Utility of In Vivo Transcription Profiling for Identifying Pseudomonas aeruginosa Genes Needed for Gastrointestinal Colonization and Dissemination

Microarray analysis of Pseudomonas aeruginosa mRNA transcripts expressed in vivo during animal infection has not been previously used to investigate potential virulence factors needed in this setting. We compared mRNA expression in bacterial cells recovered from the gastrointestinal (GI) tracts of P. aeruginosa-colonized mice to that of P. aeruginosa in the drinking water used to colonize the mice. Genes associated with biofilm formation and type III secretion (T3SS) had markedly increased expression in the GI tract. A non-redundant transposon library in P. aeruginosa strain PA14 was used to test mutants in genes identified as having increased transcription during in vivo colonization. All of the Tn-library mutants in biofilm-associated genes had an attenuated ability to form biofilms in vitro, but there were no significant differences in GI colonization and dissemination between these mutants and WT P. aeruginosa PA14. To evaluate T3SS factors, we tested GI colonization and neutropenia-induced dissemination of both deletional (PAO1 and PAK) and insertional (PA14) mutants in four genes in the P. aeruginosa T3SS, exoS or exoU, exoT, and popB. There were no significant differences in GI colonization among these mutant strains and their WT counterparts, whereas rates of survival following dissemination were significantly decreased in mice infected by the T3SS mutant strains. However, there was a variable, strain-dependent effect on overall survival between parental and T3SS mutants. Thus, increased transcription of genes during in vivo murine GI colonization is not predictive of an essential role for the gene product in either colonization or overall survival following induction of neutropenia.


Introduction
Modern molecular tools allow for analysis of levels of mRNA in bacterial cells living in different environments and many such studies have been applied to Pseudomonas aeruginosa [1,2,3,4,5]. However, apart from one study analyzing mRNA expression in P. aeruginosa present in the sputum of a single cystic fibrosis patient [6], no other studies have analyzed P. aeruginosa gene expression during in vivo infection. Part of the problem lies with recovering sufficient mRNA from bacterial cells directly isolated from infected tissues for microarray analysis. In a previously described model of murine GI colonization and dissemination following induction of neutropenia [7] we found that high levels of P. aeruginosa could be recovered from the mouse cecum, potentially identifying a source of in vivo bacterial mRNA sufficient for microarray analysis. This model mimics the morbidity and mortality of immunocompromised hosts such as patients with leukemia, severe burn wounds or recipients of organ transplants [8]. In many patients at-risk for P. aeruginosa infection (i.e. surgical patients, cancer patients receiving chemotherapy) the gastrointestinal (GI) tract is believed to be the main tissue initially colonized by this organism, often times allowing for translocation to extra-gastrointestinal sites and, in the worst cases, development of life-threatening sepsis [9,10]. In this patient group, P. aeruginosa has the highest case-fatality rate among all gram-negative pathogens [11]. The mere presence of P. aeruginosa in the GI tract of critically-ill surgical patients is associated with a 70% mortality rate, a three-fold increase over physiologically matched critically-ill patients not infected with P. aeruginosa [12].
Here we report that enough in vivo-expressed P. aeruginosa mRNA can be recovered from microbial cells in the cecum of colonized mice and determined that genes involved in two major bacterial phenotypes thought to be important in virulence, production of biofilms and elaboration of type III secretion system (T3SS) effectors, had consistent increases in mRNA levels when compared with expression levels in the bacteria living in the water used to colonize the mice. Using P. aeruginosa strains from both a non-redundant transposon insert library in strain PA14 [13] as well as deletional mutants in strains PAK and PA01, we confirmed that the genes with increased transcription that were involved biofilm production were deficient in this phenotype in vitro, but were not compromised in either their ability to colonize the GI tract in vivo or disseminate following induction of neutropenia. Mutants in T3SS effector genes were also not deficient for GI colonization. Mice infected with these mutants had an increase in time of survival after induction of neutropenia but variable, straindependent effects on overall mouse survival with different T3SS mutants. Our findings validate a system for measuring gene expression of P. aeruginosa during an important phase of the pathogenesis of human disease, but also indicate that increased mRNA expression is not necessarily predictive of a phenotype needed by this organism to induce or maintain in vivo infection or cause WT levels of pathology.

Transcription Profiling of P. aeruginosa in the Murine GI Tract
We initially identified P. aeruginosa genes that were differentially transcribed in the GI tract of eight colonized C3H mice in comparison to genes expressed in sterile water containing 1500 U penicillin G/mL. For both the basal state and colonization states, three separate biologic replicates were performed. We chose to use P. aeruginosa recovered from the drinking water as the baseline condition since water contaminated with P.aeruginosa is often cited as the source for both nosocomial and community-acquired P. aeruginosa infections. Within the hospital, P. aeruginosa is most often isolated from water sources such as sinks, drains, toilets, and showers [14] and has been identified by molecular techniques (e.g. genotyping) to be the causative pathogen in documented P. aeruginosa infections [15]. Furthermore, community-acquired P. aeruginosa infections are also often associated with exposure to contaminated water sources (i.e. swimming pools, hot tubs) [16]. Given the epidemiology of P. aeruginosa acquisition and the fact that our murine model was designed with the aim of recapitulating the pathogenesis of P. aeruginosa infection in a human host, we felt that using bacteria recovered from the drinking water was the best representative of a baseline condition. Although the P. aeruginosa arrays were developed using the PAO1 genome [17], previous studies have shown that the strain PAK transcriptome of the genes conserved between these two strains can be reliably measured with P. aeruginosa Genome Arrays (of the 5886 P. aeruginosa probe sets included on the microarray, 5678 were detected using PAK DNA) [18]. An average of the microarray hybridization intensity data (3 microarrays) for both baseline (drinking water) and colonization state (cecums) was generated and compared (Gene Expression Omnibus Accession Number GSE22665). Statistical analysis of the derived transcriptomes indicated that 1089 genes (including 321 hypothetical and unclassified genes) showed a two-fold increase in expression from the baseline to the GI colonization state, and 423 genes (including 229 hypothetical and unclassified genes) showed a two-fold decrease in expression from baseline ( Figure 1).
We found that the transcript levels of genes functionally classified as being involved in Cell Wall/LPS/Capsule were generally increased by two-fold or more over expression when the bacteria are in water: 34 of 86 genes (40%) involved in Cell Wall/ LPS/Capsule function were expressed at a higher level in the GI tract versus 2 of 86 (2%) genes expressed at a ,2-fold lower level. In contrast, relatively few genes involved in membrane protein synthesis were changed (3 of 43 increased versus 7 of 43 decreased), and genes involved in motility and attachment were roughly evenly increased or decreased (18 of 67 genes had increased mRNA levels while 20 of 67 genes had decreased mRNA levels). Of the 16 biofilm genes that we chose to study, 11 of 16 genes (69%) were expressed at .2-fold level in the GI tract compared to expression levels of bacteria in water. Most interesting, however, was that 98% of the genes (39 of 40) involved in the Type III Secretion System had increased expression and none had decreased expression (Table 1). Of the 5 T3SS genes that we studied, all of the genes (with the exception of exoU, since exoU is absent in strain PAK) showed a seven-fold or higher increase in transcript levels ( Table 1).

Selection of P. aeruginosa PA14 strains with insertions in genes needed for biofilm formation
Using the data generated from several studies [19,20,21,22], we identified genes that have been shown to be important for all stages of in vitro biofilm formation: fleR, flgK,fliP, sadB, crc, pilB, pilC, pilY1, PA3782, lasI, rhlA, and rpoN. Next, we cross-referenced the murine GI colonization transcriptome data that we generated with three published transcription profiling studies related to P. aeruginosa biofilm formation in vitro [2,4,5]. Two genes encoding hypothetical proteins (PA0713 and PA0952) were significantly increased in 2 or more datasets (Table 2), and two additional genes encoding hypothetical proteins were found to be important in biofilm formation using a signature-tagged mutagenesis analysis of factors needed for virulence in a rat P. aeruginosa pulmonary lung infection [23] model: PA0141 (25 fold increase) and PA1009 (4.8 fold increase). The transcriptome results generated from murine GI colonization for genes known to be involved in biofilm formation in vitro and for the genes encoding hypothetical proteins are provided in Table 2.

Microtiter Biofilm Assay
We then used a non-redundant transposon mutant library in P. aeruginosa strain PA14 [13], to retrieve mutant strains with transposon insertions in the candidate genes identified above ( Table 2). The disruption of the target genes by the transposon was confirmed by colony PCR. Next, the mutants were tested for biofilm formation in vitro using a 96-well microtiter plate assay [20] on two different substrates: polystyrene (PS) and polypropylene (PP). Several strains (i.e., PA14fliP, fleR and flgK) were attenuated in both in vitro biofilm formation and growth rates (Figure 2). Assuming a linear relationship between growth and biofilm formation, biofilm to growth ratios were calculated to account for differences in growth ( Figure 3). Ratios less than 1 compared to that of the wild-type strain were designated as indicative of decreased biofilm formation. The same PA14 mutants (with insertions in pilB, sadB, crc, fleR, pilY1, flgK, fliP, and pilC genes) that showed significant absolute decreases in biofilm formation also had biofilm:growth ratios less than 1. (Fig. 3). We also analyzed biofilm formation by the P. aeruginosa strain PA14galU 2 which is unable to synthesize a complete outer core lipopolysaccharide (LPS) [24], as this strain is unable to colonize the antibiotic-treated murine GI tract [7]. Interestingly, the galU mutant had a biofilm:growth ratio of 2.5 on PP, making it a better biofilm former than the WT strain PA14.
We also tested biofilm formation under anaerobic conditions (to better approximate conditions found in the human gastrointestinal tract [25]) on both PP and PS substrates. The 8 mutants (pilB, sadB, crc, fleR, pilY1, flgK, fliP, and pilC) had comparable biofilm formation and biofilm/growth ratios when grown anaerobically as those produced under aerobic growth conditions (data not shown).

Colonization and Systemic Dissemination from the Murine GI Tract by Biofilm-deficient TN-mutants
The strains identified as attenuated for biofilm formation in vitro were further investigated for colonization and dissemination phenotypes in the murine GI tract colonization and dissemination model [7]. Wild-type P. aeruginosa PA14 consistently colonized the mouse GI tract at levels comparable to those previously reported [7] for P. aeruginosa strain PAO1: PA14 (n = 8, median = 3.76610 8 cfu/g stool, first quartile = 2.84610 8 , third quartile = 5.24610 8 ) and PAO1 (n = 8, median = 3.52610 8 cfu/g stool, first quartile = 2.64610 8 , third quartile = 3.81610 8 , Fig. 4). All mutant strains colonized the murine GI tract at comparable levels to WT P. aeruginosa PA14, with the exception of the PA14crc-mutant and the negative control strain, PA14galU-, which does not colonize at detectable levels (Fig. 4). The PA14crcmutant did colonize the GI tract but at significantly lower levels (slightly more than one-log) compared to WT PA14: crc (n = 8, median = 2.93610 7 cfu/g stool, first quartile = 1.31610 7 , third quartile = 3.77610 7 ; PA14 (n = 8, median = 3.15610 8 cfu/g stool, first quartile = 3.02610 8 , third quartile=3.60610 8 , p = 0.0002 by Mann Whitney test compared to PA14Dcrc, Fig. 4). Confirmation of maintenance of the mutant genotype was obtained by performing PCR on colonies recovered from murine feces.   In additional experiments, after establishing and confirming P. aeruginosa GI colonization, mice were made neutropenic by administering a single 200 mg injection of monoclonal antibody RB6-8C5 [7,26]. One-hundred percent of mice colonized with WT PA14 (n = 8) were euthanized when moribund or died in between observation periods. Seventy-five to 100% of mice colonized with all 16 of the PA14 strains with transposon insertions (n = 8 for each mutant strain) in the genes needed for full biofilm formation also became moribund and were euthanized or died in between observation periods. Thus, there were no significant differences in mortality in mice colonized with either WT PA14 or Tn-mutant strains unable to produce biofilms in vitro (Table 3).
To verify P. aeruginosa bacteremia and dissemination had occurred in the moribund/dead animals, spleens from these mice were resected and homogenized. Homogenates were serially diluted and plated on both TSA and cetrimide agar. The presence of green, oxidase-positive colonies on cetrimide agar and absence of any heterogeneous colony types on TSA agar confirmed the presence of P. aeruginosa. All deceased mice had only P. aeruginosa recovered from the spleens, ranging in levels from 10 6 to 10 9 CFU/g spleen ( Figure 5). As with the Tn-mutant strains recovered from the feces, maintenance of the mutant genotype following dissemination was also confirmed with colony PCR performed on P. aeruginosa grown from spleen homogenates.

Colonization and Dissemination of P. aeruginosa Type III Secretion Mutants in the murine GI tract
Given that the uniform high expression levels of the genes involved in T3SS detected during GI colonization, we investigated the colonization and dissemination phenotypes of P. aeruginosa strains wherein the exoS, exoT or popB genes were deleted in strains PAO1 and PAK as well as the transposon insertional mutants in T3SS genes including exoU in the PA14 background. All three wild-type strains colonized the mouse GI tract at levels comparable to those previously reported [7]: P. aeruginosa strain PAK (n = 8, median = 1.35610 8 cfu/g stool, first quartile = 1.03610 8 , third quartile = 1.93610 8 ); strain PA14 (n = 8, median = 5.60610 7 cfu/g stool, first quartile = 2.85610 7 , third quartile = 8.77610 7 ); and strain PAO1 (n = 8, median = 1.19610 8 cfu/g stool, first quartile = 7.21610 7 , third quartile = 1.35610 8 ). All T3SS mutant strains colonized the murine GI tract at levels comparable to or even higher (in the PA14 mutants) Figure 3. Biofilm/Growth Ratios for P. aeruginosa PA14 transposon-insertion mutants listed above. Bars represent the means for three separate biological replicates (each biological replicate is the mean of seven technical replicates). Ratios ,1 indicate decreased biofilm formation relative to growth level. By definition WT biofilm/growth ratio is 1, as biofilm and growth levels were used to calculate the ratio are in % of WT. The red line indicates biofilm growth ratio of 1.0. doi:10.1371/journal.pone.0015131.g003 than their respective background strains (data not shown). Thus, the T3SS does not appear to be critical for GI colonization.
After the mice were made neutropenic by administering mAb RB6-8C5, 100% of mice colonized with WT strains PAK (n = 8), PA14 (n = 8), or PAO1 (n = 8) became moribund and were euthanized or died between observation periods. In contrast, there were significant differences in the length of survival between the WT and all their isogenic T3SS mutant strains ( Figure 6) indicating that dissemination of P. aeruginosa in the setting of neutropenia is modified by the T3SS. However, in these experiments an overall mouse survival of #37.5% is not statistically different from the 100% lack of survival following dissemination of the WT parental strains, and while this was observed for all of the T3SS mutants in strain PAO1, only the mice infected with PAKDexoT strain had an overall significantly higher survival in this strain background. While mice infected with the PAK double DexoSDexoT mutant did not survive any better than the mice infected with the parental PAK strain, the difference in survival between mice infected with the single DexoT mutant and double DexoSDexoT strains was only 1 animal, likely indicative of normal experimental variation. None of the mice infected with the PA14 T3SS-deficient strains had significantly greater overall survival compared to survival of mice infected with the parental PA14 strain.

Discussion
P. aeruginosa's pathogenic signature revolves around its ability to cause infectious in many clinical settings: ventilator-associated pneumonia, chronic lung infections in CF patients, skin and soft tissue infections of burn victims, and bacteremia and sepsis in cancer patients. The pathogenesis of P. aeruginosa bacteremia in cancer patients derives from a combination of a chronic, usually non-pathogenic state of GI colonization followed by an acute, invasive infection caused by extra-intestinal bacterial translocation and subsequent bacteremia once the immune system is compromised [9,10,27]. Using a murine model of P. aeruginosa GI colonization and neutropenia-induced dissemination, we have shown that factors critical for establishing disseminated disease in this model encompass the ability to establish GI colonization at quantitatively sufficient levels (e.g. a minimum of 10 7 cfu/g stool) to allow for dissemination from the GI tract once neutropenia is induced [7]. We hypothesized that a microarray analysis of in vivo expressed mRNA transcripts during GI colonization would identify genes and/or phenotypes predictive of essentiality for the initiation and/or maintenance of the colonization state. However, while increased transcription of genes involved in both biofilm formation and T3SS were found during GI colonization, mutations in genes needed for these phenotypes did not alter the ability of P. aeruginosa to colonize the murine GI tract. We did find that loss of T3SS factors significantly increased the length of survival of mice infected with these mutants, but there was a variable effect of loss of the T3SS on overall mouse lethality. In general, our results indicate that increased mRNA transcription during in vivo infection may not be indicative of the need for the encoded protein in the infectious process.
We chose to use the PAK strain for these transcription profiling experiments because we had the greatest experience using this particular P. aeruginosa strain for microarray analysis [18], and at the time of these initial experiments, the PA14 TN library was not available. We only decided to utilize the PA14 TN mutants subsequent to the initial transcription profiling experiments. However, in retrospect the use of a different P. aeruginosa strain background to validate the results with specific Tn mutants based on transcriptome-findings with a heterologous strain probably adds robustness to the data as strain-specific effects are less likely to have been analyzed in the follow-up studies with the Tn mutants.
In addition to genes encoding proteins needed for biofilm formation and T3SS, 40% of genes involved in Cell Wall/LPS/ Capsule production were generally increased by two-fold or more over expression when the bacteria are in water. The ability of P. aeruginosa to synthesize a complete LPS outer core [24] appears to be critical for establishing GI colonization in this murine model [7]. LPS and O-antigen synthesis have also been shown to be critical in establishing GI colonization with Yersinia pseudotuberculosis [28], Yersinia enterocolitica [29], and E. coli [30]. Thus, we did not explore these phenotypes in more depth as the need for production of an intact LPS by P. aeruginosa to establish and maintain GI colonization has been validated.
Son et al. [6] reported on P. aeruginosa transcript expression in CF sputum and found a limited number of transcripts expressed in vivo compared to those expressed by the same P. aeruginosa strain grown in vitro. This was surprising given the observation of specific phenotypes of the organisms found in vivo, such as elaboration of alginate, not correlating with increased transcription of synthetic or regulatory genes known to affect this phenotype. Further analysis indicated that the clinical isolate had increased constitutive expression of mRNA transcripts when grown in vitro when compared to the mRNA expression in the sequenced PA01 strain, indicating that many of the phenotypes that had emerged during chronic infection were based on genetic changes that were not markedly affected by transfer to the in vitro state for bacterial growth. When these findings are factored into our results, it seems that a quantitative increases in vivo expression of P. aeruginosa mRNA transcripts over that of a comparator in vitro state may have a limited role in predicting genes involved in the overall pathogenesis process.
In our experimental data set we found that 90% of the genes involved in T3SS had increased expression levels compared with P. aeruginosa in water, and in fact, almost 50% of these genes had expression levels .10-fold compared to baseline (Table 1) and none had decreased expression compared to baseline levels. However, this increased transcription of the T3SS genes did not predict a role in GI colonization but was associated with increases in the time of mouse survival following dissemination after induction of neutropenia. This suggests an importance of these effectors in the pathogenic process directed to bacterial translocation out of the GI tract and the associated ability of the T3SS effectors to systemically intoxicate host cells and tissues. Of note, there was not a uniform effect on overall survival in all three P. aeruginosa strains resulting from a loss of T3SS. With strain PAO1 all of the groups of mice infected with T3SS mutants had increased survival, whereas with strain PAK the group of mice infected with the DexoT mutant, but not the mice infected with the DexoS or double DexoS/DexoT, survived better than the parental group. With strain PA14 none of the mice infected with the T3SS mutants had better overall survival than mice infected with the parental strain. In the setting of experimental lung infection in animals evaluating the ExoU cytotoxin [31] as well as data from human P. aeruginosa lung infections [32,33,34] there is an indication of a role of the T3SS in virulence, although deletion of individual T3SS effectors in strain PAO1 had little effect on overall bacterial replication and survival in the lung [35], and only popBand exoSTY-deleted mutants had a defect in systemic spread from the lung. All of these findings, including those reported here, not only emphasize the effect of strain background on the properties of mutant phenotypes but also raise questions as to the degree to which the T3SS is a factor in P. aeruginosa virulence among clinical isolates and in which settings of infection this virulence factor is essential.
In work previously reported by this laboratory, we demonstrated using immunohistochemical examination of intestinal tissues (cecal) obtained from mice colonized with P. aeruginosa that clumps of bacteria were trapped within the mucus layer of the murine cecums, while few bacterial cells were associated with the underlying epithelial cells [36]. Thus it appears that P. aeruginosa is embedded in an extracellular matrix in vivo, reminiscent of the phenotype commonly referred to as a biofilm mode of growth [37]. Consistent with these observations are results from histochemical staining and electron microscopy of lung tissue specimens obtained at autopsy from cystic fibrosis patients that also showed that aggregates of P. aeruginosa cells are seen in vivo, and, like the cells seen in the murine GI tract, are associated with the mucus and sequestered away from the epithelial surfaces of the airways [38]. Therefore, there appears to be direct evidence indicating that P. aeruginosa forms biofilms in both the GI and respiratory tracts within the mucus layer, but our results indicate that genes identified as being needed for P. aeruginosa biofilm formation in vitro are not needed for the mode of growth of the aggregated bacterial cells observed in vivo.
To fully test whether the biofilm-related genes that had increased expression in vivo GI colonization were strong candidates for further evaluation, we compared our microarray results with those from published studies analyzing the in vitro transcriptomes of P. aeruginosa cells growing in a biofilm to try and obtain a comprehensive identification of genes with increased transcription during biofilm growth. This was predicated on the lack of a consensus from multiple in vitro studies identifying a clear set of factors contributing to the biofilm phenotype, as this phenotype itself can be generated by different in vitro conditions. When comparing the GI colonization transcriptome data with previously published studies of in vitro P. aeruginosa biofilm formation, we ended up choosing twelve P. aeruginosa strains with interruptions in genes related to biofilm production for full in vivo analysis (crc, fleR, flgK, fliP, PA3782, pilB, pilC, pilY1, sadB, lasI, rhlA, and rpoN) even though only 6 of these genes were found to have mRNA transcripts increased by two-fold or more in vivo: flgK, PA3782, pilB, pilC, pilY1, and sadB. Of note, all of the genes encoding hypothetical proteins associated with in vitro biofilm production showed expression levels that were increased by two-fold or more in vivo, ranging from 4.4 to 24.8 fold increases from baseline (Table 2). Nonetheless, even by expanding our in vivo analysis to include evaluation of the need for genes found to be involved in in vitro biofilm formation that did not have increased expression in vivo, we still could not validate a role for biofilm formation in P. aeruginosa colonization of the GI tract and dissemination following induction of neutropenia.
Most of the biofilm-defective strains chosen for in vivo analysis showed this phenotype in vitro. We chose to utilize two different types of plastic to detect biofilm formation, PS and PP, since prior investigators had shown that some mutations resulted in loss of biofilm formation only on specific substrates, [20]. We found that with the exception of PA14DrhlA, which formed WT-levels of biofilms on both PS and PP, all the remaining mutant strains were attenuated in biofilm formation on both substrates (Fig. 2). Even  Table 3 for actual numbers for each mutant strain). Mice were initially colonized with PA14 and then subsequently made neutropenic with monoclonal antibody RB6-8C5 (0.200 mg IP once). Tn, transposon-insertion. doi:10.1371/journal.pone.0015131.g005 after correcting for growth differences, the majority of mutant strains continued to show attenuated biofilm formation; the exceptions that showed WT biofilm formation with the correction for slower growth were the rhlA mutant on PS and the pilC mutant on PP (Fig. 3). What was most striking, however, was that the galU strain that we include as a negative control for murine GI colonization had a significantly increased ability to form biofilms in vitro, with a growth rate comparable to WT P. aeruginosa PA14 (Fig. 3). However, one of the limitations of the biofilm microtiter assay is that it is an endpoint assay and provides no information about biofilm developmental stages such as those that can be analyzed in flow cells. Nonetheless, we tested mutants with interruptions in both the sadB [19] and lasI [39] genes that have been shown to affect biofilm formation in dynamic flow cell assays and found they were also not attenuated in our in vivo system.
While it is not clear which in vitro assay, if any, is useful for identifying strains attenuated for biofilm formation in vivo, which in and of itself is not a defined entity, our results suggest that the most commonly used in vitro methods do not reliably identify genes needed for in vivo virulence.
Only one P. aeruginosa PA14 mutant, the Dcrc strain exhibited a defect in GI colonization, which was nearly ten-fold lower than the wild-type strain (Fig. 4). While it is unclear if differences in colonization levels are physiologically significant, especially since the crc mutant caused comparable mortality to that of the WT PA14 following induction of neutropenia, it is conceivable that in a murine host colonized with P. aeruginosa along with other indigenous microbial cells, including anaerobic species, that a one log decrease in colonization rate may prevent disseminated disease from developing. We have found that a certain threshold Figure 6. Survival curves of neutropenic mice colonized with wild-type P. aeruginosa and T3SS mutant strains. Female 6-8 wk-old C3H/ HeN mice were colonized with P. aeruginosa strains PAO1(A), PAK (B), and PA14 (C) and with various deletional or insertional mutations of the type III secretion system in the corresponding WT background and then subsequently made neutropenic with monocloncal antibody RB6-8C5 (0.200 mg IP once). Median survival of mice colonized with mutants was significantly higher than that of mice colonized with the corresponding wildtype strain (P,0.009 for PAO1 mutants, P,0.009 for PAK mutants, P,0.006 for PA14 mutants, log-rank test). Each group contained 8 mice. T3SS, type III secretion system. doi:10.1371/journal.pone.0015131.g006 level of P. aeruginosa GI colonization (e.g. ,5610 7 cfu/g stool) is needed in order for translocation to occur once neutropenia is induced (AYK and GPB, unpublished observation).
In conclusion, we have utilized several independent molecular and genetic approaches to investigate whether P. aeruginosa genes noted to be important in in vitro biofilm formation exhibited any specific phenotype in a murine model of GI colonization and neutropenia-induced dissemination. Ultimately, we found that attenuation of biofilm formation in a static in vitro microtiter well did not correlate with a significant change in GI colonization level or mortality secondary to disseminated disease. These findings, however, do not completely diminish the importance of biofilm formation by P. aeruginosa with regards to GI tract colonization. Collections of microbial cells that could be called a biofilm have previously been described in the mucus covering the intestinal epithelial wall and on particulate matter in the intestine [40,41]. It is possible that P. aeruginosa does not need any of its own gene products to make a biofilm in vivo, as the mucus layer of the intestine is believed to entrap microorganisms [42] and thus may provide the environment needed for P. aeruginosa cells to grow close together. If the microorganism can also survive the antimicrobial factors excreted by the host [43] and can proliferate faster than the mucus is being shed [44], then a microbial reservoir is created [45]. The mucus layer could potentially act as an entrapping matrix, similar to the exopolysaccharide (EPS) matrix secreted by biofilm forming organisms [46]. Thus, while all 16 genes in P. aeruginosa that were thought to be needed for biofilm formation in vitro had no appreciable phenotype in vivo, indicating that targeting these genes for disruption by therapeutic interventions likely will have little effect of the pathogenesis on infection and disease, there may still be potential interventions in the mucosal colonization process that could reduce the ability of P. aeruginosa to cause serious systemic infections. Similarly, we found that while transcription levels of T3SS genes were markedly elevated in the colonization state versus the basal state, the phenotype observed was not of decreased GI colonization but of decreased dissemination and yet there was also a variable effect from loss of the T3SS when measuring overall survival of neutropenic mice infected with mutant strains. Ultimately microarray analysis of in vivo transcriptomes to define genes needed for virulence may not be a particularly accurate tool and it seems likely better methods and experimental designs will be needed to obtain a more accurate prediction of virulence factors that impact P. aeruginosa virulence.

Bacterial strains and growth
The strains of P. aeruginosa used are listed in Table 4. The ordered transposon library in strain PA14 [13] was kindly provided by Dr. Fred Ausubel, Boston, MA. P. aeruginosa strains were grown overnight at 37uC in Luria Bertani (LB) broth (with the addition of 0.015 mg gentamicin/mL for PA14 transposoninsertion mutants), harvested by centrifugation, washed with PBS, and resuspended in PBS in preparation for in vitro biofilm assays as well as for inoculation into the drinking water to establish GI colonization in mice. The P. aeruginosa concentration was estimated using a spectrophotometer and actual cfu counts verified by enumeration on LB and cetrimide agars.

Production or RB6-8C5 monoclonal antibody and induction of neutropenia
The RB6-8C5 mAb specific for the Ly6 antigen that is highly expressed by polymorphonuclear neutrophils (PMN) was pro-duced by growth of hybridoma cells in culture (Dulbecco's modified Eagle's medium with 10% fetal calf serum (FCS)) followed by purification of antibody by protein G chromatography, as previously described [7]. A single dose of 200 mg of RB6-8C5 was administered to C3H/HeN mice to produce a severe neutropenia (absolute neutrophil count ,100) for 5 days [7].

Murine model of P. aeruginosa GI colonization and neutropenia-induced dissemination
We used a murine model of P. aeruginosa GI colonization and neutropenia-induced dissemination as previously described [7] with minor modifications. Six-to 8-week-old female C3H/HeN mice (Harlan Laboratories) were housed as groups of 4 in sterilized cages equipped with filter hoods. Mice were supplied with sterile bedding, sterile water and sterile mouse chow and maintained under specific pathogen-free conditions at the animal facilities of the Harvard Medical School in compliance with the Harvard Medical Area Institutional Animal Care and Use Committee guidelines and at the animal facilities of the University of Texas Southwestern Medical Center in compliance with the Institutional Animal Care and Use Committees of the University of Texas Southwestern Medical Center. To deplete the indigenous GI bacterial flora, mice were fed sterile water with 2 mg streptomycin/mL (Research Product International, Mt. Prospect, IL) and 1500 U penicillin G/mL (Sigma-Aldrich, St. Louis, MO) for 4 days. Stool was collected from individual mice (0.030-0.050 g per stool pellet), homogenized in 1 mL 1% proteose peptone, and 100 ml of the homogenate was spread on trypticase soy (TSA) and MacConkey agars and grown overnight at 37uC to verify reduction of indigenous GI microbial flora. P. aeruginosa strains were grown as described above, added to sterile water with 1500 Units penicillin G/mL at approximately 10 7 cfu/ml, and then administered to mice for 5 days. Bacterial water was changed every second day to maintain adequate cfu levels. After 5 days of exposure to P. aeruginosa, stool was again collected, homogenized in 1 mL 1% protease peptone, serially diluted in 1% proteose peptone and dilutions plated on TSA and LB agar (with 0.015 mg gentamicin/mL (Research Product International, Mt. Prospect, IL) for PA14 transposon insertion mutants) to measure levels of GI colonization by P. aeruginosa. Once colonization with P. aeruginosa was confirmed, 200 mg of mAb RB6-8C5 was administered to each mouse by IP injection. For mice colonized with PA14 transposon insertion mutants, sterile water with 1500 Units penicillin G/mL and 0.015 mg gentamicin/mL was initiated after RB6-8C5 administration. Mice were monitored for morbidity for 7 days. Moribund mice were euthanized and along with mice found dead between observation periods, the carcasses were frozen at minus 20uC, later thawed, spleens were resected, homogenized in 1 mL 1% protease peptone, serially diluted and 10 ml of the homogenate was drip-plated on TSA and LB (with or without 0.015 mg gentamicin/mL) agar plates. Growth media were incubated at 37uC overnight under aerobic conditions. The presence of green oxidase-positive colonies was used for confirmation of P. aeruginosa systemic dissemination.

Ethics Statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.

RNA Isolation
In order to establish a comparative transcriptome for P. aeruginosa isolated from colonized mice, we isolated P. aeruginosa RNA from bacteria in the drinking water provided to the mice. Sterile water with 1500 Units penicillin G/mL and P. aeruginosa strain PAK (10 7 cfu/ml) was stored in a sterile flask for 24 hours. Bacteria were harvested by centrifugation, and the RNA isolation protocol described below was initiated.
To obtain transcriptome information from bacteria colonizing the murine GI tract, we initiated the murine model utilizing the P. aeruginosa PAK strain. Seven days after initiating P. aeruginosa drinking water and two days after confirmation of GI colonization with PAK, eight mice were euthanized and the cecal contents were flushed with a buffer containing 10 mM TrisHCl, 1 mM EDTA and 200 mM NaCl [47] into a sterile stainless steel mortar (Fisher Scientific, Pittsburgh, PA) which was immediately immersed in liquid nitrogen. Cecal flushate contents were then ground with a sterile pestle on a stainless steel mortar immersed in liquid nitrogen, then added to 0.5 volume of boiling lysis buffer (2% SDS, 16 mM EDTA, 200 mM NaCl), homogenized, and kept at 100uC for 5 minutes with frequent mixing. A hot acid phenol:chloroform(5:1 ratio of phenol:chloroform, Ambion, Foster City, CA) extraction was carried out at 65uC. An additional 2-3 room temperature acid phenol:chloroform extractions were performed [48] followed by a final chloroform:isoamyl alcohol extraction. Isopropanol (Sigma-Aldrich, St. Louis, MO) precipitation at 220uC, overnight was followed by treatment with RQ1 DNase (Promega Corporation Madison, WI) for 1 hour at 37uC to remove contaminating DNA. After DNase treatment and additional cold acid phenol:chloroform extractions, a final chloroform:isoamyl alcohol extraction was performed. The cDNA Synthesis and Microarray Hybridization cDNA was synthesized and hybridized to micoarray chips as previously described [18]. The protocol and modifications were as follows: 12 micrograms of purified total RNA extracted from pooled murine cecal samples (8 mice) or the drinking water containing strain PAK were converted to cDNA with Superscript II Reverse transcriptase (Invitrogen, Carlsbad, CA) and a semi-random decamer, termed (NS) 5 (Invitrogen Life Technologies, custom oligonucleotide order, 59NSNSNSNSNS 39, where N = A,T,C, or G and S = C or G). As a control for cDNA production and microarray hybridization, control RNAs were spiked into the reverse transcriptase reaction. The resulting cDNA was purified, partially digested with DNase I, and end labeled with ddUTP-biotin (Enzo Life Sciences, Farmingdale, NY). The resulting targets were hybridized to GeneChip P. aeruginosa Genome Arrays (Affymetrix,). Labeled DNA was hybridized to Affymetrix P. aeruginosa Genome Arrays, and chips were washed, stained and scanned as previously described [18].

Microarray Analysis and Statistical Significance Assessment
Hybridization intensity data was collected from the scanned array images of three replicate experiments (for both baseline and GI colonization transcriptome levels), and intrachip normalizations were performed with Affymetrix Microarray Suite 5.0 software. Statistical analysis and experimental comparisons were made with GeneSpring version 4.2.1 (Agilent Technologies, Santa Clara, CA). Significance was established on the basis of three criteria: first, a Welch t-test was performed to determine whether the difference in expression for a given gene was greater between test conditions than within replicates of the same condition. For this test, a probability cutoff of 0.1 was used (p,0.1). Next, the average hybridization intensity for each probe set had to be above background in the activating condition. This determination was made on the basis of the default absolute call metric in Microarray Suite 5.0 (Affymetrix). Third, the magnitude of change between two conditions (baseline drinking water and GI colonization state) had to be greater than 2-fold [18]. Data generated from three separate microarrays were used to calculate average transcription levels. All microarray data is MIAME compliant and has been deposited into the Gene Expression Omnibus database (accession number GSE22665).

Colony PCR and gel electrophoresis
A single colony was selected from a plate, transferred into 30 ml ddH 2 O and boiled at 100uC for 5 min. Cell debris was pelleted by centrifugation (2 min at 16,0006g), and supernatants transferred into a new tube and stored on ice. The supernatant was used as DNA template for the PCR reaction, which contained 43 ml of the PCR Platinum Supermix (Invitrogen, Carlsbad, CA), 1 ml of each forward and reverse primer (10 mM) and 5 ml of DNA template combined in a PCR tube. Forward and reverse primers are listed in Table S1. PCR conditions were as follows: 94uC for 30 sec; 30 cycles of 94uC for 30 sec, 55uC for 30 sec and 72uC for 1 min.; final ext. at 72uC for 5 min. For gel electrophoresis, 1% agarose (American Bioanalytical, USA) in TAE gel was used. Samples were run at 100 V for 80 min and examined in a BioRad Gel Doc XR system (Biorad Laboratories, Hercules, CA).
All 20 of the mutants used that were retrieved from the P. aeruginosa PA14 transposon library (24) were confirmed to have the correct insert based on the results from the colony PCR. The crc mutant produced a blue/green phenotype when growing on LBagar, similar to what was observed by others [22,49].

Microtiter plate biofilm assay
The in vitro biofilm assay was performed as previously described [20] with the following modifications. A P. aeruginosa Tn-mutant strain of interest was grown overnight at 37uC on LB agar plates, then suspended in sterile water with 1500 U Penicillin G/mL and 0.015 mg gentamicin/mL to a concentration of 10 9 CFU/ml. No antibiotics were added for wild-type strains. P. aeruginosa bacterial suspensions were serially diluted to a final concentration of 1:100; 100 ml was transferred to seven replicate wells using sterile PS 96well plates (Costar 3596, Corning, Corning, NY) or PP 96-well plates (Costar 3364, Corning). Four wells in each plate were not inoculated (LB medium only) for negative control references. Plates were sealed with parafilm and incubated for 12-14 hours at 37uC. Growth was quantified by using a 96-well ELISA reader and a wavelength of 630 nm (OD 630 ) (ELx800, Bio-tek Instruments, Winooski, VT). Microtiter plate wells were emptied and washed twice in 130 ml 0.9% sterile NaCl. 200 ml 1% crystal violet was added to each well and incubated for 15 min at RT. Wells were then emptied, washed twice in 130 ml sterile 0.9% NaCl and destained with 200 ml 95% ethanol for 5 min. Biofilm formation was quantified using a 96-well ELISA Plate reader and an OD 595 wavelength.
For in vitro biofilm formation under anaerobic conditions, we utilized the same protocol as described above with the following exceptions. P. aeruginosa strains were suspended in LB broth (with or without gentamicin) containing 15 mM KNO 3 at pH 6.5 and grown anaerobically in a Coy anaerobic chamber at 37uC for 24 hours [3].

Statistical Analyses
Survival data were analyzed by Fisher's exact test and log rank test, and two-way comparisons of GI colonization levels were carried out using the Kruskal-Wallis with Dunn's multiple comparison test to a single control group using the GraphPad Prism software (San Diego, CA). When multiple comparisons or more than two groups were analyzed, Bonferroni's correction to the significance level a was invoked.