Dissecting the phase separation and oligomerization activities of the carboxysome positioning protein McdB

  1. Joseph L. Basalla
  2. Claudia A. Mak
  3. Jordan Byrne
  4. Maria Ghalmi
  5. Y Hoang
  6. Anthony G. Vecchiarelli

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    Carboxysomes enable the efficient fixation of carbon dioxide in specific bacteria. Phase separation has been invoked as a mechanism that drives the formation of carboxysomes. The current work focuses on the biophysical principles of how one of two essential specific protein components enable spatial regulation over carboxysomes. This important work highlights the connection between oligomerization via specific molecular interactions and phase separation. The work is of interest to the areas of biochemistry and carbon dioxide fixation as well as phase separation.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Across bacteria, protein-based organelles called bacterial microcompartments (BMCs) encapsulate key enzymes to regulate their activities. The model BMC is the carboxysome that encapsulates enzymes for CO 2 fixation to increase efficiency and is found in many autotrophic bacteria, such as cyanobacteria. Despite their importance in the global carbon cycle, little is known about how carboxysomes are spatially regulated. We recently identified the two-factor system required for the maintenance of carboxysome distribution (McdAB). McdA drives the equal spacing of carboxysomes via interactions with McdB, which associates with carboxysomes. McdA is a ParA/MinD ATPase, a protein family well-studied in positioning diverse cellular structures in bacteria. However, the adaptor proteins like McdB that connect these ATPases to their cargos are extremely diverse. In fact, McdB represents a completely unstudied class of proteins. Despite the diversity, many adaptor proteins undergo phase separation, but functional roles remain unclear. Here, we define the domain architecture of McdB from the model cyanobacterium Synechococcus elongatus PCC 7942, and dissect its mode of biomolecular condensate formation. We identify an N-terminal intrinsically disordered region (IDR) that modulates condensate solubility, a central coiled-coil dimerizing domain that drives condensate formation, and a C-terminal domain that trimerizes McdB dimers and provides increased valency for condensate formation. We then identify critical basic residues in the IDR, which we mutate to fine-tune condensate solubility. Finally, we find that a condensate-defective mutant of McdB has altered association with carboxysomes and influences carboxysome enzyme content. The results have broad implications for understanding spatial organization of BMCs and the molecular grammar of protein condensates.

Article activity feed

  3. Arcadia Science

    (Fig. 4D)

    I would be curious to see if there were any spatial distribution differences between the truncations before photobleaching. Did you notice any differences that “agree” or “disagree” with how the 18-hour incubated IDR + CC fragment showed exceptional recovery? How did spatial distribution of the newly formed full McdB condensates compare to the 18-hour incubated IDR + CC fragment?

  4. Arcadia Science

    (Fig. 3A)

    I really enjoyed reading about your approach to thoroughly understand the McdB-driven phase separation and oligomerization of S. elongatus carboxysomes! This was obviously a really challenging protein to work with but you did an incredible job! I noticed in Figure 3A that there is a difference in how spatially distributed the McdB proteins are before photobleaching. There seems to be a higher concentration of McdB proteins on the "outskirts" of the newly formed condensate aggregates while the "mature" aggregates have a more even distribution. Any idea as to why this is the case and/or if this could impact recovery differences after photobleaching?

  5. eLife

    Author Response

    Reviewer #1 (Public Review):

    Bacterial carboxysomes are compartments that enable the efficient fixation of carbon dioxide in certain types of bacteria. A focus of the current work is on two protein components that provide spatial regulation over carboxysomes. The McdA system is an ATPase that drives the positioning of carboxysomes. The McdB system is essential for maintaining carboxysome homeostasis, although how this role is achieved is unclear. Previous studies, by the lead author's lab, showed that the McdB system is a driver of phase separation in vitro and in cells. They proposed a putative connection between McdB phase separation and carboxysome homeostasis. The central premise of the current work is as follows: In order to understand if and how phase separation of McdB impacts carboxysome homeostasis, it is important to know how the driving forces for phase separation are encoded in the sequence and architecture of McdB. This is the central focus of the current work. The picture that emerges is of a protein that forms hexamers, which appears to be a trimer of dimers. The domains that drive that the dimerziation and trimerization appear to be essential for driving phase separation under the conditions interrogated by the authors. The N-terminal disordered region regulates the driving forces for phase separation - referred to as the solubility of McdB by the authors. To converge upon the molecular dissections, the authors use a combination of computational and biophysical methods. The work highlights the connection between oligomerization via specific interactions and emergent phase behavior that presumably derives from the concentration (and solution condition) dependent networking transitions of oligomerized McdB molecules.

    Having failed to obtain specific structural resolution for the full-length McdB as a monomer or oligomer, the authors leverage a combination of computational tools, the primary one being iTASSER. This, in conjunction with disorder predictors, is used to identify / predict the domain structure of McdB. The domain structure predictions are tested using a limited proteolysis approach and, for the most part, the predictions stand up to scrutiny affirming the PONDR predictions. SEC-MALS data are used to pin down the oligomerization states of McdB and the consensus that emerges, through the investigations that are targeted toward a series of deletion constructs, is the picture summarized above.

    Is the characterization of the oligomerization landscape complete and likely perfect? Quite possibly, the answer is no. Deletion constructs pose numerous challenges because they delete interactions and inevitably impose a modularity to the interpretation of the totality of the data.

    This is a good point and always a possibility with truncations – the protein McdB may not be as modular in nature as it seems in our tripartite model. But the deletion constructs were more so intended to be tools for identifying key regions of oligomerization and condensate formation as others have done, and for this, they were indeed useful. Additionally, we were able to strategically aim our substitution mutations based on data from the deletion constructs. These substitutions provided data consistent with the deletions, but in the context of the full-length protein (see Fig. 5 vs. Figs. 2, 4). However, we ultimately agree with the reviewer that this is always a possibility with truncations, and we have therefore mentioned this caveat in the discussion.

    Line 415 “Truncated proteins have been useful in the study of biomolecular condensates. But it is important to note that using truncation data alone to dissect modes of condensate formation can lead to erroneous models since entire regions of the protein are missing. However, data from our truncation and substitution mutants were entirely congruent. For example, deletion of the CTD or substitutions to this region caused destabilization of the hexamer to a dimer, and deletion of the IDR or substitutions to this region caused solubilization of condensates without affecting hexamer formation.”

    Accordingly, we are led to believe that the N-terminal IDR plays no role whatsoever in the oligomerization.

    Our updated data still strongly supports this interpretation. Both truncation of the IDR (Fig. 2) and the six-Q-substitution mutant in the IDR (Fig. 5) form a monodispersed hexamer in solution via SEC-MALS, as does wild-type McdB.

    Close scrutiny, driven by the puzzling choice of nomenclature and the Lys to Gln titrations in the N-terminal IDR raise certain unresolved issues. First, the central dimerization domain is referred to as being Q-rich. This does not square with the compositional biases of this region. If anything is Q/L or just L-rich. This in fact makes more sense because the region does have the architecture of canonical Leu-zippers, which do often feature Gln residues. However, there is nothing about the sequence features that mandates the designation of being Q-rich nor are there any meaningful connections to proteins with Q-rich or polyQ tracts. This aspect of the analysis and discussion is a serious and erroneous distraction.

    We changed the language here, and no longer refer to the central region as “Q-rich”. However, we would like to note that the second half of the McdB central domain is indeed enriched in glutamines (14/53 = 26.4%) to a comparable extent as the region of FUS, which has been shown to help drive condensate formation via glutamine H-bonding (14/44 = 31.8%; Murthy et al 2019). We were simply proposing that, at a molecular level, there was some insight to be gained from this comparison. We agree, however, that there is no functionally meaningful comparison between McdB and polyQ-tract proteins, as we may have previously alluded to in our discussion, and that text has been removed.

    Back to the middle region that drives dimerization, the missing piece of the puzzle is the orientation of the dimers. One presumes these are canonical, antiparallel dimers. However, this issue is not addressed even though it is directly relevant to the topic of how the trimer of dimers is assembled.

    Indeed, we were unable to resolve the orientation issue, despite much effort. The story we present is not a complete and final model of McdB structure, nor its molecular modes of oligomerization or condensate formation. However we now provide a discussion section “McdB homologs have polyampholytic properties between their N- and C-termini” that highlights this issue. We also mention the remaining dimer orientation issue at the end of the results section “Se7942 McdB forms a trimer-of-dimers hexamer”. However, we believe the data presented still provides useful initial models, which for example, allowed us to create a series of substitutions that tune McdB condensate solubility and verify that they do not affect oligomerization. We would like to further add that for other condensate forming proteins in bacteria, like the PopZ protein we mention in the text, there remains no detailed structural model beyond the resolution we provide here for McdB; despite PopZ being first identified in 2008. Over 40 publications on PopZ have progressively provided useful and more detailed models that are only now being used to develop PopZ as a tool for condensate technologies that are furthering our understanding of the biological implications of condensate formation across all cell types. The intention with our current report is therefore not to generate a finalized molecular model of this entirely unstudied class of McdB proteins. But instead, to generate useful insight into McdB biochemistry that can advance our understanding of this class of protein’s function in vivo. To this end, we now add in vivo data based on these initial models where we specifically link cellular phenotypes to McdB condensate solubility (Fig. 8). Of course, there are several follow-up studies that come from the current report, but we believe that speaks to the value of the presented research in advancing this field.

    If the trimer is such that all binding sites are fully satisfied (with the binding sites presumably being on the C-terminal pseudo-IDR), then the hexamer should be a network terminating structure, which it does not seem to be based on the data. Instead, we find that only the full-length protein can undergo phase separation (albeit at rather high concentrations) in the absence of crowder. We also find that the driving forces for phase separation are pH dependent, with pH values above 8.5 being sufficient to dissolve condensates. Substitution of Lys to Gln in the N-terminal IDR leads to a graded weakening of the driving forces for phase separation. The totality of these data suggest a more complex interplay of the regions than is being advocated by the authors.

    Thank you and we agree. As we discuss above in response #4 and below in response #7, we have changed the focus and tone of our report to say that, while the models we have generated are useful, we are aware they are incomplete at a molecular level. Furthermore, as we describe in response #6, we have added several new McdB mutants to investigate more deeply the role of the CTD, but this region was not amenable to mutagenesis as these mutants affected McdB oligomerization. Lastly, while network forming interactions are certainly important for condensate formation as the reviewer describes, so are solvent interactions. We have added new text and data related to Figs. 3, 4 that address these issues.

    Almost certainly, there are complementary electrostatic interactions among the N-terminal IDR and C-terminal pseudo IDR that are important and responsible for the networking transition that drives phase separation, even if these interactions do not contribute to hexamer formation. The net charge per residue of the 18-residue N-terminal IDR is +0.22 and the NCPR of the remainder is ≈ -0.1. To understand how the N-terminal IDR is essential, in the context of the full-length protein, to enable phase separation (in the absence of crowder), it is imperative that a model be constructed for the topology of the hexamer. It is also likely that the oligomer does not have a fixed stoichiometry.

    We agree and thank the reviewer for these comments. We have added several new substitution mutants aimed at addressing this (Figs. 5, S6). However, the C-terminus was not amenable to substitutions as the trimer-of-dimers was significantly destabilized in these mutants (Figs. 5, S7). Therefore, in this report we were unable to determine specifically how the basic residues in the IDR contribute to condensate formation. However, with the addition of new data in Fig. 8, we think we adequately show that the IDR mutants can be used to investigate McdB condensate formation in vivo, and that follow-up studies will be aimed at investigating these details. We have also added an new discussion section “McdB homologs have polyampholytic properties between their N- and C-termini” that highlight this very likely possibility suggested by the reviewer.

    Therefore, the central weakness of the current work is that it is too preliminary. A set of interesting findings are emerging but by fixating on Lys to Gln titrations within the N-terminal IDR and referring to these titrations as impacting solubility, a premature modular and confused picture emerges from the narrative that leaves too many questions unanswered.

    The work itself is very important given the growing interest in bacterial condensates. However, given that the focus is on understanding the molecular interactions that govern McdB phase behavior - a necessary pre-requisite in the authors minds for understanding if and how phase separation impacts carboxysome homeostasis - it becomes imperative that the model that emerges be reasonably robust and complete. At this juncture, the model raises far too many questions.

    We agree that our previous report was focused mainly on the molecular basis of McdB condensate biochemistry, and in that report we left the model short. In this revised version, we have added several pieces of new data that strengthen the model (Figs. 3-5), although it is still incomplete. However, in this revised version, we have also shifted the focus from a complete biochemical understanding of McdB condensates to a study that links McdB condensate formation in vitro to phenotypes in vivo. In this regard, we have added the in vivo data in Fig. 8 and somewhat changed the focus in the text.

    The MoRF analysis is distraction away from the central focus.

    The MoRF analysis has been removed.

    The problem, as I see it, is that the authors have gone down the wrong road in terms of how they have interpreted the preliminary set of results. Further, the methods used do not have the resolution to answer all the questions that need to be answered. Another issue is that a lot of standard tropes are erected and they become a distraction. For example, it is simply not true that in a protein featuring folded domains and IDRs it almost always is the case that the IDR is the driver of phase transitions. This depends on the context, the sequence details of the IDRs, and whether the interactions that contribute to the driving forces for phase separation are localized within the IDR or distributed throughout the sequence. In McdB it appears to be the latter, and much of the nuance is lost through the use of specific types of deletion constructs.

    Thank you. We have removed much of this and changed the diction on how our current model of McdB condensate formation fits into the literature in the discussion.

    Overall, the work represents a good beginning but the data do not permit a clear denouement that allows one to connect the molecular and mesoscales to fully describe McdB phase behavior. Significantly more work needs to be done for such a picture to emerge.

    Reviewer #2 (Public Review):

    In this work, Basalla et al. study the biochemical properties of the carboxysome positioning protein, McdB. Using in vitro experiments, the authors characterize McdB oligomeric states and the domains driving and modulating its phase separation. Based on bioinformatics analysis, the authors identify a putative binding recognition motif between McdB and its two-component system counterpart McdA. As McdAB-like systems emerge as spatial regulators of bacterial compartments, the data presented here may be of general interest. The study is well executed and provides exciting hypotheses to be tested in vivo.

    The authors found that McdB from S. elongatus PCC 7942 consists of three domains: an N-terminal 18 aa disordered region, a Q-rich helical domain, and a helical C-terminal domain (CTD). Analyzing these domains, the authors present three key results: (i) The Q-rich domains form dimers, and the CTD drives the formation of trimers of dimers (ii) Phase separation is pH sensitive, driven by the Q-rich domain, and modulated by basic residues in the IDR, (iii) The IDR contains a putative recognition motif that binds McdA. While these three sets of results are rich in data, they are disjointed. Relating the three datasets (oligomeric states of the protein, its phase separation behavior, and its ability to bind McdA) is required to provide a complete picture of the molecular mechanism driving McdB condensation.

    Specific comments:

    1. The main limitation of this manuscript is the lack of integration between the three areas of results. In particular: how do the IDR basic residues disrupt phase separation? Is that through interference with either the dimer or timer interface? Does the McdB IDR regulate phase separation behavior when bound to McdA? Or, in other words, is the MoRF acting both as a binding interface and as a solubility regulator, and if so, can both functions be achieved simultaneously? It seems like the MoRF includes at least three basic residues.

    Indeed, we were unable to fully resolve the specific molecular interactions that give rise to condensates versus those that give rise to oligomers, and how these two modes of self-association contribute to one another. One limitation was that, as shown in our new data, the CTD was not amenable to mutagenesis, as it caused destabilization of the trimer-of-dimers (Fig. 5, Fig. S7). Therefore, we could not dissect how the CTD contributes to oligomerization versus driving condensates. However, we did include in vivo data showing how the IDR mutations allowed us to specifically link phenotypes to McdB condensate solubility (Fig. 8). As we discuss above in responses #4, #6, and #7, we changed the focus of the revised manuscript from the molecular basis of McdB condensate formation to linking McdB condensate formation in vitro and its functionality in vivo. To this end, we think the IDR mutation set has been useful, and follow-up studies will be done to further the molecular model of McdB condensate formation. Reviewers 1 and 3 deemed the MoRF section a distraction. Therefore, MoRF analysis and discussions of McdA interactions with this potential MoRF have been removed.

    Finally, what is the effective concentration of McdB in cells, and how does that translate to the in vitro studies?

    In our previous version, we used McdB concentrations between 50-100 µM. We do not know the in vivo concentration of McdB. We have tried several antibodies against McdB, and a few were good enough to detect the presence of McdB, but not quantifiably. We therefore believe in vivo McdB levels are low (sub-micromolar), and definitely lower than the range we previously used in our in vitro studies. In our revised manuscript, we include a titration of McdB at lower concentrations, and see condensates at McdB concentrations lower than 2 µM.

    1. How general are the conclusions made here to other McdBs? The authors have published nice work surveying the commonalities and differences between homologous McdB proteins. Can you comment on the applicability of your findings to other McdB proteins?

    This is a great point, which we have added to a new discussion section titled “McdB homologs have polyampholytic properties between their N- and C-termini”.

    Additional issues:

    1. Using SEC and SEC-MALS, the authors demonstrated that the Q-rich domain forms a stable dimer and that the full-length protein forms hexamers, suggesting trimers of dimers assembly. The authors also suggest that the CTD is responsible for forming those trimers of dimers based on SEC-MALS measurements. However, Figure 2D shows that while the full length runs at 6.6x the monomer, the Q-rich+CTD runs at 5.4x the monomer. First, I could not find SEC-MALS of the full-length protein, and it is not clear whether SEC-MALS was used for all or a fraction of the constructs discussed in Figure 2D. Second, could it be that the Q-rich domain+CTD is an ensemble of hexamers and dimers? Perhaps the IDR is playing a secondary role in stabilizing the hexamer?

    We have repeated the SEC-MALS experiments and included the full-length protein (Fig. 2). Furthermore, we have included SEC-MALS for some of the key substitution mutants (Figs. 5, S7). With the additional findings, our conclusions remain the same as in our previous version of the manuscript.

    1. The analysis of the phase separation results needs to have some extra quantification. The authors show that at 100 uM protein with 10% PEG the full-length phase separates as well as IDR+Q-rich. Lines 176-178: "The CTD, on the other hand, has no effect on the Q-rich domain condensates; Q-rich+CTD condensates formed at the same protein concentration and with identical droplet morphologies at the Q-rich domain alone." It is hard to draw this conclusion solely based on the data presented in Figure 3. An alternative interpretation might be that Q-rich+CTD reduces csat. I suggest the authors include turbidity assays (as shown for pH effect) to quantitively determine csat for these different constructs and perhaps perform FRAP to determine the mobility of these different constructs. In addition, how long after the addition of PEG were these droplets imaged?

    We now include an additional figure where we characterize condensates for full-length McdB (Fig. 3), including FRAP as suggested by the reviewer. We also include additional experiments for the truncations as requested (Fig. 4), and relate the truncation data to the model we propose for the full-length protein. All condensate samples were incubated for 30 mins prior to imaging unless otherwise stated, which we have added to the methods section “Microscopy of protein condensates”.

    1. Solubility assays shown in Figures 4A, B, D, and 5C are missing error bars. Without replicates, it is difficult to assess, for example, the effect of KCl.

    We have included replicates and error bars. Apologies for the omission.

    Also, please indicate the physiological ranges of KCl and pH in Figure 6. The phase separation sensitivity to pH is intriguing. By changing basic residues to glutamines, the authors conclude that the positive charge of the IDR modulates solubility. The Q-rich domain, however, is negatively charged. Can the authors comment on the role of acidic residues in the Q-rich domain? Are they required for phase separation? Also - based on your previous bioinformatics analysis, are the charges of the IDR and the Q-rich domains conserved across McdB homologs?

    Data from this report, and as described by reviewer #1, suggest that charge in the CTD, and not the central region, may be important. Our previous report (MacCready et al., Mol Biol Evol. 2020) touches on the conservation of charge in the NTD and CTD, which we have now added to the discussion section titled ““McdB homologs have polyampholytic properties between their N- and C-termini””. However, we were unable to experimentally verify electrostatic associations between the NTD and CTD because the CTD was not amenable to mutagenesis, as shown in our new data added to the manuscript (Figs. 5, S7).

    1. In previous work, the authors showed a conserved RKR segment in the IDR is highly conserved and missing in S. elongatus PCC 7942 (MacCready et al., Mol Biol Evol. 2020). Given the current finding, it would be important to understand whether the RKR deletion carries functional implications for phase separation behavior.

    The RKR segment is not missing, but likely relates to the KKR residues from S. elongatus PCC 7942. We describe this in more detail elsewhere (MacCready et al., Mol Biol Evol. 2020). However, as we show here, these specific residue locations do not seem to be especially important for condensate formation, but instead the overall net charge of the IDR mediates condensate solubility regardless of the specific residues mutated (Fig. 6).

    1. McdB proteins with 2Q left mutated vs. 2Q middle and 2Q right seem to result in condensates with different material properties (e.g., DIC pictures show different droplet morphologies for the different constructs). Is that the case? And if so, can you comment on that?

    We have included a brief mention of this in the text. However, the overall interpretation of these results remains that regardless of the residues mutated, there is a comparable degree of condensate solubilization for constructs with the same IDR net charge (Fig. 6).

    Reviewer #3 (Public Review):

    Through a series of rigorous in vitro studies, the authors determined McdB's domain architecture, its oligomerization domains, the regions required for phase separation, and how to fine-tune its phase separation activity. The SEC-MALS study provides clear evidence that the α-helical domains of McdB form a trimer-of-dimers hexamer. Through analysis of a small library of domain deletions by microscopy and SDS-PAGE gels of soluble and pellet fractions, the authors conclude that the Q-rich domain of McdB drives phase separation while the N-terminal IDR modulates solubility. A nicely executed study in Figure 4 demonstrated that McdB phase separation is highly sensitive to pH and is influenced by basic residues in the N terminal IDR. The study demonstrates that net charge, as opposed to specific residues, is critical for phase separation at 100 micromolar. In addition, the experimental design included analysis of McdB constructs that lack fluorescent proteins or organic dyes that may influence phase separation. Therefore, the observed material properties have full dependence on the McdB sequence.

    Thank you for the kind words and this perspective. We have added a brief mention to it in the discussion section titled “McdB condensate formation follows a nuanced, multi-domain mechanism”: “Furthermore, it should be noted that the McdB constructs used in our in vitro assays were free from fluorescent proteins, organic dyes, or other modification that may influence phase separation. Therefore, the observed material properties of these condensates have full dependence on the McdB sequence.”

    Studies of proteins often neglect short, disordered segments at the N- or C- terminus due to unclear models for their potential role. This study was interesting because it revealed a short IDR as a critical regulator of phase separation. This includes experiments that remove the IDR (Fig 2 & 3) and mutate the basic residues to show their importance towards McdB phase separation. In a nice set of SDS-PAGE experiments, the authors showed that as the net charge of the IDR decreased the construct became more soluble.

    One challenge is in the experimental design when mutating residues is to assess their impact on phase separation. The author's avoided substitutions to alanine, as alanine substitutions have synthetically stimulated phase separation in other systems. The authors, therefore, have a good rationale for selecting potentially milder mutations of lysine/arginine to glutamine. A potential caveat of mutation to glutamine is that stretches of glutamines have been associated with amyloid/prion formation. So, the introductions of glutamines into the IDR may also have unexpected effects on material properties. Despite these caveats, the authors show mutation of six basic residues in the short IDR abolished phase separation at 100 mM.

    Thank you for the thoughtful consideration, and appreciation of our work! Reviewer 1 had reservations for the Gln substitutions as well. We also used Alanine in new data added to the manuscript. But as the reviewer notes, the alanine mutations artificially drove further phase separation activity, and even aggregation. We show that mutants with the introduction of glutamines, however, remain soluble in vitro and in E. coli even at very high concentrations. Furthermore, we now include SEC-MALS of the McdB variant with 6 glutamines introduced in the IDR and show that there is no impact on oligomeric state. Together the data show no amylogenic properties of these glutamine enriched mutants.

    We have added a note to this potential caveat in the discussion section “McdB condensate formation follows a nuanced, multi-domain mechanism”: “Glutamine-rich regions are known to be involved in stable protein-protein interactions such as in coiled-coils and amyloids (52, 53), and expansion of glutamine-rich regions in some proteins lead to amylogenesis and disease (54, 55). However, when we introduced glutamines into the IDR of McdB solubility was increased both in vitro and in vivo, and without any impact on hexamerization. Together, the data show that increasing the glutamine content in the IDR of McdB did not lead to amylogenesis, but rather increased solubility. Our findings therefore underpin the importance of positive charge in the IDR specifically for stabilizing McdB condensates.”

    Computational studies (Fig 7) also suggest that this short N-IDR region may play a role as a MORF upon potential binding to a second protein McdA. The formulation of this hypothesis is strengthened by the fact that for other ParA/MinD-family ATPases, the associated partner proteins have also been shown to interact with their cognate ATPase via positively charged and disordered N-termini. This aspect of understanding McdB's N-IDR as a MORF is at a very early stage. This study lacks experimental evidence for an N-IDR: McdA interaction and experimental data showing conformational change upon McdA binding. However, the computation study sets up the future to consider whether and how the phase separation activity of McdB is related to its structural dynamics and interactions with McdA.

    Based off of these comments and from Reviewer 1 comments, we have removed the MoRF analyses entirely. The MoRF analysis will be coupled to another study in the lab focused on McdB interactions with McdA.

    In summary, this study provides a strong foundation for the contribution of domains to McdB's in vitro phase separation. This knowledge will inform and impact future studies on McdB regulating carboxysomes and how the related family of ParA/MinD-family ATPases and their cognate regulatory proteins. For example, it is unknown if and how McdB's phase separation is utilized in vivo for carboxysome regulation. However, the revealed roles of the Q-rich domain and N-IDR will provide valuable knowledge in developing future research. In addition, the systematic domain analysis of McdB can be combined with a similar analysis of a broad range of other biomolecular condensates in bacteria and eukaryotes to understand the design principles of phase separating proteins.

  6. Version 3 published on bioRxiv
  7. eLife

    Evaluation Summary:

    Carboxysomes enable the efficient fixation of carbon dioxide in specific bacteria. Phase separation has been invoked as a mechanism that drives the formation of carboxysomes. The current work focuses on the biophysical principles of how one of two essential specific protein components enable spatial regulation over carboxysomes. This important work highlights the connection between oligomerization via specific molecular interactions and phase separation. The work is of interest to the areas of biochemistry and carbon dioxide fixation as well as phase separation.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

  8. eLife

    Reviewer #1 (Public Review):

    Bacterial carboxysomes are compartments that enable the efficient fixation of carbon dioxide in certain types of bacteria. A focus of the current work is on two protein components that provide spatial regulation over carboxysomes. The McdA system is an ATPase that drives the positioning of carboxysomes. The McdB system is essential for maintaining carboxysome homeostasis, although how this role is achieved is unclear. Previous studies, by the lead author's lab, showed that the McdB system is a driver of phase separation in vitro and in cells. They proposed a putative connection between McdB phase separation and carboxysome homeostasis. The central premise of the current work is as follows: In order to understand if and how phase separation of McdB impacts carboxysome homeostasis, it is important to know how the driving forces for phase separation are encoded in the sequence and architecture of McdB. This is the central focus of the current work. The picture that emerges is of a protein that forms hexamers, which appears to be a trimer of dimers. The domains that drive that the dimerziation and trimerization appear to be essential for driving phase separation under the conditions interrogated by the authors. The N-terminal disordered region regulates the driving forces for phase separation - referred to as the solubility of McdB by the authors. To converge upon the molecular dissections, the authors use a combination of computational and biophysical methods. The work highlights the connection between oligomerization via specific interactions and emergent phase behavior that presumably derives from the concentration (and solution condition) dependent networking transitions of oligomerized McdB molecules.

    Having failed to obtain specific structural resolution for the full-length McdB as a monomer or oligomer, the authors leverage a combination of computational tools, the primary one being iTASSER. This, in conjunction with disorder predictors, is used to identify / predict the domain structure of McdB. The domain structure predictions are tested using a limited proteolysis approach and, for the most part, the predictions stand up to scrutiny affirming the PONDR predictions. SEC-MALS data are used to pin down the oligomerization states of McdB and the consensus that emerges, through the investigations that are targeted toward a series of deletion constructs, is the picture summarized above.

    Is the characterization of the oligomerization landscape complete and likely perfect? Quite possibly, the answer is no. Deletion constructs pose numerous challenges because they delete interactions and inevitably impose a modularity to the interpretation of the totality of the data. Accordingly, we are led to believe that the N-terminal IDR plays no role whatsoever in the oligomerization. Close scrutiny, driven by the puzzling choice of nomenclature and the Lys to Gln titrations in the N-terminal IDR raise certain unresolved issues. First, the central dimerization domain is referred to as being Q-rich. This does not square with the compositional biases of this region. If anything is Q/L or just L-rich. This in fact makes more sense because the region does have the architecture of canonical Leu-zippers, which do often feature Gln residues. However, there is nothing about the sequence features that mandates the designation of being Q-rich nor are there any meaningful connections to proteins with Q-rich or polyQ tracts. This aspect of the analysis and discussion is a serious and erroneous distraction. Back to the middle region that drives dimerization, the missing piece of the puzzle is the orientation of the dimers. One presumes these are canonical, antiparallel dimers. However, this issue is not addressed even though it is directly relevant to the topic of how the trimer of dimers is assembled. If the trimer is such that all binding sites are fully satisfied (with the binding sites presumably being on the C-terminal pseudo-IDR), then the hexamer should be a network terminating structure, which it does not seem to be based on the data. Instead, we find that only the full-length protein can undergo phase separation (albeit at rather high concentrations) in the absence of crowder. We also find that the driving forces for phase separation are pH dependent, with pH values above 8.5 being sufficient to dissolve condensates. Substitution of Lys to Gln in the N-terminal IDR leads to a graded weakening of the driving forces for phase separation. The totality of these data suggest a more complex interplay of the regions than is being advocated by the authors. Almost certainly, there are complementary electrostatic interactions among the N-terminal IDR and C-terminal pseudo IDR that are important and responsible for the networking transition that drives phase separation, even if these interactions do not contribute to hexamer formation. The net charge per residue of the 18-residue N-terminal IDR is +0.22 and the NCPR of the remainder is ≈ -0.1. To understand how the N-terminal IDR is essential, in the context of the full-length protein, to enable phase separation (in the absence of crowder), it is imperative that a model be constructed for the topology of the hexamer. It is also likely that the oligomer does not have a fixed stoichiometry.

    Therefore, the central weakness of the current work is that it is too preliminary. A set of interesting findings are emerging but by fixating on Lys to Gln titrations within the N-terminal IDR and referring to these titrations as impacting solubility, a premature modular and confused picture emerges from the narrative that leaves too many questions unanswered.

    The work itself is very important given the growing interest in bacterial condensates. However, given that the focus is on understanding the molecular interactions that govern McdB phase behavior - a necessary pre-requisite in the authors minds for understanding if and how phase separation impacts carboxysome homeostasis - it becomes imperative that the model that emerges be reasonably robust and complete. At this juncture, the model raises far too many questions. The MoRF analysis is distraction away from the central focus.

    The problem, as I see it, is that the authors have gone down the wrong road in terms of how they have interpreted the preliminary set of results. Further, the methods used do not have the resolution to answer all the questions that need to be answered. Another issue is that a lot of standard tropes are erected and they become a distraction. For example, it is simply not true that in a protein featuring folded domains and IDRs it almost always is the case that the IDR is the driver of phase transitions. This depends on the context, the sequence details of the IDRs, and whether the interactions that contribute to the driving forces for phase separation are localized within the IDR or distributed throughout the sequence. In McdB it appears to be the latter, and much of the nuance is lost through the use of specific types of deletion constructs.

    Overall, the work represents a good beginning but the data do not permit a clear denouement that allows one to connect the molecular and mesoscales to fully describe McdB phase behavior. Significantly more work needs to be done for such a picture to emerge.

  9. eLife

    Reviewer #2 (Public Review):

    In this work, Basalla et al. study the biochemical properties of the carboxysome positioning protein, McdB. Using in vitro experiments, the authors characterize McdB oligomeric states and the domains driving and modulating its phase separation. Based on bioinformatics analysis, the authors identify a putative binding recognition motif between McdB and its two-component system counterpart McdA. As McdAB-like systems emerge as spatial regulators of bacterial compartments, the data presented here may be of general interest. The study is well executed and provides exciting hypotheses to be tested in vivo.

    The authors found that McdB from S. elongatus PCC 7942 consists of three domains: an N-terminal 18 aa disordered region, a Q-rich helical domain, and a helical C-terminal domain (CTD). Analyzing these domains, the authors present three key results: (i) The Q-rich domains form dimers, and the CTD drives the formation of trimers of dimers (ii) Phase separation is pH sensitive, driven by the Q-rich domain, and modulated by basic residues in the IDR, (iii) The IDR contains a putative recognition motif that binds McdA. While these three sets of results are rich in data, they are disjointed. Relating the three datasets (oligomeric states of the protein, its phase separation behavior, and its ability to bind McdA) is required to provide a complete picture of the molecular mechanism driving McdB condensation.

    Specific comments:

    1. The main limitation of this manuscript is the lack of integration between the three areas of results. In particular: how do the IDR basic residues disrupt phase separation? Is that through interference with either the dimer or timer interface?
    Does the McdB IDR regulate phase separation behavior when bound to McdA? or in other words, is the MoRF acting both as a binding interface and as a solubility regulator, and if so, can both functions be achieved simultaneously? It seems like the MoRF includes at least three basic residues. Finally, what is the effective concentration of McdB in cells, and how does that translate to the in vitro studies?

    2. How general are the conclusions made here to other McdB? The authors have published nice work surveying the commonalities and differences between homologous McdB proteins. Can you comment on the applicability of your findings to other McdB proteins?

    Additional issues:

    3. Using SEC and SEC-MALS, the authors demonstrated that the Q-rich domain forms a stable dimer and that the full-length protein forms hexamers, suggesting trimers of dimers assembly. The authors also suggest that the CTD is responsible for forming those trimers of dimers based on SEC-MALS measurements. However, Figure 2D shows that while the full length runs at 6.6x the monomer, the Q-rich+CTD runs at 5.4x the monomer. First, I could not find SEC-MALS of the full-length protein, and it is not clear whether SEC-MALS was used for all or a fraction of the constructs discussed in Figure 2D. Second, could it be that the Q-rich domain+CTD is an ensemble of hexamers and dimers? Perhaps the IDR is playing a secondary role in stabilizing the hexamer?

    4. The analysis of the phase separation results needs to have some extra quantification. The authors show that at 100 uM protein with 10% PEG the full-length phase separates as well as IDR+Q-rich. Lines 176-178: "The CTD, on the other hand, has no effect on the Q-rich domain condensates; Q-rich+CTD condensates formed at the same protein concentration and with identical droplet morphologies at the Q-rich domain alone." It is hard to draw this conclusion solely based on the data presented in Figure 3. An alternative interpretation might be that Q-rich+CTD reduces csat. I suggest the authors include turbidity assays (as shown for pH effect) to quantitively determine csat for these different constructs and perhaps perform FRAP to determine the mobility of these different constructs. In addition, how long after the addition of PEG were these droplets imaged?

    5. Solubility assays shown in Figures 4A, B, D, and 5C are missing error bars. Without replicates, it is difficult to assess, for example, the effect of KCl. Also, please indicate the physiological ranges of KCl and pH in Figure 6. The phase separation sensitivity to pH is intriguing. By changing basic residues to glutamines, the authors conclude that the positive charge of the IDR modulates solubility. The Q-rich domain, however, is negatively charged. Can the authors comment on the role of acidic residues in the Q-rich domain? Are they required for phase separation? Also - based on your previous bioinformatics analysis, are the charges of the IDR and the Q-rich domains conserved across McdB homologs?

    6. In previous work, the authors showed a conserved RKR segment in the IDR is highly conserved and missing in S. elongatus PCC 7942 (MacCready et al., Mol Biol Evol. 2020). Given the current finding, it would be important to understand whether the RKR deletion carries functional implications for phase separation behavior.

    7. McdB proteins with 2Q left mutated vs. 2Q middle and 2Q right seem to result in condensates with different material properties (e.g., DIC pictures show different droplet morphologies for the different constructs). Is that the case? And if so, can you comment on that?

  10. eLife

    Reviewer #3 (Public Review):

    Through a series of rigorous in vitro studies, the authors determined McdB's domain architecture, its oligomerization domains, the regions required for phase separation, and how to fine-tune its phase separation activity. The SEC-MALS study provides clear evidence that the α-helical domains of McdB form a trimer-of-dimers hexamer. Through analysis of a small library of domain deletions by microscopy and SDS-PAGE gels of soluble and pellet fractions, the authors conclude that the Q-rich domain of McdB drives phase separation while the N-terminal IDR modulates solubility. A nicely executed study in Figure 4 demonstrated that McdB phase separation is highly sensitive to pH and is influenced by basic residues in the N terminal IDR. The study demonstrates that net charge, as opposed to specific residues, is critical for phase separation at 100 micromolar. In addition, the experimental design included analysis of McdB constructs that lack fluorescent proteins or organic dyes that may influence phase separation. Therefore, the observed material properties have full dependence on the McdB sequence.

    Studies of proteins often neglect short, disordered segments at the N- or C- terminus due to unclear models for their potential role. This study was interesting because it revealed a short IDR as a critical regulator of phase separation. This includes experiments that remove the IDR (Fig 2 & 3) and mutate the basic residues to show their importance towards McdB phase separation. In a nice set of SDS-PAGE experiments, the authors showed that as the net charge of the IDR decreased the construct became more soluble.

    One challenge is in the experimental design when mutating residues is to assess their impact on phase separation. The author's avoided substitutions to alanine, as alanine substitutions have synthetically stimulated phase separation in other systems. The authors, therefore, have a good rationale for selecting potentially milder mutations of lysine/arginine to glutamine. A potential caveat of mutation to glutamine is that stretches of glutamines have been associated with amyloid/prion formation. So, the introductions of glutamines into the IDR may also have unexpected effects on material properties. Despite these caveats, the authors show mutation of six basic residues in the short IDR abolished phase separation at 100 mM.

    Computational studies (Fig 7) also suggest that this short N-IDR region may play a role as a MORF upon potential binding to a second protein McdA. The formulation of this hypothesis is strengthened by the fact that for other ParA/MinD-family ATPases, the associated partner proteins have also been shown to interact with their cognate ATPase via positively charged and disordered N-termini. This aspect of understanding McdB's N-IDR as a MORF is at a very early stage. This study lacks experimental evidence for an N-IDR: McdA interaction and experimental data showing conformational change upon McdA binding. However, the computation study sets up the future to consider whether and how the phase separation activity of McdB is related to its structural dynamics and interactions with McdA.

    In summary, this study provides a strong foundation for the contribution of domains to McdB's in vitro phase separation. This knowledge will inform and impact future studies on McdB regulating carboxysomes and how the related family of ParA/MinD-family ATPases and their cognate regulatory proteins. For example, it is unknown if and how McdB's phase separation is utilized in vivo for carboxysome regulation. However, the revealed roles of the Q-rich domain and N-IDR will provide valuable knowledge in developing future research. In addition, the systematic domain analysis of McdB can be combined with a similar analysis of a broad range of other biomolecular condensates in bacteria and eukaryotes to understand the design principles of phase separating proteins.

  11. Version 2 published on bioRxiv
  12. Version 1 published on bioRxiv