|Home | About | Journals | Submit | Contact Us | Français|
Pathogenicity islands (PAIs) are a specific group of genomic islands that contribute to genomic variability and virulence of bacterial pathogens. Using a strain-specific comparative genomic hybridization array, we report the identification of a 94-kb PAI, designated ICEPm1, that is common to Proteus mirabilis, Providencia stuartii, and Morganella morganii. These organisms are highly prevalent etiologic agents of catheter-associated urinary tract infections (caUTI), the most common hospital acquired infection. ICEPm1 carries virulence factors that are important for colonization of the urinary tract, including a known toxin (Proteus toxic agglutinin) and the high pathogenicity island of Yersinia spp. In addition, this PAI shares homology and gene organization similar to the PAIs of other bacterial pathogens, several of which have been classified as mobile integrative and conjugative elements (ICEs). Isolates from this study were cultured from patients with caUTI and show identical sequence similarity at three loci within ICEPm1, suggesting its transfer between bacterial genera. Screening for the presence of ICEPm1 among P. mirabilis colonizing isolates showed that ICEPm1 is more prevalent in urine isolates compared to P. mirabilis strains isolated from other body sites (P < 0.0001), further suggesting that it contributes to niche specificity and is positively selected for in the urinary tract.
Genomic variability between bacterial strains of the same species can result from gene gain and loss. Change in genome composition is facilitated by a variety of mechanisms and, among these, the acquisition of genes by horizontal gene transfer mediates rapid changes in the bacterial genome (1, 26, 38). An assortment of mobile genetic elements, including genomic islands, are acquired horizontally by bacteria, thereby contributing to the dynamic genome evolution observed in bacterial species (1, 26, 38). Genomic islands are 30- to 250-kb segments of DNA that are flanked by direct repeats (DRs), associated with tRNA genes, have a G+C content differing from the surrounding chromosome, and carry genes that confer a fitness advantage to the organism (13, 16). Genomic islands that encode virulence properties important for bacterial survival and pathogenicity within the host are called pathogenicity islands (PAIs) and when under positive selection continue to propagate and evolve within the chromosome (16, 17).
A subset of PAIs has been identified that can excise from the bacterial chromosome following a recombination event at the DRs flanking the island and actively transfer via a type IV secretion system (T4SS) (13, 24). The subsequent site-specific integration of the island into a tRNA gene of the new host's chromosome, as well as island excision, is dependent on a functional integrase gene present at either end of the island (24). These mobile genomic islands, identified as integrative and conjugative elements (ICEs) contain similar genetic composition and structural organization that consists of integration and replication modules, as well as a conjugation system that reflects the functional capacity to self-transfer (11). These modules are mostly syntenic among ICEs, although interspersed with genes that are specific for that particular bacterial species. These additional, or cargo genes, encode proteins important for survival in the environment of that bacterial host (24, 48).
Uropathogenic Escherichia coli, the most common cause of urinary tract infection, harbors multiple PAIs, but less is known about PAIs in other uropathogens such as Proteus mirabilis, Providencia stuartii, and Morganella morganii. Many organisms colonize the urinary tract and create biofilms on urinary catheters (28). Of these, P. mirabilis, P. stuartii, and M. morganii are of particular interest due to their high prevalence rates among long-term-catheterized individuals, as well as their propensity to cause catheter obstruction and urolithiasis (32). Catheter-associated urinary tract infections (caUTIs), the most common hospital-acquired infection, represent up to 40% of nosocomial infections (49) and can result in adverse outcomes, including sepsis and urinary stones. Once catheterized for ≥30 days, patients inevitably develop bacteriuria (51) and, in up to 77% of bacteriuric urine specimens, more than one bacterial species is isolated (28, 51). The polymicrobial etiology of caUTI suggests a common mechanism of colonization that enhances the ability of these organisms to survive in the catheterized urinary tract.
In the present study, comparative genomic hybridization (CGH) microarray analysis revealed genes common among three bacterial species isolated from polymicrobial caUTIs and therefore potentially important for urinary tract colonization. Using a strain-specific CGH to detect common genes among different bacterial genera allowed for detection of regions of the genome that were highly conserved. This approach led to the identification of a previously unknown PAI common to P. mirabilis, P. stuartii, and M. morganii that possesses the properties consistent with a mobile PAI or ICE. The newly identified PAI carries several virulence factors, including the Proteus toxic agglutinin (Pta) and the high pathogenicity island (HPI) of Yersinia spp. and is found more commonly in P. mirabilis urinary isolates than colonizing isolates from different body sites. DNA sequence of the PAI at several loci is identical between P. mirabilis, P. stuartii, and M. morganii, suggesting not only that its acquisition by these organisms was recent but also that active transfer may be occurring among these bacteria of different genera.
P. mirabilis HI4320, P. stuartii BE2467, and M. morganii TA43, as well as the other 87 caUTI isolates, were cultured from urine samples from patients catheterized for ≥30 days with bacteriuria (>105 CFU/ml) at the time of bacterial culture. These patients were residents of two long-term care facilities (LTCFs) and were not being treated with antibiotics at the time of urine culture. The urine specimens were obtained from the distal end of the urinary catheter, cultured, identified, and stored at −80°C as previously described (51). P. stuartii ATCC 25827 was obtained from the American Type Culture Collection (ATCC). As part of a separate study, colonizing strains of patients from 14 LTCFs were isolated from the anterior nares, oropharynx, groin, perianal area, or wound sources by using Culturette rayon-tipped swabs (Becton Dickinson, Inc., Cockeysville, MD) and stored at −80°C (33). A total of 33 strains were used in the present study, 14 of which came from patients that had an indwelling catheter or feeding tube. Isolates cultured from the same patient and of the same species, but isolated on different days (from either study population), were excluded from the study.
Individual colonies were inoculated into 50 ml of Luria-Bertani medium and cultured overnight at 37°C. Bacteria were collected by centrifugation (3 min, 10,000 × g, 25°C) and suspended in one-tenth volume TE (10 mM Tris-HCl, 1 mM EDTA) (pH 8.0) containing 10% sodium dodecyl sulfate and proteinase K (20 mg/ml). After 1 h of incubation at 37°C, DNA was purified by using CTAB (cetyltrimethylammonium bromide)/NaCl and phenol-chloroform extraction (35).
Genomic DNA was digested with HincII (or HindIII for P. stuartii ATCC 25827), followed by aminoallyl-labeling (2:1 aminoallyl-dUTP:dTTP). Genomic DNA was labeled with the appropriate cyanine fluorescent molecules (Cy-3 or Cy-5; GE Healthcare) and dried to completion (19). Cy-dye-labeled DNA was hybridized (50% formamide, 5× SSC [1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate], 0.1% sodium dodecyl sulfate, 0.6 μg of salmon sperm DNA/μl) to prehybridized microarray slides at 42°C for 20 h (18). P. mirabilis HI4320 was used as the reference strain in each flip-dye array pair. The P. mirabilis microarray consists of 3,719 70-mer oligonucleotide probes printed in triplicate onto UltraGAPSII glass slides (Corning) by Microarrays, Inc. Each oligonucleotide probe is unique to the 3,719 predicted open reading frames (ORFs) of the HI4320 genome and has a target melting temperature of 74°C (probes designed and synthesized by Operon Technologies, Inc.) (M. Pearson and H. Mobley, unpublished data).
Cy-3 and Cy-5 fluorescence intensities of each spot were acquired by using a Perkin-Elmer ScanArray Express microarray scanner and processed by Spotfinder (43). Background corrected spot intensities were imported into the Ginkgo CGH data analysis tool (Pathogen Functional Genomics Resource Center). Spots having intensities less than the fifth percentile of the overall intensities in the reference strain were discarded. If intensities in the test strains were undetectable, values were imputed based on the fifth percentile of intensities for the test strain in order to retain spots that did not hybridize in the test strain for further analysis. Once reference spots were removed and intensities for the test spots were imputed, in-slide replicates were merged, and the flip-dye consistency checking was performed.
GACK analysis (25) was used to determine dynamic cutoff values of intensity ratios for designating genes as present or divergent, which is a method more appropriate for highly divergent strains. Determining a cutoff value in this way allows for a different intensity ratio to be selected based on the same transition point on the curve of the log2 distributions for each hybridization experiment. Examination of the distribution of log2 intensity ratios in typical CGH experiments shows that the distributions vary between hybridizations and that the distribution is normal with a long left tail representing divergent genes (25). The distribution of our data is similar and yet, because we have a large number of probes that failed to hybridize, we saw a distribution with a long right tail. Therefore, the data were transformed by taking the inverse of all log2 ratios, which yielded a distribution with a long left tail suitable for GACK analysis. A binary analysis was used for determining the number of genes either present or absent/divergent, and a graded analysis was used for construction of the heatmap using the TreeView program (44).
Isolated colonies were cultured in Luria-Bertani broth to an optical density at 600 nm of 0.8 at 37°C with aeration. DNA was isolated from 1 ml of bacterial suspension by using a QIAquick column (Qiagen) according to the manufacturer's instructions. DNA was RNase treated and dissolved in 200 μl of the supplied elution buffer (Qiagen). PCR was conducted using primers flanking ORFs: PMI2551 (forward, CAGAAGATTACATGAATAATG; reverse, GAGAGTGTGATGAGATGTGAAT), PMI2602 (forward, GCGAATGAACTTCACCA; reverse, GCCACTAATCAGAGGGAGT), and PMI2641 (forward, GCACGCTCTGCTCCGCC; reverse, CGGGAGGTGCGTCCATG) (Invitrogen). Primers for amplification of PMI3255 (forward, CGTTGATGCTCTGATGCGTCT; reverse, GAACTACGTGTTCGCGAAGC), as a positive DNA control, were designed within the sequences of the two flanking ORFs because PMI3255 did not hybridize in the array experiments. PCRs were carried out in 25-μl reactions containing 30 ng of genomic DNA, 1× buffer, 250 μM deoxynucleoside triphosphates, 0.75 μl of Taq DNA polymerase, and 0.4 μM concentrations of forward and reverse primers. Reactions were amplified in a thermocycler at 95°C for 2 min, followed by 30 cycles of 95°C for 15 s, 55°C for 30 s, and 72°C for 60 s. Associations were determined by using the Fisher exact test (GraphPad Prism).
Purified PCR products were sequenced (ABI model 3730) and aligned by using the Wilbur-Lipman method (Lasergene). Artemis software was used for visualization of direct repeats and calculation of G+C content of the PAI (42).
Genomic DNA from catheter-associated bacteriuria isolates P. stuartii BE2467, M. morganii TA43, and fecal strain P. stuartii ATCC 25827 (test strains) were hybridized to a P. mirabilis HI4320 DNA microarray. Very few probes hybridized significantly with genomic DNA from each test strain. GACK analysis classified 365, 116, and 67 genes as conserved between P. mirabilis and P. stuartii BE2467, M. morganii, and P. stuartii ATCC 25827, respectively, whereas 3,354, 3,533, and 3,561 genes were determined to be absent or divergent (Fig. (Fig.1).1). Among the conserved genes, 110 genes were conserved in P. stuartii BE2467 and M. morganii and, of those, 20 were also conserved in P. stuartii ATCC 25827.
The majority of genes classified as conserved in the test strains are located in three contiguous regions within the P. mirabilis genome (Fig. (Fig.2D).2D). The first region (PMI0456 to PMI0538) consists of 81 phage-related genes, as well as the previously characterized uca genes, which are important in uroepithelial cell adhesion (52). A second contiguous region, PMI2549 to PMI2641, found in P. stuartii BE2467 and M. morganii but not P. stuartii ATCC 25827 genomic DNA was determined to be a PAI, designated ICEPm1. The third region (PMI3251 to PMI3284) represents ORFs encoding 28 ribosomal proteins.
To validate the findings of the CGH analyses, genes located in the beginning, middle, and end of ICEPm1 were PCR amplified. Sequence and alignment analysis of PMI2551, PMI2602, and PMI2641 revealed each gene to be 100% identical at the nucleotide level over their entire coding length between P. mirabilis, P. stuartii BE2467, and M. morganii. PCR amplification of P. stuartii ATCC 25827 genomic DNA for genes PMI2551, PMI2602, and PMI2641 yielded no products. Within the region of ribosomal proteins, PMI3255 was also PCR amplified, sequenced, and aligned in each strain used for the microarray experiments. PMI3255 was 90.5 and 90.3% similar at the nucleotide level in P. stuartii BE2467 and M. morganii, respectively, compared to the P. mirabilis sequence.
Loci PMI2549 to PMI2641 of P. mirabilis represent a previously uncharacterized PAI common to P. mirabilis HI4320, P. stuartii BE2467, and M. morganii TA43. This 94-kb region encodes 91 ORFs, has a G+C content of 44.84% (which differs substantially from that of the P. mirabilis genome [38.88%]) (Fig. (Fig.2C),2C), is flanked by 52-bp repeats, and is integrated into a tRNAPhe gene. This region (ICEPm1) carries several genes involved in DNA mobility, characteristic of PAIs, including an integrase, six transposases, and five plasmid-transfer related proteins (Fig. (Fig.2A).2A). All 91 ORFs were absent in P. stuartii fecal strain ATCC 25827 (Fig. (Fig.2D),2D), indicating that this strain lacks ICEPm1 and suggests its heterogeneous distribution in the population.
Despite being a newly recognized PAI for P. mirabilis, P. stuartii, and M. morganii, BlastP comparison (4) of all 91 ORFs showed similarity to significant portions of six well-known PAIs of other bacterial pathogens (Table (Table1):1): SPI-7 of Salmonella enterica serovar Typhi CT18 (39), HAI-2 of Pectobacterium amylovara SCRI1043 (previously Erwinia carotovora subsp. atroseptica) (6), YAPIYE of Yersinia enterocolitica 8081 (50), PAPI-1 of Pseudomonas aeruginosa PA14 (20), ICEHin1056 of Haemophilus influenzae (23), and an unnamed PAI in Photorhabdus luminescens TT01 (14). In addition, several ORFs were highly similar to a contiguous region of DNA of Serratia proteamaculans 586 that is flanked by two tRNAPhe genes, suggesting that this region is also a PAI. Although most homologous genes were syntenic within the PAIs of these other species, no one organism shared similarity with all 91 ORFs searched, highlighting the highly mosaic structure of the island (Fig. 3).
We identified a common structure among these PAIs and ICEPm1, consisting of a set of core genes, common to all PAIs discussed, supplemented by genes unique to each PAI, considered the variable region or cargo genes (Table (Table1).1). The core genes constitute putative integration, replication, and conjugation modules (Fig. (Fig.2B,2B, shaded yellow). DRs flank the ICEPm1 and, at the left-most end, the 52-bp direct repeat is located within the 5′ coding end of a tRNAPhe gene (attL). The direct repeat at the right-most end (attR) is part of the 3′ end of a truncated tRNAPhe gene.
The integration module at the left end of ICEPm1 (PMI2549 to PMI2554) encodes a site-specific integrase (PMI2549). Genes homologous to PMI2549 in the other PAIs encode site-specific recombinases of the tyrosine-like family. The module upstream of attR, the rightmost end of ICEPm1 (PMI2627 to PMI2642) encodes a topoisomerase, DNA helicase, and a chromosome-partitioning protein, suggesting its involvement in DNA replication. The chromosome-partitioning related protein is the terminal gene in the island and one of the homologues, soj (RL115) of P. aeruginosa, has been implicated in the stability of the extrachromosomal form of PAPI-1 (41). The integration and replication modules flanked by the att sites highlight the distinct boundaries of ICEPm1 from that of the surrounding chromosome.
Another important core segment found in ICEPm1 (PMI2569 to PMI2592) shows homology to a T4SS that is important for DNA transfer of ICEHin1056 (21). This 26-gene region encodes eight putative exported proteins and nine putative membrane proteins. In addition, 15 of 26 genes encode a predicted signal peptide sequence, and 14 of 26 genes encode one to three predicted transmembrane domains, common features of proteins in a T4SS (22).
The cargo genes of ICEPm1 (Fig. (Fig.2B2B [shaded gray]) are located in three variable regions: one consisting of 13 hypothetical genes, a second region of 19 genes, 13 of which are pseudogenes and, lastly, the nrp operon (PMI2596 to PMI2604). This operon encodes genes for the synthesis, transport, and uptake of the iron siderophore yersiniabactin and is homologous to the region of YAPIYE described to be the HPI common in Yersinia spp. (12).
To establish a correlation between ICEPm1 and pathogenicity, we tested 87 urinary isolates from long-term-catheterized individuals and 33 colonizing isolates of P. mirabilis, P. stuartii, and M. morganii for the presence of ICEPm1. Colonizing isolates were cultured from patients of LTCFs and isolated from the oropharynx, nasopharynx, wound, groin, or perianal area. PCR analysis of PMI2551, PMI2602, and PMI2642 (representing genes located at the beginning, middle, and end of the island, respectively) served as a proxy to estimate the ICEPm1 presence in colonizing strains. Successful PCR amplification of all three loci was considered positive for the presence of ICEPm1. Of 87 urinary isolates tested, 100% (39/39) of the P. mirabilis strains were positive for ICEPm1, whereas the prevalences in P. stuartii and M. morganii were 60% (6/10) and 28.9% (11/38), respectively (Table (Table2).2). In addition, P. mirabilis urinary isolates were significantly more likely to harbor the island 100% (39/39) compared to P. mirabilis colonizing isolates 65.2% (15/23) (P < 0.0001; odds ratio [OR] = 43.32 [1.648, 797.5]). ICEPm1 was found in 40% (2/5) of M. morganii colonizing isolates and 100% (5/5) of P. stuartii colonizing isolates, yielding no significant association between ICEPm1 and isolate origin for these species (Table (Table2).2). Using PMI2602 (nrpT) as a marker of HPI presence, we identified HPI in 100% (39/39) of P. mirabilis, 70% of (7/10) P. stuartii, and 89.5% of (34/38) M. morganii urinary isolates (Table (Table2).2). The prevalences of HPI in colonizing isolates were 69.6% (16/23), 40% (2/5), and 100% (5/5) in P. mirabilis, M. morganii, and P. stuartii, respectively. This shows a significant association (P = 0.0005; OR = 35.9 [1.9, 666.1]) between the prevalence of HPI in urine isolates compared to isolates from other body sites among P. mirabilis isolates, as well as that HPI is more common among urinary isolates than colonizing isolates (P = 0.006; OR = 49.7 [1.7, 14.5]) when all species are grouped together (Table (Table22).
PAIs are a subgroup of genomic islands that contribute not only to variation in genomic composition within species but also to the virulence and pathogenesis of microorganisms. Several recently described PAIs have been shown to excise and self-transfer to other bacteria, facilitating bacterial adaptation to surrounding niches (8, 21, 23, 31, 34, 36, 40, 41). Here, we present the discovery of a highly modular and highly conserved PAI, designated ICEPm1, of P. mirabilis, P. stuartii, and M. morganii strains that has structure similar to these recently described mobile PAIs. The 100% sequence identity at three loci within ICEPm1 among the three isolates suggests DNA transfer between these bacteria of different genera has occurred, although further DNA sequencing is required to confirm the identical element in each of these strains. In addition, we show that the ICEPm1 is significantly more likely to be present in P. mirabilis urinary isolates than in colonizing isolates from other body sites, suggesting its contribution to a common mechanism of colonization and pathogenicity among agents of caUTI.
Considering the genetic relatedness and common niche inhabited by the strains used in our study, we found it surprising that the majority of P. mirabilis microarray probes failed to hybridize significantly with P. stuartii and M. morganii genomic DNA. Upon sequence investigation, we determined the stringency of our array to be quite high, requiring sequences of approximately >93% identity for hybridization to occur. CGH is rarely applied across different genera of bacteria. In fact, to our knowledge, CGH has only been used previously to compare the nearly identical Shigella and Escherichia genomes (15). The present study is the first to describe the use of a strain-specific microarray to investigate genome variability across three disparate genera. This is most likely due to the fact that very few genes hybridize under these conditions in divergent strains rendering this technique undesirable for gene discovery applications. Nevertheless, by hybridizing genomic DNA from bacteria of different genera to a strain-specific array, we were able to quickly identify regions of very high sequence similarity between genetically divergent organisms and thus genes of interest conserved among these uropathogens.
Although P. mirabilis, P. stuartii, and M. morganii are common commensals of the human body, they are opportunistic pathogens that cause caUTIs and other diseases (29). The association between the origin of isolate and the presence of ICEPm1 in P. mirabilis isolates suggests that the island contributes to urinary tract colonization and/or pathogenicity. Organisms causing caUTI are generally of the gut microbiota and self-inoculated by patients (30). Therefore, the island could confer pathogenic potential to commensal or opportunistic organisms once introduced into the urinary tract. The presence of ICEPm1 may be a determining factor for whether or not colonization occurs and, in this case, we would expect a high prevalence of ICEPm1 in the urine isolates but a heterogenic distribution in strains from other body sites, as was observed for P. mirabilis isolates.
The core modules found in ICEPm1 show similarity to regions of other genomic islands and in these islands have been shown to be necessary for mobility and transfer. Precise excision of SPI-7, YAPIYE, and PAPI-1 has been observed (8, 36, 41, 50) and, PAPI-1, a functional integrase (xerC, RL002) and chromosome partitioning protein (soj, RL115) (located in the integration and replication modules, respectively), was important for the excision and stability of the island (41). Subsequent transfer of the islands into PAI-negative recipient strains is reported for PAPI-1 and ICEHin1056, and involvement of a T4SS has been shown to be required for transfer of ICEHin1056 in H. influenzae strains (21, 41). Phylogenetic analysis of T4SS genes in ICEHin1056, SPI-7, PAPI-1, HAI-2, and the PAI from P. luminescens TTO1 shows that the T4SS in these organisms are related and yet distinct from traditional T4SSs and have been named genomic island-like T4SSs (21). Further experiments are required to understand the excision and transfer capabilities of ICEPm1; however, the structural similarity of ICEPm1 to other ICEs, including core genes that encode a putative T4SS, suggests that it may be classified as an ICE (10, 11, 24, 48).
The variable regions, containing the cargo genes, of the PAIs described have important functions in their respective organisms (Table (Table3).3). Therefore, it is not surprising that an important virulence factor such as HPI is located in the variable region of ICEPm1. This cluster of virulence genes is involved in iron acquisition and could enhance fitness during iron limitation in the host. Consistent with this, we see that the ICEPm1 is significantly more prevalent in urinary isolates, likely because it is well known that the urinary tract is iron limited.
In our isolates, HPI is associated with ICEPm1, as well as present when ICEPm1 is not. We attribute this phenomenon to the independent transfer of HPI (7), and in strains lacking ICEPm1 we found that HPI has integrated into another region of the genome. In addition, the divergent G+C content of the HPI from the flanking regions of ICEPm1 (Fig. (Fig.2C)2C) suggests that the HPI was mobilized and integrated within ICEPm1 as a distinct event. Indeed, ICEEc1 of E. coli ECOR31 and ICEKp1 of K. pneumoniae NTUH-K2044 both harbor HPI and can transfer between bacterial strains and yet do not show similar sequence identity to other regions of ICEPm1 outside of the yersiniabactin operon (27, 46). Either of these elements could be potential elements that introduced HPI into other regions of the chromosome. The significant association observed between HPI and urinary isolates, when all species are grouped together, supports the importance of this virulence factor and thus reflects its widespread distribution within the Enterobacteriaceae family (5, 45, 47).
The importance of ICEPm1 in uropathogenicity has been shown, specifically for PMI2596 (tonB-dependent iron receptor) (37), PMI2575 (Proteus toxin agglutinin [Pta]) (2, 3), and PMI2605 (nrpG) (9), as mutant constructs of these genes are attenuated in the mouse model of ascending UTI. Furthermore, only three ORFs in the variable region (outside of HPI) share homology to members of the Enterobacteriaceae, suggesting this PAI may contain novel virulence genes that have not been previously described in this family. Future experiments will focus on the discovery of other virulence phenotypes, including investigation into ICEPm1's contribution to biofilm formation, an important virulence property in urinary catheter colonization. Characterization of ICEPm1's ability to mobilize and transfer and the conditions affecting these events will also be undertaken. The presence of known virulence factors carried on ICEPm1, taken together with the high prevalence of ICEPm1 and HPI in P. mirabilis urinary isolates, suggest positive selection for ICEPm1 among uropathogenic strains. Our description of ICEPm1 in P. mirabilis, P. stuartii, and M. morganii isolates inhabiting a similar niche highlights the importance of understanding the selective pressures positively selecting for acquisition and retention of this island, as well as those that influence horizontal transfer of this island, thus disseminating virulence factors to a broad range of bacterial hosts.
We do not have an association that may pose a conflict of interest.
This study was supported in part by Public Health Service grants AI43363 and AI59722 and an NIA K23 Career Development Award to L.M. from the National Institutes of Health.
We thank Melanie Pearson for providing guidance on the use of the P. mirabilis microarray and Christopher J. Alteri for critical reading of the manuscript.
Editor: S. R. Blanke
Published ahead of print on 17 August 2009.