Revisiting regulation of potassium homeostasis in Escherichia coli: the connection to phosphate limitation

Abstract Two‐component signal transduction constitutes the predominant strategy used by bacteria to adapt to fluctuating environments. The KdpD/KdpE system is one of the most widespread, and is crucial for K+ homeostasis. In Escherichia coli, the histidine kinase KdpD senses K+ availability, whereas the response regulator KdpE activates synthesis of the high‐affinity K+ uptake system KdpFABC. Here we show that, in the absence of KdpD, kdpFABC expression can be activated via phosphorylation of KdpE by the histidine kinase PhoR. PhoR and its cognate response regulator PhoB comprise a phosphate‐responsive two‐component system, which senses phosphate limitation indirectly through the phosphate transporter PstCAB and its accessory protein PhoU. In vivo two‐hybrid interaction studies based on the bacterial adenylate cyclase reveal pairwise interactions between KdpD, PhoR, and PhoU. Finally, we demonstrate that cross‐regulation between the kdpFABC and pstSCAB operons occurs in both directions under simultaneous K+ and phosphate limitation, both in vitro and in vivo. This study for the first time demonstrates direct coupling between intracellular K+ and phosphate homeostasis and provides a mechanism for fine‐tuning of the balance between positively and negatively charged ions in the bacterial cell.

hibitory effect of K + under these conditions. Furthermore, it has been shown that dephosphorylated enzyme IIA Ntr , which is part of the Ntr phosphotransferase system, can bind KdpD, modulates its activity and therefore links carbohydrate metabolism to K + homeostasis (Lüttmann et al., 2009).
The pho regulon comprises genes coding for proteins that are important for phosphate assimilation (Hsieh & Wanner, 2010) and the timing of their expression, as well as their production levels, are determined by the binding affinity of PhoB for the corresponding promoters (Gao & Stock, 2015). One of the targets of PhoB is the pst operon, which encodes the high-affinity phosphate transporter Pst (K m app ≈0.2 μmol/L (Rosenberg, Gerdes, & Chegwidden, 1977;Willsky & Malamy, 1980)).
The Pst transporter belongs to the ATP-binding cassette (ABC) family of transporters and comprises the periplasmic phosphate-binding protein PstS, the two transmembrane channel-forming factors PstA and PstC and the ATPase PstB. It is rather unclear how the histidine kinase PhoR senses the availability of PO 3− 4 . It is suggested that PhoR monitors the activity of the high-affinity PO 3− 4 transporter PstCAB via PhoU, which interacts with both PstB and PhoR and thereby probably modulates PhoR activity (Gardner, Johns, Tanner, & McCleary, 2014). However, PhoU is not involved in phosphate sensing in Caulobacter crescentus (Lubin, Henry, Fiebig, Crosson, & Laub, 2016). In E. coli PstCAB and PhoU have an inhibitory effect on PhoR, as the absence of any one of these components results in constitutive expression of the pho regulon (Hsieh & Wanner, 2010;Lamarche, Wanner, Crepin, & Harel, 2008). Moreover, it is also known that, besides PhoR, some noncognate histidine kinases such as ArcB, CreC, KdpD, QseC, EnvZ, and BaeS can stochastically activate PhoB (Zhou, Grégori, Blackman, Robinson, & Wanner, 2005).
In this study, we demonstrate that the histidine kinase PhoR can activate kdpFABC expression independently of KdpD, but requires functionally active KdpE to do so. Furthermore, the deletion of phoU-which is supposed to be a negative regulator of the PhoR/ PhoB system in E. coli-resulted in high kdpFABC expression in a reporter strain lacking KdpD. Using the bacterial two-hybrid (BACTH) system we show here that both the KdpD/KdpE and the PhoR/PhoB two-component systems interact with PhoU in vivo. Ultimately, we find that cross-regulation between these two systems is not just a nonphysiological curiosity, but also occurs in the presence of the partner histidine kinase under conditions of K + and phosphate limitation.

| Strains, plasmids, and oligonucleotides
Strains, plasmids, and oligonucleotides used in this study are listed in Tables 1-3. The strains LB2240ΔkdpD andLB2240ΔkdpD,kdpE D52N were constructed in two steps, using Red ® /ET ® recombination technology in combination with rpsL counterselection (Heermann, Zeppenfeld, & Jung, 2008). Briefly, in the first step, a linear DNA fragment encoding a kanamycin cassette (amplified with the primer pairs 50bpkdpD_rpsL-kan_sense + 50bpkdpD_rpsL-kan_antisense and 50bpkdpE_rpsL-kan sense + 50bpkdpE_rpsL/kan_antisense; see Table 3) was inserted into the kdpD (for LB2240ΔkdpD) and kdpE (for LB2240ΔkdpD,kdpE D52N ) genes, respectively. In the second step, the kanamycin cassette was replaced by a DNA fragment encoding either the kdpD deletion, or the kdpE D52N substitution, respectively. The DNA fragment incorporating the kdpD deletion was generated by a two-step PCR using genomic DNA of LB2240 as template and the primer pairs kdpCDforI_sense + ΔkdpD_antisense and ΔkdpD_sense + kdpE_antisense (Table 3). The fragment bearing the kdpE D52N substitution was derived from pPV-2/D52N by amplification with the primer pairs kdpE_sense + kdpE_antisense (Table 3).
LB2240ΔkdpDΔptaΔackA and LF3ΔkdpD were constructed using Quick and Easy E. coli Gene Deletion and Bac Modification Kits (Gene Bridges) as previously described . Briefly, we inserted a linear DNA fragment encoding a kanamycin cassette (obtained by amplification with the primer pairs 50bpackApta_rpsL-kan_sense + 50bpackApta_rpsL-kan_antisense or delta KdpD_up + delta KdpD_down; Table 3) as selection marker into the genes pta, ackA (parental strain LB2240ΔkdpD) and kdpD (parental strain LF3), respectively. To avoid effects of the kanamycin resistance cassette on kdpE expression levels, we removed the selection marker in LF3ΔkdpD using the pCP20 helper plasmid as described previously (Baba et al., 2006). LB2240ΔpstC, LB2240ΔkdpDΔpstC and all other LF3 deletion mutants (Table 1) were constructed by P1 transduction (Miller, 1972). Strains JW3705 (pstC::npt), JW0390 (phoR::npt) and JW3702 (phoU::npt) were used as donor strains (Baba et al., 2006). Preparation of phage lysate from donor strains and transduction to recipient strains was performed as described previously (Leder, Tiemeier, & Enquist, 1977). For double or triple deletions the kanamycin cassette was removed between steps using the helper plasmid pCP20 as previously described (Datsenko & Wanner, 2000). Successful deletion was confirmed by PCR using appropriate primers listed in Table 3.
Plasmids for the bacterial adenylate cyclase assays (BACTH) were constructed by DNA amplification using genomic DNA of E. coli MG1655 as template with primer pairs listed in Table 3, and subsequent cloning into the indicated vectors. Successful insertion was confirmed by restriction analysis with appropriate enzymes.
Plasmid pBR-Cherry pPstS was constructed by amplification of the region upstream of the pstS gene (~500 bp) using primers pPstS_ BamHI_s and pPstS_XmaI_as (Table 3) and genomic DNA of E. coli MG1655 as template. After restriction with XmaI and BamHI, the DNA fragment was ligated into pBR-Cherry. Successful cloning was confirmed by restriction and sequencing analyses.

| Growth conditions
yeast extract] was used as standard medium for strains TKR2000, LB2240 and derivatives, and Lysogeny Broth [1% (w/v) NaCl, 1% (w/v) tryptone, 0.5% (w/v) yeast extract] for MG1655, BL21 (DE3/ pLysS), LF3 and derivatives, respectively. To analyze K + -dependent growth and reporter gene expression we used a phosphate-buffered minimal medium containing the indicated K + concentrations (Epstein & Kim, 1971). For growth of cells on different PO 3− 4 and K + concentrations we used a Tris-maleic acid (TMA) minimal medium (Weiden et al., 1967), and KCl and Na 2 HPO 4 were added as indicated. Glucose was added as the carbon source at a final concentration of 0.4% (w/v).

| RNA isolation, cDNA synthesis, and qRT-PCR
At indicated time points, cells were harvested and RNA was isolated using the RNeasy Mini Kit (Qiagen) according to the manufacturer's directions. RNA concentration was adjusted to 20 μg ml −1 and treated with RNAse-free DNAse (New England Biolabs) for 60 min at 37°C to remove residual chromosomal DNA. Subsequently, DNAse was heat-inactivated for 5 min at 70°C and RNA was stored at −20°C.

| Genome sequence analysis
Quality-trimmed sequence reads were aligned to the most closely related published genome, E. coli K12 MG1655 (GenBank Accession No. CP009685) using NovoAlign (NovoCraft Technologies) and CLC Genomics Server 7.5 (Qiagen). Alignment depth was between 270 and 290. Less than 0.15% of the reads could not be aligned. The alignments were screened for differences between the sequenced E. coli LB2440 mutants and the MG1655 genome according to a previously described procedure for local realignment and SNP and indel detection (Dettman et al., 2012), adjusting the settings to suit the analyzed data. Finally, alignments were manually examined for differences between LB2240ΔkdpD and LB2240ΔkdpD* strains using the samtools pileup output (Dettman et al., 2012) and ReadXplorer (Hilker et al., 2014) for alignment visualization.
The LB2240 mutant sequence reads are publicly available in GenBank under the BioProject PRJNA322678.

| β-Galactosidase activity assays (determination of kdpFABC expression in vivo)
In vivo kdpFABC expression was analyzed using strains LF3 and derivatives (P kdpFABC ::lacZ) thereof (Table 1). Cells were aerobically grown at 37°C in minimal media containing the indicated K + concentrations (Epstein & Kim, 1971;Weiden et al., 1967) and harvested by centrifugation in late exponential phase. β-Galactosidase activity was determined as described (Miller, 1992) and is given in Miller Units.

| Cell fractionation and preparation of membrane vesicles
E. coli strain TKR2000 transformed with plasmid pPV5-3 was grown aerobically at 37°C to until OD 600 =1 in KML complex medium sup- a large proportion of the protein was found in the membrane fraction. We therefore solubilized this fraction as described previously (Jung, Tjaden & Altendorf 1997) prior to purification. Purification was performed as described before (Heermann, Altendorf, & Jung, 2003), except that 250 mmol/L imidazole was present in the elution buffer.

| Analytical procedures
The concentration of soluble proteins was determined as described by Lowry, Rosebrough, Farr, & Randall, (1951) and membrane proteins were quantified with a modified Lowry method (Peterson, 1977) using bovine serum albumin as a standard.

| Phosphorylation assay
Purified PhoR (0.2 mg ml −1 , final concentration) or membrane vesicles containing approximately 0.2 mg ml −1 KdpD (final total protein concentration 2 mg ml −1 ), respectively, were incubated in phosphoryla- continued. Aliquots were removed at different times, mixed with SDS sample buffer and subjected to SDS-PAGE. Gels were then dried and protein phosphorylation was detected by exposure of the gels to a Storage Phosphor Screen. Band intensity was quantified using ImageJ (Schindelin et al., 2015).

| E. coli requires the KdpFABC system to grow under K + limitation
In order to determine the role of the histidine kinase KdpD for K +dependent growth, we generated the E. coli strain LB2240∆kdpD, which is deleted for kdpD, as well as for trk and carries a mutated kup (kup − ), which encode the two constitutively expressed K + transporters. This strain retains a functional kdpFABC operon coding for a high-affinity uptake system, whose expression is dependent on the phosphorylation of KdpD/KdpE. We tested growth of this strain in K + -limited (0.1 mmol/L K + ) and K + -rich (115 mmol/L K + ) medium, and compared the results with those for LB2240 (the parental strain), LB2240∆kdpD/pBD5-9 complemented by a plasmid-encoded kdpD and TKV2209 (which carries an additional deletion in kdpE). All strains were able to grow in K + -rich (115 mmol/L K + ) medium (Figure 1a).
When extracellular K + levels are high, nonspecific uptake is sufficient for growth and no specific transporter is required (Laermann et al., 2013). When these strains were exposed to K + limitation (0.1 mmol/L K + ), only those carrying either chromosomally (LB2240) or plasmidencoded kdpD (LB2240∆kdpD/pBD5-9) were able to grow normally ( Figure 1b). Strikingly, however, while strain TKV2209 lacking both the kdpD and kdpE genes was unable to grow under K + limitation, exponential growth of LB2240∆kdpD abruptly set in after an initial lag phase of around 22 hr, and ultimately reached the same optical density as the kdpD + strains (Figure 1b).

| KdpE-mediated induction of kdpFABC expression relieves growth arrest in the absence of KdpD
We hypothesized that KdpE is required to rescue growth of strain We then asked how KdpE can be phosphorylated in the absence of its cognate histidine kinase KdpD. Acetyl phosphate is known to serve as a phosphodonor for KdpE in vitro (Heermann, Altendorf & Jung 2003 LB2240ΔkdpD under K + limitation (Figure 2c). The growth rate of strain LB2240ΔkdpDΔptaΔackA was lower than that of LB2240∆kdpD ( Figure 2c); however, similar effects were observed when the former was grown in K + -rich medium (data not shown). Therefore, we concluded that acetyl phosphate does not act as a phosphodonor for KdpE in vivo under these conditions. F I G U R E 1 Effects of different K + concentrations on the growth of various kdp +/− strains. LB2240 (kdpD + ), LB2240ΔkdpD/pBD5-9 (kdpD − , complemented with plasmid-encoded kdpD), LB2240ΔkdpD (kdpD − ), and TKV2209 (kdpD − , kdpE − ) were cultivated in minimal medium containing the indicated K + concentrations. (a) Growth of strains in K + -rich medium (115 mmol/L K + ). Cells were precultivated in medium containing 115 mmol/L K + , and inoculated into fresh medium at an initial OD 600 of 0.1. Growth was monitored for 24 hr. (b) Growth of strains under K + limitation (0.1 mmol/L K + ). Cells were precultivated in medium containing 115 mmol/L K + , washed with K + -free medium and transferred into medium containing 0.1 mmol/L K + at an initial OD 600 of 0.1. Growth was monitored for 52 hr. The growth curves are representative for at least three biological replicates F I G U R E 2 KdpE-P activates kdpFABC expression independently of KdpD and acetyl phosphate. (a) Growth of the indicated mutants in K + -limited (0.1 mmol/L K + ) and K + -rich (115 mmol/L K + ) minimal medium. Cells were cultivated as described in Figure 1 and samples were taken at the time points indicated. (b) Samples taken at the time points indicated in A were used for qRT-PCR. RNA was extracted and kdpA transcripts were quantified relative to expression of the gapA gene. Mean values of three technical replicates are shown, and are representative for biological duplicates. (The standard deviation was >10%). (c) Growth of strains LB2240, LB2240ΔkdpD, LB2240ΔkdpD, kdpE D52N , and LB2240ΔkdpDΔptaΔackA under K + limitation (0.1 mmol/L K + ). Cells were cultivated as described in Figure 1b. The growth curves are representative for at least three biological replicates

| Only a very small subpopulation of strain LB2240ΔkdpD survives K + limitation
Next, we wanted to know whether the whole population of LB2240ΔkdpD cells is able to adapt to K + limitation, or if only a subpopulation finds a way to induce kdpFABC expression in the absence of KdpD. LB2240ΔkdpD was cultivated as described before in minimal medium containing 115 mmol/L K + . Then about 10 8 cells were spread on plates with minimal medium containing 0.1 mmol/L K + , and incubated at 37°C. On average, five colonies grew from 10 8 cells on the K + -limited plates, whereas cells of strain LB2240 grew as a bacterial lawn (data not shown). This result provided the first hint that suppressor mutations were being generated in strain LB2240ΔkdpD during the long lag phase. If so, the isolated clones should grow under K + limitation without an extended lag phase. To test this prediction, we inoculated the parental strain in K + -limited medium and plated the outgrowing cells on agar plates containing 115 mmol/L K + . Afterward, single colonies were inoculated into K + -limited liquid minimal medium and growth was monitored over time ( Figure 3a). As expected, these single clones (from now on called LB2240ΔkdpD*) were able to grow under K + limitation without an extended lag phase (Figure 3b).

| Mutations in the phosphate transporter PstCAB result in kdpFABC expression
We then set out to identify the suppressor mutation that enables shifts the histidine kinase PhoR into the constitutive kinase "ON" state.
In order to verify the sequencing results, we deleted pstC in LB2240ΔkdpD and tested for growth of the resulting mutant under K + limitation ( Figure 4a). Indeed, we observed that, under K + limitation, the strain carrying the double deletion in kdpD and pstC (LB2240ΔkdpDΔpstC) resumed growth with no lag phase directly after inoculation, whereas the single kdpD deletion led to a growth arrest as described above (Figures 4a, 1b). All strains grew well in K + -rich medium (Figure 4b).
To confirm that a deletion in pstC would result in kdpFABC expression we performed reporter gene assays with strain LF3, in which the kdpFABC promoter is fused to the lacZ gene at the native lacZ gene locus (Fried, Lassak, & Jung, 2012)

| The histidine kinase PhoR is responsible for kdpFABC expression in the absence of KdpD
Signal perception by the histidine kinase PhoR occurs via interaction with the phosphate transporter PstCAB and the negative regulator PhoU (Gardner et al., 2014;Hsieh & Wanner, 2010;Lamarche et al., 2008). As a deletion in phoU or pstCAB shifts PhoR into the kinase   Figure 5c). Taking all these data together, we conclude that PhoR is responsible for phosphorylation of KdpE in the absence of KdpD.

| Bacterial adenylate cyclase two-hybrid experiments indicate in vivo interactions between the two-component systems KdpD/KdpE and PhoR/PhoB
To determine whether components of the two signaling systems interact with each other in vivo, we made use of the bacterial adenylate cyclase two-hybrid system (BACTH). The leucine-zipper fusion constructs zip-T18 and T25-zip from the yeast Saccharomyces cerevisiae were used as positive control and the proteins T18 and T25 alone as negative controls. In the first screen we tested for interactions on LB plates. (Figure 6). The hybrid protein T18-KdpD was found to interact strongly with T25-KdpE, T25-PhoR, and T25-PhoU on LB plates ( Figure 6). There was no detectable interaction between T18-KdpD and the noncognate response regulator hybrid T25-PhoB.
For T18-PhoR the assay revealed interaction with T25-PhoU, but not with T25-PhoB or T25-KdpE. Furthermore, the BACTH test indicated interactions between PhoB-T18 and T25-PhoU, but not between PhoU-T18 and T25-KdpE. Note that we constructed hybrids in all possible combinations, and all cases that yielded negative results were confirmed with the opposite cloning permutation (data not shown). To quantify the strengths of the interactions we determined β-galactosidase activity after culturing cells in liquid medium ( Figure 6). As expected, high β-galactosidase activities were detected F I G U R E 4 pstC deletion mutants induce kdpFABC in the absence of KdpD. (a, b) Growth curves of strains LB2240, LB2240ΔkdpD, LB2240ΔpstC, and LB2240ΔkdpDΔpstC in K + -limited (a) and K + -rich minimal medium (b). Cells were cultivated as described in Figure 1 and growth was monitored for 52 hr. Graphs are representative for two biological replicates. (c, d) β-Galactosidase activities of the reporter strains LF3 (c), LF3ΔpstC (c), LF3ΔkdpD (d), LF3ΔkdpDΔpstC (d), and LF3ΔkdpDΔpstCΔphoR (d). In all strains the native lacZ promoter region was replaced by the kdpFABC promoter region (chromosomal P kdpFABC ::lacZ fusion) and β-galactosidase activities were determined after cultivation of cells in minimal medium containing the indicated concentrations of K + . The plots show means and standard deviations for at least three biological replicates in cells producing T18-KdpD+T25-KdpE and T18-KdpD+T25-PhoR, respectively, and moderate to low activities were measured in cells producing T18-KdpD+T25-PhoU and T18-PhoR+T25-PhoU. No activity was detectable for any other combination. The only discrepancy we observed concerns the PhoB-T18/T25-PhoU pair, for which interaction was signaled when cells were grown on plates, but not in liquid culture ( Figure 6). This result implies that the interaction is very weak and becomes detectable by this assay only after long incubation times and persistent accumulation of β-galactosidase protein. Finally, we did not find any effect of the external K + and PO 3− 4 concentrations on the interaction strengths of the tested constructs (data not shown).
In summary, the BACTH assay indicated that KdpD and PhoR, KdpD and PhoU, and PhoR and PhoU can interact with each other, respectively. These data support the assumption that both PhoR and PhoU can interact with KdpD and influence its activity. In this assay we did not detect interaction of PhoR with PhoB or KdpE, which might be explained by steric hindrance due to the fused adenylate cyclase, or wrong orientation of the two halves of the adenylate cyclase.
Furthermore, it might be that the interaction between PhoR and the histidine kinase is only transient as it was shown before for other histidine kinase/response regulator pairs (Zapf, Sen, Madhusudan, Hoch, & Varughese, 2000).

| Phosphate limitation enhances kdpFABC expression
Thus far, our results indicate cross-talk between KdpE and PhoR/ PhoB in the absence of KdpD. But can it occur in an intact system? If so, it would permit functional coupling of K + and PO 3− 4 homeostasis. To answer this question, we cultivated the reporter strain LF3 in defined TMA medium supplemented with different concentration of K + and PO 3− 4 . As shown before (Fried, Lassak & Jung, 2012) there is basically no induction of kdpFABC in K + -rich medium (5 mmol/L) ( Figure 7a). Under moderate K + limitation (0.5 mmol/L K + ) kdpFABC expression is induced via KdpD. Notably, we found a threefold higher induction when cells were simultaneously exposed to K + and phosphate limitation (50 μmol/L phosphate) (Figure 7a). The kdpD deletion mutant (LF3ΔkdpD) exhibited K + -independent kdpFABC expression, which was also enhanced under phosphate limitation (Figure 7b). However, the observed upregulation was probably not solely dependent on the histidine kinase PhoR as indicated by studies with mutant LF3ΔphoR (Figure 7c). It should be noted, that overall kdpFABC expression was found to be higher in the phoR mutant than in the PhoR + strain (see also Figure 5b). We conclude that kdpFABC expression is fine-tuned under simultaneous phosphate limitation, which indicates cross-regulation between these two systems.

| K + limitation enhances expression of pstS
Having shown that phosphate limitation has an impact on kdpFABC expression, we asked whether K + limitation reciprocally affects the PhoR/PhoB system, which regulates expression of the pho regulon comprising more than 30 target genes, including the pstSCAB operon.
To analyze PhoB activity we used E. coli MG1655 cells transformed with a plasmid-based reporter system, in which the pstS promoter is fused to mcherry. Cells were cultivated in the defined TMA medium F I G U R E 5 The histidine kinase PhoR can phosphorylate KdpE. (a , b) β-Galactosidase activities of the reporter strains LF3 and corresponding deletion mutants. In all strains the native lacZ promoter region was replaced by the kdpFABC promoter region (chromosomal P kdpFABC ::lacZ fusion) and β-galactosidase activities were determined after growth of cells in minimal medium at the indicated K + concentrations. The histograms depict means and standard deviations for at least three biological replicates. (c) In vitro autophosphorylation of PhoR with [γ-32 P]ATP (time points 10 and 20 min). After 20.5 min, PhoB or KdpE was added and phosphotransfer was monitored. Phosphorylated proteins were subjected to SDS-PAGE and gels were exposed to a phosphoscreen. Each autoradiograph is representative for two independent experiments. Band intensity of phosphorylated partner and nonpartner response regulators were quantified and are indicated in percent containing different levels of K + and PO 3− 4 , and growth and fluorescence were monitored over time. Promoter activity was quantified by computing the increase in fluorescence intensity per unit time relative to the optical density of the culture (Bren et al., 2013). We observed an early activation of the pstS promoter under extreme phosphate limitation (5 μmol/L phosphate, maximal induction after 2 hr growth) and delayed activation under moderate PO 3− 4 limitation (50 μmol/L and 200 μmol/L phosphate, maximal induction after 5 and 7.5 hr of growth, respectively) ( Figure 8a). Notably, pstS promoter activity exhibited a slight additional increase when cells were simultaneously exposed to K + limitation, indicating cross-phosphorylation from KdpD to PhoB (Figure 8a). Bacterial growth was clearly determined by phosphate availability, and not by the external K + concentration (Figure 8b).
To confirm cross-regulation from KdpD to PhoB we tested autophosphorylation of KdpD in membrane vesicles and phosphotransfer in vitro. We observed phosphotransfer from KdpD-P not only to KdpE, but to PhoB as well, albeit to a comparably minor extent (Figure 8c).
These results corroborate the idea of cross-regulation between the two signal transduction systems to fine-tune the response depending on K + and phosphate availability.

| DISCUSSION
Bacteria predominantly use two-component signal transduction to adapt to changing environmental conditions (Stock, Robinson, & Goudreau, 2000). A prototypical two-component system consists of a membrane-integrated sensor kinase and a cytoplasmic response regulator that mediates the cellular response. E. coli has at least 30 two-component systems that monitor and respond to an array of environmental and cellular parameters including temperature, extracellular pH and osmolarity, and constituents such as essential nutrients.
F I G U R E 6 Bacterial adenylate cyclase two-hybrid experiments indicate interactions between the two-component systems KdpD/ KdpE and PhoR/PhoB. Fragments T18 and T25 of Bordetella pertussis CyaA were fused to proteins of interest as indicated. The hybrids with yeast leucine zipper fragments were used as a positive control and the fragments T18 and T25 alone as negative control. E. coli BTH101 was cotransformed with plasmid pairs coding for indicated hybrid proteins and cultivated under aerobic conditions. The prefixes T18 and T25 indicate fragments of CyaA N-terminally fused to the protein of interest, the suffixes refer to C-terminal T18 or T25 fusions. For plate assays cells were cultivated in LB medium overnight, washed and subsequently spotted on LB plates. All plates were supplemented with ampicillin, kanamycin, IPTG, and X-Gal as described in Experimental Procedures and were incubated at 25°C for 72 hr. For quantification of β-galactosidase activity cells were cultivated in LB medium supplemented with ampicillin, kanamycin and IPTG as described in Experimental Procedures at 25°C for 48 hr. The activity of the reporter enzyme β-galactosidase was determined and served as a measure of the interaction strength. The histograms show means and standard deviations for at least three biological replicates F I G U R E 7 Phosphate limitation enhances kdpFABC expression. (a, b) β-Galactosidase activities of the reporter strains LF3 (a), LF3ΔkdpD (b) and LF3ΔphoR (c). In all strains the native lacZ promoter region was replaced by the kdpFABC promoter region (chromosomal P kdpFABC ::lacZ fusion) and β-galactosidase activities were determined after cultivation of cells in Tris-maleic acid (TMA) minimal medium at the indicated K + and PO 3− 4 concentrations. For LF3ΔphoR 50 μmol/L phosphate was used. The histograms show means and standard deviations for at least three biological replicates A variety of receptor domains and their corresponding triggers have been identified and underline the importance of the cell's ability to sense and adapt to fluctuating conditions (for review, see (Mascher, Helmann, & Unden, 2006;Krell et al., 2010;Szurmant, White, & Hoch, 2007). However, how specificity is maintained between these signal transduction systems is not fully understood. Podgornaia & Laub (2013) suggested three key mechanisms that could serve to define the specificity of individual two-component signal transduction systems: molecular recognition, phosphatase activity, and substrate competition. According to these authors, the dominant basis for specificity mechanism is molecular recognition, that is the strong kinetic preference of a histidine kinase for its partner response regulator in vitro (Skerker, Prasol, Perchuk, Biondi, & Laub, 2005). Most histidine kinases are bifunctional enzymes, having both kinase and phosphatase activities (Willett & Kirby, 2012). The phosphatase activity is assumed to counteract unspecific phosphorylation by noncognate histidine kinases in vivo (Alves & Savageau, 2003). Although cross-talk between two-component systems has been described several times, it mainly occurs in the absence of either the cognate histidine kinase or response regulator (Fisher, Jiang, Wanner, & Walsh, 1995;Haldimann, Fisher, Daniels, Walsh, & Wanner, 1997;Silva, Haldimann, Prahalad, Walsh, & Wanner, 1998;Siryaporn & Goulian, 2008). Therefore, it is still not clear whether cross-talk between two-component systems is a widespread, but basically incidental phenomenon, or might also be of physiological importance in vivo.
In this study, we demonstrate cross-regulation between the KdpD/ KdpE and PhoR/PhoB two-component systems of E. coli. The KdpD/ KdpE system regulates expression of the high-affinity K + uptake system KdpFABC. We found that, while a mutant lacking KdpD and the two constitutively produced low-affinity K + uptake systems Trk and Kup are in principle unable to grow under K + limitation, a subpopulation emerges after a 22-hr lag phase, which subsequently expands at a virtually wild-type rate (Figure 1). Recovery of growth in these cells was dependent on a phosphorylatable KdpE and induction of kdpFABC expression. Whole-genome sequencing revealed suppressor mutations in the pstC and pstB genes and further experiments confirmed that a deletion in pstC or phoU-all of which are supposed to switch the histidine kinase PhoR into the constitutive kinase "ON" state-induced kdpFABC expression in the absence of KdpD. These observations provide compelling evidence for cross-regulation between these two systems. It has been shown previously that, in the absence of PhoR, the response regulator PhoB can be activated by noncognate histidine kinases, including KdpD (Zhou et al., 2005). In addition, an indirect link between these two-component systems has been described recently: Nonphosphorylated enzyme IIA of the Ntr-PTS system in E. coli has been shown to interact with both KdpD and PhoR, and to stimulate their activities, coupling carbon metabolism with K + and phosphate homeostasis, respectively (Lüttmann et al., 2009).
Our results thus reveal bidirectional cross-talk between the kdp-FABC and pstSCAB operons. We found that both histidine kinases can F I G U R E 8 K + limitation enhances pstS expression, and PhoB phosphorylation by KdpD. (a) pstS promoter activity in E. coli MG1655 cells carrying plasmids in which mcherry expression is under the control of the pstS promoter. Cells were cultivated in Tris-maleic acid (TMA) minimal medium containing the indicated K + and PO 3− 4 concentrations. Shown is the maximal promoter activity, which was observed after 2 hr at 5 μmol/L Na 2 HPO 4 , 5 hr at 50 μmol/L Na 2 HPO 4 , and after 7.5 hr at 200 μmol/L Na 2 HPO 4 . Shown is the mean and standard deviation of three independent experiments. (b) Corresponding growth curves of strains cultivated in Tris-maleic acid (TMA) minimal medium containing different K + and PO 3− 4 concentrations as indicated in green and blue symbols refering to K + -limited (0.5 mmol/L) and K + -rich (5 mmol/L) medium, respectively. Circles, triangles, and squares depict the indicated Na 2 HPO 4 concentration. Shown is the mean and standard deviation of three independent experiments. (c) In vitro autophosphorylation of KdpD with [γ-32 P]ATP (time points 10 and 20 min). After 20.5 min PhoB or KdpE was added and phosphotransfer was monitored. Phosphorylated proteins were subjected to SDS-PAGE and gels were exposed to a phosphoscreen. Each autoradiograph is representative for two independent experiments. Band intensity of phosphorylated partner and nonpartner response regulators were quantified and are indicated in percent reciprocally phosphorylate the corresponding noncognate response regulator in vitro, and we detected fine-tuned cross-regulation of their target genes in vivo. We therefore propose that PhoR activates KdpE and KdpD activates PhoB when cells are simultaneously exposed to both K + and phosphate limitation (Figure 9). Under phosphate limitation, PhoR autophosphorylates and transfers the phosphoryl group to PhoB, which in turn activates transcription of the pho regulon.
Moreover, PhoR also phosphorylates KdpE. However, as intra-and extracellular K + levels are sufficient, the counteracting phosphatase activity of KdpD ensures that KdpE is dephosphorylated and thus prevents kdpFABC expression. Under simultaneous K + limitation, however, KdpD is in the kinase active state and phosphorylates KdpE as well as PhoB, thereby boosting K + uptake and transcription of the pho regulon. In addition, our BACTH results suggest interactions between KdpD and PhoR, as well as between KdpD and PhoU. Formation of heterodimers between KdpD and PhoR is an attractive, but as yet untested, hypothesis that could account for these results. Besides direct cross-phosphorylation between these two two-component systems, our results also suggest that other still unknown components might play a role in the cross-regulation of these operons. While the increased kdpFABC expression in strains lacking kdpD was clearly determined by PhoR (Figure 4d and 5a), PhoR had a minor influence in KdpD + cells (Figure 5b). Moreover, cross-regulation of kdpFABC expression under concurrent phosphate and K + limitation was probably not solely dependent on PhoR (Figure 7c). Therefore, it remains still unclear whether interactions between PhoR, KdpD, and PhoU influence the phosphotransfer to the response regulators or whether other regulatory components-unrelated to the Kdp and Pho two-component systems-are responsible for regulating expression of the kdpFABC operon and the pho regulon under these stress conditions.
In 1967, it was reported that K + is important for PO 3− 4 uptake (Weiden et al., 1967), but the molecular basis for this phenomenon remained unclear. The cross-connections between the two uptake systems described here not only provide a possible explanation, but uncover an elegant cellular mechanism for fine-tuning the ratio of positively and negatively charged ions in the cytoplasm. It should be noted that during the preparation of this manuscript Moreau & Loiseau (2016) published a study about suppressor mutants generated under phosphate starvation. Interestingly one of the mutations the authors identified was located in the kdpD gene and resulted in a constitutively active KdpD protein. According to our model, a constitutively active KdpD protein rescues growth of the mutant under phosphate starvation by directly activating PhoB.
In summary, we demonstrate that cross-regulation between the kdpFABC and pstSCAB operons occurs under conditions of K + and phosphate limitation. This cross-regulation interconnects K + and phosphate homeostasis in E. coli and fine-tunes the ratio of positively and negatively charged ions within cells.
F I G U R E 9 Cross-regulation of the kdpFABC operon and the pho regulon couples K + to PO 3− 4 homeostasis in E. coli. KdpD and PhoR phosphorylate their cognate response regulators KdpE and PhoB, respectively. Under simultaneous K + and phosphate limitation an additional increase in target gene expression can be achieved by direct cross-regulation between the two-component systems and so far unknown regulatory components