain. This modification of the lipids probably evolved to adapt to the environment where the microbes reside. Here, we describe the cryoEM structures of MprF enzyme from Pseudomonas aeruginosa revealing a dimeric enzyme with a distinct architecture when compared with the homologous Rhizobium enzymes and validate this arrangement with biochemical analysis. The cryoEM maps and the models in detergent micelle and nanodisc reveal a conformational change of the terminal helix of the synthase domain, highlighting the dynamic elements in the enzyme that might facilitate catalysis. Several lipid-like densities are observed in the cryoEM maps, which might indicate the path taken by the lipids and the coupling function of the two functional domains. Thus, the structure of a well-characterised PaMprF lays a platform for understanding the mechanism of amino acid transfer to a lipid head group and subsequent flipping across the leaflet that changes the property of the membrane.

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

    We thank all the reviewers for their comments and suggestions, which has helped in revising the manuscript for a broader audience. Some of the experiments that was suggested by the reviewers has been performed and included in the revised manuscript. The response to reviewers is provided below their comments.

    Reviewer #1 (Evidence, reproducibility and clarity (Required)):

    MprF proteins exist in many bacteria to synthesize aminoacyl phospholipids that have diverse biological functions, e.g. in the defense against small cationic peptides. They integrate two functions, the aminoacylation of lipids, i.e. the transfer of Lys, Arg or Ala from tRNAs to the head group, and the flipping of these modified lipids to the membrane outer leaflet. The authors present structures of MprF from Pseudomonas aeruginosa and describe these structures in great detail. As MprF enzymes confer antibiotic resistance and are therefore highly important, studying them is significant and interesting. Consequently, their structures have been substantially characterized in recent years, including the publication of the dimeric full-length MpfR from Rhizobium (Song et al., 2021).

    While the structural work appears to be solid and carried out well on the technical part, one big criticism is how the data are presented in the manuscript, how they are analyzed and how they are put into relation to previous work. As structures of Mpfr from Rhizobium have been published, it is not required and rather distracting to explain the methodological details and the structure of Pseudomonas MprF in such great detail. Instead, the manuscript would benefit very strongly from reaching the interesting and novel parts, the comparison with the previous structures, as early as possible. Overall, the manuscript should be substantially shortened to not divert the reader's attention away from the novel parts by drowning them in miniscule description of the structural features such as secondary structure elements or lipid molecule positions where it remains completely unclear what their relevance is to the story and the message of the paper. Finally, during this revision, care should be taken to improve the language and maybe involve a native speaker in doing so.

    It is true that we have described the experimental details of PaMprF in detail including the constructs. We had reconstructed the map of dimeric PaMprF in 2020 but with the publication of the homologues structures (Song et al 2021 and the unpublished Rhizobium etli structure), we had to make sure the PaMprF dimer is not an artefact. Hence, our attempts to rule out this with different constructs and extensive testing with various detergents. Thus, we would like to keep this in the manuscript. We realise the importance of focusing on novel/interesting parts and have reshuffled sections (comparing structures and validating the dimer interface) followed by description of modelling of lipid molecules.

    Even more importantly, since the authors observe a dimer interface which strongly deviates from the previously presented arrangement of another species, the most important thing would be to properly characterize this interface and experimentally validate it, both of which has not been done sufficiently. When also taking into account that there were significant differences in the arrangement of the dimer between their structures in GDN and nanodisc, and that in the GDN structure, the cholesterol backbone of GDN appears to be involved in the interface (there should not be any cholesterol in native bacterial membranes!), there is a realistic chance that the observed dimer is an artefact. If the authors cannot convincingly rule out this possibility, all their conclusions are meaningless.

    The trials with cholesterol hemisuccinate stems more of out of curiosity (we are aware that no cholesterol is present in bacterial membranes). We had started the initial analysis of PaMprF with DDM and by itself it was largely monomeric (unpublished observation and supported by recent publication of PaMprF in DDM – Hankins et al 2025). When we observed that GDN was essential for the stability of the dimer (and not even LMNG), we asked if a combination of CHS with DDM will keep the dimer intact, which didn’t work and GDN was found to be important. The use of CHS for prokaryotic membrane protein studies has now been reported in few different systems and a recent one includes – Caliseki et al., 2025. We would like to keep the observation with CHS in the manuscript, and we have moved this figure to Appendix Fig. S3C.

    In addition, in a recent report on MgtA, a magnesium transporter (Zeinert et al., 2025), it was observed that DDM/LMNG resulted in monomeric enzyme, while GDN resulted in dimeric enzyme albeit, the dimer interface was in the soluble domain. We have added this reference and observation of MgtA in the discussion (page 13, lines 407-411).

    We like to think that the milder GDN tends to keep the membrane proteins or oligomers of membrane proteins more stable but further studies on multiple labile membrane protein systems will be required to substantiate this.

    Hence, while I think that the data presented here would be worth publishing. However, a major drawback is that the authors do not sufficiently analyse, characterise and validate the dimer interface and fail to show that the dimer is biologically relevant.

    Further major points:

    • The authors always jump between their structures in detergent and nanodisc during all the descriptions, which makes following the story even more difficult. Please first describe one of the structures and then (briefly) discuss relevant similarities and differences afterwards.

    The flow and description of the structures is now modified and the figures have now been rearranged to make it easier to follow. The panel in figure 2 describing the overlay of the GDN and nanodisc is now moved to Appendix Fig. S2B. Thus, figure 2 has only description of salient features of the structures (the interacting residues between the membrane and soluble domain) and the terminal helix.

    • The difference in dimerization between Pseudomonas and Rhizobium is the most interesting and surprising feature (if true) of the new structures. However, it is not really presented as such. The authors should put more emphasis on making clear that this is a complete rotation of the monomers with respect to each other (by how many degrees?) and they should visualize it even more clearly in Figure 4 (and label the figure so that it is possible to understand it without having to read the text or the legend first).

    We thought the colouring of the TM helices should make the difference in interface more obvious (the N and C-terminal TM helices in different colours). Now, we have also labelled the TM helices, so that it is easier to follow (this was also shown in panel E). The rotation is ~180° and this is now mentioned in the figure legend.

    • P. 10: The authors insinuate that only one of the dimer interfaces, either Pseudomonas or Rhizobium could be real, but disregard the possibility that both might be the biologically relevant interfaces of the respective species and that there might have been a switch of interfaces during evolution. They should also mention and discuss this possibility.

    We didn’t imply that one of the interfaces is real but clearly mentioned that it could also be different conformational state (page 7, lines 226-228). In the revised version, we have included a multiple sequence alignment (we had not included in the initial draft as it had been presented in several previous publications). The MSA (Appendix Fig. S6) reveals that neither of the interfaces are highly conserved.

    • Fig. 5G: The authors claim that the higher molecular band that appears in the mutant is a "dimer with aberrant migration" of >250 kDa as opposed to the expected 150 kDa. They should explain how they came to this conclusion and how they can be sure that the band does not correspond to a higher oligomer (trimer or tetramer). They could show, by extraction and purification scheme similar to the wildtype using first LMNG and then GDN, followed by at least a preliminary EM analysis, that the crosslinked mutant MprF is indeed a dimer, or use other biophysical methods to do the same, otherwise this experiment does not show much. Furthermore, they should also include a cysteine mutant in the part of Pseudomonas MprF that would be involved in a Rhizobium-like interface in their crosslinking experiments to check whether they could also stabilize dimers in this case.

    The band of the double mutant after crosslinking (or even without crosslinking) migrates at higher molecular weight than that expected for a dimer, and could potentially be a higher molecular band that a dimer. We also note that in the previous publication by Song et al 2021, the crosslinking of RtMprF also resulted in a higher molecular weight band (shown also by Western blot).

    We now substantiate the dimer of PaMprF with different approaches. We employed blue-native gel and also SDS-PAGE of the purified protein. This clearly shows that the higher molecular band after crosslinking is a dimer (Figure 4B and Fig. EV4D). In particular, in the BN-PAGE, the treatment of mutants with crosslinkers revealed a dimeric band even in the presence of SDS. Further, we have performed cryoEM analysis of the mutants - H386C/F389C and H566C. The images, classes and reconstruction show that the enzyme forms a dimer similar to the WT. Interestingly, we also observe in H566C mutant in nanodisc, a small population that has similar architecture to the Rhizobium-like interface (classes shown in Fig. EV7 and Appendix Fig. S5). This prompted us to look closely at other datasets and it is clear that during the process of reconstitution in nanodisc, we observe both kinds of dimer interface but the PaMprF dimer is predominant. We also observe higher order oligomers (tetramer) in GDN but as only few views are visible, a reconstruction could not be obtained (Appendix Fig. S5). In addition, we also introduced two cysteines on the Rhizobium-like interface and no crosslinking on the membranes were observed (Figure 4B). But it is possible that these chosen mutants are not accessible to the crosslinker. Thus, we conclude that the oligomers of PaMprF is sensitive to nature of detergents and labile.

    • As the question whether the observed interface is real or an artefact is very central to the value of the structural data and the drawn conclusions from it, the authors should make more effort to analyze and try to validate the interface. First, an analysis of interface properties (buried surface area, nature of the interactions, conservation) should be performed for the interface as observed in the Pseudomonas structure but also for a (hypothetical) Rhizobium-like interface of two Pseudomonas monomers (such a model of a dimer should be easily obtainable by AlphaFold using the available Rhizobium structures as models). Then, experimental methods such as FRET or crosslinking-MS would allow to draw more solid conclusions on the distances between potential interface residues. While these experiments are a certain effort, the question whether the dimer interface is real is so central to the paper that it would be worthwhile to make this effort.

    We have included the interface area and nature of interactions in the revised manuscript (page 7, lines 221-223).

    We attempted AlphaFold for predicting the dimeric structure of PaMprF (and included RtMprF also). Some of the attempts from the predictions is summarised in figure 1.

    The prediction of monomer is of high confidence but the oligomer (here dimer) is of low confidence (from ipTM values). Even the prediction for Rhizobium enzyme has low confidence, and gives a complete different architecture (and in some trials with lipids, it gives an inverted or non-physiological dimer). Only when the monomer of PaMprF with lipids and tRNA was given as input (requested by reviewer 2 and described below), it predicts oligomeric structure with some confidence but rest were not informative.

    • As it seems that detergents might disrupt or modify the dimer interface, it might be an alternative to solubilize the protein in a more native environment by polymer-stabilized nanodiscs using DIBMA or similar molecules.

    We have tried to use SMALPs for extraction of PaMprF. We were able to solubilise but unable to enrich the enzyme sufficient for structural studies currently and will require further optimisation.

    • Since parts of the Discussion are mostly repetitions of the Results part and other parts of the Discussion also contain a large extend of structure analysis one would usually rather expect in the Results part instead of the Discussion, the authors should consider condensing both to a combined (and overall much shorter) Results & Discussion section.

    We have rewritten much of the discussion section and removed any repetition from the results sections. We would prefer to keep the results and discussion separate.

    Minor points:

    • Explain abbreviations the first time they appear in the text, e.g. TTH

    This is now expanded in the first instance

    • Figure labels are very minimalistic. This should be improved, e.g. by putting labels to important structural features that appear in the text, otherwise the figures are not an adequate support for the text.

    The font size for the labels have been increased.

    • Figure 5: Label where the different oligomers run on the gels

    Labelled.

    Reviewer #1 (Significance (Required)):

    While the structural work appears to be solid and carried out well on the technical part, one big criticism is how the data are presented in the manuscript, how they are analyzed and how they are put into relation to previous work. As structures of Mpfr from Rhizobium have been published, it is not required and rather distracting to explain the methodological details and the structure of Pseudomonas MprF in such great detail. Instead, the manuscript would benefit very strongly from reaching the interesting and novel parts, the comparison with the previous structures, as early as possible. Overall, the manuscript should be substantially shortened to not divert the reader's attention away from the novel parts by drowning them in miniscule description of the structural features such as secondary structure elements or lipid molecule positions where it remains completely unclear what their relevance is to the story and the message of the paper. Finally, during this revision, care should be taken to improve the language and maybe involve a native speaker in doing so.

    Even more importantly, since the authors observe a dimer interface which strongly deviates from the previously presented arrangement of another species, the most important thing would be to properly characterize this interface and experimentally validate it, both of which has not been done sufficiently. When also taking into account that there were significant differences in the arrangement of the dimer between their structures in GDN and nanodisc, and that in the GDN structure, the cholesterol backbone of GDN appears to be involved in the interface (there should not be any cholesterol in native bacterial membranes!), there is a realistic chance that the observed dimer is an artefact. If the authors cannot convincingly rule out this possibility, all their conclusions are meaningless.

    Hence, while I think that the data presented here would be worth publishing. However, a major drawback is that the authors do not sufficiently analyse, characterise and validate the dimer interface and fail to show that the dimer is biologically relevant.

    Reviewer #2 (Evidence, reproducibility and clarity (Required)):

    Shaileshanand J. et al., reported the structures of Multiple Peptide Resistance Factor, MprF, which is a bi-functional enzyme in bacteria responsible for aminoacylation of lipid head groups. The authors purified MprF from Pseudomonas aeruginosa in GDN micelles and nanodiscs, and by applying cryo-EM single particle method, they successfully reached near-atomic resolution, and built corresponding atomic models. By applying structural analysis as well as biochemistry methods, the authors demonstrated dimeric formation of MprF, exhibited the dynamic nature of the catalytic domain of this enzyme, and proposed a possible model on tRNA binding and aminoacylation.

    Major comments

    1. In abstract, the authors stated 'Several lipid-like densities are observed in the cryoEM maps, which might indicate the path taken by the lipids and the coupling function of the two functional domains. Thus, the structure of a well characterised PaMprF lays a platform for understanding the mechanism of amino acid transfer to a lipid head group and subsequent flipping across the leaflet that changes the property of the membrane.' Firstly, those lipid-like densities were demonstrated in Fig 3A, since densities of lipids of purified membrane proteins often exist within regions of relatively low local resolution, or low quality, I think more detailed description on how the authors defined which part of the density belongs to lipid and how they acquired the modeling of some of the lipids is required. And the authors modeled phosphatidylglycerol into the GDN MprF, I would require additional experiment, for instance, mass spectrometry over the purified sample, to demonstrate the existence of this specific lipid with the sample. Secondly, regarding the last sentence in the abstract, how these structures lay a platform for further understanding was poorly discussed in both result section and discussion section, since the authors clearly stated 'This cavity perhaps provides a path for holding lipids...', then the statement in the next sentence 'Taken together... the vicinity to the cavities described above indicates the possible path taken by the lipids to enter and exit the enzyme' does not have a reliable evidence to support this conclusion, I would suggest the authors move these statements into discussion section, and elaborate more over this issue since it is an important part in the abstract, or make a more solid proof using other approaches, such as molecular dynamics simulation, to make these statements solid in the result section.

    The membranes of E. coli have predominantly phosphatidyl ethanolamine (PE) and phosphatidyl glycerol (PG) as the next abundant lipid with cardiolipin though smaller in number, plays an important role in functioning of many membrane proteins. In our map, the non-protein density are unambiguous and they can be observed as long density reflective of acyl chains (note that GDN used in purification has no acyl chain) and hence attributed these densities to lipids (Fig. EV4E/F and Figure 5A). Only in few of these densities, head group could be modelled and the identity of the lipid as PG at the dimer interface is based on the requirement of negatively charged lipids for oligomerisation of membrane proteins in general (for example – KcsA tetramer formation requires PG, Marius et al., 2005; Valiyaveetil et al., 2002;2004). It is true that the lipid densities are at the peripheral regions of the map but here only acyl chains have been modelled. Within the membrane domain, one reasonably ordered lipid is observed and by analogy with R. tropici structure, it is possible to build a modified-PG (in PaMprF here ala-PG). However, the density of the head group is not unambiguous (unlike lysine in the *R. tropici, *whose density stands out) and hence we have modelled it as PG alone. In the methods (page 20, lines 649-650), the identification and modelling of lipid densities is described.

    We agree that mass spectrometry analysis of purified lipids will be useful but it will not be able to tell the position of the lipid in the map (model) and for this we still require a map at higher resolution with better ordered lipids. We have recently built/developed the workflow for native MS and we plan to initiate analysis of PaMprF in the near future, which will provide details for the lipid purified with the enzyme.

    We had initiated molecular dynamics simulation during the review process, and we had included tRNA molecules (shorter version) as we felt the connection between tRNA binding and lipid modification was important. This would have also explained the path taken by lipids (performed by Hankins et al., 2025 in their publication). However, this is likely to require more work (and computing resources) and both mass spectrometry and molecular dynamics will be part of the future work.

    We have rewritten the discussion and changed the last line of the abstract to the following

    “From the structures, the binding modes of tRNA and lipid transport can be postulated and the mobile secondary structural elements in the synthase domain might play a mechanistic role”.

    (in the abstract, lines 24-26).

    Fig 2B, it seems the H566 sidechains were overlapping in the zoom-in figure of distance measurement between H566 residues, to clarify this, authors should either present another figure with rotation, to better demonstrate their relative locations, or swap this zoom-in figure with another figure with rotations. Also, could the authors briefly commenting on why they chose H566 for distance measurement specifically?

    The side chain of residue H566 in the nanodisc model face towards each other at the interface, hence this residue was chosen to shown the proximity.

    Related to previous comment, I see one additional green square in Fig. 2A and an additional green square in Fig. 2B, without any zoom-in images provided on these regions. Besides, they're focusing on two different domains with same color, any particular reason why they're there? If so, please provide the information in figure legends.

    The green squares in panels 2A and 2B are the regions that have been zoomed in panels 2D and 2E showing the interactions of the TTH. This is now made clear in the legend as well as in the figure.

    Related to previous comment, authors should also provide distance measurement over electrostatic interaction sites in Fig. 2A, since distance plays as an important factor in these forces.

    The electrostatic interactions have been included.

    For Fig. 2C, since in Fig. 1, the authors have already indicated the differences between reconstruction of the GDN and nanodisc datasets, this information provided here seems to be a bit abundant, I suggest either move this panel to Fig. 1, to make a visualization on both electron densities as well as atomic models, or move this panel to supplementary figures.

    We thank the reviewer for the suggestion. The panel, figure 2C is moved to Appendix Fig. S2B.

    Fig. 3B, some of the spheres of the lipids were also marked as red, any particular reason why they're red? Do they indicate they're phosphate heads? If so, could the authors provide evidences how they define these orientations of the lipid heads? If not, any particular reason why they're red?

    Although, there are non-protein densities (i.e., density beyond noise that remain after modelling of protein residues and found individually) have been modelled as lipids (In Fig. EV4E, these additional densities are shown). Except for few, all these densities have been modelled only as acyl chain. The lipids modelled with head group and phosphate (that have oxygen) and the fit of the density are shown in both figure 3A and EV4F. Hence, the red (oxygen) is seen in the space filling model of lipids (the density for few lipids are shown, also in the response to the comment below).

    Fig. 3C, the fitted model of lipid and its corresponding density should be added to Fig. S4, to give more detailed view on the quality of the fitting.

    The figure 3 has now been reorganised and the new figure (fig. 5) has only 3 panels. We have provided an enlarged view of the lipids in the membrane domain along with unmodelled densities in 3A. In addition, in fig. EV4F, fit of the lipid to density (select lipids) are shown.

    Fig. 4D and 4E, could the authors also indicate the RMSD values when comparing the differences of RtMprF, PaMprF, ReMprF, this information would be helpful to understand how big of a difference within these three models.

    The RMSD values of the structural comparison is given in the text.

    Fig. 6E, the coloring used for CCA-Ala were similar to the blue part of soluble domain, could the authors change the coloring a bit? Also, for Fig. 6F, I would suggest the authors provide a prediction model, such as using AlphaFold3, of this tRNA interaction site, to further validate this proposed model.

    The colour of the CCA part is changed in the revised figure. Following the suggestion of the reviewer, we used AlphaFold3 to predict the complex formation of PaMprF with tRNA (or shorter version) (Figure 2). As mentioned above in response to reviewer 1, the prediction of dimeric enzyme was of low confidence and this is also reflected when a combination of tRNA, lipids and enzyme sequence are given. Instead of full-length tRNA, if only the CCA end is provided, then the prediction program does position this in the postulated cavity. Only with the monomeric enzyme and tRNA does one get a reasonable model. With respect to the proposed model in 6F, currently we don’t have any evidence and this remains a postulate. In the revised manuscript, we have replaced this with conservation figure, which we thought is more relevant.

    In Supplementary Figures S1 and S3, the angular distribution of maps exhibited preferred orientation to certain extent, 3D FSC estimation should also be supplied for these maps, as an indication of whether the reconstructed densities were affected or not.

    We have included the 3DFSC plots for all the data sets (including the new ones in figures EV1, 2, 5, 6, 7). It is evident that the nanodisc datasets in general are slightly anisotropic.

    For Fig S3B, could the authors switch to another image with better contrast?

    This is now replaced with an image to show the particles.

    Minor comments

    1. Fig. 2E and 2F, distance measurement should also be supplied to these two panels.

    We have now included the distance measurement in both the panels, which are now Fig. 2D and 2E.

    Fig. 5D, since in Fig. 4F and 4G already mentioned the skeleton of GDN, this modeling part should be presented before exhibit it in dimer interface, the authors should rearrange the sequence over these three panels.

    The figures in the revised manuscript has been rearranged. Figure 5 (now figure 4) has been modified to include the biochemical analysis (crosslinking studies) and the panel 5D has been removed.

    In Supplementary Figure S3, which density was shown for the PaMprF local resolution estimation result? Authors should provide this information as two maps were shown in this figure.

    The local resolution is for C2 symmetrised map and this is now mentioned in the panel.

    CROSS-REFEREE COMMENTS Both Reviewer #1 and #3 made comments over technical issue, their evaluation over functional aspects of this protein is what I was lacking over my comments, also, their evaluation of the biological narrative, relevance toward previous research is also more insightful. Finally, they offer valuable suggestions on how to adjust the article to make it more readable, and better describing the biological story which I would suggest the authors to pay attention to.

    Reviewer #2 (Significance (Required)):

    Significance The authors mainly focused on the structure of MprF in Pseudomonas aeruginosa, this protein is essential for the resistance to cationic antimicrobial peptides. A combination of structural and biochemical analysis provided evidences to the dimeric formation to this enzyme, and the analysis over differences of purified proteins using GDN and nanodisc was particular interesting, which provide new insight regarding the flexible nature of this enzyme, and potentially could be beneficial to the membrane protein community, as it demonstrates the differences in detergent/nanodisc of choice could affect the assembly of the protein of interest. Still, some of the statements in the manuscript, for instance, the assignment of lipids was over-claimed and could be benefited from additional approaches to support the issue. I would suggest some refinement in the discussion section as well as some of the figures.

    My expertise: cryo-EM single particle analysis; cryo-ET; sub-tomo averaging; cryo-FIB;

    Reviewer #3 (Evidence, reproducibility and clarity (Required)):

    Jha and Vinothkumar characterize the cryoEM structure of the alanyl-phosphatidylglycerol producing multiple peptide resistance factor (MprF) of Pseudomonas aeruginosa. MprF proteins mediate the transfer of amino acids from aminoacyl-tRNAs to negatively charged phospholipids resulting in reduced membrane interactions with cationic antimicrobial peptides (produced by the host and competing microorganisms). The phospholipid modifications involve in most cases the transfer of lysine or alanine to phosphatidylglycerol. MprF proteins are membrane proteins consisting of a soluble and hydrophobic domain. Multiple functional studies have shown that the soluble domain of MprF mediates the aminoacylation of phosphatidylglycerol, while the hydrophobic domain mediates the "flipping" of aminoacylated phospholipids across the membrane, a process that is crucial to repulse or prevent the interaction of antimicrobial peptides encountered at the outer leaflet of bacterial membranes. Aside from its role in conferring antimicrobial peptide resistance, other roles of MprF have been described including more physiological roles such as improving growth under acidic conditions. Interestingly, MprF proteins are also found in Gram-negative bacteria which are already protected by an additional membrane that includes LPS. However, in Pseudomonas aeruginosa, MprF confers phenotypes that are similar to those observed in Gram-positive bacteria. Importantly, crystal structures of the soluble domain have led to important insights into aminoacyl phospholipid synthesis and recent studies on the cryoEM structure of Rhizobium tropici have confirmed functional and preliminary structural studies with other MprF proteins. The cryoEM structure from R. tropici confirmed the dimeric structure of MprF and supported a role of the hydrophobic domain in flipping lysyl-phosphatidylglycerol across the membrane. A comparison of the structures of lysyl-phosphatidylglycerol with alanyl-phosphatidylglycerol producing MprFs could reveal new insights into the mechanism of transferring aminoacyl-phospholipids from the soluble domain to the hydrophobic domain and translocation of alanyl- vs lysyl-phosphatidylglycerol across the membrane.

    Major concerns

    1. The study by Jha and Vinothkumar provides the cryoEM structure of an alanyl-phosphatidylglycerol producing MprF protein which is in principle an important milestone in gaining a better understanding of the mechanism of aminoacyl-phospholipid synthesis and flipping, including the potentially different requirements of accommodating different aminoacyl -tRNAs and aminoacyl-phospholipid species. However, this is not addressed. The authors present a "distinct architecture" compared to the structure of R. tropici- MprF, without providing functional insights and the focus of the study shifts to the role of detergents in determining MprF structures via cryoEM. Thus, after fundamental discoveries have been made with crystal structures of the soluble domain and cryoEM structure of R. tropici, this study -while valuable as a resource- seems to offer only an incremental advance in understanding the mode of action of MprF and the potential different requirements for transferring alanyl-phosphatidylglycerol to the hydrophobic domain and flipping across the membrane. The reader is left with the finding of a distinct architecture with no further explanation or hypothesis.

    We thank the reviewer for his/her comments. It is true that the crystal structures of soluble domains of MprF (from 3 species) and the cryoEM structures are now available (two Rhizobium species). However, the cryoEM maps that we have obtained has several salient features including the distinct dimeric interface and the position of the C-terminal helix of the soluble domain. This in particular is important. In the previous study, Hebecker et al 2011 had reported that the terminal helix of PaMprF was important for the activity and the construct without the TM domain can also function in modifying the lipids. The full-length cryoEM map of PaMprF in GDN now provides an idea how this occurs, with the terminal helix buried at the interface. Further, the proposed tRNA binding site (from Hebecker et al 2015, lysine amide bound structure) face other in the dimeric architecture of *R. tropici *and it is not clear how the full-length tRNA will bind without disrupting the dimer. In contrast, the dimer architecture observed for PaMprF has the tRNA binding site facing away and they can bind to the enzyme without any constraints. We think the mobile/dynamic elements (or secondary structure) of the synthase domain play a major role in interaction with substrates and mechanism. The current structures provide some evidence for this and form the basis of future studies. Instead of cartoon description, we have now included a conservation plot of the molecule in explaining the possible mechanism along with the surface representation in figure 6.

    Differences to R.tropici MprF and other studies are difficult to follow as only a topological map of the Pseudomonas MprF is provided and conserved amino acids that have been shown to be crucial in mediating synthesis and flipping are not highlighted in the text or in the figures, specifically addressed, or discussed. Conserved amino acids in the presented cryoEM structure could provide important mechanistic insights and could address substrate specificity/requirements for aminoacyl phospholipid synthesis, transfer to the hydrophobic domain and flipping.

    The conservation of residues across MprF homologues have been presented in previous published articles and hence, initially we had not included in the manuscript. We have now included multiple sequence alignment of select homologues of MprF highlighting conserved residues (Appendix Fig. S6) as well a figure (Fig. 6F) colouring the molecule with conservation scores with CONSURF. In figure 6F, zoomed in version, we highlight the many of the conserved residues in the synthase domain as they play a role in substrate selectivity.

    Authors characterize an alanyl-phosphatidylglycerol producing MprF but do not detect the lipid in the cryoEM structure. Thus, the potential path taken by alanyl-phosphatidylglycerol remains unclear. Authors model the detected lipids as phosphatidylglycerol, which may be an interesting finding as it would indicate that MprF is generally capable of flipping phospholipids (this is however not discussed). While it is plausible that MprF flippases may be able to flip phosphatidyglycerol it could have a different path and structural requirements. It is also difficult to follow what the suggested pathway of flipping is in the Pseudomonas-MprF flippase (compared to R.tropici). Authors could provide a similar overview figure as in Song et al. and indicate what the potential differences are.

    We modelled phosphatidylglycerol as the lipid as the current density doesn’t allow to model ala-PG ambiguously though it is found in the same position as the lys-PG in the R. tropici maps. The recent in-vitro assay by Hankins et al 2025 shows that PaMprF is able to flip wide range of lipids and we would also like to point out that PG from outer leaflet can be flipped, whose headgroup can be modified at the inner leaflet and flipped back. As shown by Song et al 2021 and Hebecker et al 2011, the specificity for the substrates is in the synthase domain (by mutagenesis and swapping). We don’t think there will be any difference between the lys-PG and Ala-PG path but in our opinion the positional relation between the soluble and membrane domain is the most important and has remained the focus of the manuscript along with the dimeric architecture. The figure 6 in the manuscript is descriptive of this and provides a summary of the structural observation from the presented structures.

    Minor concerns

    • Page 13: the following sentence should be rephrased: "Among the missing links in the current cryoEM maps is the lack of well-ordered density for lipid molecules on the inner leaflet closer to the re-entrant helices but it is reasonable to assume from the cluster of positive charge that there will be lipid molecules and are dynamic. "

    This is has been rephrased.

    • Page 4: Klein et al do not show that the Pseudomonas aeruginosa MprF mediates flipping

    Corrected to reflect only the modification of lipid and not flipping.

    Reviewer #3 (Significance (Required)):

    General assessment: see review

    Advance: Minor

    Audience: Specialized

  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

    Jha and Vinothkumar characterize the cryoEM structure of the alanyl-phosphatidylglycerol producing multiple peptide resistance factor (MprF) of Pseudomonas aeruginosa. MprF proteins mediate the transfer of amino acids from aminoacyl-tRNAs to negatively charged phospholipids resulting in reduced membrane interactions with cationic antimicrobial peptides (produced by the host and competing microorganisms). The phospholipid modifications involve in most cases the transfer of lysine or alanine to phosphatidylglycerol. MprF proteins are membrane proteins consisting of a soluble and hydrophobic domain. Multiple functional studies have shown that the soluble domain of MprF mediates the aminoacylation of phosphatidylglycerol, while the hydrophobic domain mediates the "flipping" of aminoacylated phospholipids across the membrane, a process that is crucial to repulse or prevent the interaction of antimicrobial peptides encountered at the outer leaflet of bacterial membranes. Aside from its role in conferring antimicrobial peptide resistance, other roles of MprF have been described including more physiological roles such as improving growth under acidic conditions. Interestingly, MprF proteins are also found in Gram-negative bacteria which are already protected by an additional membrane that includes LPS. However, in Pseudomonas aeruginosa, MprF confers phenotypes that are similar to those observed in Gram-positive bacteria. Importantly, crystal structures of the soluble domain have led to important insights into aminoacyl phospholipid synthesis and recent studies on the cryoEM structure of Rhizobium tropici have confirmed functional and preliminary structural studies with other MprF proteins. The cryoEM structure from R. tropici confirmed the dimeric structure of MprF and supported a role of the hydrophobic domain in flipping lysyl-phosphatidylglycerol across the membrane. A comparison of the structures of lysyl-phosphatidylglycerol with alanyl-phosphatidylglycerol producing MprFs could reveal new insights into the mechanism of transferring aminoacyl-phospholipids from the soluble domain to the hydrophobic domain and translocation of alanyl- vs lysyl-phosphatidylglycerol across the membrane.

    Major concerns:

    1. The study by Jha and Vinothkumar provides the cryoEM structure of an alanyl-phosphatidylglycerol producing MprF protein which is in principle an important milestone in gaining a better understanding of the mechanism of aminoacyl-phospholipid synthesis and flipping, including the potentially different requirements of accommodating different aminoacyl -tRNAs and aminoacyl-phospholipid species. However, this is not addressed. The authors present a "distinct architecture" compared to the structure of R. tropici- MprF, without providing functional insights and the focus of the study shifts to the role of detergents in determining MprF structures via cryoEM. Thus, after fundamental discoveries have been made with crystal structures of the soluble domain and cryoEM structure of R. tropici, this study -while valuable as a resource- seems to offer only an incremental advance in understanding the mode of action of MprF and the potential different requirements for transferring alanyl-phosphatidylglycerol to the hydrophobic domain and flipping across the membrane. The reader is left with the finding of a distinct architecture with no further explanation or hypothesis.

    2. Differences to R.tropici MprF and other studies are difficult to follow as only a topological map of the Pseudomonas MprF is provided and conserved amino acids that have been shown to be crucial in mediating synthesis and flipping are not highlighted in the text or in the figures, specifically addressed, or discussed. Conserved amino acids in the presented cryoEM structure could provide important mechanistic insights and could address substrate specificity/requirements for aminoacyl phospholipid synthesis, transfer to the hydrophobic domain and flipping.

    3. Authors characterize an alanyl-phosphatidylglycerol producing MprF but do not detect the lipid in the cryoEM structure. Thus, the potential path taken by alanyl-phosphatidylglycerol remains unclear. Authors model the detected lipids as phosphatidylglycerol, which may be an interesting finding as it would indicate that MprF is generally capable of flipping phospholipids (this is however not discussed). While it is plausible that MprF flippases may be able to flip phosphatidyglycerol it could have a different path and structural requirements. It is also difficult to follow what the suggested pathway of flipping is in the Pseudomonas-MprF flippase (compared to R.tropici). Authors could provide a similar overview figure as in Song et al. and indicate what the potential differences are.

    Minor concerns:

    1. Page 13: the following sentence should be rephrased: "Among the missing links in the current cryoEM maps is the lack of well-ordered density for lipid molecules on the inner leaflet closer to the re-entrant helices but it is reasonable to assume from the cluster of positive charge that there will be lipid molecules and are dynamic. "

    2. Page 4: Klein et al do not show that the Pseudomonas aeruginosa MprF mediates flipping

    Significance

    General assessment:

    The study by Jha and Vinothkumar provides the cryoEM structure of an alanyl-phosphatidylglycerol producing MprF protein which is in principle an important milestone in gaining a better understanding of the mechanism of aminoacyl-phospholipid synthesis and flipping, including the potentially different requirements of accommodating different aminoacyl -tRNAs and aminoacyl-phospholipid species. However, this is not addressed. The authors present a "distinct architecture" compared to the structure of R. tropici- MprF, without providing functional insights and the focus of the study shifts to the role of detergents in determining MprF structures via cryoEM. Thus, after fundamental discoveries have been made with crystal structures of the soluble domain and cryoEM structure of R. tropici, this study -while valuable as a resource- seems to offer only an incremental advance in understanding the mode of action of MprF and the potential different requirements for transferring alanyl-phosphatidylglycerol to the hydrophobic domain and flipping across the membrane

    Advance: Minor

    Audience: Specialized

  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

    Shaileshanand J. et al., reported the structures of Multiple Peptide Resistance Factor, MprF, which is a bi-functional enzyme in bacteria responsible for aminoacylation of lipid head groups. The authors purified MprF from Pseudomonas aeruginosa in GDN micelles and nanodiscs, and by applying cryo-EM single particle method, they successfully reached near-atomic resolution, and built corresponding atomic models. By applying structural analysis as well as biochemistry methods, the authors demonstrated dimeric formation of MprF, exhibited the dynamic nature of the catalytic domain of this enzyme, and proposed a possible model on tRNA binding and aminoacylation.

    Major comments:

    1. In abstract, the authors stated 'Several lipid-like densities are observed in the cryoEM maps, which might indicate the path taken by the lipids and the coupling function of the two functional domains. Thus, the structure of a well characterised PaMprF lays a platform for understanding the mechanism of amino acid transfer to a lipid head group and subsequent flipping across the leaflet that changes the property of the membrane.' Firstly, those lipid-like densities were demonstrated in Fig 3A, since densities of lipids of purified membrane proteins often exist within regions of relatively low local resolution, or low quality, I think more detailed description on how the authors defined which part of the density belongs to lipid and how they acquired the modeling of some of the lipids is required. And the authors modeled phosphatidylglycerol into the GDN MprF, I would require additional experiment, for instance, mass spectrometry over the purified sample, to demonstrate the existence of this specific lipid with the sample. Secondly, regarding the last sentence in the abstract, how these structures lay a platform for further understanding was poorly discussed in both result section and discussion section, since the authors clearly stated 'This cavity perhaps provides a path for holding lipids...', then the statement in the next sentence 'Taken together... the vicinity to the cavities described above indicates the possible path taken by the lipids to enter and exit the enzyme' does not have a reliable evidence to support this conclusion, I would suggest the authors move these statements into discussion section, and elaborate more over this issue since it is an important part in the abstract, or make a more solid proof using other approaches, such as molecular dynamics simulation, to make these statements solid in the result section.

    2. Fig 2B, it seems the H566 sidechains were overlapping in the zoom-in figure of distance measurement between H566 residues, to clarify this, authors should either present another figure with rotation, to better demonstrate their relative locations, or swap this zoom-in figure with another figure with rotations. Also, could the authors briefly commenting on why they chose H566 for distance measurement specifically?

    3. Related to previous comment, I see one additional green square in Fig. 2A and an additional green square in Fig. 2B, without any zoom-in images provided on these regions. Besides, they're focusing on two different domains with same color, any particular reason why they're there? If so, please provide the information in figure legends.

    4. Related to previous comment, authors should also provide distance measurement over electrostatic interaction sites in Fig. 2A, since distance plays as an important factor in these forces.

    5. For Fig. 2C, since in Fig. 1, the authors have already indicated the differences between reconstruction of the GDN and nanodisc datasets, this information provided here seems to be a bit abundant, I suggest either move this panel to Fig. 1, to make a visualization on both electron densities as well as atomic models, or move this panel to supplementary figures.

    6. Fig. 3B, some of the spheres of the lipids were also marked as red, any particular reason why they're red? Do they indicate they're phosphate heads? If so, could the authors provide evidences how they define these orientations of the lipid heads? If not, any particular reason why they're red?

    7. Fig. 3C, the fitted model of lipid and its corresponding density should be added to Fig. S4, to give more detailed view on the quality of the fitting.

    8. Fig. 4D and 4E, could the authors also indicate the RMSD values when comparing the differences of RtMprF, PaMprF, ReMprF, this information would be helpful to understand how big of a difference within these three models.

    9. Fig. 6E, the coloring used for CCA-Ala were similar to the blue part of soluble domain, could the authors change the coloring a bit? Also, for Fig. 6F, I would suggest the authors provide a prediction model, such as using AlphaFold3, of this tRNA interaction site, to further validate this proposed model.

    10. In Supplementary Figures S1 and S3, the angular distribution of maps exhibited preferred orientation to certain extent, 3D FSC estimation should also be supplied for these maps, as an indication of whether the reconstructed densities were affected or not.

    11. For Fig S3B, could the authors switch to another image with better contrast?

    Minor comments:

    1. Fig. 2E and 2F, distance measurement should also be supplied to these two panels.

    2. Fig. 5D, since in Fig. 4F and 4G already mentioned the skeleton of GDN, this modeling part should be presented before exhibit it in dimer interface, the authors should rearrange the sequence over these three panels.

    3. In Supplementary Figure S3, which density was shown for the PaMprF local resolution estimation result? Authors should provide this information as two maps were shown in this figure.

    CROSS-REFEREE COMMENTS

    Both Reviewer #1 and #3 made comments over technical issue, their evaluation over functional aspects of this protein is what I was lacking over my comments, also, their evaluation of the biological narrative, relevance toward previous research is also more insightful. Finally, they offer valuable suggestions on how to adjust the article to make it more readable, and better describing the biological story which I would suggest the authors to pay attention to.

    Significance

    Significance

    The authors mainly focused on the structure of MprF in Pseudomonas aeruginosa, this protein is essential for the resistance to cationic antimicrobial peptides. A combination of structural and biochemical analysis provided evidences to the dimeric formation to this enzyme, and the analysis over differences of purified proteins using GDN and nanodisc was particular interesting, which provide new insight regarding the flexible nature of this enzyme, and potentially could be beneficial to the membrane protein community, as it demonstrates the differences in detergent/nanodisc of choice could affect the assembly of the protein of interest. Still, some of the statements in the manuscript, for instance, the assignment of lipids was over-claimed and could be benefited from additional approaches to support the issue. I would suggest some refinement in the discussion section as well as some of the figures.

    My expertise: cryo-EM single particle analysis; cryo-ET; sub-tomo averaging; cryo-FIB;

  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

    MprF proteins exist in many bacteria to synthesize aminoacyl phospholipids that have diverse biological functions, e.g. in the defense against small cationic peptides. They integrate two functions, the aminoacylation of lipids, i.e. the transfer of Lys, Arg or Ala from tRNAs to the head group, and the flipping of these modified lipids to the membrane outer leaflet. The authors present structures of MprF from Pseudomonas aeruginosa and describe these structures in great detail. As MprF enzymes confer antibiotic resistance and are therefore highly important, studying them is significant and interesting. Consequently, their structures have been substantially characterized in recent years, including the publication of the dimeric full-length MpfR from Rhizobium (Song et al., 2021).

    • While the structural work appears to be solid and carried out well on the technical part, one big criticism is how the data are presented in the manuscript, how they are analyzed and how they are put into relation to previous work. As structures of Mpfr from Rhizobium have been published, it is not required and rather distracting to explain the methodological details and the structure of Pseudomonas MprF in such great detail. Instead, the manuscript would benefit very strongly from reaching the interesting and novel parts, the comparison with the previous structures, as early as possible. Overall, the manuscript should be substantially shortened to not divert the reader's attention away from the novel parts by drowning them in miniscule description of the structural features such as secondary structure elements or lipid molecule positions where it remains completely unclear what their relevance is to the story and the message of the paper. Finally, during this revision, care should be taken to improve the language and maybe involve a native speaker in doing so.

    • Even more importantly, since the authors observe a dimer interface which strongly deviates from the previously presented arrangement of another species, the most important thing would be to properly characterize this interface and experimentally validate it, both of which has not been done sufficiently. When also taking into account that there were significant differences in the arrangement of the dimer between their structures in GDN and nanodisc, and that in the GDN structure, the cholesterol backbone of GDN appears to be involved in the interface (there should not be any cholesterol in native bacterial membranes!), there is a realistic chance that the observed dimer is an artefact. If the authors cannot convincingly rule out this possibility, all their conclusions are meaningless.

    • Hence, while I think that the data presented here would be worth publishing. However, a major drawback is that the authors do not sufficiently analyse, characterise and validate the dimer interface and fail to show that the dimer is biologically relevant.

    Major points:

    •      The authors always jump between their structures in detergent and nanodisc during all the descriptions, which makes following the story even more difficult. Please first describe one of the structures and then (briefly) discuss relevant similarities and differences afterwards.

    •      The difference in dimerization between Pseudomonas and Rhizobium is the most interesting and surprising feature (if true) of the new structures. However, it is not really presented as such. The authors should put more emphasis on making clear that this is a complete rotation of the monomers with respect to each other (by how many degrees?) and they should visualize it even more clearly in Figure 4 (and label the figure so that it is possible to understand it without having to read the text or the legend first).

    •      P. 10: The authors insinuate that only one of the dimer interfaces, either Pseudomonas or Rhizobium could be real, but disregard the possibility that both might be the biologically relevant interfaces of the respective species and that there might have been a switch of interfaces during evolution. They should also mention and discuss this possibility.

    •      Fig. 5G: The authors claim that the higher molecular band that appears in the mutant is a "dimer with aberrant migration" of >250 kDa as opposed to the expected 150 kDa. They should explain how they came to this conclusion and how they can be sure that the band does not correspond to a higher oligomer (trimer or tetramer). They could show, by extraction and purification scheme similar to the wildtype using first LMNG and then GDN, followed by at least a preliminary EM analysis, that the crosslinked mutant MprF is indeed a dimer, or use other biophysical methods to do the same, otherwise this experiment does not show much. Furthermore, they should also include a cysteine mutant in the part of Pseudomonas MprF that would be involved in a Rhizobium-like interface in their crosslinking experiments to check whether they could also stabilize dimers in this case.

    •      As the question whether the observed interface is real or an artefact is very central to the value of the structural data and the drawn conclusions from it, the authors should make more effort to analyze and try to validate the interface. First, an analysis of interface properties (buried surface area, nature of the interactions, conservation) should be performed for the interface as observed in the Pseudomonas structure but also for a (hypothetical) Rhizobium-like interface of two Pseudomonas monomers (such a model of a dimer should be easily obtainable by AlphaFold using the available Rhizobium structures as models). Then, experimental methods such as FRET or crosslinking-MS would allow to draw more solid conclusions on the distances between potential interface residues. While these experiments are a certain effort, the question whether the dimer interface is real is so central to the paper that it would be worthwhile to make this effort.

    •      As it seems that detergents might disrupt or modify the dimer interface, it might be an alternative to solubilize the protein in a more native environment by polymer-stabilized nanodiscs using DIBMA or similar molecules.

    •      Since parts of the Discussion are mostly repetitions of the Results part and other parts of the Discussion also contain a large extend of structure analysis one would usually rather expect in the Results part instead of the Discussion, the authors should consider condensing both to a combined (and overall much shorter) Results & Discussion section.

    Minor points:

    •      Explain abbreviations the first time they appear in the text, e.g. TTH

    •      Figure labels are very minimalistic. This should be improved, e.g. by putting labels to important structural features that appear in the text, otherwise the figures are not an adequate support for the text.

    •      Figure 5: Label where the different oligomers run on the gels

    Significance

    While the structural work appears to be solid and carried out well on the technical part, one big criticism is how the data are presented in the manuscript, how they are analyzed and how they are put into relation to previous work. As structures of Mpfr from Rhizobium have been published, it is not required and rather distracting to explain the methodological details and the structure of Pseudomonas MprF in such great detail. Instead, the manuscript would benefit very strongly from reaching the interesting and novel parts, the comparison with the previous structures, as early as possible. Overall, the manuscript should be substantially shortened to not divert the reader's attention away from the novel parts by drowning them in miniscule description of the structural features such as secondary structure elements or lipid molecule positions where it remains completely unclear what their relevance is to the story and the message of the paper. Finally, during this revision, care should be taken to improve the language and maybe involve a native speaker in doing so.

    • Even more importantly, since the authors observe a dimer interface which strongly deviates from the previously presented arrangement of another species, the most important thing would be to properly characterize this interface and experimentally validate it, both of which has not been done sufficiently. When also taking into account that there were significant differences in the arrangement of the dimer between their structures in GDN and nanodisc, and that in the GDN structure, the cholesterol backbone of GDN appears to be involved in the interface (there should not be any cholesterol in native bacterial membranes!), there is a realistic chance that the observed dimer is an artefact. If the authors cannot convincingly rule out this possibility, all their conclusions are meaningless.

    • Hence, while I think that the data presented here would be worth publishing. However, a major drawback is that the authors do not sufficiently analyse, characterise and validate the dimer interface and fail to show that the dimer is biologically relevant.

  5. Version published to 10.1101/2024.08.21.608925 on bioRxiv