DrpB is a novel RcsF-independent Rcs activator

With the aim of identifying novel proteins capable of triggering the Rcs cascade in an RcsF-independent manner, we utilized a fluorescence-based genetic screen in E. coli. We transformed a plasmid library carrying 3–5 kb fragments of the E. coli chromosome into strain EAW34, carrying a reporter for Rcs activation and deleted for rcsF. The reporter used is a transcriptional fusion of the rprA promoter, an RcsB target, to the mCherry protein, referred to here as P_rprA::mCherry [18,34]. This strain normally shows low fluorescence in the absence of RcsF, consistent with a central role for RcsF in Rcs activation. The colonies were screened for increased fluorescence and 14 plasmids containing putative RcsF-independent Rcs activators were isolated. Sequencing of the plasmids revealed that 13 of them contained overlapping fragments of the E. coli chromosome that included genes encoding the YedR (DrpB) ORF along with the RseX sRNA (S1A Fig). The remaining plasmid contained genes encoding the YihA protein and the sRNAs Spot 42 and CsrC (S1B Fig). We previously reported that overexpression of Spot42 induces the Rcs pathway and while that study suggested that much of the Spot 42 effect was RcsF dependent, only a qualitative assay was done [34]. Spot 42, an Hfq-dependent regulatory RNA, regulates translation and stability of multiple mRNAs and has pleiotropic effects in the cell [35,36], making identification of a single critical target complicated.

To further validate identification of DrpB as capable of RcsF-independent induction, we assayed the growth and fluorescence of wildtype, ΔrcsF and ΔrcsB strains expressing DrpB or the known RcsF-independent activator DjlA under control of the arabinose-inducible pBAD promoter, using the same P_rprA::mCherry fusion as a reporter for Rcs activation. Overexpression of both DrpB and DjlA showed increased fluorescence, both in the presence and absence of RcsF, confirming their role as multicopy RcsF-independent Rcs activators (Fig 1B). The increased fluorescence was fully dependent on RcsB for both genes. We did not observe a significant RcsF-independent increase in the reporter expression after overexpression of YihA or rseX, and the modest induction by Spot42 in ΔrcsF was also not significant (S1C and S1D Fig). Therefore, we focused our investigation on DrpB. DrpB is a small inner membrane protein and was previously found to activate both the Rcs and the Psp stress response upon overexpression, suggesting that its overexpression leads to membrane stress [28]. We first compare DrpB to other RcsF-independent activators before examining it in more detail.

We also confirmed that the deletion of dsbA activates the Rcs pathway in an RcsF-independent manner, by measuring the fluorescence of the same P_rprA::mCherry reporter in ΔdsbA, ΔdsbA ΔrcsF, and ΔdsbA ΔrcsB strains (Fig 1C). Deletion of dsbA increased the basal P_rprA::mCherry activity independently of RcsF but not RcsB. Strains devoid of DsbA are defective in disulfide bond formation in periplasmic proteins and treatment with the reducing agent DTT, by further interfering with disulfide bond formation, exacerbates this problem [37]. Addition of DTT further induced the Rcs pathway, but only in the absence of DsbA. Unexpectedly, we found that PMBN, which normally induces the pathway in an RcsF-dependent-fashion (see WT and ΔrcsF results in Fig 1C), also induced this pathway, to a similar extent as DTT, in both ΔdsbA and ΔdsbA ΔrcsF strains (Fig 1C). Together, these results provide evidence for three RcsF-independent Rcs activation situations–lack of DsbA and overproduction of DjlA or DrpB.

We next investigated whether these inducers of Rcs depend on each other. We first measured Rcs activation in dsbA mutants in cells also deleted for the chromosomal copies of djlA or drpB (S2A Fig). Activity in the ΔdsbA ΔdjlA and ΔdsbA ΔdrpB strains was essentially unchanged from that seen in the ΔdsbA strain (see Fig 1C). Therefore, the absence of DsbA does not act by mimicking overexpression of DjlA or DrpB. In addition, overexpression of DjlA or DrpB induces activity in a dsbA deletion strain (S2B Fig), although the overexpression of DrpB led to poor growth in the absence of DsbA (see OD values, S2B Fig). We also tested if the multicopy activators DjlA and DrpB require each other to trigger Rcs signaling and found that they did not (S2C Fig). Together, these results indicate that DsbA, DjlA, and DrpB do not require one another for activity and possibly employ different mechanisms of Rcs activation. We next investigated the pathways they each use for RcsF-independent activation of Rcs.

DrpB has a unique requirement for the RcsC periplasmic domain for activation

The Rcs pathway can be activated simply by increasing the levels of RcsB [25]. If this were how the RcsF-independent factors were working, they might be independent of the histidine kinase, RcsC and the phosphotransfer protein RcsD, as observed for the effect of mutations in barA [26]. Previous work had demonstrated that multicopy DjlA and DrpB are still dependent on RcsC for signaling [23,24,28]. As seen in Fig 2, while deletion of rcsC caused an increase in the basal level of P_rprA::mCherry expression, there was no further induction in the absence of RcsC for any of these factors. Deletion of rcsC or of rcsD leads to higher basal levels of expression of the reporter, interpreted as reflecting loss of phosphatase activity combined with some phosphorylation of RcsB from other sources [18,34].

We previously showed that the RcsC periplasmic domain is not needed for the RcsF-dependent induction by PMBN [18]. We assayed Rcs activation upon deletion of dsbA or overexpression of DjlA or DrpB in strains encoding a chromosomal version of RcsC missing its periplasmic domain (rcsCΔperi). Fig 2A shows that PMBN was able to induce signaling in the rcsC Δperi strain as previously found. Deletion of dsbA in the rcsCΔperi strain increased the basal level of signaling, although not to the extent seen for the rcsC⁺ strain (ΔdsbA); DTT or PMBN further increased signaling (Fig 2A). These results demonstrate that the primary effect of ΔdsbA is independent of the RcsC periplasmic region. It seems likely that the loss of dsbA has additional effects in the periplasm, and this may contribute to the muted response in signal in the rcsCΔperi strain (Fig 2A). It is also likely that the RcsCΔperi protein is somewhat perturbed in activity, for instance somewhat reducing the balance of kinase to phosphatase, leading to less activity of the reporter. DjlA overexpression was also able to activate signaling in the rcsCΔperi strain, although at a somewhat decreased level (Fig 2B). However, upon DrpB overexpression, no Rcs activation was observed in the rcsCΔperi strain (Fig 2B). This is the first evidence of a role for the RcsC periplasmic domain in activation of the phosphorelay and provides clear evidence that DrpB acts in a manner that is distinct from DjlA overexpression or loss of DsbA.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Requirement of RcsC periplasmic domain for RcsF-independent signaling by DrpB.

A. dsbA mutants do not need the RcsC periplasmic domain for signaling: For the P_rprA::mCherry assay, the cells were grown in MOPS minimal glucose medium at 37°C. The RFU at OD 0.4 as compared to WT uninduced, set to 1, is depicted here. The cells were treated with either 1mM DTT or 20μg/ml PMBN. B. DrpB, but not DjlA, requires the RcsC periplasmic domain for Rcs activation: For the P_rprA::mCherry assay, the strains overexpressing DjlA (pBAD-DjlA/pPSG961) or DrpB (pBAD-DrpB/pDSW1977) were grown in MOPS minimal glycerol medium containing chloramphenicol (25 μg/ml) and either 0.2% glucose (-Ara) or 0.02% arabinose at 37°C. The RFU at OD 0.4 compared to the uninduced vector control is plotted. The strains used were: WT (EAW8), dsbA::kan (EAW62), rcsC::tet (EAW18), rcsC::tet dsbA::kan (EAW63), rcsCΔperi (EAW70), and dsbA::kan rcsCΔperi (EAW74). Values (mean ± SD) were statistically analyzed using multiple unpaired t-tests. Statistical significance is shown as: ns (P > 0.05; non- significant), ** (P ≤ 0.01), and **** (P ≤ 0.0001).

https://doi.org/10.1371/journal.pgen.1011408.g002

DjlA, but not DsbA and DrpB, can induce Rcs signaling in a RcsD T411A mutant

As for deletion of rcsC (Fig 2), all three factors were unable to induce Rcs activity in the absence of the phosphorelay protein RcsD (Fig 3). RcsD T411A is an allele of RcsD expressing a mutant in the cytoplasmic domain of RcsD that has a higher affinity for IgaA, blocking RcsF-dependent PMBN signaling [18] (Fig 3A). Strains carrying rcsD T411A were used to further explore the role of IgaA-RcsD interactions in RcsF-independent activation. The increase in the reporter in the absence of DsbA signaling was blocked by the rcsD T411A mutation; the basal level was reduced to that seen in rcsD T411A alone, and the DTT or PMBN increase was blocked (Fig 3A). Overexpression of DrpB also could not induce reporter activity in the rcsD T411A strain (Fig 3B). However, DjlA overexpression was still able to activate reporter expression in this strain, indicating its unique ability to overcome the effect of the rcsD T411A mutation.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Effect of RcsD T411A mutation on RcsF-independent signaling.

A. RcsD T411A mutation blocks induction by dsbA: For the P_rprA::mCherry assay, the cells were grown in MOPS minimal glucose medium at 37°C. The RFU at OD 0.4 as compared to the uninduced WT control, set to 1, is depicted here. The cells were treated with either 1mM DTT or 20μg/ml PMBN. Details of the assay are described in Materials and Methods. B. DjlA, but not DrpB, can overcome the RcsD T411A mutation for Rcs activation: For the P_rprA::mCherry assay, the strains overexpressing DjlA (pDjlA/pPSG961) or DrpB (pDrpB/pDSW1977) were grown in MOPS minimal glycerol medium containing chloramphenicol (25 μg/ml) and either 0.2% glucose or 0.02% arabinose at 37°C. The RFU at OD 0.4 compared to the vector uninduced control, set to 1, is plotted. The strains used were: WT (EAW8), dsbA::kan (EAW62), ΔrcsD (EAW19), ΔdsbA rcsD541::kan (AP13), rcsD T411A (EAW121), and rcsD T411A dsbA::kan (AP14). Statistical significance was calculated using multiple unpaired t-tests and is shown as follows: ns (P > 0.05; non- significant), * (P < 0.05), ** (P ≤ 0.01), and **** (P ≤ 0.0001).

https://doi.org/10.1371/journal.pgen.1011408.g003

We considered the possibility that DjlA could function in the rcsD T411A strain because it was the strongest inducer of the conditions tested. As seen in Fig 1B and 1C, DjlA overexpression gave around 30-fold induction, which is higher than DrpB overexpression (~17 fold), DsbA deletion (~13 fold), and the RcsF-dependent response to PMBN (~4 fold). As a test of this model, we examined the ability of RcsD T411A to block activity when RcsF is highly active. Overproduction of RcsF or expression of a derivative (RcsF^IM, carrying mutations S17D and M18Q) that localizes the protein to the inner member and constitutively activates Rcs signaling [14,38] both induce Rcs activity, with the inner-membrane localized mutant RcsF derivative, RcsF^IM, being the stronger activator, inducing better than 25-fold (S3 Fig). However, the T411A mutation was still able to block this induction (S3 Fig). These results suggest that DjlA is unique in its ability to fully overcome the rcsD T411A allele.

Thus, each of the three RcsF-independent factors has distinct patterns of induction, suggesting that they act in different manners. While all three are dependent upon RcsC and RcsD, only DrpB depends on the RcsC periplasmic domain and only DjlA can bypass the rcsD T411A allele to induce Rcs activity. Below we further examine how each of these factors works.

DsbA is needed for efficient IgaA-RcsD periplasmic interactions

DsbA resides in the periplasm and plays a central role in disulfide bond (DSB) formation in E. coli and other bacteria [reviewed in [39]]. DsbA, in concert with DsbB, introduces disulfide bonds into newly synthesized periplasmic proteins, predominantly between consecutive cysteine residues [40–42]. dsbA null mutants show a pleotropic phenotype consistent with defects in the cell envelope—they are mucoid, non-motile, sensitive to DTT as well as some drugs, show decreased fitness and attenuated virulence [37,39,40,43]. Periplasmic and membrane proteins that require disulfide bonds frequently require DsbA for proper folding. One such protein is the Rcs sensor protein, the outer membrane lipoprotein RcsF [44]. RcsF has two non-consecutive disulfide bridges and impaired disulfide bond formation leads to its degradation [5,7]. Thus, RcsF is non-functional in a dsbA mutant, supporting the idea that mucoidy in the dsbA mutant strain, likely reflecting Rcs induction, cannot be due to activation via RcsF; this is confirmed by the induction of the P_rprA::mCherry reporter in the absence of RcsF in Fig 1C. This is further validated by the absence of Rcs induction in response to Mecillinam or the MreB inhibitor A22 by the dsbA mutants, both dependent on RcsF (S4A Fig). Note that this result also distinguishes PMBN from these other inducers as uniquely still able to induce the Rcs phosphorelay in the dsbA mutant (Figs 2A and S4B; compare differences in fold induction).

Aside from RcsF, three other proteins in the phosphorelay (RcsC, RcsD, and IgaA) have a periplasmic domain. However, RcsD does not have any conserved cysteine residues within its periplasmic domain, ruling it out as a direct target of DsbA.

RcsC has two conserved cysteine residues in its periplasmic domain, but since the ΔdsbA rcsCΔperi strain is still DTT-inducible (Fig 2A), this cannot be the critical target for DsbA. However, our results do suggest that, directly or indirectly, the RcsC periplasmic domain contributes to phosphorelay activation in the absence of DsbA. While deletion of the RcsC periplasmic domain had only a modest effect on P_rprA::mCherry basal level or induction in a dsbA⁺ strain (compare WT to rcsCΔperi data, Fig 2A), levels of expression in the ΔdsbA mutants were significantly lower for ΔdsbA rcsCΔperi than for ΔdsbA rcsC⁺ (Fig 2A), with ΔdsbA 12-fold higher than WT in the absence of induction, while ΔdsbA rcsCΔperi was only 3.2-fold increased compared to WT or rcsCΔperi (Fig 2A). To determine if this reduction in induction was dependent upon RcsC disulfide bonds, we examined the behavior of a strain expressing RcsC mutant for the C111 and C154 periplasmic cysteine residues (rcsC C111A C154A, named here as rcsC C2A) in dsbA⁺ or ΔdsbA strains. As seen in S4B Fig, the rcsC C2A dsbA⁺ strain has the same pattern of expression as the WT strain or a rcsCΔperi strain (Fig 2A), responding to PMBN but not DTT. The ΔdsbA rcsC C2A strain, unlike the rcsCΔperi strain, responds to both DTT and PMBN and this response is not muted as in rcsCΔperi (compare to Fig 2A). This suggests the effect of the rcsCΔperi does not reflect a role of the RcsC periplasmic cysteine residues and is likely more indirect; this was not further explored here.

IgaA has four conserved periplasmic cysteine residues, suggesting that there are two disulfide linkages in its periplasmic domain. A study in Salmonella supported the idea that at least the first of the disulfide bonds was important for proper IgaA function [45]. Given that IgaA is essential, the dsbA mutant cannot lead to fully non-functional IgaA. However, if the disulfide bonds are not properly formed, the periplasmic interaction between IgaA and RcsD could be affected in the absence of DsbA, leading to the observed Rcs activation. We have previously shown that IgaA and RcsD interact in a bacterial two hybrid (BACTH) assay, and that the interaction detected by this assay is primarily dependent upon a periplasmic contact between the proteins [18]. This assay is based on the reconstitution of the Bordetella pertussis adenylate cyclase from its two fragments, T18 and T25, in an adenylate cyclase mutant (cya) E. coli strain BTH101 [46,47]. Fusion of each of these fragments to an interacting protein reconstitutes adenylate cyclase, assayed by the Cya-dependent synthesis of beta-galactosidase. Both orientations, IgaA-T18/RcsD-T25 as well as IgaA-T25/RcsD-T18, have been demonstrated to show a strong interaction ([18] and Fig 4A, WT strain).

We introduced the dsbA null mutant allele into the host for the BACTH assay and studied its effect on the IgaA-RcsD interaction, compared to the WT strain (Fig 4A). The interaction between IgaA and RcsD was significantly impaired in the dsbA mutant, suggesting that DsbA is needed to maintain a proper IgaA-RcsD interaction, possibly due to its role in IgaA folding. Given the localization of DsbA in the periplasm, we would expect that it is the periplasmic interaction of IgaA with RcsD that should be defective in the absence of DsbA. This was confirmed first by testing the interaction of RcsD with various domain deletion constructs of IgaA. IgaA has five transmembrane helices anchoring two cytoplasmic domains and a large periplasmic domain (schematic in Fig 4B). The IgaAΔperi (Δ384–649) protein has been shown to have a weak interaction with RcsD, confirmed here; loss of DsbA did not further decrease the interaction (Fig 4B). Deletion of the two cytoplasmic domains (IgaAΔcyt), on the other hand, did not decrease the interaction, but was very sensitive to loss of DsbA (Fig 4B). Neither of these deletions significantly affected the levels of expression of the IgaA-T18 protein, in the presence of absence of DsbA (S4D Fig). This data is best explained by a change in the IgaA periplasmic domain in the absence of DsbA, reducing the periplasmic interaction with RcsD.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. DsbA is needed for efficient IgaA-RcsD periplasmic interactions.

A. IgaA-RcsD interactions are weakened in ΔdsbA. Beta-galactosidase activity was measured in a cyaA mutant strain (BTH101 or AP 58 (BTH101 ΔdsbA)) expressing two plasmids encoding the T18 and T25 domains of adenylate cyclase fused to the proteins of interest and the expression measured compared to vector background. The IgaA/RcsD protein fusion plasmids paired with their cognate vector had very low activity; the three controls were averaged and used as “background” for normalization. The IgaA-RcsD interaction was normalized to 1 and other interactions were plotted relative to this interaction in WT. In both Fig 4A and 4B, the interaction of IgaA-RcsD was 577 units, while the vector control was 29; these units are 1000x the slope of OD420 (see Materials and Methods). This data is compiled from separate sets of assays, each normalized relative to the IgaA/RcsD signal in that experiment. The P values are relative to the dsbA⁺ values. B. Periplasmic interactions are affected in ΔdsbA and a T411A mutation strengthens this IgaA interaction. IgaA and RcsD/T411A were fused to the T18 and T25 domains, respectively. The IgaA-RcsD interaction in WT was normalized to 1 and all other interactions are plotted relative to this interaction. The RcsD T411A mutation tightens the IgaA-RcsD interaction in WT as well as ΔdsbA. The plasmids used were pEAW1 (IgaA-T18), pEAW8 (RcsD-T25), pEAW2 (IgaA-T25), pEAW7 (RcsD-T18), pAP101 (IgaA Δcyt), pEAW1peri (IgaA Δperi), and pEAW8T (RcsD T411A-T25). Values (mean ± SD) from independent experiments were statistically analyzed using multiple unpaired t-tests. Statistical significance is shown as follows: ns (P > 0.05; non- significant), *** (P ≤ 0.001), and **** (P ≤ 0.0001).

https://doi.org/10.1371/journal.pgen.1011408.g004

A crystal structure of the E. coli IgaA periplasmic domain shows two disulfide linkages, C404-C425 and C498-C504, consistent with previous work [4,45]. We mutated the Cysteine residues to Serine and tested their effect on the IgaA-RcsD interaction, either in pairs (C404S C425S or C498S C504S) or in a mutant with all four mutated (C4S) (S4C Fig). Breaking either linkage significantly weakened the interaction of IgaA with RcsD. As with the data for ΔdsbA in Fig 4B, the RcsD T411A allele overcame much (but not all) of the loss of interaction in the IgaAC4S derivative (S4C Fig). Therefore, for this interaction assay, C4S mimics the effect of loss of dsbA, supporting the idea that the disulfide linkages are required for proper folding of the periplasmic domain of IgaA and its periplasmic interaction with RcsD. Mutations in IgaA may lead to its degradation during the stationary phase [45] but the overexpressed IgaA C4S mutant was stable in the T18 fusion, suggesting DsbA primarily affects its folding (S4D Fig).

The prediction from these interaction assays is that mutations of the IgaA cys residues should also lead to increased activation of the Rcs phosphorelay, as seen for ΔdsbA in the presence of DTT. The chromosomal copy of igaA was replaced with igaAC4S in an rcsD mutant strain carrying the P_rprA::mCherry reporter. While this strain (AP168) is not fluorescent and is non-mucoid, due to lack of RcsD, when a plasmid carrying RcsD was introduced into the cells, the transformed colonies became mucoid, consistent with activation of the Rcs phosphorelay. Growth was poor and it seems likely that this C4S strain, like other strains that express the Rcs phosphorelay at high levels, rapidly picks up suppressing mutations. Therefore, we avoided purification or other manipulation of the transformed colonies. The transformed colonies were grown for assays without further purification in LB and the P_rprA::mCherry signal assayed; the expression of the reporter was higher than that seen with the ΔdsbA strain and was not affected by DTT (S4E Fig). Thus, blocking disulfide bond formation in IgaA via mutation of the cysteine residues induces Rcs expression and mimics the phenotype of the ΔdsbA strain in the presence of DTT. Because either pair of cysteine mutants abrogates the full interaction with RcsD (S4C Fig), it seems likely both disulfide bonds are important for proper IgaA folding.

DjlA can act as a cochaperone and mediate IgaA-RcsD cytoplasmic interactions

DjlA is the third member of the DnaJ domain family in E. coli. Proteins with this domain, including DnaJ and CbpA in E. coli, collaborate as part of the widely conserved DnaK/Hsp70 chaperone network. DjlA is localized at the cytoplasmic membrane with a short N-terminal region in the periplasm followed by a single transmembrane helix [48]. The conserved J-domain, needed for DnaK interaction, is in the cytoplasm at the C-terminal. A TerB-like (tellurite resistance) domain of unknown function lies between the J-domain and the membrane helix. The role played by this domain has not been clearly defined but it has been speculated to be involved in membrane/lipid-interaction, stress-sensing, DNA processing, and phage defense [49,50]. Although the precise cellular function and substrates of DjlA are unknown, it is a bonafide DnaK cochaperone capable of assisting DnaK in protein refolding [51]. DjlA is not essential for growth and a null mutant did not reveal any significant phenotype except a delayed onset of mucoidy at low temperature [24]. On the other hand, overproduction of DjlA was clearly toxic, with the cells exhibiting growth defects suggestive of a role of DjlA in cell division or membrane integrity [48]. Furthermore, strains overexpressing DjlA demonstrated increased capsule synthesis, the result of Rcs activation, and hypersensitivity to certain drugs and chelators [52]. The J-domain is not strictly required for the increased drug sensitivity; however, it is essential for its role as a cochaperone and Rcs activator [24]. DjlA-dependent activation of the Rcs phosphorelay requires DnaK and the nucleotide release factor GrpE, but not DnaJ or CbpA [24]. A functional J-domain and the transmembrane helix were both reported to be essential for Rcs induction by DjlA [24,53,54].

As a membrane-anchored cochaperone, it is possible that DjlA directs DnaK to modulate the folding and/or conformation of an Rcs component or the interaction of Rcs components. This is further supported by our observation that DjlA, uniquely, can overcome the tight interaction between IgaA and RcsD T411A (Fig 3B). Therefore, we tested the effect of DjlA overproduction on the IgaA-RcsD interaction in BACTH assays. We examined the IgaA-RcsD interaction when djlA was cloned downstream of IgaA-T18, such that its expression is controlled by the same IPTG-inducible promoter as IgaA. Fig 5A shows that there is a marked decrease in the IgaA-RcsD interactions when DjlA is overexpressed. To confirm that this effect is due to the cochaperone activity of DjlA, we also tested an inactive DjlA mutant, a H233Q mutation known to be defective for DnaK cochaperone activity [24,51]. The overexpression of this DjlA H233Q mutant did not disrupt the IgaA-RcsD interaction. Next, we asked if DjlA can weaken the tight interaction between IgaA and the RcsD T411A mutant (Fig 5B). DjlA overcame this strong interaction and could act even when the periplasmic domain of RcsD was not present (Fig 5B, T411A Δperi). This contrasts with loss of DsbA, which affected the periplasmic but not the cytoplasmic interactions between IgaA and RcsD (Fig 4), and is consistent with the ability of overproduced DjlA to cause Rcs induction even when RcsDT411A is present (Fig 3B). We also tested if any of the other Rcs components are required for DjlA to affect these interactions by carrying out the same assay in derivatives of the BTH101 strain carrying ΔrcsB (EAW1), ΔrcsC (EAW2), or ΔrcsF (EAW4) mutants. DjlA was able to weaken the interaction between IgaA and RcsD in the absence of each of these proteins (Fig 5C). This suggests that DjlA acts directly as a cochaperone, remodeling the RcsD-IgaA interaction with one or both proteins as direct targets for DjlA.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. DjlA weakens IgaA-RcsD interactions.

A. DjlA loosens IgaA-RcsD interactions: Beta-galactosidase activity was measured in a cyaA mutant strain (BTH101) in the presence of the indicated plasmids. IgaA and RcsD/RcsD T411A were fused to the T18 and T25 domains, respectively. DjlA or the H233Q mutant of DjlA, inactive as a co-chaperone, was cloned downstream of IgaA under the same promoter control. The IgaA-RcsD interaction was normalized to 1 and all other interactions are plotted relative to this interaction. In both Fig 5A and B, the interaction of IgaA-RcsD was 545 units, while the vector control was 18; these units are 1000x the slope of OD420 (see Materials and Methods). B. DjlA can disrupt cytoplasmic interactions of IgaA and RcsD: The IgaA-RcsD interaction was normalized to 1 and all other interactions are plotted relative to this interaction. C. RcsB, RcsC, RcsF are not needed for DjlA to weaken interactions. The IgaA-RcsD interaction was tested in BTH101 (WT), EAW 1 (BTH 101 rcsB::tet), EAW2 (BTH 101 rcsC::tet) and EAW4 (BTH 101 rcsF::cat). Plasmids used in this set of experiments were pEAW1 (IgaA-T18), pEAW8 (RcsD-T25), pEAW8T (RcsD T411A-T25), pAP804 (RcsD T411A Δperi -T25), pAP1401 (IgaA-T18 + DjlA), and pAP1402 (IgaA-T18 + DjlA H233Q). The WT interaction here was 533 units, with the vector control as 22 units. Statistical significance was calculated using multiple unpaired t-tests and is shown as follows: ** (P ≤ 0.01), *** (P ≤ 0.001), and **** (P ≤ 0.0001).

https://doi.org/10.1371/journal.pgen.1011408.g005

Previous studies speculated that RcsC is the target of DjlA, likely via transmembrane domain interactions, although that work was done before RcsD and IgaA were known or characterized [53]. As shown above, the periplasmic domain of RcsC was not needed for DjlA induction, but the role of the RcsC transmembrane helices had not been tested. We had previously demonstrated Rcs activation in cells expressing a RcsC_MalF strain containing the RcsC cytoplasmic domain fused to the first two transmembrane (TM) helices of the maltose-transporter membrane protein MalF [18]. RcsC_MalF increased the basal Rcs signal of the strain, but it responded to a RcsF-dependent signal like PMBN [18]. DjlA overexpression also activated Rcs signaling in this strain, demonstrating that DjlA does not need to interact with the RcsC TM helices (S5A Fig).

We further examined the role of the DjlA TM helix in DjlA-dependent Rcs activation. Point mutants in the TM helix of DjlA have been shown to completely block Rcs activation without affecting its stability or localization [53]. Furthermore, the DjlA TM helix is a dimerization domain and contains conserved glycine residues important for Rcs induction [55]. We replaced the DjlA TM helix with the first two MalF TM helices such that the cytoplasmic portion of DjlA is anchored to the membrane with the MalF helices (DjlA_MalF; S5B Fig). We tested P_rprA::mCherry induction upon overexpression of this DjlA_MalF variant in a ΔdjlA ΔrcsF strain. We observed that this chimeric DjlA_MalF variant is capable of inducing Rcs activity, unlike the DjlA H233Q mutant or the DjlA ΔTM variant lacking any TM helix (S5B Fig). These results lead us to conclude that the DjlA TM helix does not have an active role in Rcs induction, but that membrane localization is indeed essential. Previous reports did not observe any Rcs induction with either DjlA ΔTM or a MalF-DjlA chimeric variant [24,53,55]. The discrepancy could be due to the somewhat different DjlA_MalF protein construct used by us or may reflect the increased sensitivity of our P_rprA::mCherry fluorescent reporter as compared to the cps-lacZ beta-galactosidase assay used previously. We further tested the ability of DjlA ΔTM and DjlA_MalF to break the IgaA-RcsD interaction using the BACTH assay. S5C Fig shows that both of these DjlA constructs can disrupt the interaction. This is consistent with the DjlA TM helix not having a specific role in recognizing its Rcs target protein. T18-IgaA levels did not decrease upon overexpression of DjlA (S5D Fig). Surprisingly, DjlA ΔTM was also capable of disrupting the IgaA-RcsD interactions in BACTH even though it was not functional in Rcs activation. Possibly this is due to the difference in protein levels of IgaA/RcsD/DjlA between these two assays. Thus, the TM helix of DjlA is important for inducing activity but may be less important for recognizing its Rcs substrate. We note that others have found that a very similar construct of DjlA lacking the transmembrane region is stable and has co-chaperone function [24,54], consistent with our construct showing function in the BACTH assay.

DjlA was also able to disrupt the interaction of IgaAΔcyt with RcsD (S5C Fig). We have previously shown that the periplasmic interaction of IgaA and RcsD drives the signal in the BACTH assay [18], consistent with the high interaction seen in S5C Fig for IgaAΔcyt with RcsD. Assuming DjlA, with the J-domain in the cytoplasm, interacts with cytoplasmic domains of its substrates, the observation that full-length DjlA can disrupt IgaAΔcyt-RcsD interactions would suggest that DjlA acts on the PAS-like domain of RcsD, leading to conformational changes that not only overcome the T411A interaction with IgaA but also lessen the strong periplasmic contacts between RcsD and IgaA. Interestingly, the DjlA ΔTM was less effective than full length DjlA in disrupting IgaAΔcyt-RcsD interactions.

If multicopy DjlA weakens IgaA-RcsD interactions, it could be speculated that it has a role in modulating these interactions even in single copy under appropriate physiological conditions. However, we failed to find a single copy phenotype of ΔdjlA for Rcs signaling. A strain devoid of DjlA responded like a wild-type strain to varying doses of PMBN (S6A Fig) and responded to A22 and mecillinam (S6B Fig). It has been shown that deletion of djlA rendered a dnaJ deletion strain unable to grow at 40° and above [54], suggesting a possible role for DjlA at high temperatures. However, even at 42°C, no effect of deleting djlA on Rcs basal or induced activity was detected (S6C Fig). While we were unable to detect a role for single copy DjlA for Rcs induction, BACTH assays of IgaA-RcsD interactions showed a somewhat better interaction in the absence of DjlA, consistent with its proposed role in chaperoning this interaction (S6D Fig). Consistent with a role for DjlA for the cytoplasmic interaction of these proteins, an increased interaction of RcsDΔperi with IgaA was also seen in the ΔdjlA strain. RcsD T411A, with a tighter cytoplasmic interaction with IgaA, was unaffected by the absence of DjlA. Based on ribosome profiling, djlA is expressed at a modest level at 37° [56]; little is currently known about expression of this protein under other growth conditions. Therefore, we cannot be sure whether DjlA is significantly expressed under these assay conditions. Overall, these results are consistent with the idea that DjlA plays a role in the cytoplasmic signaling interactions of IgaA and RcsD, but under conditions not yet identified.

DrpB activates Rcs independently of its cell division role

In contrast with DsbA and DjlA, induction by DrpB overexpression has a requirement for the RcsC periplasmic domain for Rcs activation (Fig 2). This observation strongly suggests the possibility that there is signaling to the Rcs phosphorelay via the RcsC periplasmic domain, akin to that seen for many other sensor histidine kinases (reviewed in [57]). Whether there is a direct interaction between DrpB and RcsC for the signaling to occur or the RcsC periplasmic domain senses a signal generated by DrpB overexpression is not yet known. DrpB is a small (100 aa) protein with two transmembrane helices and a small periplasmic region (see schematic in S7A Fig). There are no additional recognizable domains. However, DrpB was identified as a multicopy suppressor of the growth defect of a ΔftsEX mutant at low ionic strength [58]. FtsEX is an ABC transporter that coordinates peptidoglycan remodeling and cell division, allowing cell separation. FtsEX helps in recruitment of downstream cell division proteins to the Z-ring, cell constriction, and activation of periplasmic peptidoglycan amidases [59–61]. FtsEX mutants display chaining defects and have low viability in low-osmolarity medium. DrpB localizes to the divisome and overexpression allowed the growth of an ΔftsEX strain at low osmolarity [58]. Activation of amidases by FtsEX is not strictly required for survival in low salt medium, so it is likely that overproduction of DrpB rescued the growth defect by improving recruitment of downstream components of the divisome [58,61]. Although no cell division or cell morphology defects were observed in a ΔdrpB strain, a double deletion strain with another cell-division protein, DedD, was filamentous and DrpB was found to interact with multiple divisome-associated proteins [58].

RcsB positively regulates a promoter of ftsZ, and multicopy RcsF is a known suppressor of the temperature sensitive ftsZ84 mutant [10]. Although DrpB did not rescue the ftsZ(Ts) phenotype [58], it seemed possible that the ability of DrpB to act as a weak suppressor of the ftsEX mutant is due to its role as an Rcs activator. To test this hypothesis, we asked whether RcsB was needed for the ability of DrpB to suppress the growth defect of ΔftsEX cells in the absence of salt. We overexpressed DrpB from a plasmid-encoded arabinose-inducible promoter in ΔftsEX and ΔftsEX ΔrcsB strains and assayed colony formation on LB medium with and without NaCl. As seen in Fig 6A, multicopy DrpB can rescue the growth defect of ftsEX mutants even in cells deleted for rcsB, demonstrating that the suppression of ΔftsEX is not via Rcs activation. We also asked whether activation of Rcs by DrpB required ftsEX, using the P_rprA::mCherry reporter in a ΔftsEX strain. We found that DrpB indeed induces Rcs in this strain (Fig 6B). Therefore, the role of DrpB in cell growth in the absence of ΔftsEX appears to be independent of Rcs activation and vice versa.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. DrpB signaling is independent of ftsEX.

A. Role of DrpB as a ftsEX suppressor is independent of its role as an Rcs activator: Strains transformed with the pBAD33 vector or pDrpB (pDSW1977) were grown overnight in LB Miller media with chloramphenicol at 37°C. The cultures were diluted to an OD₆₀₀ of 1 and 4 ul dilutions were spotted on LB Miller (permissive), LB without NaCl, and LB with 0.2% arabinose with no NaCl. All plates contained chloramphenicol (25 μg/ml). Plates were imaged after 16h incubation at 37°C. The strains used are: WT (EC251), rcsB::kan (AP158, ΔrcsB), ΔftsEX (EC1215), and ΔftsEX rcsB::kan (AP159, ΔftsEX ΔrcsB). B. DrpB can activate Rcs signaling in ΔftsEX: For the P_rprA::mCherry assay, the WT (EAW8) and ftsE::kan (AP154, ΔftsEX) strains transformed with the pBAD33 vector or pDrpB (pDSW1977) were grown in MOPS minimal glycerol medium (with 150 mM NaCl) containing chloramphenicol and either 0.2% glucose or 0.02% arabinose at 37°C. The RFU at OD 0.4 is plotted relative to uninduced WT with the vector, set to 1. Statistical significance was calculated using multiple unpaired t-tests and is shown as follows: ns (P > 0.05; non- significant),** (P ≤ 0.01), *** (P ≤ 0.001).

https://doi.org/10.1371/journal.pgen.1011408.g006

A set of drpB mutants were constructed in the pBAD plasmid and tested both for Rcs induction and for ability to rescue the growth of ΔftsEX in the absence of salt (S7 Fig). Since the periplasmic domain of RcsC is needed for DrpB signaling, it could be hypothesized that there is a periplasmic interaction between these proteins. We constructed a DrpB mutant with a deletion of its periplasmic region (Δperi_48-57); this deletion had no effect on the ability of DrpB to activate Rcs (S7A Fig). However, DrpB Δperi_48-57 did not rescue the growth of the ΔftsEX mutant, suggesting it is important for this function (S7B Fig). An interaction between the transmembrane regions of DrpB and RcsC also seemed possible. Based on the sequence similarity with other DrpB homologs, we generated point mutants in four highly conserved amino acid residues in and around each of the TM helices of DrpB. S7A Fig shows the effect of these mutants on Rcs activity in a strain deleted for RcsF. Only DrpB T89A had lost the ability to activate Rcs. This allele only slightly decreased suppression of ΔftsEX (S7B Fig). The DrpB R38F had decreased but not absent activity for both induction of Rcs and suppression of ΔftsEX (S7B Fig). We cannot rule out that these mutants are less active due to mislocalization or decreased stability. However, overall, the results confirm the existence of two separable functions of DrpB, with the periplasmic region critical for suppression of ΔftsEX but not for Rcs induction, and T89 critical for Rcs induction but less important to suppress ΔftsEX.

The BACTH assay was used to test if DrpB directly interacts with any of the Rcs components. However, full-length RcsC has previously been found to be non-functional in the BACTH assay [18], and was also negative for interaction with DrpB (S8A Fig). We observed variable and weak (not significant) interactions of DrpB with both IgaA and RcsD (S8A Fig). In a previous study, DrpB interacted with multiple divisome proteins as well as with MalF (used as a control), suggesting the possibility of false-positives in BACTH assays of DrpB with other membrane proteins [58]

As with DjlA, we were not successful in defining an Rcs phenotype for deletion of drpB. The ΔdrpB strain responds to PMBN, A22 and mecillinam (S8B Fig), which is consistent with a previous observation that DrpB is not needed for the Rcs response to osmotic shock or ethanol stress [28]. Multicopy DrpB also was previously found to induce the psp (phage shock) signaling pathway [28]. Rcs induction by DrpB was unaffected by the absence of psp components pspC and pspF (S8C Fig), ruling out induction of Rcs indirectly via induction of Psp. Ribosome profiling suggests a modest expression of DrpB from the chromosome [56], but very little is known about the roles and expression level of the single copy DrpB protein. More studies are needed to ascertain if multicopy DrpB activates the Rcs cascade by interacting directly with RcsC or more indirectly. More broadly, we do not know whether there is a physiological condition during which DrpB made from the single copy gene acts to promote Rcs activation, or whether the multicopy effect triggers a stress that leads to activation, but we would expect a critical role for the RcsC periplasmic domain in either case.

Similar to DrpB, a group of small cytosolic proteins in the ycgz-ymgABC operon, which is involved in biofilm formation and acid resistance, have been demonstrated to trigger the Rcs cascade [30,62]. AriR/YmgB, in complex with YcgZ and YmgA, was reported to target RcsC to induce the Rcs pathway [63]. We tested the Rcs activation by YmgB in our strains and found it to be RcsF independent and blocked by the RcsD T411A mutation (S9 Fig). However, in contrast with DrpB, it is independent of the periplasmic domain of RcsC, indicating a different mode of Rcs activation.

As noted above, the BarA phosphorelay protein was shown to regulate RcsB in Salmonella, independent of RcsF [26]. While our identification of the mode of action of DsbA, DrpB and DjlA would suggest that they act upstream of RcsB, we tested whether BarA was necessary for these factors to induce Rcs. First, the effect of a barA mutant on expression of our reporter under normal growth was tested (S10A Fig). The basal expression of the reporter was modestly reduced in the barA mutant, suggesting that, as in Salmonella, BarA may signal to regulate RcsB; a reduction was also seen in strains deleted for rcsC or rcsD, but not for the rcsB deletion, consistent with an effect on RcsB. Overexpression of DjlA or DrpB were each still able to induce the reporter in a barA mutant (S10B Fig, compare to Fig 1B). Similarly, deletion of dsbA induced signaling in the absence of barA (S10C Fig). Therefore, these three RcsF-independent mechanisms are also independent of BarA.

Interrupting the interactions of RcsD and IgaA induce Rcs signalling

Loss of DsbA or DjlA overproduction each activate the phosphorelay by weakening the interactions between IgaA and RcsD, affecting their folding/conformation via distinct modes. We have previously shown that IgaA and RcsD interact both through their periplasmic domains and their cytoplasmic domains ([18], Fig 1A). The periplasmic strong contact provides an anchor that keeps signaling from rising too high. Based on the identification of the RcsDT411A mutant in the cytoplasmic PAS domain of RcsD that blocks RcsF-dependent induction, the cytoplasmic contact was defined as the regulatory switch region [18]. The loss of DsbA impacts the periplasmic contacts, likely due to its role in formation of disulfide bonds in IgaA, and therefore IgaA folding. This is reflected in a decrease in the interaction of IgaA and RcsD in the absence of DsbA and in elevated Rcs activation, further increased when cells are treated with DTT. Presumably in a strain devoid of DsbA, a fraction of IgaA is in a reduced state with one or both disulfide bonds disrupted; addition of DTT leads to the disruption of any remaining disulfide bonds. Mutating the IgaA periplasmic cysteines predicted to be involved in these disulfide bonds mimics the dsbA mutant, decreasing the interaction of IgaA and RcsD (S4C Fig) and increasing signaling (Fig 1C). RcsF itself requires disulfide bonds for proper folding and function [7,44]. Therefore, it is not surprising that dsbA mutants no longer recognize RcsF-dependent cues such as A22 and mecillinam (S4A Fig), since RcsF is non-functional in this strain. Similar observations have been made in Salmonella; Salmonella dsbA mutants activate Rcs, independent of RcsF, and this is increased upon DTT addition, while a strain carrying a bamB deletion induces Rcs, dependent upon RcsF and is unaffected by DTT treatment [64].

We might have also expected PMBN, which normally induces the Rcs system in an RcsF-dependent fashion (see Fig 1C) to be ineffective in the dsbA mutant. However, it induces in much the manner that DTT does, independently of rcsF (Fig 1C). This result suggests that PMBN causes some redox stress, in addition to the LPS/membrane damage that leads to RcsF-dependent induction. When the Dsb network is functional, DTT cannot induce Rcs activation and any PMBN induction requires RcsF, suggesting that Rcs probably does not monitor disulfide bond formation in the periplasm as part of its normal function. Instead, the disulfide bonds are necessary for proper folding of IgaA to allow the periplasmic interaction with RcsD. We cannot rule out the possibility that growth conditions exist under which IgaA disulfide bonds might be disrupted sufficiently to induce Rcs signaling. A study of the roles of the periplasmic cysteines of IgaA in Salmonella also found that disulfide bonds were critical both for proper repression of Rcs and, to some extent, for viability of the cells [45].

Overproduction of DjlA, a membrane-localized DnaJ family protein, also loosens the interaction of IgaA and RcsD (Fig 5A) and turns up phosphorelay activity independently of RcsF (Fig 1B); these effects depend on an active J domain (mutated in DjlA H233Q). However, the primary effect of DjlA overexpression is likely on the cytoplasmic domains of RcsD, IgaA, or both. This is suggested both by the observation that DjlA overproduction but not loss of DsbA overcomes the RcsD T411A mutation (Fig 3) and by the location of the J domain in the cytoplasm. The ability of DjlA overproduction to induce signaling also allows us to conclude that the interaction of the RcsD PAS-like domain with the IgaA cytoplasmic domain can regulate activation by both RcsF-dependent and RcsF-independent signals.

DjlA was the only RcsF-independent activator which overcame the heightened interaction between RcsD T411A mutant and IgaA. This would be consistent with DjlA acting as a chaperone to mediate these interactions. Molecular chaperone systems can assist protein-protein interactions by loosening and altering protein conformations (reviewed in [65]). Unlike DnaJ and CbpA, the other E. coli DnaJ family proteins capable of working with DnaK, DjlA does not carry a specific substrate-binding domain, so it is not known how it identifies its clients [53]. Since only DjlA, and not the other two DnaJ family proteins, induces Rcs, it seemed possible that the membrane insertion of DjlA is critical in recruiting DnaK to the sites of IgaA/RcsD, reinforced by the lack of Rcs activation from a cytosolic soluble form of DjlA ([24,53], S5B Fig). However, we found that DjlA fused to the MalF transmembrane domains can activate Rcs, strongly suggesting this helix is not required for Rcs interaction. Our results are most consistent with RcsD as the target of DjlA function, based on the ability of DjlA to disrupt interaction of IgaAΔcyt and RcsD (S5C Fig). Because the periplasmic interaction of RcsD and IgaA is sufficient for a strong interaction signal in the BACTH assay, we can also conclude that the action of DjlA on the cytoplasmic domain of RcsD must lead to a change in the RcsD periplasmic domain.

The effects here were seen when DjlA was overproduced. We do not know whether or when the chromosomally encoded DjlA is able to modulate the IgaA-RcsD interaction. Deletion of djlA strengthened the interaction of IgaA and RcsD in the BACTH assay (S6D Fig) but did not have any effect on Rcs induction under a variety of conditions (S6 Fig). Not much is known about the physiological signals regulating DjlA expression; DjlA-stimulation of Rcs might be relevant under specific growth conditions or might be important for particular inducing signals not tested here. There is also evidence that DjlA acts as a negative regulator of Rcs [12,66]. Those studies demonstrated a stimulation of the Rcs activation upon deletion of DjlA, particularly in mutant strains such as mdo and rfa which are already activated for Rcs in an RcsF-dependent manner. Lack of DjlA could increase the envelope stress caused by mutants in mdo and rfa independently of the RcsF-independent effect seen here. Undoubtedly, DjlA has substrates other than the Rcs system, and possibly the opposing effects (controlling envelope stress to negatively regulate Rcs and overcoming the RcsD/IgaA protein to induce Rcs) balance each other, making it more difficult to define a single copy effect.

Rcs activation via histidine kinase RcsC

Uniquely among the three proteins we investigated, DrpB required the RcsC periplasmic domain for Rcs activation. RcsC is a member of a broad family of histidine kinases; in many of these, the periplasmic domains are critical for sensing environmental signals and regulating histidine kinase activity [67–69]. In our previous work, we found that IgaA interacted with RcsD, the phosphorelay protein, rather than RcsC and Rcs inducing treatments were independent of the RcsC periplasmic domain [18]. While RcsC activity (kinase vs. phosphatase) must be adjusted by RcsF-dependent induction, there was no previous evidence for a role of RscC in signal sensing. Thus, the identification of DrpB and its dependence on the RcsC periplasmic region demonstrates the likelihood that some signals likely act in the “classical” way seen for other histidine kinases; if so, such signals would be independent of RcsF. While we do not know if DrpB acts directly or indirectly on RcsC, it is tempting to speculate that the RcsC periplasmic domain is sensing a signal created by DrpB overexpression, allowing RcsC to activate signaling. Interestingly, a ribosome profiling study revealed DrpB as one of the proteins upregulated during severe acid stress in E. coli and this may be linked to its Rcs function [70]. Alternatively, the overproduction of DrpB itself is a source of stress which is detected by RcsC. DrpB overproduction also leads to the induction of another envelope stress response, the Psp response pathway [28]. The Psp response primarily detects and mitigates stress at the inner membrane (reviewed in [71]). While we find that the Psp pathway is not necessary for Rcs induction by DrpB (S8 Fig), it will be of interest to see whether other inducing signals shared by these two systems also share dependence upon the RcsC periplasmic domain.

Further evidence for RcsC as a sensor comes from studies on the protein YmgB. YmgB is a cytoplasmic protein involved in biofilm formation and acid-resistance and it stimulates the Rcs pathway, reportedly by interacting with the RcsC cytoplasmic domain [62,63]. In our hands, and consistent with the model suggested by DrpB overexpression, YmgB-dependent induction is RcsF-independent, but, unlike DrpB, is also independent of the periplasmic domain of RcsC (S9 Fig). This is consistent with previous work on YmgB and raises the possibility that RcsC in fact senses different signals with its cytoplasmic and periplasmic domains.

A number of other proteins have been found to affect Rcs activity. BarA plays a role in the induction of the Rcs system in Salmonella, via effects on RcsB synthesis [26]. YpdI, a lipoprotein of unknown function, induces mucoidy in the absence of RcsF [29]. Activation of Rcs in a double mutant of yqjA and yghB, both membrane proteins and putative lipid flippases, is partially dependent on RcsF [33,72]. Deciphering their mechanisms could add to our knowledge about Rcs regulation.

Overall, our results emphasize the versatility and flexibility of the Rcs pathway. RcsF senses stress at the outer cell surface and in the periplasm. The RcsF-independent signaling pathways studied here suggest that, in addition, RcsC may play a role in sensing stressors at the cytoplasmic membrane, and possibly in the cytoplasm as well. Possibly this sensing by RcsC represents the remnants of the original version of this system, before RcsD, IgaA and RcsF evolved to sense a broader set of inducing signals. Clearly there is still much to be learned about this system, across species, and in different natural environments. We look forward to better understanding this unique system in the future.

Bacterial growth conditions and strain construction

The strains were grown in LB medium with appropriate antibiotics (ampicillin 100 μg/ml, kanamycin 30–50 μg/ml, chloramphenicol 10 μg/ml for cat-sacB allele or 25 μg/ml for chl^R plasmids, tetracycline 25μg/ml, zeocin 50μg/ml). Glucose at a final concentration of 1% was added to reduce the basal expression from the plasmids containing pBAD and pLac promoters.

Strains, plasmids, primers and synthetic DNA fragments (gBlocks) used in this study are listed in S1–S4 Tables respectively. Oligonucleotides and gBlocks were from IDT DNA, Coralville, IA. Strains were constructed by recombineering or P1 transduction with selectable markers (S1 Table). Recombineering was done in strains carrying a chromosomal mini-λ Red system (miniλ::tet) or a plasmid-borne Red system (pSIM27). Some strains were generated by direct P1 transduction from the corresponding mutant strains in the Keio collection [73]. Plasmids were constructed by the Gibson assembly method using the In-fusion HD Cloning kit (Takara Bio USA) [74]. PCR products were purified using column purification (Qiagen) and transformed into NEB DH5α F’lacIQ cells. Site-directed mutagenesis in the genes was carried out using the QuikChange Site-directed mutagenesis kit (Agilent). Sequencing was done to confirm the chromosomal and plasmid modifications.

Screening for multicopy RcsF-independent activators

A pBR-based plasmid library [75] was transformed into EAW34 electrocompetent cells. Aliquots of the transformed mixture were plated on LB plates containing 100 μg/ml ampicillin and incubated overnight at 37°C. Subsequently, the plates were imaged and the colonies displaying high fluorescence were restreaked on LB ampicillin plates and purified. The plasmid DNA was isolated from these colonies and retransformed into EAW34 to validate their phenotype. The positive candidate plasmids were isolated, sequenced and the reads mapped to the E. coli K-12 MG1655 genome to identify the inserts.

Fluorescence reporter assays

Growth and fluorescence assays for Rcs activation were carried out in 96-well plates in a Tecan Spark microplate reader. All the strains carried a P_rprA::mCherry transcriptional fusion at the ara locus as a reporter for Rcs signaling [18]. For strains carrying overexpression plasmids expressing DjlA, DrpB, or RcsF, the overnight cultures were grown in MOPS minimal glucose (0.2%) medium (Teknova). For the P_rprA::mCherry assays for these strains, they were diluted to OD₆₀₀ 0.03–0.05 in MOPS minimal glycerol media (0.05% glucose, 0.5% glycerol) with 0.02% arabinose (induced) or 0.2% glucose as an uninduced control. Both the optical density (OD₆₀₀) and mCherry fluorescence were monitored every fifteen minutes for 12 hours. Polymyxin B nonapeptide (PMBN; Sigma) was used at 20 μg/ml, to induce the Rcs system. A22 (an MreB inhibitor) was used at 5 μg/ml, and mecillinam was used at 0.3 μg/ml. Each assay was performed in technical duplicates in the microtiter plate, with the biological replicates performed on different days. The fluorescence values at equivalent OD₆₀₀ values (0.4 ± 0.04) for each strain were converted to bar graphs and plotted as signal fold change of fluorescence signal over the uninduced wild type strain. Error bars represent the standard deviation of at least 3 assays. For strains that did not reach an OD₆₀₀ value of 0.4, the OD is noted on the bar graph. For the P_rprA::mCherry assays for the strains not carrying any plasmids, MOPS minimal glucose media was used for growing overnight cultures and performing the fluorescence assays. For the assays pertaining to dsbA mutants, the overnight cultures were grown in LB medium. Subsequently, cells were washed with and diluted to OD₆₀₀ 0.05 in MOPS minimal glucose media to carry out the P_rprA::mCherry assay. To activate the Rcs system, Dithiothreitol (DTT, Sigma) was added at a final concentration of 1mM or Polymyxin B nonapeptide (PMBN; Sigma), was used at 20 μg/ml. All assays were performed at 37°C, unless otherwise indicated. Values derived from at least three independent experiments were plotted to show the mean with error bars indicating standard deviation. These were statistically analyzed by multiple unpaired t-tests using GraphPad Prism 10 software. The primary data for all graphs are found in S1 Data.

Bacterial adenylate cyclase two hybrid assay (BACTH)

For the bacterial adenylate cyclase two hybrid assay (BACTH), an adenylate cyclase mutant strain (BTH101) or a derivative if this strain was used. The proteins whose interactions were being examined were cloned into the T18 and T25 portions of adenylate cyclase [46,47]. The reconstitution of adenylate cyclase upon their interaction allows production of cAMP, which interacts with CRP to activate the lac operon, assayed as beta-galactosidase activity. In the absence of interaction, T18 and T25 do not get reconstituted to form a functional adenylate cyclase and the strain does not express beta-galactosidase. The proteins used in this study were fused at their C-terminal to the Cya fragments. Plasmids expressing IgaA or RcsD as T18/T25 fusion constructs were cotransformed into BTH101, plated on LB-agar medium containing 100 μg/ml ampicillin and 50 μg/ml kanamycin and incubated at 30°C for 2 days. To measure the beta-galactosidase activity, the resulting colonies were inoculated and grown overnight in LB medium containing 100 μg/ml ampicillin, 50 μg/ml kanamycin, and 0.5 mM IPTG at 30°C. The beta-galactosidase assay was performed in 96-well plates. The activity was analyzed by measuring the kinetics of ONPG degradation by monitoring the OD 420 nm at 28°C for 40 min at intervals of 1 min in a Tecan Spark microplate reader. The beta-galactosidase activity was calculated using the slope of OD 420 divided by their OD 600. Each protein fusion paired with the cognate vector produces very little activity and these were used as negative controls. Every graph is compiled from at least 3 separate sets of assays, and the beta-galactosidase activity was plotted relative to the IgaA/RcsD interaction in that particular experiment.

To test the effect of DsbA activity on the various IgaA/RcsD constructs, the BACTH assays were carried out in BTH101 and a dsbA mutant derivative of BTH101 (AP58). To test the effect of DjlA on the IgaA/RcsD interactions, djlA or mutants of djlA were cloned downstream of IgaA under the control of the same promoter. The assay was carried out in BTH101 or its derivatives EAW1, EAW2, and EAW4 as above and interactions plotted relative to the IgaA/RcsD interaction in WT, set to 1. The primary data for all graphs are found in S1 Data.

Western blots

For S4D Fig, plasmids were transformed into EAW8 and EAW62 and for S5D Fig, the plasmids were transformed into E. coli DH5α competent cells (NEB). The colonies were inoculated into LB media containing Amp (100 μg/ml) and 1 mM IPTG and then grown for 8 hours at 30°C. One ml of culture was precipitated using TCA, washed in acetone and resuspended in LDS sample buffer standardized to OD. 8 μl volumes were loaded on 4–12% NuPage gradient gels (Invitrogen, CA) and then transferred onto nitrocellulose membranes in an iBlot2 transfer device as per manufacturer’s specifications (Invitrogen, CA). The membranes were blocked with Casein PBS and then incubated in CyaA-T18 mouse monoclonal primary antibody at 1:10,000 (Santa Cruz Biotechnology, CA) or/and a mouse monoclonal anti-EF-Tu at 1:10,000 (Hycult Biotech, PA) overnight at 4°C. Secondary fluorescent antibody anti-mouse DyLight800 (Bio-Rad, CA) was used at 1:10,000. Imaging was done on a ChemiDoc MP imaging system (Bio-Rad).

Viability testing of ftsEX mutants

The wildtype and ftsEX deletion strains were transformed with the pBAD33 vector or a DrpB containing plasmid and plated on LB Miller-agar medium at 30°C containing 25 μg/ml chloramphenicol. Overnight cultures were grown in LB Miller medium 25 μg/ml chloramphenicol at 30°C. Cells were harvested from 1 mL of culture and diluted to OD₆₀₀ 1.0 in LB Miller or LB0N (LB without any NaCl) medium. Serial dilutions (10-fold) were made and 4 μl was spotted onto LB Miller-agar or LB0N-agar plates containing 25 μg/ml chloramphenicol. Arabinose was added at 0.2% for induction in LB0N. Plates were imaged after 16 h of growth at 37°C.