Transcriptome of Proteus mirabilis in the Murine Urinary Tract: Virulence and Nitrogen Assimilation Gene Expression^,

Abstract

The enteric bacterium Proteus mirabilis is a common cause of complicated urinary tract infections. In this study, microarrays were used to analyze P. mirabilis gene expression in vivo from experimentally infected mice. Urine was collected at 1, 3, and 7 days postinfection, and RNA was isolated from bacteria in the urine for transcriptional analysis. Across nine microarrays, 471 genes were upregulated and 82 were downregulated in vivo compared to in vitro broth culture. Genes upregulated in vivo encoded mannose-resistant Proteus-like (MR/P) fimbriae, urease, iron uptake systems, amino acid and peptide transporters, pyruvate metabolism enzymes, and a portion of the tricarboxylic acid (TCA) cycle enzymes. Flagella were downregulated. Ammonia assimilation gene glnA (glutamine synthetase) was repressed in vivo, while gdhA (glutamate dehydrogenase) was upregulated in vivo. Contrary to our expectations, ammonia availability due to urease activity in P. mirabilis did not drive this gene expression. A gdhA mutant was growth deficient in minimal medium with citrate as the sole carbon source, and loss of gdhA resulted in a significant fitness defect in the mouse model of urinary tract infection. Unlike Escherichia coli, which represses gdhA and upregulates glnA in vivo and cannot utilize citrate, the data suggest that P. mirabilis uses glutamate dehydrogenase to monitor carbon-nitrogen balance, and this ability contributes to the pathogenic potential of P. mirabilis in the urinary tract.

INTRODUCTION

The urinary tract is a common site of infection (for a recent review see reference 57). Recent strides have been made in understanding the pathogenesis of uropathogenic Escherichia coli (UPEC), the most common cause of urinary tract infection (UTI) in otherwise healthy individuals. There are now studies defining UPEC transcription during experimental UTI in mice (74), asymptomatic bacteriuria in humans (70), and human UTI (22). However, far less is known about bacterial agents of complicated UTI, which occurs in catheterized patients or patients with structural or functional abnormalities of the urinary tract and is frequently polymicrobial (28). UTI is the most common type of nosocomial infection, causing an estimated 424,000 cases and 13,000 deaths in U.S. hospitals in 2002 (34). Proteus mirabilis is one of the major causes of complicated UTI (81, 82) and is associated with particularly severe disease due to its ability to cause nephrolithiasis (kidney stones) and block urinary catheters (32, 41, 54).

P. mirabilis, a member of the Enterobacteriaceae, is well known for its ability to swarm across solid surfaces (55, 64). Urease, which mediates the hydrolysis of urea into ammonia and carbon dioxide, is perhaps the most potent virulence factor of P. mirabilis. Urease activity increases the pH of urine, leading to the precipitation of minerals and stone formation (53); it is required for P. mirabilis virulence during experimental UTI (33). P. mirabilis carries genes that encode 17 putative fimbrial operons (65), 5 of which have been named: MR/P (mannose-resistant Proteus-like) (9, 60), PMF (P. mirabilis fimbriae) (50), UCA (uroepithelial cell adhesin) (15, 85), PMP (P. mirabilis P-like pili) (11), and ATF (ambient-temperature fimbria) (47, 48). Two of these, MR/P (8) and PMF (49, 86), have been shown to play a role in virulence during experimental UTI and have been the focus of vaccine efforts (39, 66). Other known virulence factors for this organism include flagella (3, 51), toxins (HpmAB) (77, 80), proteases (Zap [83] and Lon [13]), autotransporters (Pta [2] and AipA and TaaP [1]), iron (27, 42) and zinc (59) uptake, and capsule (16). Signature-tagged mutagenesis studies have also highlighted the importance of central metabolism in the disease process (12, 26). Recent microarray studies examined total gene transcription during swarming (64) and iron restriction (27).

In this article, we investigate for the first time the transcriptome of P. mirabilis during infection, using a mouse model of ascending UTI. This is also the first study that investigates global transcription over time during UTI. The results have many parallels with UPEC gene expression during murine and human UTI, but also some important differences. Probing the distinctions in P. mirabilis and UPEC gene expression led to the discovery of an unexpected novel role for glutamate dehydrogenase (GDH) in P. mirabilis metabolism.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

P. mirabilis HI4320, isolated from the urine of an elderly, long-term-catheterized woman, has been previously described (54), sequenced, and annotated (65). E. coli CFT073 was isolated from the blood and urine of a patient with acute pyelonephritis (52) and has also been sequenced and annotated (84). The P. mirabilis ureC mutant has been previously reported (33). E. coli DH5α was used as the host for cloning mutational constructs. Bacteria were routinely cultured at 37°C with aeration in LB broth (10 g/liter tryptone, 5 g/liter yeast extract, 0.5 g/liter NaCl) or on LB solidified with 1.5% agar. The urease mutant was also cultured in filter-sterilized, pooled human urine from healthy adult volunteers. For studies on the effect of carbon source on P. mirabilis growth and gene transcription, minimal A medium (10) was utilized with 1% glucose, glycerol, citrate, or acetate as the carbon source.

Mouse model of ascending UTI.

To investigate in vivo gene transcription by P. mirabilis during UTI, 30 female CBA/J mice were transurethrally inoculated with approximately 1 × 10⁷ CFU of P. mirabilis HI4320, using a modification (32) of a previously published method (23). Briefly, an overnight culture of wild-type P. mirabilis HI4230 was adjusted to an estimated density of 2 × 10⁸ CFU/ml (optical density at 600 nm [OD₆₀₀] of 0.2). Six- to 8-week-old female CBA/J mice were inoculated transurethrally with 50 μl of the suspension (i.e., 1 × 10⁷ CFU) via a sterile polyethylene catheter, using an infusion pump (Harvard Apparatus). At 1, 3, and 7 days postinfection, urine was collected from mice directly into 14-ml polypropylene round-bottom tubes (Falcon, Becton-Dickinson) containing RNA Protect (Qiagen). At least 4 ml urine was collected per time point per experiment. Two milliliters of RNA Protect was used per ml of urine collected. The RNA-stabilized urine was placed on ice for at least 1 h, after which the fluid was separated from a crystalline precipitate that had formed (74). The fluid was centrifuged (10 min, 7,000 × g, 4°C), and the bacterial pellet was stored at −20°C before RNA extraction. The University of Michigan University Committee on Use and Care of Animals approved all mouse protocols.

RNA extraction and microarray analysis.

RNA was isolated with the RNeasy minikit (Qiagen). One column was used per ml of urine collected. RNA was treated with DNase (Ambion) and concentrated with an RNeasy MinElute column (Qiagen). The entire RNA sample (2 to 4 μg) was converted to cDNA and labeled with Cy5 CyDye (GE Healthcare) as previously reported (64). As a comparison condition, RNA was extracted from a mid-logarithmic-phase (OD₆₀₀ of 1.0) P. mirabilis HI4320 LB broth culture, converted to cDNA, and labeled with Cy3. The in vivo and in vitro cDNA samples were hybridized with a P. mirabilis HI4320-specific microarray as previously described (64). Slides were scanned with a ScanArray Express microarray scanner (Perkin Elmer) at 10-μm resolution and quantified using ScanArray Express software. The resulting data were normalized by total intensity, and median spot intensities were identified by using MIDAS (v. 2.22) software (71). The statistical analysis of microarray (SAM) algorithm (79) was conducted using the MeV program (v. 4.5.1) (71). Genes were considered to be significantly differentially regulated in vivo compared to in vitro if they were identified by SAM, had a median fold change value of at least 2-fold, and had a median spot intensity above background fluorescence. Genes that were differentially regulated over time were identified by using the Bayesian estimation of temporal regulation (BETR) algorithm (6) in MeV.

qRT-PCR.

RNA was converted to cDNA by using the Superscript first-strand synthesis system (Invitrogen) according to the manufacturer's protocol. The resulting cDNA was confirmed to be free of chromosomal DNA contamination by PCR on samples with or without reverse transcriptase (RT) treatment. Samples were adjusted to a uniform concentration (4 or 6 ng/μl) by using a Nanodrop ND-1000 spectrophotometer. Each quantitative RT-PCR (qRT-PCR) was set up in duplicate and consisted of 30 ng cDNA template, 150 nM each primer, and 12.5 μl 2× SYBR green PCR master mix (Stratagene). Target genes were amplified with an Mx3000P thermal cycler (Stratagene). Melting curve analysis was used to confirm a lack of primer dimers, and no-template controls were used to ensure there was no genomic DNA contamination of reagents. Data were normalized to rpoA (RNA polymerase A) or rpmC (encodes 50S ribosomal protein L29) as specified and analyzed by the threshold cycle (2^−ΔΔCT) method (43). Statistical analysis, including t tests and linear regression, was performed on log₂-transformed data by using Prism software (GraphPad). Primer sequences will be provided upon request.

Mutant construction.

An isogenic gdhA mutant was constructed by insertion of a kanamycin resistance cassette into the gdhA (PMI3008) coding sequence by the Targetron method (Sigma) as previously described (63). Basically, a group II intron was retargeted by PCR to specifically insert into gdhA. Intron insertion in gdhA was confirmed by PCR. The mutant was maintained on LB agar supplemented with kanamycin (25 μg/ml).

Preparation of crude protein extracts.

P. mirabilis cells were grown in 1-liter flasks under the specified conditions for each experiment, harvested by centrifugation at 6,000 × g for 15 min, and resuspended in 5 ml buffer (for urease assays, 20 mM sodium phosphate buffer [pH 7.0], 1 mM EDTA, and 5 mM β-mercaptoethanol; for GDH and GS assays, 50 mM sodium phosphate buffer [pH 7.0]). Cells were disrupted by two 20,000-lb/in² passages through a French pressure cell press (American Instruments Company) and immediately placed on ice. The sample was cleared by centrifugation at 30,000 × g for 30 min at 4°C to remove insoluble material. The resulting cell extract was used as the enzyme source for activity measurements. Total protein concentration was determined using the bicinchoninic acid (BCA) assay (Thermo Scientific) according to the manufacturer's instructions, using bovine serum albumin as the standard.

Enzyme assays.

Urease (EC 3.5.1.5) activity was determined by measuring released ammonia by the Berthelot reaction (24). Calibration curves used NH₄Cl as a standard. The reaction mixture contained 30 mM urea, 1 mM EDTA, 10 mM β-mercaptoethanol, and 20 mM sodium phosphate buffer (pH 7.5), in a total volume of 240 μl. The reaction was initiated by addition of 10-μl aliquots of various enzyme concentrations. After allowing the reaction to proceed for 5 to 10 min at room temperature, it was stopped by addition of 500 μl phenol-nitroprusside solution, 500 μl alkaline hypochlorite, and 2.5 ml of water, and color was allowed to develop for 5 min at room temperature. One unit of urease activity was defined as the amount of enzyme required to produce 1 μmol NH₃/min/ml. Jack bean urease (Sigma) was used as a positive control. Crude extracts of the P. mirabilis ΔureC strain grown under the same conditions and similarly processed were used to measure background activity. Absorbance at 570 nm was measured with a BioSpec-1601 spectrophotometer (Shimadzu).

Glutamine synthetase (GS; EC 6.3.1.2) activity was assayed using the γ-glutamyltransferase assay as previously described (75). Activity was measured in imidazole buffer both in the presence of 0.4 mM Mn²⁺ at pH 7.38 (both adenylylated and unadenylylated enzymes are active under these conditions) and in the presence of 60 mM Mg²⁺ and 0.4 mM Mn²⁺ at pH 7.2 (only unadenylylated enzyme is active). The reaction buffer contained 0.15 mM glutamine, 4.1% neutral NH₂OH, 0.4 mM Na⁺-ADP, and 2% KAsO₄ (pH 7.0) in 160 mM imidazole buffer (pH 7.4), and the divalent cation(s) in the concentrations specified above. The pH of the assay mix was adjusted according to the reaction. Briefly, 490-μl reactions were initiated by addition of 10-μl dilutions of crude P. mirabilis protein extract and allowed to proceed for 30 min at 37°C. The reaction was halted by addition of a stop mixture (3.3% FeCl₃, 2% trichloroacetic acid [TCA], 0.25 N HCl), and samples were briefly centrifuged at 4,000 × g to remove any precipitate. The production of γ-glutamylhydroxamate was determined at 540 nm in a Shimadzu BioSpec-1601 spectrophotometer. One unit of enzyme activity was defined as the amount of enzyme producing 1 μmol γ-glutamylhydroxamate/min/mg protein in the transfer reaction. Glutamine synthetase from E. coli (Sigma) was used as a positive control.

Glutamate dehydrogenase (EC 1.4.1.4) activity was measured in the forward (reductive amination activity) direction as described elsewhere (72). The reaction followed the oxidation of NADPH₂ spectrophotometrically at 340 nm and was monitored for 3 min at room temperature in a Genesys 10-S spectrophotometer (Thermo Scientific). The reaction mixture contained 100 mM Tris-HCl (pH 8.0), 100 mM NH₄Cl, 10 mM α-ketoglutarate, 0.1 mM NADPH₂, and appropriate dilutions of cell extracts in a total volume of 1 ml. Specific enzyme activity was calculated by using the NADPH₂ extinction coefficient of 6220 cm⁻¹ M⁻¹. One unit of enzyme activity was defined as 1 μmol of coenzyme oxidized/min/ml. Crude extracts of the P. mirabilis gdhAΩkan strain grown under the same conditions and similarly processed were used to measure background activity. Glutamate dehydrogenase from bovine liver (Sigma) was used as a positive control.

In vivo cochallenge.

Overnight cultures of wild-type P. mirabilis HI4320 and the isogenic gdhA mutant strain were individually adjusted to an estimated density of 2 × 10⁸ CFU/ml (OD₆₀₀ of 0.2). The gdhA mutant culture was mixed in a 1:1 ratio with the wild-type parent strain. Mice were transurethrally infected with a 50-μl suspension of the mixture (i.e., 1 × 10⁷ CFU) as described above. At 7 days postinoculation, bladders, kidneys, and spleens were removed from euthanized mice, homogenized in 3 ml phosphate-buffered saline (PBS) using an OMNI mechanical homogenizer (OMNI International), and plated on LB medium or LB medium supplemented with kanamycin by using a spiral plater (Autoplate 4000; Spiral Biotech) to assess bacterial burden. Colony counts were enumerated using a Qcount (Spiral Biotech). Wild-type HI4320 infection was determined by subtracting the number of colonies on the kanamycin plate from the number of colonies on the plain LB plate. Statistical significance was assessed using the two-tailed Wilcoxon matched-pairs test. The competitive index (CI) was calculated by dividing the ratio of mutant to wild-type output CFU to the ratio of mutant to wild-type input CFU. CI values below 1 indicate outcompetition of the mutant by the wild-type parent.

Growth curves.

Growth of bacterial strains was assessed with a Bioscreen C growth curve analyzer (Growth Curves, Ltd.). The optical density at 600 nm for triplicate cultures was measured at 15-min intervals. Cultures were maintained at 37°C with continuous shaking.

Microarray data accession number.

Microarray data have been deposited in the Gene Expression Omnibus (GEO) database under series GSE25977.

RESULTS

Transcriptome of Proteus mirabilis in the murine urinary tract.

Mice were transurethrally inoculated with P. mirabilis, and the infection was allowed to progress for 1 week. At 1, 3, and 7 days postinfection, transcription of bacteria in pooled voided mouse urine was measured by microarray and compared to transcription by P. mirabilis cultured to the logarithmic phase in LB broth. LB broth was selected for the in vitro condition, despite its complexity, for three reasons. LB broth represents a common laboratory medium for culture of enteric bacteria, the use of this medium allowed comparison to the only other in vivo transcriptome of a uropathogen, E. coli (74), and culture of P. mirabilis in urine ex vivo is cidal for the urease-producing bacterium.

At 7 days postinfection, there is a typical median of 1 × 10⁶ to 1 × 10⁷ CFU/ml in urine. Three independent microarrays were analyzed for each time point. To understand the overall pattern of gene expression in vivo, all nine microarrays were analyzed together. There were 471 genes significantly upregulated in vivo compared to in vitro, and there were 82 genes significantly downregulated in vivo. Table 1 lists the top 50 genes upregulated in vivo, Table 2 lists the 50 nonribosomal genes most highly expressed in vivo, and Table 3 lists the 50 genes most downregulated in vivo. The complete lists of differentially regulated genes are found in Tables S1 and S2 in the supplemental material. The genes encoding mannose-resistant Proteus-like (MR/P) fimbriae were very highly expressed in vivo; the nine genes comprising the mrp operon (40) were the nine most upregulated genes in vivo (Table 1). Indeed, mrpA, the gene encoding the major structural subunit of MR/P fimbriae, was the ninth most highly expressed nonribosomal gene in vivo (Table 2). Major classes of genes upregulated in vivo encode amino acid and carbohydrate transporters, energy production and conversion proteins, and ion transport (predominantly iron) proteins (Table 4). Genes encoding cell motility proteins were the most highly downregulated genes in vivo; 19 of the 50 most downregulated genes encoded flagellar components and chemotaxis proteins (Table 3). Major classes of genes downregulated in vivo, of which cell motility genes are the largest defined category, are depicted in Table 4.

Table 1.

The 50 genes most highly upregulated in vivo in this study

ORFc	Gene	Annotation	Ranka		Fold changeb
			In vivo	In vitro
PMI0263	mrpA	Major mannose-resistant/Proteus-like fimbrial protein	37	3,040	1,3806.33
PMI0264	mrpB	Fimbrial subunit	138	3,474	1,672.09
PMI0266	mrpD	Fimbrial chaperone protein	146	3,496	1,358.22
PMI0268	mrpF	Fimbrial subunit	218	2,905	1,227.47
PMI0271	mrpJ	Fimbrial operon regulator	191	2,688	904.14
PMI0269	mrpG	Fimbrial subunit	255	2,804	693.49
PMI0267	mrpE	Fimbrial subunit	272	3,018	616.84
PMI0265	mrpC	Fimbrial outer membrane usher protein	230	3,124	598.01
PMI0270	mrpH	Fimbrial adhesin	458	2,891	514.01
PMI3233		Putative membrane protein	350	2,194	344.22
PMI1706	dmsB	Anaerobic dimethyl sulfoxide reductase chain B	530	2,818	250.99
PMI0173		Hypothetical protein	547	2,892	232.47
PMI2673	gntK	Thermoresistant gluconokinase	537	3,125	220.23
PMI1205		Putative anaerobic dimethyl sulfoxide reductase chain B	533	2,487	210.95
PMI3234		Putative membrane protein	558	2,280	204.00
PMI1688		Putative di-/tripeptide transporter	546	2,490	183.19
PMI2969		Major facilitator superfamily transporter	503	2,286	167.77
PMI0730		Conserved hypothetical protein	514	1,973	161.35
PMI0249	betU	Putative secondary glycine betaine transporter	738	2,723	153.81
PMI0361		Putative phosphodiesterase	645	3,614	153.01
PMI2134	agaA	N-Acetylglucosamine-6-phosphate deacetylase	652	3,233	148.09
PMI1705	dmsA	Dimethyl sulfoxide reductase chain A	881	2,851	140.24
PMI0391		Putative sigma 54 modulation protein	493	2,108	131.78
PMI3232		Putative peptidase	403	1,924	122.47
PMI0973		Putative heat shock protein	930	2,728	122.19
PMI0706	focA	Probable formate transporter	568	1,898	122.12
PMI0187	dsdA	d-Serine dehydratase	909	1,990	121.54
PMI1568		Putative exported protease	620	2,266	119.64
PMI1406	gadC	Probable glutamate/γ-aminobutyrate antiporter	413	1,871	118.51
PMI2946		ABC transporter, permease protein	1,133	2,874	113.19
PMI3515		Phosphotransferase system IIBC component	1,003	3,455	108.08
PMI2672	gntU	Low-affinity gluconate transporter	797	2,149	102.07
PMI2948		Probable methyltransferase	661	2,215	100.26
PMI0287		Putative amidohydrolase/metallopeptidase	764	2,288	100.05
PMI2192		LuxR family transcriptional regulator	697	2,632	99.95
PMI0272		Probable transport protein	1,008	2,662	98.42
PMI3674		Probable aminohydrolase	813	2,423	97.48
PMI1601		Putative sodium:alanine symporter	664	2,278	96.94
PMI1725		Putative phage holin (lysis protein)	1,063	3,227	96.93
PMI3675		Putative C4-dicarboxylate transporter	931	2,866	92.96
PMI2292	ptsG	Phosphotransferase system, glucose-specific IIBC component	479	2,153	88.71
PMI2970		Putative membrane protein	854	2,925	88.61
PMI1707	dmsC	Anaerobic dimethyl sulfoxide reductase chain C	655	2,641	87.94
PMI2846	dppB	Dipeptide ABC transporter, permease protein	759	2,116	87.06
PMI2466		Putative plasmid-related protein	1,286	3,103	83.09
PMI0144		Putative ATP-binding protein	1,204	2,462	82.19
PMI2847	dppA	Dipeptide ABC transporter, substrate-binding protein	1,241	2,463	81.34
PMI1945	ireA	Putative TonB-dependent ferric siderophore receptor	722	2,667	78.99
PMI0434	gltL	Glutamate/aspartate ABC transporter, ATP-binding protein	1,048	2,580	74.90
PMI2529	rpiB	Putative ribose 5-phosphate isomerase	159	1,396	72.28

^aMedian expression rank across nine microarrays (including ribosomal genes): 1 is most highly expressed, and 3,719 is most weakly expressed.

^bMedian fold change in vivo compared to LB broth across nine microarrays.

^cORF, open reading frame.

Table 2.

The 50 nonribosomal genes most highly expressed in vivo in this study

ORFc	Gene	Annotation	Ranka		Fold changeb
			In vivo	In vitro
PMI1035		Translation initiation factor IF-3	9	3	−1.33
PMI0765	ompF	Outer membrane porin	13	3	−1.25
PMI3251	tufB	Elongation factor Tu	13	6	−1.23
PMI3275	secY	Preprotein translocase SecY subunit	23	20	1.03
PMI3280	rpoA	DNA-directed RNA polymerase alpha chain	26	19	−1.32
PMI2792	tufB	Elongation factor Tu	28	29	−1.16
PMI0856		Conserved hypothetical protein	29	36	−1.85
PMI0862	acpP	Acyl carrier protein	33	16	−1.66
PMI0263	mrpA	Major mannose-resistant/Proteus-like fimbria	37	3,040	13,806.33
PMI0785	ompA	Outer membrane protein A	37	50	1.21
PMI1676	cspC	Cold shock-like protein	39	9	−2.09
PMI2771	hupA	DNA-binding protein HU-alpha (HU-2)	52	60	−1.39
PMI3375	priB	Primosomal replication protein N	52	34	−1.61
PMI2793	fusA	Elongation factor G (EF-G)	52	80	1.17
PMI0716	ihfB	Integration host factor beta-subunit	57	53	−1.25
PMI0384	rimM	16S rRNA processing protein	59	59	−1.28
PMI2284	tsf	Elongation factor Ts	59	75	−1.00
PMI3059	atpE	ATP synthase C chain	59	57	−1.47
PMI1213	ahpC	Alkyl hydroperoxide reductase subunit C	62	165	6.14
PMI3062	atpA	ATP synthase alpha chain	62	68	−1.07
PMI3415	secG	Protein export membrane protein	66	71	−1.05
PMI0073	ahpC	Putative alkyl hydroperoxide reductase	67	104	1.83
PMI3063	atpG	ATP synthase gamma chain	68	70	−1.14
PMI0385	trmD	tRNA (guanine-N1)-methyltransferase	69	65	−1.69
PMI0585	pal	Peptidoglycan-associated lipoprotein	71	46	−2.60
PMI3060	atpF	ATP synthase B chain	71	60	−1.39
PMI1504	gapA	Glyceraldehyde 3-phosphate dehydrogenase A	72	81	1.33
PMI0118	hupB	DNA-binding protein HU-beta	75	76	−1.70
PMI3064	atpD	ATP synthase beta chain	75	69	−1.18
PMI3065	atpC	ATP synthase epsilon chain	78	84	−1.22
PMI3061	atpH	ATP synthase delta chain	81	83	−1.15
PMI2544	groS	10-kDa chaperonin	82	66	−1.26
PMI3057	atpI	ATP synthase protein I	82	85	−1.35
PMI3296		Conserved hypothetical protein	83	96	−1.11
PMI0418	cspE	Putative cold shock protein	84	69	−2.52
PMI0861	fabG	3-Oxoacyl-[acyl-carrier protein] reductase	85	79	−1.67
PMI1039	pheT	Phenylalanyl-tRNA synthetase beta chain	93	103	1.06
PMI2784	rpoC	DNA-directed RNA polymerase beta′ subunit	93	95	−1.15
PMI1898	grcA	Autonomous glycyl radical cofactor	95	788	19.18
PMI3622	fis	DNA-binding protein Fis	96	90	−1.37
PMI1034	thrS	Threonyl-tRNA synthetase	96	90	−1.27
PMI0114	tig	Trigger factor	98	98	−1.15
PMI1828	ptsH	Phosphotransferase system phosphocarrier protein	100	683	16.77
PMI1488	hns	DNA-binding protein (histone-like structuring protein)	100	93	−1.41
PMI3364	miaA	tRNA delta(2)-isopentenylpyrophosphate transferase	102	137	−1.11
PMI2833		Putative endoribonuclease	103	124	−1.02
PMI0842		TonB-dependent receptor	104	913	17.18
PMI3058	atpB	ATP synthase A chain	106	77	−1.66
PMI0243	fbaA	Fructose-bisphosphate aldolase class II	110	144	1.61
PMI0691	infA	Translation initiation factor IF-1	111	101	−1.60

^aMedian expression rank across nine microarrays (including ribosomal genes): 1 is most highly expressed, and 3,719 is most weakly expressed.

^bMedian fold change in vivo compared to LB broth across nine microarrays.

^cORF, open reading frame.

Table 3.

The 50 genes most highly downregulated in vivo in this study

ORFc	Gene	Annotation	Ranka		Fold changeb
			In vivo	In vitro
PMI1629	fliE	Flagellar hook-basal body complex protein	2,748	492	−88.25
PMI0994		Putative lipoprotein	2,785	506	−70.63
PMI1654	flgB	Flagellar basal-body rod protein	2,254	465	−47.91
PMI1618	fliA	RNA polymerase sigma factor for flagellar operon	1,448	136	−39.10
PMI1651	flgE	Flagellar hook protein	2,232	439	−35.30
PMI1877	pmfA	Major fimbrial subunit	671	66	−29.64
PMI1650	flgF	Flagellar basal-body rod protein	1,943	435	−23.85
PMI1096		2-Oxoglutarate and Fe(II)-dependent oxygenase superfamily protein	1,794	243	−21.27
PMI1636	fliL	Flagellar protein FliL	2,049	503	−20.33
PMI1649	flgG	Flagellar basal-body rod protein (distal rod protein)	1,716	502	−20.03
PMI0993		Putative lipoprotein	2,330	540	−19.98
PMI1653	flgC	Flagellar basal-body rod protein	2,525	703	−19.72
PMI1644	flgL	Flagellar hook-associated protein 3 (hook-filament junction protein)	2,323	674	−19.41
PMI0488		Putative phage membrane protein	1,245	394	−19.10
PMI1941		Conserved hypothetical protein	2,460	1,084	−18.60
PMI2109	mgtE	Magnesium transporter	1,769	412	−15.85
PMI1638	fliN	Flagellar motor switch protein	2,409	710	−15.41
PMI1637	fliM	Flagellar motor switch protein	2,410	679	−15.38
PMI2814		RpiR-family transcriptional regulator	2,364	873	−15.02
PMI3177	kbl	2-Amino-3-ketobutyrate coenzyme A ligase	1,871	471	−12.99
PMI1631	fliG	Flagellar motor switch protein	2,449	704	−12.48
PMI1469	fim8A	Fimbrial subunit	1,787	593	−11.97
PMI0022		Putative cell killing protein	1,859	441	−11.88
PMI1595	pagP	Putative antimicrobial peptide resistance and lipid A acylation protein	1,662	550	−10.92
PMI1622	fliS	Flagellar protein FliS	1,825	905	−10.83
PMI0291	treB	Phosphotransferase system, trehalose-specific IIBC component	2,792	941	−10.65
PMI0596		Putative membrane protein	1,338	174	−10.58
PMI1656	flgM	Negative regulator of flagellin synthesis	1,614	718	−10.24
PMI3178	tdh	l-Threonine 3-dehydrogenase	1,464	482	−10.21
PMI1943		Major facilitator superfamily transporter	1,415	207	−10.15
PMI1620	flaA/fliC1	Flagellin 1	750	108	−9.59
PMI1666	cheD	Methyl-accepting chemotaxis protein	2,481	936	−9.53
PMI1645	flgK	Flagellar hook-associated protein 1	2,057	807	−9.29
PMI0044		Putative outer membrane protein	1,442	363	−8.33
PMI1621	flaD/fliD	Flagellar hook-associated protein 2	2,306	1,261	−8.29
PMI3211	glpF	Glycerol uptake facilitator protein	1,171	217	−7.52
PMI1506	mipA	MltA-interacting protein precursor	566	107	−7.37
PMI1879	pmfD	Fimbrial chaperone protein	1,504	659	−6.98
PMI1489	wbnF	Probable nucleotide sugar epimerase	1,730	387	−6.98
PMI1617	fliZ	FliZ protein	979	240	−5.85
PMI1061	fim5A	Fimbrial protein	1,945	752	−5.51
PMI2203		Putative isochorismatase	1,828	799	−5.42
PMI0580		Putative thioesterase	1,432	550	−4.99
PMI2093	speB	Agmatinase	853	172	−4.98
PMI2094	speA	Biosynthetic arginine decarboxylase	898	285	−4.93
PMI0582	tolR	Biopolymer transport protein	1,280	379	−4.77
PMI1958		Putative DNA-damage-inducible protein	1,436	658	−4.31
PMI2930	glpD/glyD	Aerobic glycerol-3-phosphate dehydrogenase	458	97	−3.96
PMI0682	artP	Arginine ABC transporter, ATP-binding protein	470	187	−3.74
PMI1286	osmB	Osmotically inducible lipoprotein	1,692	925	−3.68

^aMedian expression rank across nine microarrays (including ribosomal genes): 1 is most highly expressed, and 3,719 is most weakly expressed.

^bMedian fold change in vivo compared to LB broth across nine microarrays.

^cORF, open reading frame.

Table 4.

Categories of genes differentially regulated in vivo

Category	No. of genes:
	Upregulated in vivo	Downregulated in vivo
Amino acid transport and metabolism	58	7
Carbohydrate transport and metabolism	37	4
Cell cycle control, cell division, and chromosome partitioning	6	0
Cell motility	0	19
Cell wall/membrane/envelope biogenesis	19	9
Coenzyme transport and metabolism	17	3
Defense mechanisms	1	0
Energy production and conversion	50	4
Fimbriae	10	4
Function unknown	57	12
General function prediction only	42	4
Inorganic ion transport and metabolism	41	2
Intracellular trafficking, secretion, and vesicular transport	2	1
Lipid transport and metabolism	6	0
Nucleotide transport and metabolism	17	0
Phage and mobile element	20	1
Posttranslational modification, protein turnover, and chaperones	24	2
Replication, recombination, and repair	19	1
Secondary metabolite biosynthesis, transport, and catabolism	4	1
Signal transduction mechanisms	5	0
Transcription	25	3
Translation, ribosomal structure, and biogenesis	12	5

Virulence gene expression.

Regulation of genes previously identified in P. mirabilis virulence was specifically examined (Fig. 1). As mentioned, the mrp operon is both highly upregulated (Fig. 1) and highly expressed in vivo. However, pmfA, which encodes the major structural subunit of Proteus mirabilis fimbriae (PMF), was downregulated in vivo despite having a role in bladder (49, 86) and kidney (86) colonization in mice. Genes coding for two other fimbriae, previously identified as virulence factors by signature-tagged mutagenesis (12, 26), were either downregulated (fim8A) or upregulated (fim14A), although neither was highly expressed in vivo. Genes encoding three P. mirabilis toxins, zapA (44, 83), hemolysin hpmA (77), and pta (Proteus toxic agglutinin) (2), were expressed at approximately background levels in vivo and were inconsistently upregulated. The entire urease operon, which is induced in the presence of urea, was both highly expressed and highly upregulated in vivo. Peptide transporter genes dppA and oppA, which have been implicated in virulence for both P. mirabilis (64) and UPEC (4), were upregulated in vivo; iron uptake genes (exbBD, sitABC, hmuST, ireA, nrpY, feoAB, PMI0331, PMI0363, PMI0842, PMI1424, PMI1437, and PMI2957 to PMI2960) were similarly upregulated.

Fig. 1.

Heat map of expression data for specific virulence-associated genes, depicting the ratio of expression in LB broth versus in vivo. The legend at the top indicates the color associated with log₂ fold change: red, upregulated in vivo; green, downregulated in vivo; black, not differentially regulated.

Other genes of interest that were upregulated in vivo included pyruvate metabolism genes (aceEF, pta, ackA, poxB, pflB, and dsdA), genes likely involved in osmoprotection (betU, putP, gltS, gltL, and gadC), and eight genes within the previously identified mobile integrated and conjugative element ICEPm1 (PMI2568, PMI2573, PMI2595, PMI2598, PMI2608, PMI2627, PMI2628, and PMI2629) (18).

Temporal gene regulation.

Microarrays were also analyzed separately at each time point (1, 3, and 7 days postinfection), and some temporal regulation was noted. At 1 day postinfection, 93 genes were significantly upregulated in vivo; 5 genes were downregulated. At 3 days postinfection, 294 genes were upregulated and 6 were downregulated. At 7 days postinfection, 204 genes were upregulated and 1 gene was downregulated (data not shown). The BETR algorithm identified three groups of genes that were differentially regulated over time (Fig. 2). The first group consists of genes that were highly induced in vivo at day 1 but by day 7 were less upregulated (Fig. 2A). This group of seven genes included three genes from the mrp operon (mrpA, mrpE, and mrpH). Three other genes in this group are predicted to encode transporters (PMI0272 and PMI2947), including a putative di-/tripeptide transporter (PMI1688). The second group of genes was repressed at day 1 but by day 7 had similar expression compared to genes from cells cultured in LB broth (Fig. 2B). Of the 10 genes in group 2, 5 encode flagellar components (fliN, flgD, fliH, flhD, and flgL). Three genes in this group encode putative lipoproteins (PMI0977, PMI0994, and PMI1840). The third group consisted of genes that were either slightly repressed or not differentially regulated at day 1 but were induced at day 7 (Fig. 2C). There were eight genes in this group, including the transcriptional regulator hexA.

Fig. 2.

P. mirabilis genes are temporally regulated in vivo. Nine microarrays were analyzed for differential regulation of genes over time using the BETR algorithm. (A) Group 1 genes were highly induced in vivo at day 1 but less highly expressed by day 7. (B) Group 2 genes were repressed in vivo at day 1 but were not differentially expressed compared to in vitro at day 7. (C) Group 3 genes were either slightly repressed or not differentially regulated at day 1, but were induced at day 7.

Validation of the microarrays by qRT-PCR.

Expression of 15 genes that were found to be differentially regulated by microarray in vivo compared to LB was also measured by qRT-PCR (Fig. 3A). Urine collected from mice 3 days postinfection was used as the in vivo source of RNA. Microarray and qRT-PCR data were in agreement for 13 of the 15 genes, including the very strong in vivo upregulation of mrpA. Two genes predicted to be upregulated in vivo by microarray, ipaD and PMI2192, were either downregulated (ipaD) or not differentially regulated (PMI2192) by qRT-PCR. To test the concordance of microarray and qRT-PCR data, the log₂-transformed fold changes were plotted against one another (Fig. 3B). Linear regression calculations confirmed the significant agreement between microarray and qRT-PCR data (r² = 0.5186, P = 0.0025).

Fig. 3.

qRT-PCR validation of in vivo microarrays. (A) Expression of 15 genes in vivo compared to during logarithmic-phase growth in LB culture as measured by qRT-PCR. (B) Concordance of microarray and qRT-PCR data for the 15 genes depicted in panel A. The log₂-transformed fold change values are plotted, with linear regression shown. Median fold change at 3 days was used for the microarray data points. The r² value is 0.5186, and the slope has a significant deviation from zero (P = 0.0025).

Glutamate dehydrogenase is upregulated during urinary tract infection.

Ammonia is the preferred nitrogen source for enteric bacteria. The two primary ammonia incorporation systems, gdhA (PMI3008; glutamate dehydrogenase [GDH]) and glnA (PMI2882; glutamine synthetase [GS]), were inversely regulated in vivo, with gdhA upregulated 10.75-fold and glnA downregulated 2.78-fold (Fig. 1). The opposite result was observed for UPEC in vivo, in which gdhA was slightly downregulated (1.75-fold) and glnA was upregulated 12-fold (74). Glutamine synthetase is the primary ammonia uptake system when nitrogen levels are limited, while glutamate dehydrogenase confers ammonia uptake when ammonia levels are high or, correspondingly, when energy is limited (25).

Nitrogen assimilation appears to be handled differently by P. mirabilis than UPEC during UTI based on microarray and qPCR data assessing transcription. However, because nitrogen assimilation can be regulated posttranslationally as well as transcriptionally, we attempted to correlate the transcript levels of urease, GDH, and GS with the enzyme activities of these gene products (see below).

GDH is not upregulated in excess ammonia alone.

To define in vitro culture conditions that yielded in vivo gene expression patterns, we conducted a series of experiments. Upregulation of gdhA and repression of glnA in vivo by P. mirabilis suggested that ammonia incorporation drives the expression of these genes in P. mirabilis. To test the effect of nitrogen availability on gdhA and glnA gene expression, P. mirabilis HI4320 was cultured in LB broth with or without 500 mM NH₄Cl. As assessed by qRT-PCR (Fig. 4A), gdhA was not upregulated; however, as expected, glnA was downregulated. Similar results were obtained by using NH₄Cl concentrations ranging from 100 mM to 1 M (data not shown). As assessed by enzyme activity (Fig. 4B), GDH was also not significantly upregulated when P. mirabilis was cultured in 500 mM NH₄Cl. GS enzyme activity was downregulated under this condition. Therefore, in this experiment, transcription, as assessed by microarray and qRT-PCR, serves as a proxy for enzyme activity. NH₄Cl, however, in this experiment was supplied exogenously, while in vivo, urea in urine is hydrolyzed to ammonia by intracellular urease. Therefore, P. mirabilis was also cultured in LB broth supplemented with 400 mM urea (within the normal human physiological concentration range for urine). Again, gdhA (Fig. 4A) and GDH enzyme activity (Fig. 4B) were not upregulated compared to those in the cells cultured in LB broth, although urease gene ureA expression (mean of 43.5-fold) and urease enzyme activity were highly upregulated during culture in urea (Fig. 4A and B). The pH of the urea culture was approximately 8.5 at the time bacteria were harvested, as a consequence of urease activity.

Fig. 4.

Exogenous nitrogen is not the driving force behind in vivo gdhA and glnA expression. (A) RNA was isolated from P. mirabilis HI4320 cultured in LB supplemented with either 500 mM NH₄Cl or 400 mM urea. No significant changes in gdhA or glnA were found by qRT-PCR in comparison to unsupplemented LB. However, addition of urea significantly upregulated expression of ureA. (B) Enzyme activities were determined for whole-cell lysates of P. mirabilis HI4320 cultured under identical conditions to those described for panel A. Glutamate dehydrogenase (GDH), total glutamine synthetase (GS), and urease (URE) enzyme activities (average of five technical replicates) were measured and expressed as fold change relative to preparations from P. mirabilis cultured in LB broth. (C) A ureC mutant was cultured in LB broth or pooled human urine with or without 100 mM NH₄Cl, and expression of gdhA and glnA was measured by qRT-PCR. (A and C) Expression was normalized to rpoA. At least three independent experiments were conducted for each condition tested. Error bars represent standard errors of the mean. *, P < 0.05.

To assess whether a component in urine might contribute to the differential expression of gdhA and glnA, we used a ureC mutant that is unable to make functional urease (33) and therefore does not increase the pH of urea-containing cultures. This mutant was cultured to the logarithmic phase in pooled human urine with or without 100 mM NH₄Cl supplementation. Compared to ureC cells cultured in LB, gdhA was significantly upregulated in urine with or without NH₄Cl (38.6-fold and 24.5-fold, respectively) (Fig. 4C). However, glnA was not differentially regulated. These data suggest that upregulation of gdhA in vivo was not due to ammonia excess per se and that some other undefined factor was regulating gdhA and glnA.

High osmolarity alone does not induce GDH.

Because gdhA and glnA were expressed differentially between P. mirabilis and UPEC in vivo, we sought to determine additional differences in the physiology of these species that might provide clues regarding this pattern of gene expression. Urine is generally hyperosmotic, averaging 500 to 800 mosmol/kg in humans and 1,040 mosmol/kg in female CBA mice (21), which can interfere with bacterial growth (7). In addition, glutamate, a product of gdhA upregulation, may be used to buffer against osmotic stress (reviewed in reference 45). Thus, P. mirabilis HI4320 and E. coli CFT073 were each cultured in LB supplemented with NH₄Cl over a range of 0 to 1 M. While growth was similar for P. mirabilis and E. coli at lower salt concentrations, E. coli was unable to grow at 1 M NH₄Cl (Fig. 5A). In contrast, P. mirabilis growth in 1 M NH₄Cl was only slightly retarded compared to that at lower salt concentrations (Fig. 5A). However, because gdhA is not upregulated in 500 mM NH₄Cl (Fig. 4A) or 1 M NH₄Cl (data not shown), osmotic buffering was ruled out as the sole cause for differential gene expression in P. mirabilis and E. coli.

Fig. 5.

Comparison of growth of P. mirabilis HI4320 (Pm) and E. coli CFT073 (Ec) under high-osmolarity conditions or with specific carbon sources. (A) LB broth supplemented with either 100 mM or 1 M NH₄Cl; (B) minimal A medium with 1% glycerol, citrate, or acetate as the carbon source; (C) P. mirabilis HI4320 and HI4320 gdhAΩkan cultured in LB broth or minimal A medium with glycerol or citrate as the carbon source. At least three independent experiments were conducted for each condition tested; one representative experiment with three technical replicates is shown.

P. mirabilis uses citrate as a carbon source, while UPEC does not.

Citrate, secreted in human urine at a rate of 3 to 20 mg/kg/day (5), is also a key intermediate of the TCA cycle and was a candidate carbon source that upregulates gdhA. Most P. mirabilis strains can use citrate as the sole carbon source (29), while most E. coli strains cannot. To test whether P. mirabilis HI4320 and E. coli CFT073 differ in their ability to use specific carbon sources, both strains were cultured in minimal medium supplemented with 1% glycerol, citrate, or acetate. P. mirabilis HI4320 was able to use all three carbon sources, while UPEC strain CFT073 was unable to utilize citrate (Fig. 5B).

A gdhA mutant has a growth defect in minimal medium containing citrate as the sole carbon source.

Opposing regulation of gdhA in P. mirabilis and E. coli, coupled with differing abilities to utilize citrate, led us to hypothesize that GDH might play a unique role in central metabolism for P. mirabilis. Therefore, gdhA was inactivated by insertion of a targeted transposon, and the ability of the mutant to grow in rich and minimal media was compared to that of the wild-type isogenic parent strain (Fig. 5C). The gdhA mutant displayed a slight growth defect in minimal medium with glycerol as a carbon source but a more pronounced defect when citrate was the sole carbon source (Fig. 5C). As with E. coli (68), there was no difference in growth rates of wild-type P. mirabilis and the gdhA mutant in LB broth.

Culture in citrate recapitulates in vivo nitrogen gene expression patterns.

Since P. mirabilis gdhA expression is increased in vivo, and a gdhA mutant has a growth defect when cultured in minimal medium with citrate as a carbon source, we postulated that citrate could be directly affecting expression of ammonia incorporation genes gdhA and glnA. Therefore, wild-type P. mirabilis was cultured to the mid-logarithmic phase in LB or minimal medium, and gene expression of gdhA and glnA was measured by qRT-PCR. Expression of gdhA (Fig. 6A) and GDH enzyme activity (Fig. 6B) were significantly increased in all three carbon sources tested (glucose, glycerol, and citrate) compared to the levels in LB broth, but glnA expression was significantly decreased only when citrate was used as the carbon source (Fig. 6A). GS enzyme activity was also lower in citrate than in glucose or glycerol (Fig. 6B). Thus, culture in citrate mimics in vivo conditions with respect to the expression pattern of gdhA and glnA.

Fig. 6.

Citrate as the sole carbon source drives in vivo expression patterns of gdhA and glnA. (A) P. mirabilis HI4320 was cultured in LB or minimal A medium with glucose, glycerol, or citrate as the carbon source. qRT-PCR analysis of RNA revealed upregulation of gdhA in minimal A medium with all three carbon sources, but gdhA was only downregulated when citrate was used as the carbon source. At least three independent experiments were conducted for each condition tested. Expression of rpoA, routinely used for normalization of gene expression by P. mirabilis, was decreased in minimal medium relative to LB; therefore, rpmC was used to normalize expression in these experiments. *, P < 0.05. (B) Enzyme activities were determined for whole-cell lysates of P. mirabilis HI4320 cultured under identical conditions to those described for panel A. Glutamate dehydrogenase (GDH), total glutamine synthetase (GS), and unadenylylated (active) GS enzyme activities (average of five technical replicates from each of two independent preparations) were measured and expressed as fold change relative to preparations from P. mirabilis cultured in LB.

gdhA contributes to in vivo fitness of P. mirabilis.

Based on the experiments described above, we hypothesized that GDH plays a central metabolic role during infection of the urinary tract. To test this, the gdhA mutant was mixed 1:1 with the isogenic parent strain and transurethrally inoculated into the bladders of 10 mice. After 7 days, the bacterial burdens of the gdhA mutant and wild type were enumerated in the bladders, kidneys, and spleens of these mice. As assessed by the competitive indices, the gdhA mutant was defective in colonization of all three sites tested (P < 0.05) (Fig. 7), suggesting that GDH is necessary for P. mirabilis fitness in the urinary tract, perhaps serving as a checkpoint for intracellular carbon and nitrogen levels.

Fig. 7.

HI4320 gdhAΩkan is less fit in vivo than wild-type HI4320. Ten female CBA/J mice were transurethrally infected with a 1:1 mixture of HI4320 and HI4320 gdhAΩkan. After 7 days, bacteria were enumerated in the bladder, kidneys, and spleen. The competitive index (CI) of the gdhA mutant compared to the wild type was plotted; each dot represents the CI from one mouse. The mutant was significantly outcompeted by the wild type in all three organs. Bars indicate median values.

DISCUSSION

This is the first microarray analysis of transcription by an agent of complicated UTI during an infection, as well as the first study examining global temporal control of bacterial gene expression during UTI. Previous studies have examined gene expression by E. coli during UTI in mice (74), in humans with asymptomatic bacteriuria (70), and in patients with UTI (22). There are several similarities in gene expression by P. mirabilis and E. coli during UTI. Both P. mirabilis and E. coli activate systems involved in iron uptake, pyruvate catabolism, peptide uptake, and osmoprotection. Both species have robust metabolism in vivo, as indicated by high expression of ribosomal genes. P. mirabilis and UPEC each upregulate a specific fimbrial operon in vivo (mrp and fim, respectively), although the upregulation of mrp is much stronger than that of fim (13,806-fold versus 7.82-fold for mrpA and fimA, respectively). Likewise, both pathogens downregulate flagellar and chemotaxis genes in vivo.

P. mirabilis in vivo global gene expression not only was verified by qRT-PCR but also agrees with previous reports on P. mirabilis virulence. Genes encoding MR/P fimbriae were the most highly upregulated genes in vivo, with the major structural gene mrpA upregulated 13,800-fold (median). Previous studies have shown that increased expression of MR/P fimbriae correlates with higher levels of colonization in the mouse bladder (36), the majority of bacteria in the bladder and ureters express MR/P fimbriae (31), and an mrp mutant is attenuated in the mouse model of ascending UTI (8). Furthermore, vaccine trials in mice using intact MR/P fimbriae (39) or recombinant components of MR/P fimbriae (38, 39, 73) reduce bacterial burden after challenge. Expression of MR/P fimbriae has not been tested during human infection; this experiment will be critical to future vaccine efforts since UPEC highly expresses type I fimbriae in mice (56, 74), and these fimbriae are required for virulence in mice (14, 46), but type I fimbriae are not well expressed or possibly transiently expressed during human UTI (22). The last gene of the mrp operon, mrpJ, encodes a transcriptional repressor of flagella (40). The high expression of this gene in vivo is consistent with reduced expression of flagella during murine UTI. Interestingly, the mrp operon upregulation is highest at days 1 and 3 postinfection (mrpJ median of 1,189-fold upregulated), and less at day 7 (mrpJ median of 126-fold upregulated); at the same time, the repression of flagellar genes may be easing by day 7. (The flagellar master regulator flhD is downregulated a median of 12.5-fold at days 1 and 3 and upregulated a median of 1.59-fold at day 7.) Additionally, another regulator of flagellum-mediated motility, hexA (64), was upregulated at day 7. These data suggest a potential increased role for flagella and corresponding lesser role for MR/P fimbriae at this later stage in the disease process. During experimental UPEC UTI, flagellar genes are transiently upregulated 4 h postinfection, when the bacteria are ascending the ureters from the bladder to the kidneys (35). It will be useful to examine P. mirabilis fimbrial and flagellar gene expression at early time points (less than 24 h), as well as to study global temporal gene transcription during UPEC UTI.

Other genes found to be upregulated in vivo also agree with single-gene-knockout virulence studies. Iron uptake systems upregulated in vivo are required for colonization (hmuR2) (42) and virulence (pbtA) (27) in mice, and many are immunogenic (PMI0409, PMI0842, hmuR2, ireA, and PMI2596) (58). Zinc uptake systems are also required for virulence (59), are immunogenic in mice (58), and were upregulated in vivo. Likewise, urease, a well-documented virulence factor for P. mirabilis (12, 20, 32, 33, 54), was upregulated. Toxin genes (hpmA, zapA, and pta) were a notable exception; these genes were expressed at or below background levels both in vitro and in vivo. It is possible that expression of these genes is more localized to specific sites in the urinary tract or that subtle changes in transcript level lead to much larger changes in protein activity. Two classes of genes which are involved with virulence were downregulated in vivo: flagella and PMF fimbriae. Flagella are energetically expensive to construct and maintain; they also activate innate immune response via Toll-like receptor 5 (17). Thus, flagellar regulation is likely tightly controlled in vivo, much as it is temporally controlled during experimental UPEC UTI (35). The role of PMF fimbriae is unknown at this time, although knockout studies have shown these fimbriae are required for full virulence (49, 86, 87), and components of PMF fimbriae have been successfully used in experimental vaccines (73). It is possible these fimbriae are required for adherence to specific tissues, which prevents PMF-expressing bacteria from being shed into the urine. Alternatively, like flagella, PMF fimbriae are only required during a limited time during UTI progression and may be tightly regulated in vivo.

There were several differences between P. mirabilis and E. coli transcription in vivo. Although both P. mirabilis and UPEC upregulate pyruvate catabolism pathways in vivo, there are more, and more highly upregulated, genes leading to acetate secretion by P. mirabilis. Increased acetate secretion, which lowers pH, might be a useful counter to urease-induced alkalinization. In addition to peptide uptake, surprisingly, P. mirabilis also upregulates genes involved in glucose uptake (crr and ptsG) and glycolysis (pgi, pfkA, and gpmAB) in vivo. Certainly, both E. coli and P. mirabilis can utilize glucose. With the exception of uncontrolled diabetes, generally glucose is not present in urine. Thus, the apparent ability of P. mirabilis to obtain glucose during UTI warrants further investigation.

With respect to nitrogen assimilation, there are some intriguing differences between P. mirabilis and UPEC transcription in vivo. The potent urease encoded by P. mirabilis, but absent in UPEC, ensures an abundant source of ammonia in the urinary tract (Fig. 8). Thus, while UPEC is starved for nitrogen in vivo (22, 74), the carbon-nitrogen balance for P. mirabilis may be flipped. Likewise, the genes encoding a portion of the TCA cycle (gltA, acnB, and icd), comprising the steps from acetyl coenzyme A (acetyl-CoA) formation to α-ketoglutarate synthesis, are upregulated by P. mirabilis in vivo (Fig. 8), while the rest of the TCA cycle genes in P. mirabilis were not differentially regulated in vivo. No such upregulation occurs in UPEC during UTI. (Indeed one gene, acnA, is even downregulated in vivo.) At this point in the P. mirabilis cycle, α-ketoglutarate can continue in the TCA cycle to make succinyl-CoA or be converted by incorporation of ammonia to generate l-glutamate via glutamate dehydrogenase (GDH). We initially hypothesized that expression of gdhA and glnA differs in P. mirabilis and UPEC in the urinary tract because of nitrogen availability.

Fig. 8.

In vivo upregulated genes suggest carbon-nitrogen balance drives P. mirabilis central metabolism. The intersection of the TCA cycle with pyruvate catabolism and ammonia assimilation is depicted. Genes upregulated in vivo >2-fold are shown in red; fold increases (blue) are indicated in parentheses.

Indeed, regulation of nitrogen uptake and assimilation is complex, with multiple levels of transcriptional and posttranslational control (reviewed in references 37 and 67). The carbon/nitrogen ratios for P. mirabilis and UPEC in the urinary tract are certainly not the same. Continuous high levels of GDH activity in vivo would lead to diminished α-ketoglutarate pools (30), and therefore, upregulation of upstream TCA cycle enzymes would be required to maintain the full TCA cycle (i.e., to convert α-ketoglutarate to succinyl-CoA) (Fig. 8). Although inverse in vivo regulation of gdhA and glnA by P. mirabilis and UPEC initially was suspected to be solely due to urease activity in P. mirabilis, this was not the case. Addition of exogenous ammonium (NH₄Cl) or intracellular ammonia (catalyzed by urease-mediated urea hydrolysis) in culture did not lead to the differential gene expression observed in vivo. However, choice of carbon source, such as citrate, a major component of urine, had a profound effect on gdhA expression. Rich media (19), the presence of specific amino acids in minimal medium (68), or a greater carbon/nitrogen ratio (69) have been reported to repress gdhA. Only when citrate was used as the carbon source, which forces the TCA cycle to operate, did we see a concomitant decrease in glnA expression. However, although P. mirabilis HI4320 can utilize citrate as a carbon source and E. coli cannot, there are more favorable carbon sources available in the urinary tract. Additionally, not all P. mirabilis strains utilize citrate as the sole carbon source (29), and future studies on such strains will be informative.

Glutamate dehydrogenase plays a critical metabolic role during P. mirabilis UTI. gdhA mutants, however, generally have only a subtle phenotype, if any, in other species (68). For example, an E. coli gdhA mutant only showed a defect when placed in direct competition with the wild-type parent strain in batch culture during glucose limitation (25). Neisseria meningitidis has also been shown to require gdhA for full virulence in an infant rat model (76), although this species lacks glutamate synthase (gltBD) (62, 78) and therefore cannot synthesize glutamate via the glutamine synthetase-GOGAT system (67). Furthermore, one of two N. meningitidis gdhA transcripts is regulated by adjoining gene gdhR (61), which is not found in enteric bacteria.

Global analysis of pathogens within the host continues to yield new insights into pathogenesis and central metabolism. This analysis of transcription during UTI will provide a baseline for future studies of early events in transcription, proteomics, tissue-specific expression, and mixed-species infections.