The phosphohistidine phosphatase SixA dephosphorylates the phosphocarrier NPr

Histidine phosphorylation is a posttranslational modification that alters protein function and also serves as an intermediate of phosphoryl transfer. Although phosphohistidine is relatively unstable, enzymatic dephosphorylation of this residue is apparently needed in some contexts, since both prokaryotic and eukaryotic phosphohistidine phosphatases have been reported. Here we identify the mechanism by which a bacterial phosphohistidine phosphatase dephosphorylates the nitrogen-related phosphotransferase system, a broadly conserved bacterial pathway that controls diverse metabolic processes. We show that the phosphatase SixA dephosphorylates the phosphocarrier protein NPr and that the reaction proceeds through phosphoryl transfer from a histidine on NPr to a histidine on SixA. In addition, we show that Escherichia coli lacking SixA are outcompeted by wild-type E. coli in the context of commensal colonization of the mouse intestine. Notably, this colonization defect requires NPr and is distinct from a previously identified in vitro growth defect associated with dysregulation of the nitrogen-related phosphotransferase system. The widespread conservation of SixA, and its coincidence with the phosphotransferase system studied here, suggests that this dephosphorylation mechanism may be conserved in other bacteria.

Histidine phosphorylation is a posttranslational modification that alters protein function and also serves as an intermediate of phosphoryl transfer. Although phosphohistidine is relatively unstable, enzymatic dephosphorylation of this residue is apparently needed in some contexts, since both prokaryotic and eukaryotic phosphohistidine phosphatases have been reported. Here we identify the mechanism by which a bacterial phosphohistidine phosphatase dephosphorylates the nitrogen-related phosphotransferase system, a broadly conserved bacterial pathway that controls diverse metabolic processes. We show that the phosphatase SixA dephosphorylates the phosphocarrier protein NPr and that the reaction proceeds through phosphoryl transfer from a histidine on NPr to a histidine on SixA. In addition, we show that Escherichia coli lacking SixA are outcompeted by wild-type E. coli in the context of commensal colonization of the mouse intestine. Notably, this colonization defect requires NPr and is distinct from a previously identified in vitro growth defect associated with dysregulation of the nitrogen-related phosphotransferase system. The widespread conservation of SixA, and its coincidence with the phosphotransferase system studied here, suggests that this dephosphorylation mechanism may be conserved in other bacteria.
SixA (Signal Inhibitory factor X) is a well-conserved bacterial protein (38) belonging to the histidine phosphatase superfamily (1,39,40). The "histidine" in this superfamily name refers not to substrate specificity but rather to the conserved residue in the active site that is essential for enzyme activity and that becomes phosphorylated during catalysis. SixA was first reported to be a phosphohistidine phosphatase that dephosphorylates the histidine-containing phosphotransfer domain of the E. coli sensor kinase ArcB (32,41), and the activity was shown to be dependent on the conserved activesite histidine (His8) of SixA. However, beyond the two reports that proposed this activity 2 decades ago, we are unaware of any other publications describing SixA regulation of ArcB. In fact, at least two subsequent studies failed to find an effect of SixA on ArcB-regulated transcription (33,42). Recently, analysis of a growth defect of a SixA-null strain identified a different regulatory pathway that appears to be a physiological target of SixA (33): the nitrogen-related phosphotransferase system.
In contrast to the classical carbohydrate phosphotransferase systems, the nitrogen-related phosphotransferase system does not participate in sugar phosphorylation and uptake, but, like the carbohydrate systems, it regulates a diverse set of cellular processes, including potassium homeostasis, nitrogen and carbon metabolism, the stringent response, and twocomponent signaling (10,43). In most cases, regulation is mediated by phosphorylation-dependent protein-protein interactions (2,10,(43)(44)(45). In E. coli, the nitrogen-related phosphotransferase system consists of a phosphotransferase, EI Ntr , and two phosphocarriers, NPr and EIIA Ntr (Fig. 1A). Together, these three proteins perform reversible phosphoryl transfer reactions via histidine residues in a similar fashion to their counterparts in carbohydrate phosphotransferase systems: a phosphoryl group is first transferred from phosphoenolpyruvate to EI Ntr , next from EI Ntr to NPr, and finally from NPr to EIIA Ntr .
Whereas carbohydrate phosphotransferase systems donate phosphoryl groups to incoming carbohydrates or acceptor proteins, the fate of the phosphoryl groups that pass through the nitrogen-related phosphotransferase system is not understood in most bacteria. Two exceptions are Acinetobacter baumannii, in which phosphoryl groups can be transferred from NPr to a serine residue of the protein GigB and subsequently removed by the phosphoserine phosphatase GigA (46), and Sinorhizobium meliloti, in which an aspartic acid residue A C B D Figure 1. E. coli NPr is hyperphosphorylated when SixA is absent. A, schematic of phosphoryl transfer by the nitrogen-related phosphotransferase system, with a focus on sources of NPr phosphorylation and dephosphorylation. Solid arrows denote paths of phosphoryl transfer by EI Ntr , NPr, EIIA Ntr , and SixA, which all occur though phosphohistidine intermediates. The genes for each protein of the phosphotransferase system are indicated in parentheses. Cross talk refers to phosphorylation of NPr and EIIA Ntr by other pathways. B, representative anti-E-Tag western blot of cell lysates separated by native gel electrophoresis. The asterisk indicates additional NPr-E-Tag-dependent bands (see text for discussion). Strains used in this experiment were MG1655, JES287, JES288, JES314, and JES315. C, band intensities of NPr and NPr-P from (B) and two additional blots were quantified with ImageJ (74); %NPr-P = 100 × NPr-P/(NPr + NPr-P). Columns of the bar graph indicate average %NPr-P for strains from the three independent experiments, error bars show one standard deviation, and symbols represent values from individual cultures. D, representative anti-E-Tag western blot of cell lysates from strains expressing NPr-E-Tag and either wild-type SixA or a catalytic mutant of SixA. Expression of SixA proteins was from leaky (i.e., uninduced) transcription from the trc promoter on a plasmid. Strains used were JES288 (ΔsixA NPr-E-Tag + ) transformed with empty vector pTrc99a, SixA-expressing plasmid pSixA, or SixA(H8A)-expressing plasmid pSixA(H8A).
near the phosphorylation site of NPr (called HPr in this organism) is proposed to heighten the autohydrolysis rate (47). While it is possible that autohydrolysis also plays a role in E. coli, recent work suggests that SixA provides another mechanism for dephosphorylating the nitrogen-related phosphotransferase system (33). In the absence of SixA, EIIA Ntr becomes hyperphosphorylated. In addition, the reduction of EIIA Ntr phosphorylation by SixA requires NPr. These observations, together with epistasis studies of a ΔsixA growth defect, led to the proposal that phosphorylated NPr (NPr-P) is a substrate for SixA.
Here we establish through in vivo and in vitro experiments that SixA is a phosphohistidine phosphatase for NPr-P. The NPr-P dephosphorylation reaction proceeds through an unstable phosphohistidine intermediate of SixA. We also show that the SixA-NPr interaction is important for E. coli to colonize its natural niche, the mammalian intestinal tract. These results establish a substrate for a highly conserved prokaryotic phosphohistidine phosphatase and identify a mechanism for modulating the activity of the nitrogen-related phosphotransferase system.

SixA regulates NPr phosphorylation in vivo
To determine the effect of SixA on NPr phosphorylation, we analyzed strains expressing a tagged NPr (E-Tag) by western blot after native gel electrophoresis. Under nondenaturing conditions, the phosphorylated and nonphosphorylated forms of the protein are resolved into two bands, with the phosphorylated form corresponding to higher mobility (48), (Fig. 1B, lanes 2-5). Comparison of the results for the sixA + and ΔsixA strains indicates that the fraction of NPr protein that is phosphorylated is increased when SixA is absent (Fig. 1B, lanes 2 and 3; Fig. 1C). We also note that there are several additional bands (with lower mobility) in the lanes with E-tagged NPr that are absent in the lane with untagged NPr (Fig. 1B, compare lanes 2-5 with lane 1). We do not know the identity of these additional bands, but they may indicate complexes of NPr with other proteins.
NPr is phosphorylated by EI Ntr and also by reverse phosphotransfer from EIIA Ntr , which can be phosphorylated by cross talk from other phosphotransferase systems in mutant backgrounds (49) (Fig. 1A). Cross talk to NPr directly is also plausible (50) but, to our knowledge, has never been investigated in vivo. We therefore examined the effect of deleting sixA on NPr phosphorylation in strains that lack EI Ntr and EIIA Ntr . A ΔptsP ΔptsN strain (EI Ntr -null, EIIA Ntr -null) had no detectable NPr-P (Fig. 1B, lane 4). In contrast, the tripledeletion ΔsixA ΔptsP ΔptsN showed considerable NPr phosphorylation (Fig. 1B, lane 5). These results establish that NPr can be phosphorylated by other pathways (possibly carbohydrate phosphotransferase systems) when EI Ntr , EIIA Ntr , and SixA are absent and further establish that SixA does not require EI Ntr and EIIA Ntr to modulate NPr phosphorylation.
We also tested whether SixA's active-site histidine is required for the protein to affect NPr phosphorylation by comparing the effects of expressing wild-type SixA or a Six-A(H8A) mutant. Wild-type SixA expression eliminated NPr-P, whereas expression of SixA(H8A) did not (Fig. 1D). Collectively, the above results show that SixA modulates NPr phosphorylation independently of EI Ntr and EIIA Ntr , and they support the hypothesis that SixA directly dephosphorylates NPr-P.
SixA dephosphorylates NPr-P in vitro and forms a transient phosphohistidine intermediate To study the activity of SixA against NPr-P in vitro, we used purified NPr that consisted of the complete amino acid sequence of the protein plus eight additional residues at the Cterminus (see Experimental procedures). We will refer to this protein as "wild-type" NPr. We phosphorylated NPr by incubating with EI Ntr and phosphoenolpyruvate. As expected (51,52), phosphorylation required His16 of NPr ( Fig. 2A, compare lanes 4 and 8). Curiously, the H16A mutant ran as a doublet on native gels; we will discuss this observation further below. Addition of purified SixA to NPr-P resulted in complete dephosphorylation of the protein, whereas the catalytic mutant SixA(H8A) had no observable effect on NPr-P level (Fig. 2B).
For the experiments described above and below, the polyhistidine tag used to purify NPr was cleaved from the protein.
We had noticed that phosphorylation reactions of NPr-(His) 6 , a protein consisting of the complete amino acid sequence of NPr plus the eight additional residues LEHHHHHH, produced multiple bands ( In addition, incubation with SixA eliminated these bands, leaving only the band corresponding to unphosphorylated NPr-(His) 6 ( Fig. 2D). To determine whether these multiple bands correspond to multiple phosphorylation states of the protein, we analyzed NPr-(His) 6 phosphorylation and mockphosphorylation samples by mass spectrometry. Monoisotopic masses of phosphorylated NPr-(His) 6 measured by electrospray ionization MS matched the masses for singly and doubly phosphorylated protein (Fig. S1).
Unfortunately, LC-MS/MS produced spectra of insufficient quality to assign phosphorylation to specific residues (data not shown), despite efforts to preserve acid-labile histidine phosphorylation. Since the extra bands appear only for the construct with the polyhistidine tag, and since these bands were eliminated after incubation with SixA ( Fig. 2D), we suspect that one or more histidines in the (His) 6 tag were phosphorylated. We further suspect that the (His) 6 tag is phosphorylated by intramolecular phosphoryl transfer, because, first, the site of NPr phosphorylation, His16, is adjacent to the C-terminus of NPr (53,54), and, second, NPr(H16A)-(His) 6 shows no evidence of phosphorylation (Fig. 2C, lane 7). We are unaware of previous reports of histidine phosphorylation of polyhistidine tags, but we speculate that this reaction may occur if the polyhistidine tag is near a site of histidine phosphorylation of the native protein.
As noted above, NPr(H16A) protein preparations ran as doublets on native gels ( Fig The substrate of a phosphohistidine phosphatase forms of the protein because NPr(H16A) runs as a single band on SDS-PAGE (Fig. S2) and analysis of NPr(H16A) by electrospray ionization MS identified only a single species (data not shown). Although wild-type NPr protein preparations primarily ran as a single band with mobility similar to the higher mobility band of NPr(H16A), a faint lower mobility band for the wild-type protein was also discernable (for example Fig. 2A, lanes 1-3; Fig. 2C, lanes 1-4). The doublet on native gels may therefore indicate that our recombinant NPr has two stable (or quasi-stable) conformations. Indeed, previous work has also reported evidence for two conformations of recombinant NPr (54). Based on NMR spectroscopy, it was suggested that the two protein forms could be due to interactions with a disordered C-terminal tail containing additional residues not found in native NPr. Thus, it is possible that the doublet we observe on native gels is also due to the presence of additional residues at the C-terminus of the NPr proteins described above.
To follow the time course of the NPr-P dephosphorylation reaction, we turned to denaturing gels so that reactions could be quenched with denaturing loading dye. To detect phosphorylated NPr, we used antibodies against phosphohistidine. Histidine can be phosphorylated on either of the two nitrogen atoms of its imidazole ring, forming phosphoramidates called π-phosphohistidine (π-pHis) and τ-phosphohistidine (τ-pHis) (Fig. 3) (23,24,26). Based on western blots with phosphohistidine-isomer-specific antibodies, NPr-P consists primarily of π-pHis (Fig. 3). Following addition of SixA, NPr-P showed progressive loss of the π-pHis band (Fig. 3, 3 min-27 min). This decrease in π-pHis required addition of SixA and was not observed for mutant SixA(H8A) (Fig. 3, lanes 2 and 3). In addition, a band at the expected location for SixA appeared on the anti-τ-pHis blot at the time points corresponding to NPr-P dephosphorylation, and this band disappeared as NPr-P was dephosphorylated (Fig. 3, 3 min-27 min). These results indicate that NPr-P dephosphorylation by SixA proceeds via a transient SixA phosphohistidine intermediate. The anti-τ-pHis blot also showed a faint band at the expected location of NPr (Fig. 3, asterisk). In addition, the anti-π-pHis blot shows very faint bands at the position for NPr-P even at the latest time points in the blot. These weak bands were still detectable when samples were boiled to eliminate phosphohistidine (Fig. S3), suggesting that they may reflect weak antibody binding to unphosphorylated NPr rather than true phosphohistidine signal. Interestingly, we do not detect a comparable signal when the sole histidine residue of NPr is mutated to alanine (Fig. S3), which could indicate weak crossreactivity of the antibodies with unphosphorylated histidine on NPr.
From the in vitro results presented above, we conclude that SixA directly dephosphorylates NPr-P, and the reaction proceeds by phosphoryl transfer from π-pHis on NPr to τ-pHis on SixA. The substrate of a phosphohistidine phosphatase A sixA deletion has an NPr-dependent colonization defect in the mouse intestine Since animal intestinal tracts are one of the primary niches for E. coli, we wondered whether SixA is important for E. coli fitness in this environment. We took advantage of a previously developed system for studying E. coli intestinal colonization (55) that uses a mouse E. coli commensal isolate, MP1. This strain is genetically tractable and does not require continuous antibiotic treatment for long-term persistence in the mouse intestine. We found that the ΔsixA strain had a colonization defect in competition with wild-type (Fig. 4A). Six weeks after orogastric gavage of a suspension of wild-type and mutant bacteria, counts of the ΔsixA strain were below the detection limit of our assay for all mice. In contrast, the wild-type strain stably colonized mice for 12 weeks.
Based on the results above, we hypothesized that the ΔsixA colonization defect is due to hyperphosphorylation of NPr (encoded by ptsO). We therefore tested whether deletion of sixA also affects colonization of an NPr-null strain. In competition experiments, counts of ΔptsO and ΔptsO ΔsixA strains were comparable for four of the five mice 12 weeks after gavage (Fig. 4B). These results are consistent with our conclusion above that SixA dephosphorylates NPr-P and also with the hypothesis that the colonization defect is due to aberrant NPr-P dephosphorylation.
Previously, we showed that the absence of SixA causes a growth defect in minimal medium but not in rich medium (33) (see also Fig. S4). The slow-growth phenotype is suppressed by deleting ycgO, which encodes a putative cation-proton antiporter that has been proposed to be regulated by the nitrogenrelated phosphotransferase system (56). To test whether the ΔsixA mouse colonization defect also involves YcgO, we performed competitions between ΔycgO and ΔycgO ΔsixA strains. Six weeks after gavage, counts of the ΔycgO ΔsixA strain were The substrate of a phosphohistidine phosphatase below the detection limit of our assay, whereas the ΔycgO strain continued to stably colonize mice (Fig. 4C). From these results we conclude that the ΔsixA mouse colonization defect is not related to the YcgO-dependent slow-growth phenotype of ΔsixA E. coli in minimal medium. Taken together, our results suggest that the colonization defect arises from the absence of SixA dephosphorylation of NPr-P. The resulting hyperphosphorylation of one or more members of the nitrogen-related phosphotransferase system then leads to misregulation of a protein target other than YcgO.

Discussion
Here we have demonstrated that SixA dephosphorylates the phosphocarrier NPr, establishing a phosphatase-based mechanism for controlling the phosphorylation state of the nitrogen-related phosphotransferase system. Across different bacteria, this phosphotransferase system has been shown to regulate diverse metabolic pathways and to affect host colonization by pathogens and symbionts (10,43,57,58). Our mouse experiments, which revealed that eliminating SixA impaired E. coli colonization though an NPr-dependent mechanism, provide another example of a role for the nitrogen-related phosphotransferase system in host-microbe interactions. We note that this ΔsixA mouse colonization defect is characterized by a gradual decrease in mutant bacteria over time, and we cannot rule out the possibility that this phenotype is due to a subtle growth defect that also exists outside of an animal host. However, this colonization defect is distinct from a previously described growth defect of the ΔsixA mutant in vitro. This latter defect, which we observe in minimal medium but not in rich medium, is suppressed by deleting ycgO (33), a gene that is predicted to encode a protoncation antiporter. The mouse colonization defect, on the other hand, is not affected by ΔycgO (Fig. 4C).
SixA homologs are distributed across many phyla and are especially well represented among Proteobacteria, Cyanobacteria, and Actinobacteria (38). Within Proteobacteria, SixA sequences tend to cluster together by taxonomic class (Fig. 5). In addition, some bacterial genomes that encode SixA homologs appear to lack genes encoding phosphocarrier proteins (NPr, HPr, or proteins containing an NPr/HPr-like domain), e.g., Bdellovibrio bacteriovorus and members of Epsilonproteobacteria (Fig. 5) as well as Cyanobacteria and Mycobacterium species. Therefore, at least some SixA homologs, if they function as phosphatases, must target substrates other than phosphocarrier proteins. Of course, it is also possible that E. coli SixA has other targets in addition to NPr.
Using antibodies specific to either τ-pHis or π-pHis isomers, our western blot analyses indicate that SixA dephosphorylates π-pHis of NPr (Fig. 3). Studies of NPr homologs have previously observed that these proteins are phosphorylated at the πposition (59, 60). Our results also revealed that during the dephosphorylation reaction, SixA reacts with τ-pHis antibody (Fig. 3). This finding is in agreement with predictions that catalysis by SixA proceeds through a τ-pHis intermediate (38,39). Together, these two observations indicate that phosphoryl transfer from NPr to SixA involves switching from π-pHis (on NPr) to τ-pHis (on SixA). Interestingly, several other phosphoryl transfer reactions between histidine residues also involve switching between phosphohistidine isomers (τ-pHis → π-pHis, or π-pHis → τ-pHis). For example, phosphoryl transfer through the bacterial carbohydrate phosphotransferase system proceeds via the sequence: τ-pHis (EI) → π-pHis (HPr) → τ-pHis (EIIA) (see (45) for references and additional phosphotransferase system examples). In addition, in mammalian cells, π-pHis → τ-pHis phosphoryl transfer is observed when the phosphohistidine phosphatase PGAM5 dephosphorylates NDPK-B (31) and also when NDPK-B phosphorylates the potassium channel KCa3.1 (12). More examples can be found in (61,62). From our survey of the phosphohistidine literature, we were unable to find examples of π-pHis → π-pHis or τ-pHis → τ-pHis phosphoryl transfer. It is an interesting question whether the above pattern of The substrate of a phosphohistidine phosphatase switching between the π-pHis and τ-pHis isomers arises from a general mechanistic constraint on pHis → pHis phosphoryl transfer or instead reflects some aspect of the evolution of the associated pathways.
Although SixA was discovered 2 decades ago, phosphohistidine phosphatases remain relatively understudied. In addition to the target for SixA described here, physiological targets of the mammalian phosphohistidine phosphatases PHPT1, PGAM5, and LHPP have only recently been uncovered (21,(29)(30)(31). Further study of these four proteins, as well as the discovery of additional phosphohistidine phosphatase substrates, will reveal new biological processes that exploit histidine phosphorylation in both prokaryotes and eukaryotes.

Growth conditions
Bacterial cultures were grown at 37 C with aeration in Lysogeny Broth (LB) Miller medium (66), which contains (per Figure 5. Patterns of co-occurrence between SixA and NPr homologs in Proteobacteria species. SixA protein sequences from a selection of Proteobacteria were used to construct a tree as described in Experimental procedures. The black and white squares on the right indicate whether a species has proteins with homology to NPr/HPr-presence (black) or absence (white) of NPr/HPr-like proteins. The Proteobacteria class of each species is indicated at the far right by the Greek letters (alpha, beta, gamma, delta, epsilon) or, for Oligoflexia, the letter "O." The substrate of a phosphohistidine phosphatase liter) 10 g tryptone, 5 g yeast extract, and 10 g NaCl (catalog no. BP1426-500; Fisher Scientific). Antibiotics were used at the following concentrations (in μg/ml): ampicillin, 100; kanamycin, 25 or 50; chloramphenicol, 12; and tetracycline, 15. Strains expressing proteins from the trc promoter (P trc ) were grown in the absence of inducer (IPTG; isopropyl-β-D-thiogalactopyranoside); expression was based on leaky transcription from P trc .

Strain construction
Antibiotic-resistance cassettes flanked by FLP recombination target (FRT) sites were excised from bacterial genomes using FLP recombinase expressed from pCP20 (67). Phage transduction was performed using P1 vir (66). Recombineering was performed following previously published protocols (68) with PCR fragments and strains transformed with pKD46, as described in more detail below.
JES286 is a strain with a C-terminal epitope sequence (E-Tag) fused to the NPr gene ptsO at its chromosomal locus in E. coli MG1655. A PCR fragment, containing the end of ptsO fused to a sequence encoding the E-Tag (GAPVPYPD-PLEPR) and also followed by a FLP-recombinase-excisable cat gene, was constructed by PCR with the template pKD3 and with primers npr-Etag-lred-F/R. The resulting DNA segment was electroporated into strain MG1655/pKD46, and cells were selected on chloramphenicol. The ptsO locus of JES286 was confirmed by DNA sequencing. To ensure selection of both the ptsN deletion and the ptsO E-Tag fusion during construction of JES314 and JES315, transductants were selected on agar plates containing both chloramphenicol and kanamycin.
MP1 deletion strains were constructed by electroporating MP1/pKD46 with PCR products and selecting for mutants on kanamycin. Electrocompetent MP1/pKD46 was prepared as previously described (55), except cells were washed with an ice cold solution containing 10% glycerol and 1 mM unbuffered 3-(N-morpholino) propanesulfonic acid (MOPS). To construct the sixA-deletion strain, MP203, a PCR fragment containing FRT-kan-FRT was amplified from strain JW2337 with primers sixA-del-U1/L1. To construct the ptsO-deletion strain, MP263, a PCR fragment containing FRT-kan-FRT was amplified from pKD13 with primers MP1-npr-lred-u/d2. Finally, to construct the ycgO-deletion strain, MP264, a PCR fragment containing FRT-kan-FRT was amplified from pKD13 with primers MP1-ycgO-lred-u/d2. Deletions in MP1 were confirmed by PCR and moved into the MP1 derivatives by P1 transduction as indicated in Table S1. Out of convenience for strain construction, the ptsOand ycgO-deletion strains used in mouse competitions were not constructed in the order that would ensure the closest possible common ancestor between the sixA + and sixAderivatives (Table S1).

Assays of NPr phosphorylation in vivo
Cell pellets from 2 ml of stationary-phase cultures were used to prepare cleared cell lysates by either (1) resuspending pellets in 0.5 ml of loading dye (10% glycerol, 40 mM glycine, 5 mM Tris HCl, 0.005% bromophenol blue, pH 8.8), sonicating suspensions on ice, and centrifuging cell lysates to pellet cell debris or (2) resuspending pellets in 0.25 ml B-PER (Bacterial Protein Extraction Reagent; catalog no. 78248; Thermo Fisher Scientific), incubating at room temperature for 15 min, and mixing cleared lysate with an equal volume of loading dye. Samples were then analyzed by native gel electrophoresis on 4% to 20% Mini-PROTEAN TGX precast protein gels (catalog no. 4561096; Bio-Rad Laboratories) with running buffer consisting of 25 mM Tris, 192 mM glycine, pH 8.3. Electrophoresis was performed at room temperature or 4 C until the loading dye reached the end of the gel. Protein transfer to 0.45μm-pore-size Immobilon-P PVDF (polyvinylidene difluoride) membrane (catalog no. IPVH00010; Millipore) was performed with transfer buffer containing 20% methanol, 25 mM Tris, 192 mM glycine, pH 8.3.

Expression plasmid construction
To construct the NPr-(His) 6 expression plasmid pJS43, primers NdeI-ptsO-F and XhoI-ptsO-R were used to amplify the ptsO gene from MG1655 genomic DNA. The resulting DNA segment was digested with NdeI and XhoI and cloned into pET-22b(+), which was digested with the same enzymes. A plasmid expressing the H16A ptsO mutant was constructed by site-directed mutagenesis: plasmid pJS43 was amplified with primers npr-H16A-F/R, the ends were phosphorylated with T4 polynucleotide kinase, and the resulting DNA was circularized by blunt-end ligation with T4 DNA ligase, generating pJS47. NPr-(His) 6 and NPr(H16A)-(His) 6 expressed from these plasmids consist of the complete NPr sequence plus the additional residues LEHHHHHH at the C-terminus.
To purify NPr and SixA proteins with cleavable polyhistidine tags, genes were cloned into the pET-41 vector backbone using NEBuilder HiFi DNA Assembly (New England BioLabs). Each fusion protein contains the complete protein sequence plus the additional residues GTEN-LYFQGSTMDHHHHHHHH at the C-terminus. After removing the polyhistidine tag with TEV (tobacco etch virus) protease, the residues GTENLYFQ remain at the C-terminus.
The insertions in all engineered plasmids were verified to be correct by DNA sequencing.

Protein purification
Proteins were expressed in BL21(DE3) and purified using Ni-NTA agarose (catalog no. 30210; QIAGEN) following the manufacturer's protocols. Protein expression was induced with 1 mM IPTG at mid-to late-exponential phase and grown for an additional 3 to 4 h at 37 C. Cells were lysed by sonication on ice or by suspension in B-PER. Cleavable histidine tags were removed by incubation with His-tagged TEV protease (catalog no. T4455; Sigma), and tag-free protein was purified from the reaction by incubating with Ni-NTA agarose and recovering the flow-through. Proteins were transferred into storage buffer (50 mM Tris HCl, 150 mM NaCl, 1 mM DTT, 10% glycerol, pH 8) by either dialysis or with centrifugal filter units (3 kDa cutoff; Amicon UFC800308, Millipore-Sigma). The storage buffer for NPr proteins also contained 5 mM MgCl 2 . Protein concentrations were estimated with the Bradford method (Bio-Rad Protein Assay 500-0006, Bio-Rad Laboratories) using a standard curve with BSA. Proteins were stored at −80 C.

NPr phosphorylation assay
To prepare phosphorylated NPr, phosphoenolpyruvate, NPr, and EI Ntr (a gift from Alan Peterkofsky) were first incubated individually in reaction buffer (50 mM Tris HCl, 5 mM MgCl 2 , 2 mM DTT, pH 8) at 30 C for 15 min (46). After preincubation, reaction components were then mixed together and incubated for 30 min at 30 C. Protein and phosphoenolpyruvate concentrations in phosphorylation reactions are indicated in the figure captions.

NPr-P dephosphorylation assay
For the dephosphorylation reactions shown in Figure 2, aliquots of NPr phosphorylation reactions were mixed with SixA, SixA(H8A), or buffer alone and incubated at 30 C for an additional 20 min before analysis by native gel electrophoresis.
Quenched samples were separated by electrophoresis on duplicate gels as described above, except that denaturing conditions were used with SDS-PAGE running buffer (0.1% SDS, 25 mM Tris, 192 mM glycine, pH 8.3). After protein transfer, the two membranes were blocked in 3% milk TBST. Each membrane was probed with only one of the two antiphosphohistidine antibodies (catalog no. ZRB1330, RRID: AB_2868462; catalog no. ZRB1352, RRID: AB_2868463; Millipore-Sigma), which were diluted in TBS (TBST lacking Tween) to 0.6 μg/ml. ECL detection was performed as described above.

Mouse colonization
All animal studies were performed in accordance with animal protocols approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. Ten-weekold female C57BL/6 mice (The Jackson Laboratory) were raised under standard conditions. Bacterial suspensions for mouse inoculation were prepared as described previously (55). The suspension, containing a 1:1 mixture of two bacterial strains, was used to inoculate five cohoused mice by orogastric gavage with approximately 10 9 bacteria. The ratio of ΔsixA to sixA + bacteria in the solution used to gavage mice ranged between 0.7 and 1.1. Colony forming units (CFUs) were determined by stool serial dilution plating on selective agar media, as previously described (55). In brief, stool dilutions were plated on LB agar plates containing tetracycline to induce fluorescent protein production. Colonies were imaged with a home-built fluorescence imaging system, and the competing strains were distinguished by mCherry or GFP fluorescence. When ΔsixA strains were less abundant at later time points, stool suspensions were also plated on LB agar plates containing both tetracycline and kanamycin to count the GFP + kanamycin-resistant colonies of ΔsixA strains.

SixA phylogenetic tree
SixA phylogenetic tree MEGA-X (69) and NCBI BLAST (70) were used to prepare a maximum likelihood tree of SixA The substrate of a phosphohistidine phosphatase sequences as described in (71). The tree was visualized and annotated using iTOL (72). SixA homologs were identified based on the amino acid sequence determinants described in (38). NPr/HPr homologs or NPr/HPr-like domains in fusion proteins were identified based on the well-conserved signature sequence surrounding the active-site histidine residue (73). Table S4 contains the strains, NCBI taxonomic identifier, and protein accession information for SixA homologs and NPr/ HPr homologs, if present. In the cases where more than one SixA homolog or NPr-like protein is present in a bacterial strain, only one representative sequence was taken.

Data availability
Requests for data are to be addressed to the corresponding author (Mark Goulian, goulian@sas.upenn.edu).
Acknowledgments-We thank Alan Peterkofsky for providing us with purified EI Ntr protein. We also wish to thank Kushol Gupta, Hsin-Yao Tang, Thomas Beer, Steve Seeholzer, and Lynn Spruce for their assistance with biochemical and biophysical analyses of NPr proteins.  Conflicts of interest-The authors declare that they have no conflicts of interest with the contents of this article.