A frameshift in Yersinia pestis rcsD alters canonical Rcs signalling to preserve flea-mammal plague transmission cycles

  1. Xiao-Peng Guo
  2. Hai-Qin Yan
  3. Wenhui Yang
  4. Zhe Yin
  5. Viveka Vadyvaloo
  6. Dongsheng Zhou
  7. Yi-Cheng Sun

  • Curated by eLife

    eLife logo

    eLife assessment

    This valuable study substantially advances our understanding of the phenotypic divergence of the plague-causing bacterium, Yersinia pestis, from a closely related species, Yersinia pseudotuberculosis, which causes much milder disease. The authors provide convincing evidence that a frameshift mutation in the Y. pestis rcsD gene changes a signaling pathway that contributes to the flea-mammal transmission of plague. The work shows how small genetic differences can alter pathogenicity and stress survival.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Multiple genetic changes in the enteric pathogen Yersinia pseudotuberculosis have driven the emergence of Yesinia pestis , the arthropod-borne, etiological agent of plague. These include developing the capacity for biofilm-dependent blockage of the flea foregut to enable transmission by flea bite. Previously, we showed that pseudogenization of rcsA , encoding a component of the Rcs signalling pathway, is an important evolutionary step facilitating Y. pestis flea-borne transmission. Additionally, rcsD, another important gene in the Rcs system, harbours a frameshift mutation. Here, we demonstrated that this rcsD mutation resulted in production of a small protein composing the C-terminal RcsD histidine-phosphotransferase domain (designated RcsD-Hpt) and full-length RcsD. Genetic analysis revealed that the rcsD frameshift mutation followed the emergence of rcsA pseudogenization. It further altered the canonical Rcs phosphorylation signal cascade, fine-tuning biofilm production to be conducive with retention of the pgm locus in modern lineages of Y. pestis . Taken together, our findings suggest that a frameshift mutation in rcsD is an important evolutionary step that fine-tuned biofilm production to ensure perpetuation of flea-mammal plague transmission cycles.

Article activity feed

  1. Version published to 10.7554/elife.83946 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    The paper addresses an interesting question - how genetic changes in Y. pestis have led to phenotypic divergence from Y. pseudotuberculosis - and provides strong evidence that the frameshift mutation in rcsD is involved. Overall, I found the data to be clearly presented, and most of the conclusions well supported by the data. The authors convincingly show that (i) the frameshift mutation in rcsD alters the regulation of biofilm formation, (ii) this effect depends upon expression of a small protein that corresponds to the C-terminal portion of RcsD, and (iii) the frameshift mutation in rcsD prevents loss of the pgm locus. I felt that the discussion/conclusions about what phosphorylates/dephosphorylates RcsB and how this impacts biofilm formation are overstated, as there are no experiments that directly address this question. I also felt that the authors' model for what phosphorylates/dephosphorylates RcsB in Y. pestis should be more clearly articulated, even if it is only presented as speculation. Lastly, the authors propose that full-length RcsD is made in Y. pestis and contributes to phosphorylation of RcsB, but the evidence for this is weak (faint band in Figure 2d). It may be that the N-terminal domain of RcsD is functional. I recommend either softening this conclusion or testing this hypothesis further, e.g., by introducing an in-frame stop codon early in rcsD after the frame-shift.

    Thanks for your comments. We have provided a model and revised the discussion about phosphorylation/dephosphorylation of RcsB and how this impacts biofilm formation (Figure 8 and Supplementary Figure 4). In addition, we have introduced an in-frame stop codon in rcsD before the frameshift and showed that full-length RcsD is only made in wildtype Y. pestis but not in the rcsDpe-stop mutant (Supplementary Figure 1g).

    Reviewer #2 (Public Review):

    Guo et al. have investigated the consequences of a frameshift mutation in the rcsD gene in the Yersinia pseudotuberculosis progenitor that is conserved in modern Y. pestis strains. Interestingly, they identify a start codon with a ribosome binding site that enables production of an Hpt-domain protein from the C-terminus in Y. pestis. Targeted deletion of this Hpt-domain increased biofilm production in Y. pestis. They find that the ancestral RcsDpstb (full length) is a positive regulator of biofilm in Y. pestis while the Hpt-domain version (RcsDYP) represses biofilm in vitro. When fleas were infected with Y. pestis expressing the ancestral RcsDPSTB protein, there was no difference in bacterial survival or rate of proventricular blockage. This strain also killed mice the same rate (in a different Y. pestis strain background). However, replacing RcsDYP with RcsYPTB dramatically increases the frequency of pgm locus deletion (containing Hms ECM and yersiniabactin genes) during flea infection. The authors predict that this would reduce the invasiveness of the bacteria in mammals and/or flea blockage in subsequent flea-rodent-flea transmission cycles. They also measured global gene expression differences between RcsDPSTB compared to the wild-type strain. They argue that the frameshift of RcsD maintaining the Hpt-domain (RcsDYP) was needed to regulate biofilm while limiting loss of the pgm locus.

    Loss of the pgm locus was not tested in the Y. pestis rcsD mutant strain (lacking the entire gene or just the C-terminal Hpt domain). Therefore, the claim that maintaining the Hpt-domain protein was important lacks convincing evidence. Additionally, it is possible that the population of rcsDpe::rcsDpstb after in vitro growth for 6 days would still be proficient at infecting and blocking fleas, even though many of the bacteria would have lost the pgm locus. Production of Hms polysaccharide by pgm+ could trans-complement those that are pgm-. The nature of the pgm locus loss is assumed to be due to recombination between IS elements. This is certainly the likeliest explanation but not the only one. The authors checked for pgm loss by phenotype (CR binding) and by two sets of primers, one targeting the hmsS gene and another set that is unspecified. Loss of the entire pgm (especially yersiniabactin genes) should be clarified.

    Thanks for your comments. We have now provided the data to show that deletion of RcsD-Hpt resulted in increased loss of the pgm locus (Figure 5d) to strengthen the claim that maintenance of the Hpt-domain is significant for retention of the pgm locus. We also agree that 6-day old cultures of a mixture of pgm+ and pgm- rcsDpe::rcsDpstb will still be capable of infecting and blocking fleas. However, these strains will be less efficient at causing disease in the vertebrate host in the absence of the pgm locus. We agree that recombination between IS elements might not be the only cause of loss of the pgm locus. To verify the loss of the pgm locus, we have used two sets of primers. One set targets the hmsS gene and another set targets the upstream and downstream sequences of the pgm locus (Supplementary Table 3). We have clarified this in the revised manuscript (Line 610-613).

    Reviewer #3 (Public Review):

    The Rcs phosphorelay plays an important role in regulating gene expression in bacteria; most of the current knowledge about the Rcs proteins is from E. coli. Yersinia pestis, carrying mutations in two central components of the Rcs machinery, provides an interesting example of how evolution has shaped this system to fit the life cycle of this bacteria. In bacteria other than Y. pestis, most Rcs activating signals are sensed via the outer membrane lipoprotein RcsF; from there, signalling depends on inner membrane protein IgaA, a negative regulator of RcsD. Histidine kinase RcsC is the source of the phosphorylation cascade that goes from the histidine kinase domain of RcsC to the response regulator domain of RcsC, from there to the histidine phosphotransfer (Hpt) domain of RcsD, and finally to the response regulator RcsB. RcsB, alone or with other proteins, regulates transcription of many genes, both positively and negatively. These authors have previously shown that RcsA, a co-regulator that acts with RcsB at some promoters, is functional in Y. pseudotuberculosis but mutant in Y. pestis, and that this leads to increased biofilm in the flea. The authors also noted that rcsD in Y. pestis contains a frameshift after codon 642 in this 897 aa protein; in theory that should eliminate the Hpt domain from the expressed protein. However, they found evidence that the frame-shifted gene had a role in regulation. This paper investigates this in more depth, providing clear evidence for expression of the Hpt domain (without the N-terminal domain), and demonstrating a critical role for this domain in repressing biofilm formation. The Y. pseudotuberculosis RcsD does not express a detectable amount of the Hpt domain nor does it repress biofilm formation. The ability of the Hpt domain protein to keep biofilm formation low explains most of what is observed for the full-length frame-shifted protein.

    1. The authors provide a substantial amount of data supporting the expression of the C-terminus of RcsD is sufficient and necessary for low biofilm levels, and that this is dependent upon the active site His in the RcsD Hpt domain (H844A) as well as other components of the basic phosphorelay (RcsC and RcsB). However, it is only possible to see this protein by Western blot in 100-fold "Enriched" lysates (Figure 2). No small protein was detected in the RcsDpstb strain, although the enriched lysate was not shown for this. Without that experiment, it is not possible to evaluate whether the small protein is also made from the rcsDpstb gene. Either answer would be interesting, and would allow other conclusions to be drawn. Is the RBS and start codon the same for the HPT region of this rcsD gene (it could be added to Supplementary Table 6). If the small protein is made, is its ability to function blocked by the excess full length protein in terms of interactions with RcsC? Or is the expression of the small protein dependent upon loss of overlapping translation from the upstream start?

    The small Hpt protein may be produced from expression of the epitope tagged rcsDpstb gene as it can be detected in an enriched isolation of this sample (Supplementary Figure 1f). Because only a small amount of the RcsD-Hpt is produced from the rcsDpstb substitution, it might only function at low levels in the presence of large amounts of RcsDpstb. The RBS and start codon are the same for the RcsD-Hpt in Y. pestis and Y. pseudotuberculosis, we have added them in the Supplementary Table 6. In addition, we have provided a model to show the function and regulation of RcsD and Hpt (Supplementary Figure 4).

    1. In many phosphorelays, the protein kinase also acts as a phosphatase, and which direction P flows is critical for regulation. It is often difficult to follow what the model for this is in this paper, and that is important to understand for evaluating the results. Most of this paper uses two assays, biofilm formation and crystal violet staining (also related to biofilm formation) to assess the functioning of the Rcs phosphorelay. Based on the behavior of the rcsB mutant, it would seem that functional Yersinia pestis Rcs (RcsDpe) represses this behavior, and this correlates with RcsB phosphorylation (Figure4). What is the basis (Line 443-44) for saying that RcsD phosphorylates RcsB while RcsDHpt dephosphorylates? Yersinia pseudotuberculosis RcsD(pstb) shows no difference with the rcsB mutant. Doesn't that suggest that RcsDpstb is no longer repressing (phosphorylating)? In the presence of the RcsDpstb as well as multicopy RcsF, an activating signal in other organisms, RcsDpstb seems able to phosphorylate. This all suggests that the full-length protein, like the Hpt domain, is capable of phosphorylating, but that it may be doing nothing in the absence of signal (or dephosphorylating). Given these results, saying that RcsDpstb is positively regulating biofilm formation (Fig.1 title, and elsewhere) is somewhat misleading. What it presumably does is prevent the Hpt domain, expressed from the chromosomal locus in Figure1b, from signalling to RcsB. By itself, it is not clear it is doing anything. Understanding this clearly is important for interpreting this system and the tested mutants. A clear model and how phosphate is flowing in the various situations would help a lot. Currently Supplementary Figure3 seems to reflect the appropriate directional arrows, but the text does not. Moving the rcsB data earlier in the paper (after Figure1, 2, or maybe earlier, before Figure3) would certainly help.

    RcsD dephosphorylates RcsB while RcsD-Hpt phosphorylates RcsB. Expression of RcsDpstb in the wild type strain and the N-term deletion mutant resulted in increased biofilm, indicating RcsB is less phosphorylated (Figure 1b and 1c). While over-expression of RcsD-Hpt resulted in decreased biofilm formation, indicating RcsB is more phosphorylated. In addition, the Phos-tag experiments showed that the RcsDpstb strain has a lower level of phosphorylated RcsB (Figure 4b). Expression of RcsDpstb in the wild type strain showed similar results as a rcsB mutant indicating a lower level of phosphorylated RcsB in the presence of RcsDpstb.

    It is possible that the RcsDpstb interferes with the ability for RcsD-Hpt to phosphorylate RcsB. However, plasmid expression of the rcsDpstb-H844A mutant in the Y. pestis rcsDN-term deletion mutant formed significantly less biofilm than wild type rcsDpstb indicating H844 might be important for RcsD to dephosphorylate RcsB (Supplementary Figure 2b and Line 180-183). In addition, it is known that RcsD plays a dual role in phosphorylation and dephosphorylation of RcsB in other organisms (Majdalani N, et al., 2005, J. Bacteriol. https://doi.org/10.1128/JB.187.19.6770-6778.2005; Wall EA, et al., 2020, Plos Genetics, https://doi.org/10.1371/journal.pgen.1008610; Takeda S., et al., 2001, Mol. Microbiol., https://doi: 10.1046/j.1365-2958.2001.02393.x). We therefore think it is safe to say that the full length RcsD might function to dephosphorylate RcsB. We have modified the model in the revised manuscript (Supplementary Figure 4 and Figure 8). Regulation of RcsB has been investigated previously. The main finding of our manuscript is regulation of RcsB by the mutated RcsD (RcsD-Hpt). Thus, we have moved the known rcsB deletion mutant data to Figure 1 in the revised manuscript as suggested. We kept the rest of data in Figure 4 the same. We think it might be better to first show the mutation of rcsD alters Rcs signaling and then show how this occurs (by affecting RcsB phosphorylation).

    1. The authors show (in their pull-down) that there is a bit of full-length RcsD even in the frame-shifted protein. Is there any clear evidence this does anything here? Does the N-terminus (truncated after the frame-shift) have a function?

    We have introduced a stop codon in rcsDpe and showed that full-length RcsD is made by rcsDpe but not by rcsDpe with the stop codon (Supplementary Figure 1g). RcsDN-term seems do not have a function in our tested condition (Figure 1e).

    1. While the RNA seq data is useful addition here, it is difficult to interpret without a bit more data on the strain used for the RNA seq, including the biofilm phenotypes of the WT and mutant derivatives, as well as the relevant rcsD sequences, and maybe expression of a few genes or proteins (Hms or hmsT). Are these similar in the parallel strains used earlier in the paper and the one for RNA seq, in WT, rcsB- and the RcsDpstb derivative? It would appear that rcsB- and rcsDpstb have opposite effects, at least at 25{degree sign}C, while in Figure4, these two derivatives have similar effects on biofilm. Is this due to temperature, strains, or biofilm genes that are not shown here? It is certainly possible that the ability of the full-length RcsD changes its kinase/phosphatase balance as a function of temperature, or dependent on other differences in these Y. pestis strains.

    The strain used for RNA seq is a derivative of the biovar Microtus strain 201 which has a similar in vitro phenotype as the strain KIM6+ (Line 297-298). We used this strain for RNA seq because it has the virulence plasmid pCD1 and we wanted to analyze the gene expression of this plasmid, which is required for virulence, as well. RNAseq data showed that rcsB- and rcsDpstb have opposite effects on mRNA level of some genes. However, no significant change in expression of biofilm genes was noted in the RNAseq data set. In fact, our previous data has shown that the biofilm related (hmsT and hmsD) genes are only moderately (Less than 2-fold change between wild type and rcsB mutant) regulated by RcsB based on RT-PCR and β-gal analysis (Sun YC, et al., 2012, J. Bacteriol. https:// doi: 10.1128/JB.06243-11and Guo XP, et al., 2015, Sci. Rep. https://doi: 10.1038/srep08412 and Figure 4c).

  3. eLife

    eLife assessment

    This valuable study substantially advances our understanding of the phenotypic divergence of the plague-causing bacterium, Yersinia pestis, from a closely related species, Yersinia pseudotuberculosis, which causes much milder disease. The authors provide convincing evidence that a frameshift mutation in the Y. pestis rcsD gene changes a signaling pathway that contributes to the flea-mammal transmission of plague. The work shows how small genetic differences can alter pathogenicity and stress survival.

  4. eLife

    Reviewer #1 (Public Review):

    The paper addresses an interesting question - how genetic changes in Y. pestis have led to phenotypic divergence from Y. pseudotuberculosis - and provides strong evidence that the frameshift mutation in rcsD is involved. Overall, I found the data to be clearly presented, and most of the conclusions well supported by the data. The authors convincingly show that (i) the frameshift mutation in rcsD alters the regulation of biofilm formation, (ii) this effect depends upon expression of a small protein that corresponds to the C-terminal portion of RcsD, and (iii) the frameshift mutation in rcsD prevents loss of the pgm locus. I felt that the discussion/conclusions about what phosphorylates/dephosphorylates RcsB and how this impacts biofilm formation are overstated, as there are no experiments that directly address this question. I also felt that the authors' model for what phosphorylates/dephosphorylates RcsB in Y. pestis should be more clearly articulated, even if it is only presented as speculation. Lastly, the authors propose that full-length RcsD is made in Y. pestis and contributes to phosphorylation of RcsB, but the evidence for this is weak (faint band in Figure 2d). It may be that the N-terminal domain of RcsD is functional. I recommend either softening this conclusion or testing this hypothesis further, e.g., by introducing an in-frame stop codon early in rcsD after the frame-shift.

  5. eLife

    Reviewer #2 (Public Review):

    Guo et al. have investigated the consequences of a frameshift mutation in the rcsD gene in the Yersinia pseudotuberculosis progenitor that is conserved in modern Y. pestis strains. Interestingly, they identify a start codon with a ribosome binding site that enables production of an Hpt-domain protein from the C-terminus in Y. pestis. Targeted deletion of this Hpt-domain increased biofilm production in Y. pestis. They find that the ancestral RcsDpstb (full length) is a positive regulator of biofilm in Y. pestis while the Hpt-domain version (RcsDYP) represses biofilm in vitro. When fleas were infected with Y. pestis expressing the ancestral RcsDPSTB protein, there was no difference in bacterial survival or rate of proventricular blockage. This strain also killed mice the same rate (in a different Y. pestis strain background). However, replacing RcsDYP with RcsYPTB dramatically increases the frequency of pgm locus deletion (containing Hms ECM and yersiniabactin genes) during flea infection. The authors predict that this would reduce the invasiveness of the bacteria in mammals and/or flea blockage in subsequent flea-rodent-flea transmission cycles. They also measured global gene expression differences between RcsDPSTB compared to the wild-type strain. They argue that the frameshift of RcsD maintaining the Hpt-domain (RcsDYP) was needed to regulate biofilm while limiting loss of the pgm locus.

    Loss of the pgm locus was not tested in the Y. pestis rcsD mutant strain (lacking the entire gene or just the C-terminal Hpt domain). Therefore, the claim that maintaining the Hpt-domain protein was important lacks convincing evidence. Additionally, it is possible that the population of rcsDpe::rcsDpstb after in vitro growth for 6 days would still be proficient at infecting and blocking fleas, even though many of the bacteria would have lost the pgm locus. Production of Hms polysaccharide by pgm+ could trans-complement those that are pgm-. The nature of the pgm locus loss is assumed to be due to recombination between IS elements. This is certainly the likeliest explanation but not the only one. The authors checked for pgm loss by phenotype (CR binding) and by two sets of primers, one targeting the hmsS gene and another set that is unspecified. Loss of the entire pgm (especially yersiniabactin genes) should be clarified.

  6. eLife

    Reviewer #3 (Public Review):

    The Rcs phosphorelay plays an important role in regulating gene expression in bacteria; most of the current knowledge about the Rcs proteins is from E. coli. Yersinia pestis, carrying mutations in two central components of the Rcs machinery, provides an interesting example of how evolution has shaped this system to fit the life cycle of this bacteria. In bacteria other than Y. pestis, most Rcs activating signals are sensed via the outer membrane lipoprotein RcsF; from there, signalling depends on inner membrane protein IgaA, a negative regulator of RcsD. Histidine kinase RcsC is the source of the phosphorylation cascade that goes from the histidine kinase domain of RcsC to the response regulator domain of RcsC, from there to the histidine phosphotransfer (Hpt) domain of RcsD, and finally to the response regulator RcsB. RcsB, alone or with other proteins, regulates transcription of many genes, both positively and negatively. These authors have previously shown that RcsA, a co-regulator that acts with RcsB at some promoters, is functional in Y. pseudotuberculosis but mutant in Y. pestis, and that this leads to increased biofilm in the flea. The authors also noted that rcsD in Y. pestis contains a frameshift after codon 642 in this 897 aa protein; in theory that should eliminate the Hpt domain from the expressed protein. However, they found evidence that the frame-shifted gene had a role in regulation. This paper investigates this in more depth, providing clear evidence for expression of the Hpt domain (without the N-terminal domain), and demonstrating a critical role for this domain in repressing biofilm formation. The Y. pseudotuberculosis RcsD does not express a detectable amount of the Hpt domain nor does it repress biofilm formation. The ability of the Hpt domain protein to keep biofilm formation low explains most of what is observed for the full-length frame-shifted protein.

    1. The authors provide a substantial amount of data supporting the expression of the C-terminus of RcsD is sufficient and necessary for low biofilm levels, and that this is dependent upon the active site His in the RcsD Hpt domain (H844A) as well as other components of the basic phosphorelay (RcsC and RcsB). However, it is only possible to see this protein by Western blot in 100-fold "Enriched" lysates (Figure 2). No small protein was detected in the RcsDpstb strain, although the enriched lysate was not shown for this. Without that experiment, it is not possible to evaluate whether the small protein is also made from the rcsDpstb gene. Either answer would be interesting, and would allow other conclusions to be drawn. Is the RBS and start codon the same for the HPT region of this rcsD gene (it could be added to Supplementary Table 6). If the small protein is made, is its ability to function blocked by the excess full length protein in terms of interactions with RcsC? Or is the expression of the small protein dependent upon loss of overlapping translation from the upstream start?
    2. In many phosphorelays, the protein kinase also acts as a phosphatase, and which direction P flows is critical for regulation. It is often difficult to follow what the model for this is in this paper, and that is important to understand for evaluating the results. Most of this paper uses two assays, biofilm formation and crystal violet staining (also related to biofilm formation) to assess the functioning of the Rcs phosphorelay. Based on the behavior of the rcsB mutant, it would seem that functional Yersinia pestis Rcs (RcsDpe) represses this behavior, and this correlates with RcsB phosphorylation (Fig. 4). What is the basis (line 443-44) for saying that RcsD phosphorylates RcsB while RcsDHpt dephosphorylates? Yersinia pseudotuberculosis RcsD(pstb) shows no difference with the rcsB mutant. Doesn't that suggest that RcsDpstb is no longer repressing (phosphorylating)? In the presence of the RcsDpstb as well as multicopy RcsF, an activating signal in other organisms, RcsDpstb seems able to phosphorylate. This all suggests that the full-length protein, like the Hpt domain, is capable of phosphorylating, but that it may be doing nothing in the absence of signal (or dephosphorylating). Given these results, saying that RcsDpstb is positively regulating biofilm formation (Fig.1 title, and elsewhere) is somewhat misleading. What it presumably does is prevent the Hpt domain, expressed from the chromosomal locus in Fig. 1b, from signalling to RcsB. By itself, it is not clear it is doing anything. Understanding this clearly is important for interpreting this system and the tested mutants. A clear model and how phosphate is flowing in the various situations would help a lot. Currently Supplementary Fig. 3 seems to reflect the appropriate directional arrows, but the text does not. Moving the rcsB data earlier in the paper (after Fig. 1, 2, or maybe earlier, before Fig. 3) would certainly help.
    3. The authors show (in their pull-down) that there is a bit of full-length RcsD even in the frame-shifted protein. Is there any clear evidence this does anything here? Does the N-terminus (truncated after the frame-shift) have a function?
    4. While the RNA seq data is useful addition here, it is difficult to interpret without a bit more data on the strain used for the RNA seq, including the biofilm phenotypes of the WT and mutant derivatives, as well as the relevant rcsD sequences, and maybe expression of a few genes or proteins (Hms or hmsT). Are these similar in the parallel strains used earlier in the paper and the one for RNA seq, in WT, rcsB- and the RcsDpstb derivative? It would appear that rcsB- and rcsDpstb have opposite effects, at least at 25{degree sign}C, while in Fig. 4, these two derivatives have similar effects on biofilm. Is this due to temperature, strains, or biofilm genes that are not shown here? It is certainly possible that the ability of the full-length RcsD changes its kinase/phosphatase balance as a function of temperature, or dependent on other differences in these Y. pestis strains.

  7. Version published to 10.1101/2022.10.24.512971v1 on bioRxiv