Proteomic Investigation of the Signal Transduction Pathways Controlling Colistin Resistance in Klebsiella pneumoniae

Direct role of PhoPQ and the indirect role of signal transduction from PhoPQ through PmrD to PmrAB in Arn protein production and colistin resistance caused by mgrB mutation.A collection of six K. pneumoniae spontaneous single-step mutants were selected in the laboratory using Muller-Hinton agar containing 32 μg · ml⁻¹ of colistin, and in each case, PCR and whole-genome sequencing confirmed only mutations in mgrB or upstream. Three different mutations were seen, causing the following changes: Gln29STOP in MgrB (found in four colistin-resistant mutants here, represented by mutant P21); an A to G transition at −31 relative to the mgrB start codon, weakening a putative second −10 promoter sequence (Fig. S1) (represented by mutant P22); a deletion comprising the region between −19 relative to the start codon, to remove the first 41 amino acids of MgrB (represented by mutant P23).

Colistin MICs against these three representative mutants were determined and, in all cases, colistin resistance was confirmed (Table 1). Envelope proteomics identified 45 proteins significantly differentially regulated in all three mgrB mutants relative to the parent strain, Ecl8 (33 up- and 12 downregulated; Table S1). These included the ArnABCDT operon proteins known to be responsible for modification of lipopolysaccharide by the addition of 4-amino-4-deoxy-l-arabinose (5). Figure 1 shows that Arn protein production is highest in the mutant P23, where MgrB is entirely lost. The mgrB nonsense mutant truncated at 29 amino acids, P21, and the mgrB promoter mutant P22 both have significantly lower Arn protein production than in P23, implying some residual repressive activity of MgrB in both cases. Overall, P22 had the lowest Arn protein production of the three (Fig. 1). Nonetheless all three mutants are colistin resistant though, as expected based on Arn protein production levels, the highest MIC is against P23 and the lowest against P22 (Table 1). This leads to the conclusion that once Arn protein production increases above a certain threshold, colistin resistance is conferred, but that as protein production increases further (P23 > P21 > P22), MIC also increases.

• Open in new tab
• Download powerpoint

FIG 1

Arn protein abundance in parent strain Ecl8 versus three mgrB mutants. Strains were grown in CAMHB and raw envelope protein abundance data for each Arn protein in a sample are presented, normalized using the average abundance of the 50 most abundant proteins in that sample. Data for three biological replicates of parent (Ecl8) and mgrB mutants (P21, P22 and P23) are presented as mean ± standard error of the mean. All mutants have statistically significantly increased production of all Arn proteins relative to Ecl8 based on a t test (P < 0.05).

It has been proposed that, in K. pneumoniae, arn operon gene expression is increased upon mgrB mutation through activation of the two-component system PhoPQ, and that, in addition, there is a secondary increase through activation of the PmrAB two-component system, with PmrD being necessary for transducing PhoPQ activation into PmrAB activation (5). Using mgrB mutant P23 as a starting point, we disrupted phoP, pmrD, and pmrA to test the effects of these mutations on Arn protein production and colistin MICs.

Observed Arn protein abundance changes in these regulatory mutants demonstrated the primacy of PhoPQ activation in colistin resistance driven by mgrB loss. Arn protein abundance returned to wild type (Fig. 2) and the colistin MIC fell below the resistance breakpoint (Table 1) upon disruption of phoP in the mgrB mutant P23. In contrast, our proteomics analysis (Fig. 2) showed that PmrD and PmrAB play only a minor role in increased Arn protein production seen in an mgrB mutant. Only ArnC significantly reduced in abundance following disruption of pmrA. There was a larger effect following disruption of pmrD (four out of five Arn proteins significantly reduced in abundance), but in neither mutant did any of the Arn proteins fall in abundance significantly below levels seen in the mgrB mutants P21 and P22 (Fig. 1) and, not surprisingly therefore, the pmrA and pmrD mutant derivatives of mgrB mutant P23 remained colistin resistant (Table 1). These data add support for our conclusion that there is a threshold of Arn protein abundance required for colistin resistance. It seems clear that PhoPQ activation alone can support an abundance above this threshold even without any additional effects caused by PmrAB activation.

• Open in new tab
• Download powerpoint

FIG 2

Arn protein abundance in mgrB loss-of-function mutant P23 versus its phoP, pmrD, and pmrA loss-of-function mutant derivatives. Strains were grown in CAMHB and raw envelope protein abundance data for each Arn protein in a sample are presented, normalized using the average abundance of the 50 most abundant proteins in that sample. Data for three biological replicates of parent (P23) and mutants where phoP, pmrD, or pmrA had been insertionally inactivated (PhoP⁻, PmrD⁻, or PmrA⁻) are presented as mean ± standard error of the mean. All PhoP⁻ mutants have statistically significantly reductions in production of all Arn proteins relative to P23 based on a t test (P < 0.05). For PmrD⁻ and PmrA⁻ mutants, significant changes relative to P23 were seen, respectively, for ArnD, -A, -C and -B, or ArnC only.

PhoPQ-regulated proteins identified following mgrB mutation and PhoQ activation.Of 45 proteins (i.e., including the Arn proteins) differentially regulated in all three mgrB mutants (Table S1), 18 of those upregulated in the mgrB mutant P23 returned to wild-type levels upon disruption of phoP (Table 2). These included the five Arn proteins, the response regulator PhoP itself, LpxO, and two Mg²⁺ transporters, including MgtA, SlyB and MacA. Transcripts representing all these proteins have been seen to be upregulated in mgrB loss-of-function and in PhoQ activatory (PhoQ*) colistin-resistant clinical isolates relative to colistin-susceptible control isolates through transcriptomics (20). Our proteomics analysis reinforces the definition of a core PhoPQ regulon, but a secondary observation is that the majority (27/45) of the protein abundance changes seen in the mgrB mutant P23 (Table S1) do not occur via activation of PhoPQ, i.e., they were not reversed following disruption of phoP (Table 2). The implication of this finding, made here by comparing otherwise isogenic pairs of strains, is that MgrB interacts with regulatory networks other than PhoPQ in K. pneumoniae, though these additional effects are not important for colistin resistance, which was completely reversed following disruption of phoP (Table 1).

In order to further investigate the role of direct PhoPQ activation in colistin resistance, we turned to a colistin-resistant, PhoQ* (activatory) mutant that we selected from K. pneumoniae clinical isolate KP47 (Table 1). Whole-genome sequencing identified the mutation causes a Tyr89Asn change in PhoQ. Proteomics comparing KP47 with the PhoQ* mutant derivative revealed that levels of Arn protein production in the PhoQ* mutant were not significantly different from those in the mgrB loss-of-function mutant P23 (Fig. 3). Indeed, despite starting with a different parent strain, significant upregulation of 15/18 proteins seen to become downregulated when phoP was disrupted in the Ecl8-derived mgrB mutant P23 were also upregulated in the PhoQ* mutant relative to its parent, KP47. This further focused down onto the core K. pneumoniae PhoPQ regulon, which is shown in Table 2.

• Open in new tab
• Download powerpoint

FIG 3

Arn protein abundance in mgrB mutant P23 versus clinical isolate KP47 and PhoQ* or PmrB* (activatory) mutant derivatives. Strains were grown in CAMHB and raw envelope protein abundance data for each Arn protein in a sample are presented, normalized using the average abundance of the 50 most abundant proteins in that sample. Data for three biological replicates of parent (KP47) and mutants where PhoQ or PmrA have been activated, compared with the mgrB loss-of-function mutant P23 are presented as mean ± standard error of the mean. All Arn proteins, and all Arn proteins except ArnB, were significantly upregulated relative to KP47 in its PhoQ* mutant and PmrB* mutant, respectively, based on a t test (P < 0.05). For the PmrB* mutant, significantly lower abundances relative to the PhoQ* mutant were seen for ArnA and ArnB.

Proteins upregulated following activation of PmrAB.It has been reported that PmrAB activation directly by mutation can confer colistin resistance (23, 24) and we were able to select a colistin-resistant mutant of clinical isolate KP47 with an activatory mutation in PmrB (Table 1). The mutation identified using whole-genome sequencing was Thr157Pro. Arn protein abundance in this PmrB* mutant was significantly elevated relative to KP47 in all cases except for ArnB. For all except ArnA and ArnB, the extent of abundance increase was like that seen in the PhoQ* derivative of KP47, and the mgrB mutant P23 (Fig. 3). In total, however, only 7/45 proteins significantly up- or downregulated in the mgrB mutant P23 were significantly up- or downregulated in the PmrB* mutant. As well as the Arn proteins (except ArnB), these were SlyB, LpxO, and one Mg²⁺ transporter, which are all part of the core PhoPQ regulon (Table S1). Transcripts representing all seven of these PrmAB/PhoPQ dual regulated proteins, plus ArnB, have also been seen to be upregulated in clinical isolates with activatory mutations in CrrB (20), which indirectly activates PmrAB (26, 27).

The fact that ArnB abundance did not increase significantly above the level seen in wild-type KP47 was surprising since the PmrB* mutant is colistin resistant (Table 1). This suggested that either significant upregulation of ArnB is not essential for colistin resistance or there is another mechanism involved in colistin resistance in the PmrB* mutant. To identify a possible additional mechanism, we searched the 65 proteins differentially regulated in the PmrB* mutant relative to KP47. Three of those most strongly overproduced were PmrA, PmrB, and PmrC (Fig. 4). Transcription of pmrC (also known as eptA) is known to be positively controlled by PmrAB in K. pneumoniae (24). In Salmonella spp. it encodes a phosphoethanolamine transferase, responsible for modifying lipopolysaccharide by decorating it with phosphoethanolamine, and this increases the MIC of colistin (25).

• Open in new tab
• Download powerpoint

FIG 4

PmrA, B, C, and E protein abundance in mgrB mutant P23 versus clinical isolate KP47 and PhoQ* or PmrB* (activatory) mutant derivatives. Strains were grown in CAMHB and raw envelope protein abundance data for each Pmr protein in a sample are presented, normalized using the average abundance of the 50 most abundant proteins in that sample. Data for three biological replicates of parent (KP47) and mutants where PhoQ or PmrA have been activated, compared with the mgrB loss-of-function mutant P23 are presented as mean ± standard error of the mean. PmrA, B, and C were significantly upregulated relative to KP47 in its PmrB* mutant based on a t test (P < 0.05). No other differences were statistically significant.

Little evidence for PhoPQ-PmrD-PmrAB-mediated signal transduction for in vitro colistin resistance in K. pneumoniae.As shown above, disruption of pmrD or pmrA in an mgrB loss-of-function mutant did not reduce Arn protein production below a threshold required for colistin resistance (Fig. 2, Table 1). It was also interesting to find that PmrA, B, or C were not produced above our limit of detection (around 100 times less than the level seen in the PmrB* mutant) in any mgrB loss-of-function mutant or in the PhoQ* mutant (Fig. 4). These two findings lead us to conclude that cross activation of PmrAB (the direct regulator of pmrC) following activation of PhoPQ is very limited under the growth conditions used for our analysis.

Our observation that the phosphoethanolamine transferase PmrC is unique to the PmrAB-mediated branch of the colistin resistance-mediating regulatory system may well explain why, of multiple studies monitoring the impact of mgrB loss-of-function on lipopolysaccharide modification in K. pneumoniae (11–13, 29), only one has reported elevated levels of phosphoethanolamine modification (29). Indeed, even in this case, contrary to expectations, the observed modification, and the observed upregulation of pmrC expression were both apparently dependent on PhoPQ but not PmrAB, suggesting it was not caused by PhoPQ-PmrD-PmrAB cross-regulation at all (29). Furthermore, the authors showed that pmrC disruption in an mgrB loss-of-function mutant background only had a small impact on survival in the presence of colistin (29). This implies that even when rarely seen, phosphoethanolamine modification by PmrC has only a minor role in colistin resistance in K. pneumoniae. Indeed, an absence of phosphoethanolamine modification has been advocated as a way to identify mutational colistin resistance (as opposed to mcr-mediated resistance, which does cause this modification), whether due to mgrB loss-of-function mutation, PhoPQ activation, or PmrAB activation in K. pneumoniae (30).

The reason why other earlier seminal reports have placed far higher importance on the cross-activation of PmrAB by PhoPQ via PmrD in K. pneumoniae (14, 18) could be that they used growth media that caused greater basal activation of PmrAB. It is important to remember that the PmrD linker protein has only experimentally been shown to increase the activation of PmrAB once PmrAB has been activated by an external signal, not to activate PmrAB from the basal state (14, 18). Since PmrAB activation is affected by iron concentration and pH, it may be that the medium used for colistin susceptibility testing—and used by us here—does not activate PmrAB in the first place, so there is nothing that PmrD can do to enhance activation, effectively silencing the cross-regulatory pathway. Indeed, it is worth noting that in some Enterobacteriaceae species, e.g., Enterobacter cloacae, there is no pmrD gene and PhoPQ and PmrAB work independently to regulate arn operon transcription (31).

Another explanation for differences between the conclusions of our own and previous work is that previous work relied on measurements of transcript levels. In some cases, small changes in protein abundance are associated with phenotypically relevant changes in antimicrobial susceptibility, as we have shown previously in K. pneumoniae, such as with loss-of-function mutations in ramR, which, despite having <5-fold effects on OmpK35 porin and AcrAB-TolC efflux pump production, have large effects on susceptibility to a range of antimicrobial agents from different classes (32). But in some cases, large changes in gene expression are required to have a phenotypic effect when a gene is not highly expressed in the wild type, for example in the case of OqxAB efflux pump production as controlled by OqxR in K. pneumoniae, which needs to increase >10,000-fold to have a phenotypic effect on resistance (33). One major advantage of proteomics is that comparisons of protein abundance can be drawn between different gene products, which is not always the case when transcript levels are measured, since the kinetics of DNA hybridization can have major influences on signal. This advantage is exemplified here in the context of PmrE production, which we find to be constitutive in K. pneumoniae and not part of the PhoPQ or PmrAB regulons (Fig. 4). In contrast, it is controlled by both regulators in Salmonella spp. (5). In the context of PmrB activation in K. pneumoniae, constitutive expression of pmrE has also been shown by reverse transcriptase quantitative PCR (qRT-PCR) (26) and in the context of CrrB activation—leading to PmrAB activation—this has been shown through transcriptomics (20). The added value of proteomics is that we can conclude that PmrE, which is an enzyme responsible for driving the committed step for the biosynthesis of 4-amino-4-deoxy-l-arabinose required for colistin resistance (5), is present at levels in wild-type cells similar to the levels of Arn protein produced in colistin-resistant mutants, rather than its production being constitutive but at low levels (Fig. 3 and 4).

Conclusions. We conclude that colistin resistance caused by PhoPQ activation in conditions defined for colistin susceptibility testing, i.e., in most clinical cases (28), is due almost exclusively to direct Arn protein upregulation. While PmrAB activation can also cause Arn protein upregulation and colistin resistance, there is no evidence for significant activation of PmrAB via PmrD as an additional mechanism for upregulating Arn proteins following PhoPQ activation under these growth conditions. We also conclude that the level of Arn protein upregulation dictates colistin MICs. It is interesting to note, therefore, that clinical isolates with multiple mutations activating PhoPQ can be found, where there is an additive effect on colistin MIC (28). The implication is that real-world colistin usage in the clinic in some cases selects for mutations or combinations of mutations that confer colistin MICs above the currently defined resistance breakpoint.