Genetic determinants of Pseudomonas aeruginosa biofilm establishment

Global screen to identify genetic determinants of biofilms

The aim of this study was to identify P. aeruginosa PA14 mutants exhibiting an altered biofilm phenotype in a global screen. While multiple genetic screens for modified biofilm formation have been performed in a broad range of bacterial pathogens, including P. aeruginosa (Espinosa-Urgel et al., 2000; Heilmann et al., 1996; Loo et al., 2000; O'Toole & Kolter, 1998a, b; Pratt & Kolter, 1998; Recht & Kolter, 2001; Tu Quoc et al., 2007; Watnick & Kolter, 1999), this is believed to be the first report of a microscopic screen for mutants forming altered biofilms.

An absolute pre-requirement for the applicability of a 96-well plate-based optical method to serve as a reproducible and robust screen for mutants forming altered biofilms is that multiple equivalent biofilms are formed in the individual wells of the microtitre plate. Fig. 1⇓ shows 3D biofilm images and the c.f.u. counts of PA14 grown in eight individual wells of one 96-well plate as well as the processed data (total biovolume) of the replicates. No major differences were noted when comparing images, c.f.u. counts or processed data in the replicates of the PA14 reference strain.

Next, we assessed the robustness and discriminating power of our image-based method to determine the ability of mutants to form biofilms at the bottom of 96-well plates using a confocal laser scanning microscope after live and dead staining of the cells. Three PA14 NR mutants (pqsA, pqsE and pqsL) harbouring a transposon insertion within genes involved in alkyl-quinolone signalling, [described elsewhere as playing a role in biofilm formation (Diggle et al., 2003)], and an flgF mutant derived from the PA14 wild-type, were grown in at least 11 replicates in one 96-well plate for 48 h. Images of the biofilm stacks were acquired and physical parameters (see Methods) were determined and used for hierarchical clustering. The cluster tree in Fig. 2(a)⇓ demonstrates the successful discrimination of the different mutant biofilms. Representative images of the various biofilm morphologies are shown in Fig. 2(b)⇓. The applied optical screen clearly proved to be sufficient to discriminate high (flgF mutant) from low (pqsA mutant) biofilm producers. However, whereas all replicates of the high biofilm-producing and wild-type-like mutants (pqsL mutant) were found within the same cluster, there were two pqsE mutant and one pqsA mutant biofilms which did not cluster with the other replicates of the respective mutant group. This intra-strain variability can often be explained by image-conversion errors, e.g. during the Otsu-thresholding of overexposured images or of low-contrast images in general, which has a stronger impact on low-biofilm producers. Nevertheless, the intra-strain variability was much smaller than the inter-strain variability between high and low biofilm-producing strains, as Fig. 2(a)⇓ clearly demonstrates.

We next monitored the biofilm life cycle of the pqsA mutant, which we selected as a control strain for poor biofilm formation, and of the flgF mutant, selected as a control strain for enhanced biofilm production, in comparison over time with that of the PA14 wild-type. As depicted in Fig. 3⇓, the initial bacterial lawn which covered the bottom of the wells after 24 h of incubation became more and more structured and accentuated over time, whereas after 96 h bacteria began to disperse. Because the characteristics of the wild-type, poor and enhanced biofilms were most pronounced at 72 h, we decided to screen the PA14NR library for altered biofilm formation at this time point of incubation.

In order to ascertain the sensitivity and specificity of the screening method and to further address the inter-plate variability, PA14 wild-type and poor-biofilm controls were included in each of the 96-well screening plates, whereas the enhanced-biofilm control was added to roughly half of the microtitre plates. A 2D projection based on singular value decomposition of the 16 discriminative parameters of all 5833 mutants in addition to 536 internal controls (180 wild-type, 268 pqsA and 88 flgF mutants) is shown in Fig. 4⇓. The boundaries of the three groups, including mutants with reduced, normal and enhanced biofilms, were defined following k-means clustering, and were set very conservatively in order to strictly minimize the identification of false-positive mutants with reduced or enhanced biofilm production. However, this resulted in higher rates of false-negative mutants, probably leading to a strong underestimation of the number of factors important for biofilm formation, a fact which might explain why our screen missed some known biofilm-forming factors.

In our genome-wide screen, the conservative grouping classified none of the poor-biofilm controls in the enhanced-biofilm group and vice versa. In addition, none of the 35 mutants with the same mutant ID that appear twice in the library was found in both the high- and the poor-biofilm clusters. Nevertheless, we found that inactivation of 12 out of 1025 genes, where more than one transposon was inserted (at a different position within the gene), resulted in either an enhanced- or a poor-biofilm phenotype. This finding probably reflects the limitation of the screen, where overexposure, cross-contamination, incomplete dye penetration, foreign particles in the field, image-processing errors and false assignment of the PA14NR mutants might lead to misinterpretations.

In Supplementary Table S1, the PA14 mutant subgroup exhibiting poor biofilms as compared with the wild-type is listed. This subgroup comprises 7 % of all tested mutants, 81.7 % of all poor-biofilm controls, only 11.6 % of the PA14 wild-type and none of the PA14 enhanced-biofilm controls. Supplementary Table S2, on the other hand, lists the PA14 mutant subgroup which comprises roughly 5 % of all mutants exhibiting an enhanced-biofilm phenotype. This mutant subgroup included 50 % of all enhanced-biofilm controls, only 8.9 % of the PA14 wild-type and none of the PA14 poor-biofilm controls. The relatively high rate of misinterpretation clearly demonstrates the limitation of our screen, especially in the detection of enhanced biofilm-producing strains, where the conservative grouping, chosen to assure high specificity, led to very low sensitivity (only 50 %).

In order to estimate the effect of growth on biofilm formation we searched for the 5 % of the mutants which showed the lowest growth in planktonic cultures (as identified in the study of Dötsch et al. (2009) and determined their distribution among the poor- and enhanced-biofilm categories, respectively. We found 4.2 % of the slow-growing mutants within the list of enhanced biofilm-forming mutants and 10.9 % in the poor biofilm-forming group. This indicates that slow-growing mutants tend to be, but are not necessarily, poor biofilm producers.

Interestingly, among the mutants with defective biofilm formation, 64.7 % of the affected genes encoded hypothetical proteins, whereas 69.1 % of the enhanced biofilm-forming mutants encoded hypothetical proteins. These data indicate that there are many yet to be discovered regulatory cascades required for biofilm establishment.

It should be noted that the assignment of a single PA14NR mutant to the wild-type, poor- or enhanced-biofilm group should be interpreted with care. The assignment of the PA14NR mutants might be erroneous, false-positive and especially false-negative mutants cannot be excluded (see above), and the screening of the 5833 mutants was only done once, and although the biofilm growth conditions within the microtitre plates were highly standardized, deviations in, e.g. inoculum size, microscope settings and staining procedures, might lead to misinterpretations.

Thus, in consideration of a potential miscategorization of single PA14NR mutants, we will concentrate in this manuscript on the discussion of sets of functionally related mutants that were found to be affected in biofilm formation.

Rare detection of previously identified genetic determinants of biofilm formation

Our approach uncovered 394 (Supplementary Table S1) and 285 (Supplementary Table S2) genetic determinants that contribute to the biofilm phenotype of P. aeruginosa PA14. Only very few of these genes have previously been detected in global screens for factors participating in biofilm formation in P. aeruginosa or other bacterial pathogens (Espinosa-Urgel et al., 2000; Heilmann et al., 1996; Loo et al., 2000; O'Toole & Kolter, 1998a, b; Pratt & Kolter, 1998; Recht & Kolter, 2001; Tu Quoc et al., 2007; Watnick & Kolter, 1999). Furthermore, we observed a discrepancy not only in the identity of the contributing factors but also in the number of mutants detected. The use of an optical detection method in this study identified a proportion of poor biofilm producers that exceeded by far those of most previous genetic screens. The main reason seems to be that, whereas previous screens focused on the identification of genes involved in the initial steps of adhesion, we analysed mature biofilms.

The first global genetic screen for mutants that form altered biofilms in P. aeruginosa was performed by O'Toole & Kolter (1998a). Their screen uncovered 15 surface attachment deficient (sad) mutants (Kukavica-Ibrulj et al., 2008); however, only one of these mutants (sadB) was detected within our screen. Instead, we observed that sad mutants with defects in swimming motility produced more dense biofilms than the wild-type. This apparent inconsistency strongly supports the earlier finding that the requirement for flagella-driven motility for the establishment of biofilms varies depending on environmental conditions (Klausen et al., 2003). Whereas O'Toole and Kolter used minimal medium to cultivate pseudomonads in microtitre plates at the air–liquid interface (O'Toole & Kolter, 1998a, b), we used a rich medium (LB broth) to promote biofilm formation at the bottom of the microtitre plate under low-oxygen conditions.

Since 1998, various additional factors have been associated with the biofilm lifestyle in P. aeruginosa, including GacA (Parkins et al., 2001), CupA (Vallet et al., 2001), LasI (Davies et al., 1998), the catabolite repression control protein Crc (O'Toole et al., 2000), PslA (Matsukawa & Greenberg, 2004) and AlgR (Whitchurch et al., 2002a). With the expection of a cupA3 and a pslG PA14NR mutant, none of the other mutants was identified in our screen.

Furthermore, it has become increasingly apparent that the intracellular signalling molecule cyclic di-GMP (c-di-GMP) is involved in the regulation of the transition from the planktonic to the surface-attached biofilm mode of growth in many Gram-negative bacterial pathogens (Jenal & Malone, 2006; Römling & Amikam, 2006). c-di-GMP is synthesized by a family of enzymes called diguanylate cyclases (DGCs) and degraded by c-di-GMP phosphodiesterases (PDEs). Although nine P. aeruginosa mutants in genes encoding both DGCs and proteins which possess both GGDEF and EAL domains have been reported to exhibit abolished or reduced biofilm formation capability, and six mutants enhanced biofilm formation capabilities (Kulasakara et al., 2006), we only identified two mutants with a transposon insertion within a gene encoding a DGC as forming reduced biofilms (PA0290 and PA5487). Interestingly, a transcriptional analysis of biofilm-grown P. aeruginosa PAO1 has indicated expression of PA0290 to be significantly enhanced (Waite et al., 2005).

Genetic determinants promoting survival under microaerophilic conditions are essential for efficient P. aeruginosa biofilm formation

Among the 394 biofilm-deficient mutants, we identified the two main components of the rhl quorum-sensing system, rhlR and rhlI. The rhl system is known to be highly expressed under biofilm (Singh et al., 2000) and microaerophilic conditions (Alvarez-Ortega & Harwood, 2007), and P. aeruginosa has been shown to form robust anaerobic biofilms, the viability of which requires the rhl quorum-sensing system (Yoon et al., 2002). Rhl mutants exhibit high levels of toxic NO, which seems to account for premature cell death (Yoon et al., 2002). We also identified an oprF mutant as forming poor biofilms in our screening assay. OprF has been shown to be highly expressed in clinical CF P. aeruginosa isolates (Sriramulu et al., 2005). A proteomic analysis has revealed a marked upregulation of OprF under anaerobic conditions (Platt et al., 2008), and OprF has been suggested to be important for NO detoxification (Yoon et al., 2002) and thus for the establishment of biofilms.

We also identified two poor-biofilm mutants affected in genes with predicted terminal oxidase activity. One of them has been described to be preferentially expressed in the stationary phase of growth (CoxA) (Schuster et al., 2004), and the other, a cyanide-insensitive bd-type oxidase (cio), has been predicted to have a high affinity for oxygen and shown to contribute together with the cbb₃-1 and cbb₃-2 oxidases to sustaining respiration when oxygen levels are low (Alvarez-Ortega & Harwood, 2007). We did not identify mutants of the cbb₃ oxidases to exhibit a reduced-biofilm phenotype. However, we observed a clear defect in biofilm formation in the cioB mutant, although a cioA mutant has been described to exhibit biofilm formation capabilities similar to those of the wild-type when grown in a flow chamber (Alvarez-Ortega & Harwood, 2007). Interestingly, those authors showed that expression of the hydrogen cyanide synthase genes was activated about fivefold in a transcriptome analysis under microaerophilic conditions and suggested that the cyanide-insensitive oxidase exerts the additional role of protecting the cells from hydrogen cyanide toxicity during microaerophilic growth. This scenario might be of major importance under our experimental conditions, and might also explain why cultivation of a cioA mutant within the flow chamber does not show a biofilm-defective phenotype, because HCN is more efficiently eliminated.

Furthermore, it is interesting to note that we identified additional poor-biofilm mutants that were affected in genes described to be pronouncedly upregulated under microaerophilic/anaerobic growth conditions, including hutU, encoding a urocanase (Alvarez-Ortega & Harwood, 2007), PA4142, a conserved hypothetical gene within the same operon that encodes a predicted 4Fe–4S ferredoxin transmembrane protein and a homologue of the catalytic subunit of cbb₃ oxidase (CcoN) (Alvarez-Ortega & Harwood, 2007), PA5481 (Alvarez-Ortega & Harwood, 2007), azu (Vijgenboom et al., 1997), nirE, nirJ, nosL and ptxR (Gaines et al., 2007).

Arginine metabolism is essential for P. aeruginosa biofilms

When oxygen availability becomes limited, a central metabolic process to generate ATP in P. aeruginosa is arginine fermentation via the arginine deaminase (ADI) pathway (Lu et al., 2004). Although we did not find mutants affected in the ADI pathway to exhibit a reduced biofilm formation phenotype, it is interesting to note that we identified several mutants with insertions within genes involved in the pyrimidine nucleotide biosynthetic pathway (pyrBC, carAB). The first step in this pathway is the formation of carbamoyl phosphate by carbamoyl-phosphate synthase, which is encoded by the carAB genes. Carbamoyl phosphate is also required for the biosynthesis of arginine from ornithine via the arginine metabolic pathway, which involves argFGH (Lu et al., 2004). We identified not only mutants affected in the pyrimidine nucleotide biosynthetic pathway but also an argG and argH mutant affected in the formation of arginine from carbamoyl phosphate and ornithine, and two mutants (gabT and aruG) affected in arginine and ornithine degradation, respectively. It therefore seems that P. aeruginosa PA14 requires a functional arginine metabolism in order to establish robust biofilms; accordingly, arginine probably serves as a source of carbon, energy and nitrogen.

Furthermore, we identified a mutant with an insertion within the cbrA gene which exhibited a reduced-biofilm phenotype. It has been shown that CbrAB is a two-component system essential for the expression of ArgR-dependent pathways (Nishijyo et al., 2001). Since ArgR is the major regulator of arginine and glutamine metabolism (Lu et al., 2004), this finding underscores the importance of arginine metabolism for the establishment of P. aeruginosa biofilms.

Most interestingly, in Vibrio parahaemolyticus (Enos-Berlage et al., 2005), a carA transposon mutant has been shown to form only thin pellicles at the air–medium interface, and a comparison of the previously published transcriptome profiles of S. aureus (Beenken et al., 2004), Legionella pneumophila (Hindré et al., 2008) and Escherichia coli (Schembri et al., 2003) revealed a common set of highly expressed genes under biofilm conditions that are involved in the pyrimidine nucleotide biosynthetic pathway as well as arginine metabolism. PyrRPBC, carAB, pyrFE, argDABC and argGH are highly expressed in S. aureus biofilms (Beenken et al., 2004); pyrGH, carA and argD in L. pneumophila biofilms (Hindré et al., 2008); and artJ, pyrB, pyrI, gltB, glnA, gltD, argC and argF in E. coli biofilms (Schembri et al., 2003), indicating that a functional arginine metabolism is essential for biofilm formation in various bacterial species. Arginine has been shown to enhance the formation of P. aeruginosa biofilms (Caiazza & O'Toole, 2004). With the aim of evaluating the effect of arginine on biofilm formation under our experimental conditions, we exogenously added arginine to the PA14 biofilm cultures and clearly observed enhanced biofilm formation at arginine concentrations of 0.4 % (w/v) in the wild-type and the argG and argH mutants (data not shown).

pH homeostasis is essential for P. aeruginosa biofilms

One of the most extensively differentially regulated genes, which was identified in the first screen of the global P. aeruginosa gene expression profile under biofilm growth conditions, is ureB, encoding a urease (Whiteley et al., 2001). This enzyme has been described to be markedly upregulated in Streptococcus salivarius (Li et al., 2000) and S. aureus (Beenken et al., 2004) biofilms, and it has been speculated that it is involved in the maintenance of pH homeostasis in biofilm cultures. Ureases hydrolyse urea to two molecules of ammonia and one of carbon dioxide, which results in efficient alkalinization of the environment as an alternative to the generation of ammonia from the ADI pathway. Our findings indicate that pH homeostasis is crucial to PA14 biofilm formation in our experimental setting. We found not only various mutants affected in arginine metabolism (see above) that exhibited poor biofilms but also a ureB mutant, which produced low levels of biofilms. Furthermore we identified a kdpD mutant as developing poor biofilms. In addition to the production of ammonia, cation transport ATPases, such as the high-affinity K⁺-specific transport system encoded by the kdp operon, can also contribute to pH homeostasis through the exchange of K⁺ for H⁺. In E. coli, KdpD and KdpE, proteins that control expression of the kdpFABC operon, are members of the class of sensor kinase/response regulator proteins (Voelkner et al., 1993).

In order to evaluate the impact of the pH on biofilm development under our experimental conditions, we grew PA14 wild-type biofilms in HEPES-buffered LB medium at various pH values. As compared with non-buffered controls, biofilm establishment was seriously affected under buffered LB medium conditions (data not shown).

Genes essential for the repair and maintenance of DNA are required for efficient biofilm formation

The screen for biofilm-deficient mutants uncovered the importance of six genes involved in DNA repair mechanisms (recA, recB, recG, recJ, recN and uvrD). RecA has very recently been described to affect biofilm development in Streptococcus mutans (Inagaki et al., 2009). Most interestingly, in P. aeruginosa it has been demonstrated that double-stranded DNA break repair is required for biofilm-mediated diversity and that an endogenous oxidative stress triggers a DNA repair mechanism that in turn generates genetic variants with a wide range of fitness characteristics (Boles et al., 2004; Boles & Singh, 2008). It therefore seems that mutants with defects in the generation of self-induced diversity do not withstand the intrinsic and extrinsic stresses that act upon biofilm bacteria in our 96-well format system.

Interbacterial signalling contributes to biofilm formation

In addition to the mutants within the rhl quorum-sensing system, we also identified a mexH and mexI mutant as forming deficient biofilms. The MexH protein was recently described to be upregulated at the protein level during biofilm development and to play a critical role in biofilm maturation (Southey-Pillig et al., 2005). This might be due to the influence of the MexGHI-OpmD pump on interbacterial communication, since a recent report shows that mutation of genes encoding the pump results in the inability to produce N-(3-oxododecanoyl)-l-homoserine lactone (3-oxo-c12-hsl) and alkyl-quinolones (Aendekerk et al., 2005). Our screen identified pqsA and pqsC mutants as poor biofilm producers, which is in agreement with the findings of Diggle et al. (2003), who showed that diminished alkyl-quinolone production results in low biofilm production. Unexpectedly, the pqsD mutant was categorized within the group of enhanced-biofilm builders; however, this mutant produced alkyl-quinolones at levels comparable with those of the PA14 wild-type, indicating a misassignment. Interestingly, mutations within the pyrimidine pathway (carB, pyrB and pyrD) have previously been described to exhibit a reduced production of the alkyl-quinolones, which could be complemented in both growth and alkyl-quinolone production by the exogenous addition of UMP (Bredenbruch et al., 2005). Although the low alkyl-quinolone production may contribute to the low biofilm production of the pyrimidine mutants, the finding that enzymes of the pyrimidine biosynthetic pathway are upregulated under biofilm growth conditions in various bacterial pathogens (see above) suggests that other mechanisms, such as the requirement for arginine, also contribute to the poor biofilm formation.

Conclusions

The study of biofilm development in various bacterial pathogens by the use of genetic screens and the application of transcriptional profiling under biofilm growth conditions have uncovered several complex regulatory systems required for surface-attached growth of the respective pathogens and revealed some general strategies for the initiation of biofilms (Beloin & Ghigo, 2005; Kuchma & O'Toole, 2000; Lazazzera, 2005). In this study we show the power of combining a genetic screen with microscopic analysis to uncover several novel genetic determinants of P. aeruginosa biofilm formation and maintenance. Although the relatively high amount of false positives and false negatives is clearly a drawback of the automated technique used in this study, it seems that some of the identified genetic determinants of biofilm formation, such as the requirement for arginine metabolism, pH homeostasis and survival under microaerophilic growth conditions, are common themes for the establishment and maintenance of biofilms that have also emerged from earlier analyses of global biofilm transcriptional profiles in various pathogens.