Generation and growth assessment of sugar transporter mutants in Proteus mirabilis

Glycolysis and related carbohydrate pathways have been shown to play a central role in P. mirabilis pathogenesis during urinary tract infection (UTI) [17,18]. To further define the contribution of sugar import and metabolism to bacterial fitness, we constructed a panel of targeted mutants in genes predicted to encode sugar transporters. Candidate genes were selected using multiple complementary approaches. First, phosphotransferase system (PTS) components were prioritized due to their known roles in carbohydrate uptake. In parallel, bioinformatic tools including KEGG [34], Transporter Classification Database (TCDB 2.0) [35], NCBI BLAST against sugar transporters encoded by UPEC CFT073 [36], NCBI Conserved Domains Database [37], Phyre2 [38], and the Bacterial and Viral Bioinformatics Resource Center (BV-BRC) [39] were used to identify additional sugar transport genes from ones annotated in the P. mirabilis HI4320 genome [40]. To cast a wide net, we used a broad definition of sugar such that a prediction for sugar transport in any one database merited inclusion. This yielded a final list of 50 targets (Table 1). Mutants were generated using the targetron system for site-specific group II intron insertion [41,42]. Of the 50 genes targeted, we successfully created 47 mutants, each confirmed by PCR.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Candidate P. mirabilis sugar transporter genes.

https://doi.org/10.1371/journal.ppat.1014324.t001

We examined the growth of all 47 sugar transporter mutants in LB and in Minimal A medium supplemented with either 0.2% glucose or 0.2% glycerol as the sole carbon source (Fig 1). Doubling time during logarithmic phase was calculated and compared to wild-type HI4320 (Fig 1A). As expected, all mutants exhibited robust growth in LB, indicating that disruption of sugar-specific transport systems did not impair viability under nutrient-rich conditions. In contrast, four mutants displayed a statistically significant increase in doubling time when cultured in Minimal A with glucose as the carbon source (Fig 1B), and all of them were in PTS-related genes. The ptsG (PMI2292) mutant, encoding the EIIBC component of the glucose-specific PTS system [43], showed reduced growth (doubling time 52.63 min vs. 36.22 min for wt, P < 0.0001), consistent with its role in glucose import. Similarly, mutants lacking crr (PMI1830, 53.41 min, P < 0.0001) and ptsI (PMI1829, 51.23 min, P = 0.0005), which encode the carbohydrate repression resistance protein and Enzyme I of the PTS [44], respectively, were attenuated under these conditions. Notably, the ptsH (PMI1830) mutant, which lacks the general phosphocarrier protein HPr, exhibited impaired growth in both glucose- (57.75 min, P < 0.0001) and glycerol- (49.28 min vs. 39.15 min for wt, P = 0.0441) containing Minimal A medium (Fig 1B-1C). While these growth defects in glucose were expected because glucose is a well-characterized PTS substrate, the defect in glycerol for the ptsH mutant was not anticipated. We also observed a modest decrease in growth parameters for the ptsI mutant in glycerol (Fig 1C), but this was not statistically significant (47.08 min, P = 0.2959). Growth of all four PTS mutants was restored in both glucose and glycerol following genetic complementation (S1 Fig). The ptsH growth defect in Minimal A was partially rescued by ptsH in trans, but fully restored with ptsHI, consistent with their expression as an operon. These findings align with the functional hierarchy of the PTS, as PtsI and PtsH comprise the phosphorelay for multiple sugar import systems and serve as central nodes for carbohydrate uptake and regulation in P. mirabilis, while Crr is paired with several substrate-specific importers including PtsG).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Growth parameters of 47 sugar transporter mutants.

(A) Doubling time measurements. Predicted gene names and annotations may be cross-referenced in Table 1. Left, LB; middle, Minimal A with 0.2% glucose as the carbon source; right, Minimal A with 0.2% glycerol as the carbon source. Vertical lines show means. *P < 0.05, ****P < 0.0001; one-way ANOVA vs. wild type HI4320 with Dunnett’s multiple comparisons test. B-C, Growth curves for the four PTS mutants with growth defects in Minimal A. (B) Minimal A with 0.2% glucose as the carbon source. (C) Minimal A with 0.2% glycerol as the carbon source. Error bars = SEM (n ≥ 3).

https://doi.org/10.1371/journal.ppat.1014324.g001

In vivo screening identifies xapB, ptsH, and ptsI as fitness factors

Previous work established that the bottleneck for P. mirabilis HI4320 in the murine model of ascending UTI limits reliable input pools to approximately 25 mutants [45]. Based on this constraint, we divided our panel of 47 sugar metabolism mutants into two pools for in vivo testing (Table 1). Pool 1 consisted of 24 mutants from the ATP-binding cassette (ABC) and major facilitator superfamily (MFS) transporter families. Pool 2 included 23 mutants from the phosphotransferase system (PTS) and other single-gene predicted sugar transporters (Fig 2A). Each pool was independently inoculated into the bladders of female CBA/J mice, and infection was allowed to proceed for 24 hours. Urine, bladder, and kidney samples were then harvested and individually barcoded for downstream sequencing (S1 Table). The abundance of each mutant in the inoculum and recovered organs was determined using an insertion sequencing (In-seq) strategy (S1 and S2 Datasets; GEO accession #GSE244606). An aliquot of each pool was plated to quantify the total input CFU and was also sequenced to identify variation that could have occurred from plating and outgrowth; no notable variation was observed in these control samples (S1 and S2 Datasets, input “I” data).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Identification of sugar importers that contribute to fitness during UTI.

(A) Experimental workflow for In-seq screening P. mirabilis sugar transporter mutants in mice. B-D, 1:1 wild type:mutant co-challenge competitive indices (7 dpi) for mutants identified in the In-seq screen. (B) wild type vs. xapB (n = 10); (C) wild type vs. ptsH (n = 15); (D) wild type vs. ptsI (n = 15). Horizontal lines show medians. Dashed line indicates equal fitness of wild type and mutant (log CI = 0). *P < 0.05; **P < 0.01; exact value given for 0.10 > P > 0.05, one sample Wilcoxon test vs. a theoretical median of 0.

https://doi.org/10.1371/journal.ppat.1014324.g002

From Pool 1, the xapB (PMI1570) mutant emerged as the only strain with a statistically significant fitness defect (S2 Fig). This mutant was underrepresented exclusively in the kidneys, suggesting an organ-specific role (S2C Fig). xapB encodes a putative xanthosine permease of the MFS family, based on homology to the E. coli XapB transporter as predicted by KEGG [46,47]. From Pool 2, both ptsH and ptsI mutants were depleted in the bladder, and ptsH also showed a defect in the kidneys (S3 Fig). The 24-hour time point for In-seq was selected based on documented bottleneck effects in the murine UTI model [48,49]. In this pooled approach, mutants were assessed in the presence of 22–23 other mutants without a wild-type control. To validate the in vivo In-seq findings, each mutant (xapB, ptsH, and ptsI) was individually assessed in the established 7-day murine co-challenge model [50]. For each strain, mice were inoculated transurethrally with a 1:1 mixture of mutant and wild-type HI4320, and competitive indices (CI) were calculated for urine, bladder, kidneys, and, to gauge dissemination into the bloodstream, spleens. The xapB mutant displayed a significant colonization defect in the kidneys (median log₁₀ CI -0.4177, P = 0.0137) (Figs 2B, S4A), consistent with the original In-seq data. The ptsH mutant exhibited significant defects in the bladder, kidney, and spleen (median log₁₀ CI -1.459, -2.261, and -1.469, and P = 0.0322, 0.0034, and 0.0312, respectively), while ptsI was significantly outcompeted in the bladder and spleen (median log₁₀ CI -1.538 and -0.6682, and P = 0.0093 and 0.0469, respectively) (Figs 2C-2D, S4B–S4C). Collectively, these experiments screened 47 sugar transport mutants for in vivo fitness and identified organ-specific defects that were reproducible in the traditional ascending UTI murine model.

Functional characterization of putative xanthosine transporter XapB

To investigate the function of xapB and the other In-seq hits (ptsH and ptsI), we performed targeted phenotypic analyses guided by predicted functions of each transporter. We first focused on PMI1570, the putative xanthosine permease (xapB). When Minimal A medium was used with 0.1% xanthosine as the sole carbon source, HI4320 failed to grow (Fig 3A). We therefore screened growth across Biolog Phenotype MicroArray carbon and nitrogen source plates to identify additional potential substrates in an unbiased fashion (S3 Dataset). No growth was observed when xanthosine was provided as a sole nitrogen source after 20 hours of incubation, indicating that strain HI4320 was unable to utilize xanthosine in this context (Fig 3B). These findings suggest that XapB may not function as a xanthosine importer in P. mirabilis under the tested conditions. However, ruling xanthosine out as a carbon or nitrogen source does not exclude possibility for import.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. PMI1570 is likely misannotated as a xanthosine transporter.

(A) Growth curves for wild-type HI4320 in Minimal A with no carbon source, 0.2% glycerol, or 0.1% xanthosine as a sole carbon source (n = 3; error bars show SD). (B) Growth on xanthosine as sole nitrogen source (Biolog plate PM3). (C) The genetic organization of the putative xap locus in P. mirabilis (top) is missing elements from the xap locus in E. coli, where it has been characterized.

https://doi.org/10.1371/journal.ppat.1014324.g003

Further analysis of the genomic context revealed notable differences in operon structure compared to the E. coli MG1655 xap locus (Fig 3C). In P. mirabilis, both xapA, encoding a xanthosine phosphorylase (PNP-II), and the regulatory gene xapR are absent. This single-gene organization raises the likelihood that XapB transports an alternate substrate. Beyond KEGG and BV-BRC’s annotations as a xanthosine transporter, other databases predict xapB to encode an undefined nucleoside transporter (BLAST and BioCyc). While P. mirabilis HI4320 XapB is 49% identical and 70% similar to E. coli MG1655 XapB, it is also 44% identical and 65% similar to the nucleoside:H⁺ symporter NupG. Biolog data suggested differential growth for wild-type HI4320 and the xapB mutant on guanosine (S5A–S5B Fig); however, the increase in optical density appeared to be due to a chemical precipitate. To confirm the turbidity was due to precipitate and not outgrowth, we plated the contents of the guanosine wells after the experimental endpoint and no CFU were recovered. There were no other notable differences in growth across 251 unique substrates between wild type and xapB. Nevertheless, we hypothesized that guanosine might enter through XapB.

Our group previously reported that a guaA mutant, which is deficient in GMP biosynthesis, exhibited impaired growth on murine organ agar, a severe defect in vivo, altered swarming motility, and a nutritional defect that could be rescued by exogenous RNA [45]. As predicted, HI4320 was unable to use guanosine as either a sole nitrogen or carbon source in Minimal A medium (Fig 4A, 4B). Despite this, exogenous guanosine successfully complemented the growth defect of the guaA mutant in Minimal A with glycerol, confirming the presence of a functional guanosine uptake system (Fig 4C). A modest slowing of growth in the presence of guanosine was not due to the DMSO solvent, nor could this slowing be mitigated by using a lower concentration of guanosine (S5C–S5D Fig). To determine whether XapB was responsible for this activity, we constructed a guaA/xapB double mutant. As with the single guaA mutant, growth defects were rescued by guanosine supplementation (Fig 4D). These data indicate that guanosine can enter the cell independent of XapB. The actual substrate of XapB, which must be available in the urinary tract and contributes to P. mirabilis fitness, remains to be determined.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. PMI1570 is not likely to be a guanosine transporter.

(A) Prediction of biochemical pathways encoded by P. mirabilis HI4320 including xanthosine and guanosine. P mirabilis is not predicted to be able to convert guanosine or related molecules into central metabolism (denoted by dashed outline). (B) Growth curves in Minimal A indicated that wild-type HI4320 can’t use guanosine (G, 0.25 mg/mL) as a sole carbon (C) or nitrogen (N) source. C-D, Growth curves in Minimal A containing 0.2% glycerol. (C) HI4320 can import guanosine to chemically complement the guaA mutant (wild type data are the same as in B). (D) A guaA xapB double mutant can still be chemically complemented by guanosine, indicating that xapB probably doesn’t transport guanosine. In these experiments, guanosine was added at a final concentration of 0.05 mg/mL. For sole nitrogen source growth curves, n = 2. For all other conditions, n = 3. Error bars show SD.

https://doi.org/10.1371/journal.ppat.1014324.g004

Dissecting the role of PTS components in P. mirabilis urinary tract fitness

The phosphotransferase system (PTS) is essential for carbohydrate uptake and consists of a relay of proteins that transfer phosphate from phosphoenolpyruvate (PEP) to sugar-specific permeases. In P. mirabilis, the PTS system appears to be organized similar to other Enterobacterales: ptsI encodes Enzyme I, which initiates the phospho-transfer cascade, and ptsH encodes HPr, a central phosphocarrier that relays phosphate to multiple specific import systems [51]. P. mirabilis HI4320 is predicted to encode nine of these substrate-specific PTS importers (Fig 5). Our in vivo co-challenge studies showed that both ptsI and ptsH single mutants were significantly outcompeted by wild type (Fig 2). These broad defects suggested that loss of general PTS function impairs fitness, but the specific transporter systems, and thus substrates, responsible remained unclear. Of the nine importers, only glucose transporter ptsG had been experimentally confirmed for P. mirabilis HI4320 (Fig 1B). To identify other PTS substrates that could contribute to the ptsH and ptsI fitness defects, we assessed growth of wild-type HI4320 and the ptsH mutant on 190 carbon sources and 95 nitrogen sources (S3 Dataset).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. KEGG predictions for PTS-family transporters in P. mirabilis HI4320.

Nine substrate-specific importers, shown here spanning the cell membrane depicted as parallel lines, are predicted to rely on the PtsHI phosphorelay. Crr (PTS enzyme IIA component) is predicted to interact with three substrate-specific importers and is highlighted in black. Arrows indicate two sets of non-crr-dependent triple mutants (red and yellow) based on in vivo expression data shown in Fig 7A that were tested in mice.

https://doi.org/10.1371/journal.ppat.1014324.g005

In addition to confirming glucose as a PTS substrate, we found the ptsH mutant had, as expected, greatly diminished growth on N-acetyl-D-glucosamine, N-acetyl-D-galactosamine, and trehalose (Fig 6A–6D, S1 Table). Predicted substrates of sucrose, maltose, and D-cellobiose were unable to support the growth of wild-type HI4320 (Fig 6E–6G), suggesting possible misannotation. D-galactose did facilitate robust growth, but no differences were observed between wild type and mutant (Fig 6H), which is consistent with the prediction that galactose is not a PTS substrate for P. mirabilis. Finally, the ptsH mutant displayed a growth defect, compared to wild type, in the presence of chondroitin sulfate C (Fig 6I). This is interesting because chondroitin sulfate C is not a predicted PTS substrate; however, chondroitinase activity has been historically described across the Proteus genus, including P. mirabilis and P. vulgaris, resulting in cleavage of the substrate into glucuronic acid and PTS substrate N-acetylgalactosamine [52–55]. More recent work has further shown that clinical P. mirabilis isolates can degrade chondroitin sulfate under multiple growth conditions [56,57]. Collectively, we confirmed four predicted PTS substrates and did not identify any additional unpredicted PTS substrates that supported growth of P. mirabilis HI4320 (S2 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Notable carbon source growth curve results for wild type vs. the ptsH mutant.

A-H, Predicted PTS substrates. A-D, Predicted PTS substrates that showed expected growth defects for the ptsH mutant. E-G, Predicted PTS substrates that did not support growth by wild-type HI4320, suggesting that their transporters might be misannotated. (H) P. mirabilis utilizes galactose but it is not a PTS substrate. (I) Chondroitin sulfate C was not predicted as a PTS substrate.

https://doi.org/10.1371/journal.ppat.1014324.g006

Remarkably, we detected growth defects of the ptsH mutant on various nitrogen sources (S6 Fig). Both ammonia and urea are preferred nitrogen sources that are particularly relevant to the urinary tract environment, and we observed a large growth defect of the ptsH mutant, compared to wild type, for both (S6A–S6B Fig). Similarly, L-amino acids such as leucine, arginine, histidine, phenylalanine, and cysteine also resulted in growth defects for the ptsH mutant (S6C–S6G Fig). Interestingly, nitrogen sources of serine and methionine did not yield differences between strains (S6H–S6I Fig). The nitrogen-dependent phenotypes we observed correlate with the growth defects of the ptsH and ptsI mutants in Minimal A with glycerol as the carbon source and provide an explanation why ptsH and ptsI had more severe defects with glucose as the carbon source compared with ptsG or crr mutants (Fig 1B–1C). Specifically, glycerol was not expected to be imported via PTS, and reduced ptsH and ptsI growth in glycerol is likely due to a defect in taking up the nitrogen source (ammonium). Importantly, growth in glycerol was restored when ptsH and ptsI expression was restored (S1 Fig). We propose the greater reduction of growth for ptsH and ptsI in glucose compared with ptsG or crr reflects a cumulative effect of impaired glucose import and defective nitrogen assimilation.

P. mirabilis swarming motility is induced by amino acid cues [58,59]. In addition, swarming defects often correlate with fitness defects during experimental UTI [60]. Furthermore, a previous study in Bacillus cereus reported swarming defects for a ptsH mutant [61]. For these reasons, we tested swarming motility and found that the ptsH mutant exhibited a modest but statistically significant reduction in swarming motility compared to wild-type HI4320 (S7A Fig). Because MR/P fimbriation and urease are also crucial for P. mirabilis virulence [62], we measured mrp promoter invertible element orientation and qualitatively assessed urease activity and observed no notable differences (S7B–S7C Fig).

To narrow down the specific substrate sugar import pathways that contribute to UTI fitness, we first excluded systems associated with Crr (PMI1830), an Enzyme IIA component not identified as a fitness factor in our In-seq screen (S3, S5 Fig). We then grouped the remaining PTS systems into categories of significantly high (Fig 5, red arrows) or low expression/no (Fig 5, yellow arrows) induction during UTI based on previous in vivo transcriptomic data [17] (Fig 7A). A triple mutant targeting the uninduced transporters bglF, chbB, and agaV, with predicted substrates β-glucoside, cellobiose, and N-acetyl-galactosamine, did not exhibit any fitness defect in co-challenge experiments (S8 Fig), suggesting these importers are dispensable in the urinary tract.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. PTS multi-mutant co-challenges.

(A) Published in vivo differential expression was used to prioritize mutant construction [17]. Individual fold-change data points from 9 microarrays are shown, and genes that were significantly differentially regulated in the prior publication are indicated with an asterisk. Red and yellow bars correlate with the arrows in Fig 5. (B) a combinatorial triple mutant constructed based on higher induction during experimental UTI (red bars) did not have a competitive disadvantage. (C) quadruple mutant with crr recapitulated ptsH phenotype. (D) quadruple mutant with ptsG also was similar to ptsH phenotype. (E) removing nagE preserved the ptsH phenotype. (F) an scrA ptsG double mutant had a much weaker defect, emphasizing the combinatorial contribution of different PTS transporters to fitness during experimental UTI. Horizontal lines show medians. Dashed line indicates equal fitness of wild type and mutant (log CI = 0). *P < 0.05; **P < 0.01; one sample Wilcoxon test vs. a theoretical median of 0.

https://doi.org/10.1371/journal.ppat.1014324.g007

We next tested a triple mutant targeting the highly induced genes nagE, scrA, and ulaC, which encode predicted transporters for N-acetylglucosamine, sucrose, and L-ascorbate, respectively. This strain also showed no significant in vivo defect (Fig 7B). However, the addition of a fourth mutation in crr resulted in a composite fitness defect, with significantly reduced recovery from urine and kidneys (median log₁₀ CI -2.55, and -1.87 and P = 0.0312 and 0.0156, respectively) (Fig 7C). Of the importers predicted to work with Crr, glucose-specific EIIBC component ptsG transcript was the only one induced during experimental UTI in mice (Fig 7A). Replacing crr with a ptsG mutation produced a similar result, with significant attenuation in urine, kidneys, and spleen (median log₁₀ CI -2.60, -1.47, and -1.74, and P = 0.0039, 0.0156, and 0.0312, respectively) (Fig 7D). Both quadruple mutants phenocopied the fitness loss seen for the ptsH mutant (Fig 2C), indicating that the substrates imported by these systems collectively contribute to in vivo fitness. CFU recovery for all PTS mutant co-challenges are shown in S9 Fig.

To determine the minimal functional set of PTS importers needed for urinary tract fitness, we progressively removed individual genes from the nagE/scrA/ulaC/ptsG quadruple mutant. After restoring wild type nagE, the least induced gene in vivo (Fig 7A), the defective phenotype of triple mutant scrA/ulaC/ptsG was not altered (significant median log₁₀ CI of -2.22 and -1.42 and P = 0.0078 and 0.0156 in the urine and kidneys, respectively) (Fig 7E). However, further removal of the ulaC mutation reduced the magnitude of the competitive defect to ≤1 log, although the mutant remained significantly attenuated in urine, kidneys, and spleen (median log₁₀ CI -0.59, -1.00, and -0.77, and P = 0.0273, 0.0391, and 0.0078, respectively) (Fig 7F). Import mediated by ulaC, scrA, and ptsG, collectively, phenocopies the PTS-dependent in vivo defect first observed with the ptsH and ptsI mutants (Fig 2C–2D). Taken together, this suggests the substrates transported by these three systems all contribute to P. mirabilis fitness in the urinary tract and further demonstrates the coordination of the general PTS phosphorelay mediated by PtsH and PtsI with substrate-specific enzyme II PTS importers.

Although we identified ulaC, scrA, and ptsG as important contributors to in vivo fitness, only PtsG had an experimentally confirmed substrate in P. mirabilis (glucose, Fig 1B). UlaC and ScrA are predicted to import L-ascorbate and sucrose, respectively. However, wild-type HI4320 failed to utilize 10 mM L-ascorbate as a carbon source under aerobic conditions (S10A Fig) and instead showed steadily declining optical density compared to control media lacking a carbon source. E. coli has been reported to ferment L-ascorbate under anaerobic conditions [63]. However, an anaerobic atmosphere did not allow P. mirabilis growth on ascorbate (S10B Fig), further suggesting that this substrate is not imported by HI4320. While the PMI1775–1777 operon appears to encode a PTS IIA, IIB, IIC importer, closer examination of this locus in other databases did not yield viable alternative substrates. Specifically, TransportDB listed fructose as the substrate, but HI4320 did not grow on D-fructose in the Biolog PM1 carbon source panel (S2 Table).

Likewise, we found that sucrose was not a functional carbon source for strain HI4320 (Fig 6E), raising questions about the annotated role of ScrA. The gene encoding scrA, PMI3515, appears to be part of a four-gene operon comprising PMI3514–17 (S10C Fig). The protein encoded by PMI3514 has 47% identity and 67% similarity to E. coli MG1655 MurQ, which is an N-acetylmuramic acid (MurNAc) 6-phosphate etherase. In E. coli, the next gene encodes MurP, a PTS enzyme IICB component that imports MurNAc and shares 53% similarity to HI4320 ScrA. PMI3516 encodes a protein that looks like transcriptional regulator MurR, although in E. coli, murR is divergently transcribed from the mur operon. However, the last genes in each operon (PMI3517 or yfeW) are dissimilar. We therefore hypothesized that ScrA imports MurNAc. Despite the similarities, HI4320 was unable to grow using 0.2% MurNAc as a sole carbon source (S10D Fig). Although E. coli encodes machinery to import via PTS and shuttle MurNAc into glycolysis and other central metabolic pathways, it is possible that P. mirabilis takes up MurNAc via PTS but is unable to utilize this substrate as a sole carbon source. Thus, although both ulaC and scrA contributed to P. mirabilis fitness and are induced during experimental UTI, the substrates for both remain to be determined. These results highlight the need for experimental validation of sugar transporter function in P. mirabilis, as many annotations based on E. coli homology may not accurately reflect substrate specificity in this organism.

Glucosuria enhances P. mirabilis colonization and increases the severity of infection

The in vivo defects observed with ptsH and ptsI mutants reinforce prior studies demonstrating that carbohydrate metabolism is critical for P. mirabilis fitness during urinary tract infection [17,18]. Although sugars are not typically considered a major carbon source in urine, they are clearly accessible to P. mirabilis during experimental UTI. We hypothesized that increasing urinary sugar levels would intensify the fitness disadvantage observed for the ptsH mutant. We accomplished this using the sodium-glucose cotransporter 2 (SGLT2) inhibitor dapagliflozin, which reduces renal glucose reabsorption leading to increased glucose concentrations in urine [64].

As expected, administration of dapagliflozin via drinking water increased glucosuria by more than 20-fold within 24 h in female CBA/J mice, raising the mean glucose from 375 to 14,288 mg/dL, P = 0.0225 (Fig 8A). Elevated glucosuria remained for the duration of the 7-day experiment, falling slightly to 8,311 mg/dL by day 7. Following infection with wild-type HI4320, urinary glucose levels remained relatively stable in mice that received normal drinking water over 7 days (Fig 8B). Interestingly, over time, two of the mice inoculated with P. mirabilis showed glucose levels that declined by more than half over the course of the experiment; one animal strikingly had an over 90% reduction from baseline by day 3 (77.5 vs. 7.21 mg/dL) that further fell below the limit of detection by day 7. Dapagliflozin-treated mice showed increased urinary glucose similar to the pilot study, marked by a stark reduction in urinary glucose on day 3 post-inoculation, suggestive of bacterial glucose consumption (Fig 8B). Additionally, 20% (n = 2) of dapagliflozin-treated animals reached humane endpoints before or at the scheduled termination of the study. This was consistent with higher median bacterial burdens in the urine, particularly on day 3 post-inoculation (Fig 8C). At 7 days post-inoculation, kidney colonization was significantly increased in the dapagliflozin group (Fig 8D, median 1.98 x 10⁷ vs. 5.91 x 10⁵ CFU/g, P = 0.0397); bladder colonization was also increased, although not statistically significant (median 2.11 x 10⁸ vs. 2.15 x 10⁶ CFU/g, P = 0.0635, n = 4 in the treated group). The lack of significance is likely due to reduced sample size from increased morbidity and mortality at day 7 post-infection. These results demonstrate that SGLT2 inhibition effectively induces glucosuria in CBA/J mice and promotes enhanced colonization and disease severity during P. mirabilis UTI.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Experimental P. mirabilis UTI during SGLT-2i treatment.

(A) Measurement of urinary glucose. Administration of SGLT-2 inhibitor dapagliflozin to female CBA/J mice via drinking water on “day -1” resulted in increased glucose excretion via urine. *P < 0.05, **P < 0.01 vs. day -1, two-way ANOVA with Dunnett’s multiple comparisons test. B-D, Experimental inoculation of P. mirabilis HI4320 on day zero in mice with or without dapagliflozin treatment (n = 5/group). Of the five mice in the dapagliflozin group, one mouse died at day 6 and one was moribund at day 7 (unable to collect enough urine for glucose measurement). (B) Urinary glucose. Lines connect data from individual mice. Glucose levels fell somewhat in infected mice, and the effect was more apparent in the dapagliflozin-treated mice. # indicates one control mouse with undetectable glucose at day 7. ***P < 0.001, ****P < 0.0001 for dapagliflozin-treated mice vs. day -1, mixed-effects analysis with Dunnett’s multiple comparisons test. (C) Bacterial burden in urine. Median CFU recovery was higher in dapagliflozin-treated mice, although with 5 mice the difference was not statistically significant. (D) Bacterial burden in organs at experimental endpoint (7 dpi). Dashed line indicates limit of detection. C and D, *P < 0.05; for 0.05 < P < 0.1, exact value shown; Mann-Whitney test. Horizontal lines denote medians.

https://doi.org/10.1371/journal.ppat.1014324.g008

Due to the increased morbidity observed during infection of dapagliflozin-treated mice, we next tested the fitness of the ptsH mutant in a 3-day co-challenge experiment, timed to coincide with peak glucose utilization and elevated urinary colonization (Fig 8B–8C). As expected, urinary glucose levels remained elevated in dapagliflozin-treated mice and declined by day 3, again suggesting bacterial consumption (Fig 9A). Consistent with previous findings, dapagliflozin-treated mice exhibited higher overall urinary colonization (S11A Fig), and the fitness defect of the ptsH mutant was significant on day 2 in treated mice but not in controls (median log₁₀ CI -0.64 vs. -0.38 control, P = 0.0059 and 0.4258, respectively) (Fig 9B). By day 3, the ptsH mutant showed a significant fitness defect in the urine of both groups (median log₁₀ CI -1.54 vs. -0.80 control, P = 0.0078 and 0.0039, respectively), and this was also observed in the bladder (-1.59 vs. -0.83 control, P = 0.0020 and 0.0039, respectively) (Fig 9C). Interestingly, although the fitness defect of the ptsH mutant in dapagliflozin-treated mice was not significantly different from control mice in the kidneys (median log₁₀ CI -1.04 vs. -1.71 control) and spleens (median log₁₀ CI -0.99 vs. -1.22 control), in both organs, the CI reached statistical significance only for dapagliflozin-treated mice. This is because bacteria were recovered from all mice in the dapagliflozin group, whereas many mice in the control group did not have measurable ascending (n = 5; 50%) or disseminated (n = 7; 70%) infection (S11A–S11B Fig). Adding up the total CFU (wild type plus ptsH) recovered from each mouse showed that overall bacterial burden was significantly higher in mice treated with dapagliflozin (S11C Fig). When glucose levels were correlated with bacterial burdens on day 3, a strong inverse relationship was observed in control mice across urine, bladder, and kidney samples (R² = 0.28, 0.64, and 0.84, respectively) (Fig 10A, 10C, 10E). No such trend was seen in the dapagliflozin-treated group (Fig 10B, 10D, 10F), suggesting that hyperglucosuria alters the physiological responses of P. mirabilis during experimental UTI. These findings support the use of dapagliflozin to model glucosuria in CBA/J mice and demonstrate that elevated urinary glucose exacerbates P. mirabilis colonization and infection severity.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 9. Wild type vs. ptsH co-challenge during SGLT-2i treatment.

Twenty mice, half receiving dapagliflozin and half receiving normal water, were administered a 1:1 mixture of wild type and ptsH mutant bacteria. (A) urinary glucose. Lines connect urinary glucose levels over time in individual mice. *P < 0.05, **P < 0.01, ***P < 0.001 control vs. dapagliflozin, mixed-effects analysis with Šidák’s multiple comparisons test. B-C, competitive indices. Horizontal lines show medians. *P < 0.05, **P < 0.01 vs. theoretical median of log CI = 0, one sample Wilcoxon test. Comparisons between control and dapagliflozin groups were not significant (Multiple Mann-Whitney tests with Holm-Šídák correction). (B) urinary competitive index measured over time. (C) organs at experimental endpoint (3 dpi).

https://doi.org/10.1371/journal.ppat.1014324.g009

Download:

• PNG
  larger image
• TIFF
  original image

Fig 10. Inverse correlation between glucosuria and CFU recovery (A, C, E) but not during SGLT-2i-induced hyperglucosuria (B, D, F) at 3 dpi.

Data were compiled from wild type single challenge and wild type vs. ptsH co-challenge (total CFU; limit of detection = 100 for organs). Lines indicate simple linear regression.

https://doi.org/10.1371/journal.ppat.1014324.g010

Discussion

The metabolic strategies employed by bacterial pathogens to colonize the urinary tract differ by species, reflecting adaptations to specific host niches. UPEC relies primarily on amino acid metabolism in vivo, with carbohydrate utilization playing a minor role [11,18,21]. In contrast, prior transcriptomic studies have shown that P. mirabilis upregulates glycolytic enzymes and sugar transporters during UTI, in addition to amino acids, suggesting a greater reliance on sugar-derived carbon sources [17]. Here, we directly assessed the importance of sugar import systems in P. mirabilis by generating 47 targeted transporter mutants and evaluating both their in vitro growth and in vivo fitness. Several mutants displayed in vivo fitness defects despite lacking observable phenotypes in standard growth conditions, indicating that specific import systems contribute to pathogenesis in a context-dependent manner.

In vivo pooled mutant screens are inherently limited by physiological bottlenecks in the ascending UTI model, including for P. mirabilis [45]. Although our In-seq experiments were designed to remain within established bottleneck constraints, we nonetheless observed clear founder effects in a subset of mice, where 1–3 mutants disproportionately dominated sequencing reads from individual organs. Similar stochastic population dynamics have been reported in UPEC infection models, even under carefully controlled conditions [48,65,66]. Despite these limitations, all fitness determinants identified by In-seq were confirmed in a traditional 7-day co-challenge model, exceeding validation rates reported in comparable E. coli studies [48]. These findings underscore both the challenges of pooled in vivo approaches and the robustness of our experimental design and follow-up strategy.

The Major Facilitator Superfamily (MFS) comprises a large and diverse group of membrane transport proteins characterized by 12 transmembrane helices and driven by electrochemical gradients. MFS transporters are present in both prokaryotes and eukaryotes; for example, glucose uptake in humans is mediated by members of this family [67]. In this study, xapB, an MFS transporter, was identified as important for P. mirabilis fitness during infection. Although XapB is annotated as a xanthosine permease based on homology to E. coli [47], its substrate specificity in P. mirabilis remains unverified. Prior studies support a role for nucleoside metabolism in P. mirabilis polymicrobial pathogenesis; loss of xanthine-guanine phosphoribosyltransferase (gpt) and purine-nucleoside phosphorylase (deoD) impaired in vivo fitness during co-infection with Providencia stuartii [19]. However, we found that strain HI4320 could not use xanthosine as a sole carbon or nitrogen source under the conditions we tested, and guanosine uptake occurred independently of XapB, validated in a guaA/xapB double mutant. P. mirabilis HI4320 encodes nine additional genes that are annotated as nucleoside importers, suggesting there are multiple routes of entry for these molecules.

Importantly, the xapB mutant displayed a significant defect during experimental UTI, suggesting that the imported substrate(s) is likely non-redundant. In this study, we did not initially set out to study nucleoside import; xapB was included in our 47 mutant panel because it was annotated as a generic sugar importer by TransportDB at the start of the project. Interestingly, over the course of this work, TransportDB updated the annotation for xapB at least twice, first to generic nucleoside transport, and then to melibiose importer melB. Melibiose is a plant disaccharide that did not support growth of HI4320 in our carbon source testing, leading us to conclude that this revised annotation is also likely incorrect. These findings highlight a recurring challenge in microbiological research: reliance on gene annotations inferred from E. coli often fail to recapitulate actual function.

Similar annotation discrepancies were observed for multiple phosphotransferase system (PTS) family transporters in our study. PtsH and PtsI are conserved upstream components of the PTS, a multi-component phosphorelay essential for sugar uptake and carbon catabolite repression [27]. Both ptsH and ptsI mutants exhibited reproducible defects in a murine model of ascending UTI, consistent with a central role for glycolysis in P. mirabilis pathogenesis and contrasting with UPEC and other uropathogens [17,18,68]. Even so, because PtsH and PtsI interact with many sugar-specific components, these findings did not identify the specific substrates responsible for fitness.

Here, we found that the three most-induced substrate-specific PTS genes during experimental P. mirabilis UTI in mice, ptsG, scrA, and ulaC, combined to recapitulate the ptsH mutant phenotype. However, in vitro substrate validation for two of these transporters was inconsistent with predicted annotations. HI4320 was unable to use either sucrose, predicted to be imported by ScrA, or L-ascorbate, predicted to be imported by UlaC, as a carbon source. Nor did ScrA support growth on MurNAc in follow-up experiments despite homology with the E. coli mur locus. Likewise, HI4320 failed to grow on predicted PTS substrates maltose, cellobiose, and β-glucosides. Incidentally, although E. coli was initially shown to utilize cellobiose, later work indicated the more physiologically likely substrate is chitobiose [69]. Because the chbB mutant was not found to be important for P. mirabilis fitness during experimental UTI (S8 Fig), we did not investigate chitobiose further. Overall, only 4/9 of the KEGG-predicted PTS substrates for P. mirabilis verified experimentally. Notably, our findings are consistent with Biolog data for P. mirabilis isolates obtained from broiler chickens [70]. It is possible that imported sugars or sugar derivatives are used by P. mirabilis but are not connected to central metabolism in a way that would allow use as a sole carbon or nitrogen source. However, most PTS substrates would typically be accessible to metabolic pathways such as glycolysis and other sugar interconversion pathways, and carbohydrate substrate prediction is, at the current time, frequently unreliable. We therefore conclude that the most likely explanation for the mismatch between PTS substrate prediction and experimental growth results is misannotation. Teasing apart the combinatorial contributions of UlaC, ScrA, and PtsG to UTI fitness will be greatly aided by identifying the substrates for the first two transporters.

Beyond sugar transport, the PtsHI phosphorelay appears to influence nitrogen metabolism. The ptsH mutant showed defects when grown on ammonia, urea, and several amino acids, suggesting cross-regulation between carbon and nitrogen pathways. Likewise, both ptsH and ptsI mutants had an initially unexpected decrease in growth on glycerol and displayed a more pronounced defect on glucose compared with the more substrate-specific crr and ptsG mutants. In E. coli, the PTS^Ntr system coordinates nitrogen assimilation by sensing glutamine and α-ketoglutarate [31,71], but has not been characterized in P. mirabilis. Our findings raise the likelihood of similar regulatory integration.

While the urinary tract is generally sugar-poor, elevated glucose levels occur in specific conditions such as diabetes or treatment with SGLT2 inhibitors. Diabetic individuals have increased UTI susceptibility, and glucosuria is common in both humans and mouse models [68,72]. SGLT2 inhibitors, such as dapagliflozin, reduce renal glucose reabsorption, leading to elevated urinary glucose. Although debated in clinical studies [73,74], preclinical studies consistently show worsened UTI outcomes under glucosuric conditions, including increased bacterial burden and dissemination [25,75]. We observed similar results for P. mirabilis infection, where dapagliflozin treatment led to elevated colonization and exacerbated disease. Despite these findings, the ptsH mutant did not demonstrate an enhanced competitive defect in urine under glucosuric conditions, contrary to our prediction. One explanation for this result is that additional sugar import systems contribute to fitness during UTI, thereby partially masking the effect of ptsH loss under glucosuric conditions. Consistent with this idea, growth of the ptsH mutant in minimal medium with glucose was slowed but not abolished, indicating that glucose can still enter the cell through alternative pathways. Several transporters, including crr, exhibited borderline effects on in vivo fitness in the In-seq screen, suggesting that their contributions may become apparent only in specific metabolic contexts. It is also possible that loss of ptsH induces compensatory metabolic or regulatory responses that mitigate fitness defects when glucose availability is increased.

The connection between glucosuria and UTI severity likely involves more than nutrient availability. Hyperglycemia impairs innate immunity, including reduced neutrophil responses and cytokine signaling in the kidney [25,76]. We observed an inverse correlation between urinary glucose and bacterial burden in untreated mice, but this trend was lost in dapagliflozin-treated animals, likely reflecting both increased glucose consumption and altered host-pathogen interactions. Mouse background also matters; others have reported that C57BL/6J mice treated with dapagliflozin had similar bladder and urine colonization by UPEC [26]. C3H/HeOuJ mice had to be administered 10 mg/kg of dapagliflozin to see increased UPEC burden in bladder and kidneys, and results were dependent on bacterial strain [26]. Mice genetically engineered to develop type 2 diabetes (db/db) also have increased UTI risk that is correlated with altered innate immune markers [77]. However, dapagliflozin-treated CBA/J mice were shown to have increased UPEC and K. pneumoniae colonization at chronic and acute timepoints [25]. Interestingly, while UPEC infections in this model could be sustained for 7 days, P. mirabilis infection resulted in increased morbidity at this time point, consistent with this species’ increased contribution to UTI in complicated backgrounds. Last, streptozotocin-treated ICR mice, a model of type 1 diabetes, carry higher P. mirabilis burden in bladders and kidneys, and glucose has been found to increase P. mirabilis adherence to cultured kidney cells [78]. Further investigation into P. mirabilis pathogenesis in diabetic models is warranted.

In conclusion, our findings provide mechanistic insight into the metabolic requirements of P. mirabilis during UTI. Sugar import systems, particularly PTS transporters, are essential for in vivo fitness, and substrate annotations based on E. coli must be experimentally validated. Our data reveal regulatory crosstalk between carbon and nitrogen metabolism and demonstrate that host metabolic conditions, such as glucosuria, exacerbate infection severity. These results have implications for managing UTI in individuals with metabolic disease and highlight the need for continued investigation into the roles of uncharacterized transporters and their regulation in P. mirabilis pathogenesis. They also underscore a broader gap in our understanding of the P. mirabilis genome and the limitations of relying on comparative functional annotation. Future work will focus on identifying the specific substrates imported by fitness-contributing transporters, confirming the presence of these metabolites in the urinary tract, and determining how P. mirabilis accesses these nutrient sources during infection.

Ethics statement

Animal experiments were approved by the University of Michigan Medical School Institutional Animal Care and Use Committee, protocol number PRO00010856. During catheterization procedures, mice were anesthetized by intraperitoneal injection of ketamine/xylazine. Mice were euthanized by inhalant isoflurane anesthetic overdose prior to organ removal.

Bacterial strains and culture conditions

P. mirabilis strain HI4320 was isolated from the urine of an elderly female nursing home patient with a long-term (≥30 days) indwelling catheter [3,40,79]. E. coli TOP10 (Thermo Fisher) was used for plasmid construction and maintenance. Bacteria were routinely cultured at 37°C in lysogeny broth (LB; per liter: 10 g tryptone, 5 g yeast extract, 0.5 g NaCl) with aeration or on LB solidified with 1.5% agar. As needed, antibiotic selection was used as follows (µg/mL): kanamycin 25, ampicillin 50, chloramphenicol 20. For experiments requiring minimal, chemically defined media, Minimal A [MinA; per liter: 10.5 g of K₂HPO₄, 4.5 g of KH₂PO₄, 0.47 g of sodium citrate, 1.0 g of (NH₄)₂SO₄; autoclave to sterilize and add 1 mL of 1 M MgSO₄, 10 mL of 20% glycerol (or other carbon source as specified), and 1 mL of 1% nicotinic acid] was used [80]. The carbon source in Minimal A was 0.2% glycerol unless otherwise specified. All strains used in this study are listed in S3 Table.

Mutant construction

All mutants were constructed using a P. mirabilis-tailored version of targetron insertional mutagenesis [42,81]. Briefly, stable chromosomal mutations were constructed using a synthesized 353 bp group II intron fragment (eBlocks, Integrated DNA Technologies) that specifically targeted each gene designed using the ClosTron prediction algorithm [82]. Reprogrammed intron fragments were cloned into pACD4K-CloxP using NEBuilder HiFi DNA Assembly master mix (New England Biolabs) with primers designed to amplify vector or intron templates and confirmed by DNA sequencing (Eurofins). Targetron-containing plasmids and a source for T7 polymerase, pAR1219 [83], were introduced into P. mirabilis HI4320 using electroporation and induced to jump into the specified genes. Insertional mutations in kanamycin-resistant mutants were confirmed using PCR. To construct multiple mutations in the same background, the kanamycin resistance gene in the initial insertion was removed using cre/lox recombination to create a markerless mutant [42,84,85]. Targetron sequences are listed in S4 Table.

Murine model of ascending UTI

Bacterial fitness during UTI was assessed using a well-established mouse model [50,86,87]. Briefly, overnight cultures of P. mirabilis were diluted in LB to OD₆₀₀ = 0.092-0.094 (~2 × 10⁸ CFU/mL). For co-challenge experiments, wild type and mutant bacteria were mixed 1:1. Ten female CBA/J mice, aged 5–6 weeks (Jackson Laboratory), were transurethrally inoculated with 50 µL of this 1:1 mixture (10⁷ CFU/mouse) over 30 s using a Harvard pump. At 7 days post-inoculation, urine was collected; mice were euthanized; and bladders, kidneys, and spleens were harvested. Organs were homogenized and plated to quantify CFU; mutants were distinguished from wild-type colonies using kanamycin. Competitive indices were calculated for each site by comparing the ratio of output wild type and mutant to the ratio of input bacteria [50,88]. For sites with no recovered CFU, the limit of detection for urine was set to 20 and for organs 100. Statistical significance of competitive indices was calculated using the Wilcoxon signed rank test.

Pooled transporter mutant murine challenge

To measure the relative contributions of sugar transporters to UTI, 47 targetron mutants were individually cultured and the density adjusted as described above. 23 (ABC + MFS, group 1) or 24 (PTS + Others, group 2) strains were mixed together in equal volume, and 15 mice/group were transurethrally inoculated with 50 µL (~10⁷) CFU of either mixture. A 1 mL aliquot of each input was directly pelleted for DNA purification and a second aliquot was plated to quantify CFU (group 1, 2.51 × 10⁸ CFU/mL; group 2, 2.68 × 10⁸ CFU/mL). After 24 h, urine was collected and diluted to 250 µL, mice were euthanized, and bladders and kidneys were collected. Organs were homogenized (Omni International) in 2 mL phosphate-buffered saline (PBS); a portion of organs and urine was dilution-plated to determine output CFU, and the remainder was spread-plated on LB agar containing kanamycin. The next day, colonies from each plate were swabbed into 10 mL of PBS, pelleted, and frozen. Colonies from the plated input samples were also collected as a control for growth on agar (input spiral, insp). Chromosomal DNA was purified from inputs and urine, bladder, and kidney outputs using the Wizard Genomic DNA Purification Kit (Promega) and quantified using a Qubit fluorometer (Invitrogen).

Insertional site sequencing (In-seq)

Targetron junctions were enriched and prepared for high-throughput sequencing with modification of established protocols [89,90]. Briefly, we used the NEBNext Ultra II FS DNA Library Prep with Sample Purification Beads kit to fragment DNA into 200–450 bp lengths and ligate sequencing adaptors with sequences modified to be in line with lower %GC content in P. mirabilis (TA_Adaptor_Top and TA_Adaptor_Bottom). Targetron-gene junctions were enriched by PCR using primers Targetron_enrich_For and Transposon_enrich_Rev (S5 Table). Samples (n = 78) were barcoded by indexing PCR to label each specific library using NEBNext Multiplex Oligos for Illumina (Dual Index Primers Set 1) (S1 Table). PCR products were measured by Qubit and submitted to the University of Michigan Advanced Genomics Core for Illumina sequencing (150 nt, paired-end).

In-seq analysis

Mutant fitness in mice was assessed by quantifying the proportion of each of the 47 targetron-gene junctions in the input and output sequences. Identification and quantification of targetron junctions was conducted by the University of Michigan Medical School’s Bioinformatics Core. Filtering, trimming, and deduplication of reads was accomplished using an established Tn-seq pipeline [19,91], and subsequent steps were adapted for the relatively small number of targetron insertions. BLASTN with the 3′ end of the targetron as query was used to identify targetron-containing sequences consisting of at least 100 nt at 100% identity on the plus strand. The 19 nt immediately following the targetron sequence were used to uniquely identify the locus of gene insertion for each targetron (i.e., 47 input mutants). A counts matrix was generated from the 1,739,134 unique 19 nt reads. Features which were members of each individual group (1 or 2) were input into EdgeR, a limma-based R package which is able to deal with group sizes of only one sample in a differential enrichment calculation [92]. Deduplicated library sizes were used for depth normalization.

Genetic complementation

To complement growth defects for PTS mutants, genes were complemented in trans using the pGEN vector backbone [93] and expressed under their native promoters as predicted using the BPROM module in SoftBerry [94]. Specifically, ptsH, ptsI, ptsHI together, and crr were all expressed from the ptsH predicted promoter, while ptsG was expressed from its monocistronic promoter. Genes and promoters were PCR-amplified from P. mirabilis HI4320 genomic DNA and cloned using the Gibson method (NEBuilder, New England Biolabs). Clones were constructed in E. coli TOP10 cells and first screened by PCR then confirmed by DNA sequencing. Plasmids were then introduced to relevant P. mirabilis strains using electroporation and selection on ampicillin. Primers used for cloning are listed in S5 Table and complementation plasmids are shown in S6 Table.

Diabetic UTI murine model

To assess P. mirabilis UTI during glucosuria, we adapted the method from Nishitani et al. [95]. The SGLT-2 inhibitor dapagliflozin (MedChem Express) was dissolved in ethanol (125 mg/ml), then diluted in Ann Arbor city water to a final concentration of 0.02 mg/ml. Female CBA/J mice were administered dapagliflozin via drinking water ad libitum beginning 24 h before bacterial inoculation. Bacteria were prepared as described above. Urine was collected at specified intervals and glucose was quantified using an Infinity glucose hexokinase assay (Thermo Fisher) according to the manufacturer’s instructions. Aliquots of urine from the same time points were diluted and plated to quantify CFU, and bacterial burden in organs was assessed as described above.

Growth curves

Overnight cultures were started from a single colony in lysogeny broth incubated at 37°C with aeration. For downstream assays, cultures were diluted 1:100 (v/v) into sterile media. Bacterial growth over time (24 h) was measured in triplicate by recording the optical density at 600 nm (OD₆₀₀) at 15 min intervals using a Bioscreen C set to 37°C with continuous shaking. Growth curves with genetically complemented mutants were performed without ampicillin, as pGEN is stable without selection [93]. Chemical complementation of growth defects was performed in Minimal A medium with 0.2% glycerol as the carbon source unless otherwise specified. Anaerobic growth curves were conducted in an anaerobic chamber (Coy Lab Products, Grass Lake, MI) at 37°C under an atmosphere of 5% H₂, 5% CO₂, and 90% N₂. OD₆₀₀ of 96-well plates was measured every 10 mins for 48 h in a microplate stacking device (BioStack 2WR) with coupled absorbance reader (Powerwave HT, BioTek Instruments). Doubling times for the 47 transporter mutants were calculated from three biological replicates using the AMiGA software package in [R] [96] from 0.5 to 12 hours. Doubling time was computed as ln [2] multiplied by the inverse of the maximum specific growth rate. To identify potential sugar transporter substrates, Biolog Phenotype MicroArray plates PM1–2 (carbon sources) and PM3 (nitrogen sources) were inoculated with wild-type or mutant P. mirabilis and assayed for growth every 10 min for 24 h using a LogPhase 600 Microbiology Reader (BioTek) with shaking. Biolog cultures were conducted using the manufacturer’s recommended medium with the respiration substrate (tetrazolium-based dye) omitted.

Swarming motility

Swarming motility experiments were conducted as previously described [97]. Briefly, 5 µl of a logarithmic-phase culture was added to the center of an LB (10 g/L NaCl) agar plate, allowed to dry, and incubated at 30°C for 16 h, after which the swarming radius was measured.

Invertible element assay

Orientation of the mrp promoter invertible element (IE) was measured as previously described [98]. Briefly, strains were cultured in LB to log phase with aeration (a naturally IE-OFF condition) or statically for 48h (a naturally majority-ON condition). Bacteria were adjusted to OD₆₀₀ = 0.4, incubated at 95°C for 10 min, then subjected to PCR using primers Pm IE P1 and Pm IE P2 (S5 Table). Amplicons were digested with AflII (NEB) and electrophoretically separated on a 2% agarose gel. The proportion of ON and OFF IE orientation was assessed using densitometry (Bio-Rad Image Lab v6.1.0).

Urease test

Christensen’s urea agar, with tryptone substituted for peptone, was used to qualitatively assess urease activity [99]; per liter: 1g tryptone, 1g dextrose, 5g NaCl, 2g KH₂PO₄, 20g urea, 0.012g phenol red, 15g agar. The first six ingredients were filter-sterilized then added to autoclaved agar before dispensing into slants. A 10 µl loop of overnight LB culture was added to each slant, which was then incubated at 37°C and observed from 0.5-24 h for a color change from yellow to pink, indicating the release of ammonia and subsequent rise in pH.

Statistical analysis

All graphs were plotted, and in most cases, statistics calculated using GraphPad Prism 10. P values for In-seq data were generated using edgeR’s exactTest function [100]. Statistical tests and significance values used for each experiment are indicated in figure legends. Error bars show SD unless otherwise indicated.