Genome-wide analysis and literature-based survey of lipoproteins in Pseudomonas aeruginosa

Pseudomonas aeruginosa

The Gram-negative bacterium Pseudomonas aeruginosa is an opportunistic pathogen and a notorious nosocomial infective agent. This bacterium infects immunocompromised persons, such as AIDS patients and people with severe burn wounds, and is a serious risk for intensive care unit patients (Lyczak et al., 2000; Meynard et al., 1999; Pirnay et al., 2003). P. aeruginosa is also the predominant pathogen found in the airways of cystic fibrosis (CF) patients. In CF lungs, P. aeruginosa switches to a mucoid alginate exopolysaccharide producing phenotype, which causes chronic infections that correlate with high mortality rates in CF patients (Lyczak et al., 2002; Ratjen & Doring, 2003). The multiple occurrence of infections caused by this bacterium is largely due to its high degree of antibiotic resistance, which is the result of the synergistic effect of the low permeability of the outer membrane (OM) and the existence of numerous multidrug efflux pumps, which secrete the drugs directly out of the cell (Schweizer, 2003). A class of acylated proteins, called lipoproteins, has been shown to play a role in many fundamental cellular processes and in the pathogenesis of several bacterial infections. In P. aeruginosa, microarray analysis has revealed that there is a prominent induction of lipoprotein-encoding genes during mucoid conversion (Firoved et al., 2004). Lipoproteins are also capable of eliciting a strong host immune response. They are recognized by Toll-like receptors present on the surface of several antigen-presenting cells, which leads to the activation of a signal transduction cascade that initiates the expression of pro-inflammatory cytokines (Liang et al., 2005). The genome of P. aeruginosa comprises about 6200 kb and 3.3 % of this genome is predicted to encode 185 different lipoproteins (Lewenza et al., 2005). However, despite the abundance and importance of this class of proteins, very little is known about the numerous P. aeruginosa lipoproteins. Almost 60 % of the predicted lipoproteins are classified as hypothetical proteins in the Pseudomonas Genome Database (Winsor et al., 2009).

Bacterial lipoproteins

Lipoproteins are synthesized in the cytoplasm as precursors with an N-terminal signal sequence and are translocated across the inner membrane (IM) via the Sec pathway (or in rare instances by the Tat pathway). The signal sequence always ends in a consensus sequence called the lipobox. The lipobox consists of four amino acids, namely Leu-Ala/Ser-Gly/Ala-Cys, with Cys being the first amino acid of the mature lipoprotein (Hayashi & Wu, 1990). In the first step, a diacylglyceryl group is attached to the thiol group of the Cys via a thioether linkage. This step is catalysed by the enzyme phosphatidylglycerol/prolipoprotein diacylglyceryl transferase (Lgt). Next, the signal sequence is cleaved off just before the Cys by the prolipoprotein signal peptidase (LspA or signal peptidase II), which is specific for lipoproteins and can be inhibited by globomycin (Kiho et al., 2003). Finally, the cysteine is also aminoacylated by the phospholipid/apolipoprotein transacylase (Lnt) to create the mature lipoprotein with three acyl chains attached to its N-terminal Cys residue (Sankaran & Wu, 1994). Although their maturation occurs at the IM, most lipoproteins are located at the periplasmic side of the OM. Work of H. Tokuda and co-workers revealed that the transport of these OM lipoproteins is mediated by the Lol system, which is composed of five proteins, LolABCDE, also present in P. aeruginosa (Tanaka et al., 2007). LolCDE releases them out of the IM in an ATP-dependent manner and subsequently a water-soluble complex is formed with the periplasmic chaperone LolA in a 1 : 1 ratio, and finally lipoproteins are transferred from LolA to LolB, which itself is a lipoprotein (Taniguchi et al., 2005). The homologous proteins composing the lipoprotein biogenesis machinery and Lol system in P. aeruginosa are indicated in Table 1⇓. In Escherichia coli, lipoproteins are sorted to the OM by default, unless they have an aspartate residue at position 2 after the cysteine (Asp⁺²), which acts as the IM retention signal (Masuda et al., 2002; Seydel et al., 1999; Yamaguchi et al., 1988). However, the E. coli rule that lipoproteins with Asp⁺² stay in the IM, while all lipoproteins with another residue at position 2 are sorted to the OM, is applicable only in the family Enterobacteriaceae (Lewenza et al., 2008; Narita & Tokuda, 2007).

How many lipoprotein genes are there in P. aeruginosa PAO1?

Lewenza et al. (2005) predicted 185 lipoprotein-coding genes in P. aeruginosa PAO1, which represent 3.3 % of the genome. This prediction was made by looking for the presence of a lipoprotein signal peptide containing a lipobox. Analysis of the 185 predicted lipoproteins with LIPOP (Juncker et al., 2003) and LIPO (Berven et al., 2006) together with a manual inspection of all signal peptides predicted by these authors led us to eliminate some of these putative lipoproteins (10 in total). This is the case for PA1228 (putative signal MVTHFLSESGQAAC), PA1373 (MSKRIVVTGMGAVSPLGC), PA1864 (MKKIRQRNLQLILDAAC), PA2137 (MNRRLPGTLLIALC) and PA5382 (MALSFRQLQIFCAVARC), which have a very short signal lacking a clear h-region composed of hydrophobic amino acids; and PA1465 (MEVVALALALAACLGLAAAC), PA3894 (MVGSFVGFLVVFSAISGC) and PA4371 (MSLPSPSMPLACLLTALLLGGC) because the positively charged n-region of the signal was missing. Furthermore, for some of these dubious lipoproteins a cytoplasmic localization can be predicted, as for PA1373 (FabF2 3-oxoacyl carrier protein), PA1864 (TetR regulator), PA2137 (a response-regulator), PA5382 (LysR regulator). ArgE (N-acetylornithinase, PA5206) is also unlikely to be a lipoprotein since it is involved in arginine biosynthesis and the corresponding E. coli enzyme is cytoplasmic (Meinnel et al., 1992). This gives us a total of 175 lipoproteins in P. aeruginosa PAO1, corresponding to 3.2 % of the annotated protein-encoding genes (versus 3.3 % as previously proposed by Lewenza et al., 2005).

For two lipoproteins, PA1812 (MPPQTRKTPDLDALARAVRVSILLIAGALAGC) and PA1969 (MPLVRWRSQKIRGGGMPLKQFSSALVLAALLAGC), the predicted N-terminal sequence is very long and therefore unlikely to be a signal peptidase II-dependent signal, but analysis of the DNA sequence suggests that translation begins at the second ATG (or GTG) since it is preceded by a good ribosome-binding site, resulting in a shorter signal (14 and 18 residues, respectively, italicized), which fits perfectly with a canonical lipoprotein signal peptide. The possibility also exists that some lipoprotein genes are not correctly annotated. This could be the case for PA1712 (exsB), which is predicted to be a lipoprotein although the signal lacks an n-region with one or two positively charged amino acids. Looking at the nucleotide sequence upstream of the ATG, one finds a GTG followed by an AGG (Arg), TGC (Cys) and TGG (Trp), meaning that indeed a positively charged amino acid could be present. However, this protein was not retained in our list. The list of P. aeruginosa PAO1 lipoproteins and their localization, when known, is presented in Supplementary Table S1, available with the online version of this paper.

How many lipoproteins are retained in the inner membrane?

The majority of lipoproteins are destined for the OM, but some stay in the IM because of a LolCDE avoidance signal, which in Enterobacteriaceae is an Asp residue after the Cys (Asp⁺²). However, this appears not to be the case for P. aeruginosa since only four lipoproteins have Asp at position +2: the glycosylase MltA (PA1222), one hypothetical lipoprotein encoded by PA1592, the PpiC peptidyl-prolyl isomerase (PA3262) and the NosL NO reductase (PA3396). This anomaly compared to the E. coli situation stimulated Lewenza et al. (2008) to search for alternative IM retention signals. For example, MexA of P. aeruginosa is an IM lipoprotein that forms part of the MexAB-OprM drug efflux pump and possesses Gly and Lys residues at positions 2 and 3, respectively (Yoneyama et al., 2000). To investigate the lipoprotein sorting signals in P. aeruginosa, chimeric lipoproteins consisting of various regions of MexA and OprM (an OM lipoprotein) were created and their membrane localization was determined. This analysis revealed that, in contrast to the E. coli rule, specific combinations of the residues at position 3 and 4 determine the final membrane destination of the P. aeruginosa lipoproteins (Narita & Tokuda, 2007). For instance, Lys⁺³-Ser⁺⁴ was shown to be a potent IM retention signal (Narita & Tokuda, 2007). By fusing the signal peptide and the first four amino acids of the P. aeruginosa OM lipoprotein OmlA to the red fluorescent protein mCherry, Lewenza et al. (2008) created a very sensitive method to determine the membrane localization of the chimeric lipoprotein by fluorescence microscopy. By mutagenesis of the residues at positions 2, 3 and 4, they confirmed Lys⁺³-Ser⁺⁴ as a Lol avoidance signal and identified several other combinations of residues that act as an IM retention signal, such as Lys⁺²-Val⁺³-Glu⁺⁴, Gly⁺²-Gly⁺³-Gly⁺⁴, Gly⁺²-Asp⁺³-Asp⁺⁴ and Gln⁺²-Gly⁺³-Ser⁺⁴ (Lewenza et al., 2008). These IM retention signals are found in approximately 5 % of the P. aeruginosa lipoproteins, but given the lack of conservation it is unclear how these sequences act as Lol avoidance signals. It was speculated that the secondary structure formed by these residues might play an important role and that the periplasmic loops of LolC and/or LolE could be involved in the recognition of the sorting signal at positions 3 and 4 of P. aeruginosa lipoproteins (Tokuda, 2009). In their analysis, Lewenza et al. (2008) discovered new retention signals, bringing the number of IM lipoproteins to 13. Assuming that some lipoproteins, such as membrane fusion protein (MFP) components of efflux pumps, and some periplasmic enzymes are IM localized, we can predict more retention signals based on the four amino acids after the Cys residue (Table 2⇓). Seven MFPs are predicted to have a lipoprotein signal peptide with an IM retention signal: PA0156 (TriA), PA0157 (TriB), PA2019, PA2493 (MexE), PA3523, PA3677 and PA4599 (MexC) (Narita & Tokuda, 2007). During the last few years, other IM lipoproteins have been described and their localization confirmed as shown in Table 2⇓. MliC is a periplasmic lysozyme inhibitor, first described by Callewaert et al. (2008) in E. coli, and its homologue in P. aeruginosa (PA0867) has been confirmed to be an IM lipoprotein (Lewenza et al., 2008). The same authors confirmed the IM localization of PscJ (PA1723), a type III secretion protein, the glycosylase MltD (PA1812), the GntK (PA2321) gluconate kinase, the metalloproteinase IcmP (PA4370), the type IV pili biogenesis lipoprotein PilP (PA5041), and two hypothetical lipoproteins, encoded by PA3677 and PA5414, respectively. The following lipoproteins are also likely to be IM proteins: AbpE (PA2993), which has been described to be involved in the biosynthesis of thiamine in Salmonella (Beck & Downs, 1999), NlpD (PA3623), an accessory lipoprotein of the OM protein assembly machinery (Uehara et al., 2009), the PA4065 FtsX protein involved in cell division (Reddy, 2007), the PA4367 BifA cyclic-di-GMP phophodiesterase (Kuchma et al., 2007) and the transglycosylase MltB1 (Reid et al., 2006). In total, three transglycosylases have an IM lipoprotein anchor: MltA (PA1222), MltD (PA1812) and MltB1 (PA4444).

A total of 24 lipoproteins are predicted be inserted in the IM, and we propose to include 16 others in this list (Table 2⇑), which would bring the percentage of IM lipoproteins to 23 %. Since this analysis was done taking into account the known or predicted biological function, hypothetical proteins, the function of which could not be predicted, have not been included, so the percentage of IM lipoproteins might still be underestimated. When comparing the sequences of the four residues after the cysteine, it is striking that many IM lipoproteins have a G at the +1 position, often with a negatively charged amino acid (D or E). This is the case for TriA and TriB, which are MFPs involved in the efflux of triclosan, and are associated with TriC antiporter and OpmH OM porin (Mima et al., 2007).

Proposed inner-membrane lipoproteins and their predicted function

Two peptidyl-prolyl isomerases are predicted to be IM lipoproteins, PA0699 (PpiC) and PA3262, the latter having a D residue at the position +1 after the cysteine. Some enzymes are involved in carbohydrate metabolism, such as l-sorbose dehydrogenase (PA2414), TreA trehalase (PA2416) and GPCD (PA4792), a glycerophosphodiester phosphodiesterase. NirF (PA0516) is involved in the biosynthesis of haem D1, and PA0541 and PA5328 encode c-type cytochromes. QuiP is an acylase encoded by the PA1032 gene, which has been shown to modulate quorum sensing in P. aeruginosa by de-acylation of the signal molecules N-acylhomoserine lactones (Huang et al., 2006). One predicted IM lipoprotein is a thioredoxin (PA0953), one is involved in thiamine biosynthesis (PA2993), like the already mentioned AbpE, and one is a putative amidase (PA5485). Two lipoproteins are involved in type VI secretion (PA1048 and PA2364), one is a predicted peptidase (PA4016), and the last one is a sensor protein (PA3044).

Families of outer-membrane lipoproteins in P. aeruginosa

Although the function of many lipoproteins is unknown and cannot be predicted at the moment, the characteristics of some of them have been described and these can be grouped into different families.

Conservation of lipoproteins in the genus pseudomonas and some evolutionary aspects in P. aeruginosa

Among the 175 lipoprotein-encoding genes identified in PAO1, 62 have an orthologue in all the sequenced Pseudomonas genomes, while 27 are common to and found only in all four sequenced P. aeruginosa genomes (PAO1, PA14, PA7 and LES) (Fig. 2⇓). According to the membrane localization (OM, identified IM or proposed IM), the numbers of orthologues do not differ significantly (Table 4⇓), showing the same conservation of the OM and IM lipoproteins in the genus Pseudomonas. Fig. 2⇓ shows clearly two major peaks, one corresponding to the genes found in all four P. aeruginosa, and the peak corresponding to the conserved lipoprotein genes, which are present in all analysed representatives, representing the ‘core’ lipoprotein genome.

The already mentioned lipoprotein gene oprI has been previously described to be present in all representatives of the genus Pseudomonas sensu stricto since the oprI coding sequence could be PCR-amplified from different species of Pseudomonas, but not from other species (De Vos et al., 1997). The oprI gene has also been used to study the phylogeny of pseudomonads (De Vos et al., 1998). It should be noted that the oprI gene is not annotated in the P. aeruginosa PA14 and Pseudomonas fluorescens Pf5 genomes although a blastx search reveals the presence of the gene in these two genomes, which means that oprI is also a lipoprotein gene of the core genome. Another lipoprotein gene of diagnostic value is oprL, which has been used for the specific detection of P. aeruginosa from clinical specimens (De Vos et al., 1997; Pirnay et al., 2003; Rao et al., 2009). The specificity of PCR detection of oprL relies on the fact that the N-terminal portion of the lipoprotein is more divergent, since oprL can be amplified from different Pseudomonas species using primers targeting the more conserved C-terminal part of the protein (unpublished results). An interesting case is the PA1382 gene encoding a type II secretion OM secretin, since this gene is not present in the other P. aeruginosa genomes, but is found in P. fluorescens SBW25 (Pflu_4070), P. syringae 1448a (PSPPH_3054), and P. syringae B728a (Psyr_3141).

The G+C content at the third position of codons can be used to detect a recent lateral transfer from an organism that has a different GC3 content (Muto & Osawa, 1987), especially in the genus Pseudomonas (Bodilis & Barray, 2006). The GC3 content of the lipoprotein-encoding genes is high (87.1±5.6 mol%) and does not differ significantly according to the membrane localization (OM, identified IM or proposed IM), showing a weak impact of recent lateral transfers in the evolutionary history of the lipoprotein-encoding genes in P. aeruginosa (Supplementary Table S1).

It has previously been shown that the extracellular and OM protein encoding genes in P. aeruginosa evolve faster than cytoplasmic protein encoding genes (Dotsch et al., 2010; Julenius & Pedersen, 2006). As expected, we found average values of the non-synonymous versus synonymous mutations ratio (d_N/d_S), using the modified Nei–Gojobori method (Tamura et al., 2007), much more important for the lipoprotein genes (0.21) than for the ribosomal genes (0.06) (Bodilis et al., 2009). However, we did not find significant differences either between the d_N/d_S ratios or between the overall mean variabilities (P_i) of the OM and IM lipoprotein genes (Table 4⇑). Because periplasmic-facing proteins are probably under less selective pressure than surface-exposed proteins, we can assume that most of the lipoproteins in P. aeruginosa are exposed to the periplasm whatever their membrane localization, as in E. coli (Narita et al., 2004).

We also compared the topologies of the phylogenetic trees (neighbour-joining tree with Kimura two-parameter correction) (Tamura et al., 2007) built from each cluster containing the four P. aeruginosa orthologues (Table 4⇑). An unrooted tree built from four P. aeruginosa genes has three possible topologies. The topology that groups PAO1 and LESB58 together is the most represented (72.5 %) among lipoprotein genes, with no significant difference according to the membrane localization (Table 4⇑ and Supplementary Figure S1).

Interestingly, this topology is the right one, i.e. the topology of the organism's phylogeny, PA7 being an outlier (Roy et al., 2010). This observation confirms that the lipoprotein-encoding genes are probably little impacted by lateral transfers. It is worth mentioning that a phylogeny with an atypical topology does not necessarily mean that the gene has undergone lateral transfer since this could also be explained by a variation of the evolution rate, i.e. a high d_N/d_S ratio (Bodilis & Barray, 2006).

Regulation of lipoprotein genes in P. aeruginosa

Given the large variety of different lipoproteins, it is not surprising that they have different functions and are differently regulated. Some lipoprotein genes are considered as constitutively expressed, such as oprI and oprL, while others are induced by different factors. Therefore different transcriptome results from the literature were analysed. The factor that has the greatest influence on the expression of lipoprotein genes is exposure to human airway epithelial cells (Chugani & Greenberg, 2007; Frisk et al., 2004). Nineteen lipoprotein genes were found to be upregulated upon contact with epithelial cells, while 10 were downregulated. Among the upregulated lipoproteins, one finds PA1592, PA0737, PA3691 and PA3692. The same genes, plus PA0062 and PA5526, are also induced as a result of the conversion of P. aeruginosa to mucoidy due to the production of alginate extracellular polysaccharide and are known to be dependent on the AlgU (σ^Ε) sigma factor (Firoved et al., 2002, 2004). The LptF (lipotoxin F) gene has recently been described to be also dependent on the AlgU alternative sigma factor, and LptF contributes to survival in CF lungs (Damron et al., 2009). Another lipoprotein gene upregulated by exposure to epithelial cells and by conversion to mucoidy is osmE (PA4876), which is also induced by a wide range of other environmental changes, such as exposure to d-cycloserine, and osmotic shock; it is also regulated by the sensor PhoQ, and is quorum sensing induced (Aspedon et al., 2006; Gooderham et al., 2009; Schuster et al., 2003; Wood et al., 2006). d-Cycloserine exposure results in higher expression of several lipoprotein genes: a type VI secretion protein gene (PA1048), the alginate secretion lipoprotein gene algK (PA3543), PA3962, mexC (PA4599) and PA5107–PA5108 (Keiski et al., 2010; Wood et al., 2006).

Some RND efflux OM lipoprotein genes (oprM, opmQ, oprN, opmD, opmG) are also induced or repressed by different conditions. The oprM (PA0427) gene is downregulated by exposure to epithelial cells, while opmQ (PA2391) and oprN (PA2495) are induced. The OpmQ protein is the OM component of the PvdRT-OpmQ efflux pump, which has recently been shown to be involved in the recycling and re-excretion of the siderophore pyoverdine in the periplasm after release of iron (Imperi et al., 2009) and, as expected, these genes are iron-regulated (Ravel & Cornelis, 2003). OpmD is the OM protein of the MexGHI-OpmD efflux pump involved in the quorum sensing network of P. aeruginosa, and the opmD gene is induced by quorum sensing, belongs to the stationary-phase sigma factor RpoS regulon, and is SoxR-dependent and upregulated by the antibiotic azithromycin, known to interfere with the quorum sensing system in P. aeruginosa (Aendekerk et al., 2005; Dietrich et al., 2006; Nalca et al., 2006; Narita et al., 2004; Schuster et al., 2004). The opmE gene is induced upon exposure to copper, suggesting that it could be involved in the efflux of this metal (Teitzel et al., 2006). Interestingly, a number of lipoprotein genes are upregulated by the recently described PpyR (PA2663) regulator, including mexE and oprN, which are also under the control of MexT (Attila et al., 2008; Tian et al., 2009). The MexT regulon also includes PA2759, encoding a hypothetical lipoprotein. As already mentioned, some lipoprotein genes are iron regulated since they are repressed under condition of iron sufficiency by the ferric uptake regulator Fur; these include the already mentioned opmQ gene, omlA (PA4765), icmP (PA4370) and PA4372 (Ochsner et al., 2002; van Oeffelen et al., 2008).

Conclusion

P. aeruginosa is a highly adaptable bacterium with a rather large genome (>6 Mb) and previous in silico predictions and experimental data suggested that in PAO1 185 genes encoded lipoproteins, many of them of unknown function. The present analysis suggests that the number of lipoproteins should be reduced to 175, while more lipoproteins can be predicted to be localized in the IM (23 %). OM porins from RND systems involved in the efflux of antibiotics are lipoproteins, while others belonging to the same family are not, as shown by the present phylogenetic analysis. The reason for a supplementary lipoprotein anchorage for these efflux porins rather than via only transmembrane β-sheets domains remains to be elucidated. The present analysis of the conservation of lipoprotein genes in the genomes of different pseudomonads showed 62 genes to be conserved, forming the core of lipoprotein genes, while no evidence for large horizontal gene transfer could be found. Finally, lipoprotein genes are mostly regulated by exposure to epithelial cells, conversion to mucoidy, and treatment with membrane-damaging antibiotics such as d-cycloserine.