| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
LMDF (Laboratoire de Microbiologie Du Froid), UPRES 2123, ABISS (Atelier de Biologie, Informatique, Statistique et Sociolinguistinque), Université de Rouen, 76821 Mont Saint Aignan, France
Correspondence
Josselin Bodilis
josselin.bodilis{at}univ-rouen.fr
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
; Jaouen et al., 2004; Orange, 1994; Rawling et al., 1995; Worgall et al., 2005). This protein has only been found in this genus (but see Rediers et al., 2004) and may be considered as a diagnostic protein for Pseudomonas sensu stricto (de Mot et al., 1994; Vermeiren et al., 1999; Aagot et al., 2001). OprF shares C-terminal similarity with OmpA of Escherichia coli and thus is a member of the OmpA superfamily of porins (de Mot et al., 1994).
OprF is an oligomer, possibly a trimer, that is associated with both LPS and peptidoglycan (Nikaido, 2003; Tamber & Hancock, 2004). The protein consists of three domains: the N-terminal domain, a central linker region, and the C-terminal domain. The N-terminal domain of OprF (the first 160 amino acids) forms an eight-stranded 
-barrel (Brinkman et al., 2000). A central linker region (amino acids 161–209) contains a variable sequence of multiple proline-alanine repeats followed by a much-conserved sequence with two disulfide bonds that is absent in some Pseudomonas species (Vermeiren et al., 1999; Bodilis et al., 2004). The C-terminal region of OprF (amino acids 210–326) shares considerable similarity with that of OmpA (56 %), including the fact that its structure has not been completely elucidated (Pautsch & Schulz, 1998). Two folding models have been proposed (Nikaido, 2003; Tamber & Hancock, 2004). In the first model, OprF forms a 16--stranded membrane, since the C-terminal half is exposed and available to monoclonal antibodies (von Specht et al., 1995; Rawling et al., 1995). In the second model, the C-terminal half is a globular domain that lies in the periplasm since it is involved in peptidoglycan binding (Rawling et al., 1998). Finally, to explain the variations in channel size observed in OprF (Bellido et al., 1992), it has been suggested that the C-terminal domain can translocate between the periplasm and the surface (Nikaido, 2003; Tamber & Hancock, 2004).
In addition to having a complex and partially elucidated structure, OprF is a multifunctional protein. It is a non-specific porin permitting passive diffusion of small polar nutrients (Orange, 1994). The OprF pore changes its channel size according to the growth conditions, and this could affect outer-membrane permeability in the three major species of the genus Pseudomonas: P. fluorescens, P. putida and P. aeruginosa (Dé et al., 1997; Jaouen et al., 2004). OprF is also involved in maintaining cell shape and in growth in a low-osmolarity environment (Woodruff & Hancock, 1989; Rawling et al., 1998). Moreover, this protein probably plays a role in adhesion to roots in P. fluorescens strains isolated from the rhizosphere (de Mot et al., 1992) and in adhesion to fibronectin in P. fluorescens (Rebière-Huët et al., 1999). For P. aeruginosa, it has recently been reported that the binding of human interferon-
to OprF results in the expression of the PA-I lectin, a quorum-sensing-dependent virulence determinant (Wu et al., 2005).
In spite of the extensive studies of this protein and its potential taxonomic, ecological and medical applications, only two phylogenetic studies have been reported (Vermeiren et al., 1999; Bodilis et al., 2004). In the latter, we showed important differences in tree topologies between the OprF and 16S rDNA phylogenies obtained from 69 Pseudomonas isolates or reference strains (Bodilis et al., 2004). The incongruence is that both the clusters, the one containing P. fluorescens strains and the one containing P. putida strains, each divides into two parts (termed o-clusters); no cross horizontal transfer was observed between the fluorescens o-clusters and the putida o-clusters (Fig. 1). Since the fluorescens 1 o-cluster contained most of the bulk soil (non-rhizospheric) isolates and the fluorescens 2 o-cluster contained all the clinical isolates and most of the rhizospheric isolates, this dichotomy of OprF seemed to be correlated with the ecological niche.
|
): (i) an ancient horizontal transfer of oprF gene (after the P. fluorescens–P. putida speciation), (ii) an ancient duplication of the gene (before the speciation) followed by several losses, or (iii) artifacts in phylogeny reconstructions. These different possibilities and more generally the evolutionary history of the oprF gene have been investigated in this study. |
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
(i) The 7-sequence sets (or 6-sequence set for the cobA gene) were retrieved from the finished genomes of P. aeruginosa PAO1, P. putida KT2440, P. syringae pv. tomato DC3000, P. syringae pv. syringae B728a, P. fluorescens Pf0-1 and P. fluorescens Pf-5 (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi), and the unfinished genome of P. fluorescens SBW25, sequenced with about an eightfold coverage (http://www.sanger.ac.uk/Projects/P_fluorescens/). The sets corresponded to the sequences of oprF, 16S rDNA, concatenated ribosomal genes, genes upstream of oprF (ppsA, estX, menG, cmaX, crfX, cmpX, sigX), concatenated upstream genes and cobA (only six sequences because of a deletion in strain Pf-5). For the set of ribosomal genes, 46 ribosomal genes, ubiquitous in the seven sequenced Pseudomonas genomes, were aligned individually and concatenated, leading to 18 045 unambiguously aligned nucleotide positions.
(ii) The 20-sequence sets (for oprF and 16S rDNA genes) were composed of the 7-sequence sets plus the 13 sequences for which the accession numbers are given in Table 1.
|
), minus the hybrid oprF of strains MFY71 and MFY140, and minus the oprF of strain MFY78 because it occurs at a flawed position between the putida 1 o-cluster and the fluorescens 1 o-cluster. The remainder of the 69-sequence set included the oprF gene of strains DC3000, B728a and Pf-5. This set included the set of 20 sequences and, therefore, the set of 7 sequences.
Phylogenetic analyses.
For the three different sets of sequences, protein alignments (and nucleotide alignment for the 16S rDNA) were generated using CLUSTAL X version 1.81, with default parameters (Thompson et al., 1997). These were truncated to the same size as the shortest sequence. The sequences in the 20- and 69-sequence sets were only partial sequences: amino acid residues 6–328 for OprF of strain PAO1 (there are 350 amino acids in the complete protein, with the peptide signal) and nucleotide positions 113–1363 for the 16S rDNA of strain PAO1 (there are 1536 nucleotides in the complete gene). Although oprF and 16S rDNA are complete in the sequenced genomes, the same partial sequences were used for the 7-sequence sets, to compare the results more easily. All alignments were optimized visually and all ambiguous positions and positions with gaps were removed for all subsequent analyses. In addition, positions corresponding to the central domain of OprF, which contained a low-complexity region (proline rich) and several gaps, were eliminated. Altogether, amino acid residues 1–5, 79–84, 125–126, 168–171, 187–233, 308 and 328–350 were removed for the OprF of the PAO1 strain. Finally, all nucleotide alignments were deduced from corrected protein alignments.
Phylogenetic neighbour-joining trees were computed using MEGA v2.0 (Saitou & Nei, 1987; Kumar et al., 2001). Mutiple substitution events were taken into account using Poisson correction, Kimura two-parameter correction (Kimura, 1980) and Jukes–Cantor correction (Jukes & Cantor, 1969), for protein sequences, nucleotide sequences and synonymous positions, respectively. Bootstrap confidence levels were determined by randomly resampling the alignment positions 1000 times.
Maximum-likelihood (ML) analyses were performed using PAUP software version 4.0b10 (D. L. Swofford, Smithsonian Institution, Sunderland, MA, USA). The GTR model was used with an among-site rate heterogeneity approximated by a discrete gamma distribution with four divisions (nCat parameter) plus a proportion of invariant sites (pInvar parameter). The best tree was sought by three successive iterations with TBR (tree bisection and reconnection) branch swapping: for each iteration, all parameters (GTR parameters, gamma shape and pInvar) were estimated on the best tree of the previous iteration and were fixed for the heuristic search. The tree used for parameters of the first iteration was a neighbour-joining tree with Kimura two-parameter correction.
The topologically constrained searches were performed in the same way as the unconstrained search except for the option ‘enforce constraints’. For example, for the topology constraint 3 (Fig. 2), the constraint was: (PAO1,KT2440,[SBW25,Pf0-1,Pf-5,(DC3000,B728a)]). While only the defined monophyletic groups were enforced (i.e. DC3000+B728a and SBW25+Pf0-1+Pf-5+DC3000+B728a), all the other nodes were free during the search for the best tree. Topologies 4 and VIII correspond to the species phylogeny expected (i.e. monophylies of species), for the 7- and 20-sequence sets, respectively. Topologies 8 and III correspond to the OprF phylogeny, for the 7- and 20-sequence sets, respectively. The other topologies tested correspond to intermediate and alternative topologies. The intermediate topologies, for example 1, 2 and 3 for topology 4, are less resolved. In other words, topology 4 is a nested topology constraint of the topology constraints 1, 2 and 3 (Fig. 2). The alternative topologies (9, 10, IX and X) are, a priori, unlikely.
|
). For each tree, the likelihood was calculated with parameters estimated on the unconstrained tree. If a constrained tree had the same likelihood as the unconstrained tree (i.e. it is noted in the ‘Best topologies' column of Table 2), we considered that the monophylies enforced by the topology constraint were present in the unconstrained tree. In other words, the topologies of the constrained and unconstrained trees were the same. If the likelihood of a constrained tree was different from that of the unconstrained tree, we considered that some of the monophylies enforced by the topology constraint were not present in the unconstrained tree. By using the Shimodaira–Hasegawa test, the likelihood of this constrained tree was significantly different (i.e. it is noted in the ‘Topologies significantly different’ column of Table 2) or not significantly different (i.e. it is noted in the ‘Topologies not significantly different’ column of Table 2).
|
For the 7-sequence sets, codon usage was estimated using the codon adaptation index (CAI). CAI is a measure of the relative adaptation of the gene codon usage towards the codon usage of highly expressed genes for that organism (Sharp & Li, 1987). To estimate this index, we used as a reference pool for each organism (i.e. a reference codon usage table) the 46 concatenated ribosomal genes, since it is impossible for us to determine which genes are highly expressed (especially on the unfinished genome of P. fluorescens SBW25). Highly expressed ribosomal protein genes are commonly used in genome analyses where experimental expression data are limited or unavailable (Lafay et al., 2000; Palacios & Wernegreen, 2002; Herbeck et al., 2003). The EMBOSS programs cusp and cai were used to estimate the reference codon usage table of a strain and the CAI of a gene, respectively (Rice et al., 2000).
Synonymous and nonsynonymous rates of nucleotide substitution.
The synonymous rate (Ks) and nonsynonymous rate (Ka) were calculated using the modified Nei–Gojobori method implemented in the MEGA v2 0 software (Zhang et al., 1998). The transition to transversion ratio was fixed at 2 and the Jukes–Cantor correction was used to account for multiple substitutions at the same site. The transition to transversion ratios estimated from oprF and ribosomal genes, were 1·2 and 2·3, respectively. The net mean synonymous or nonsynonymous distance between groups of taxa (dA) was calculated using MEGA v2.0 software and was given by dA=dXY–(dX+dY)/2, where dXY is the mean distance between groups X and Y (i.e. the arithmetic mean of all pairwise distances between taxa in the inter-group comparisons), and dX and dY are the mean within-group distances.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
) and that have no duplicated copies. The ribosomal sequences were aligned individually and concatenated, leading to 18 045 unambiguously aligned nucleotide positions. The distance tree (Kimura two-parameter correction) of our concatenated gene set and bootstrap values are shown in Fig. 1(c). Alternatively, an analysis with deduced protein sequences was conducted; bootstrap values are also reported on the tree. While the monophyly of P. fluorescens is only supported with nucleotide sequences, the monophylies of P. syringae and P. syringae plus P. fluorescens are supported by maximum boostrap values (100 %) with both protein and nucleotide sequences.
Similar analyses with the oprF gene, however, led to a different topology. Two P. fluorescens strains (SBW25 and Pf-5) are separated from the third P. fluorescens strain (Pf0-1) and the two P. syringae strains by a node with a bootstrap value of 99 % (Fig. 1b). Analyses on a broader set of sequences (Table 1) were also performed and led to similar results, with the P. fluorescens strains splitting into two separate groups, previously termed fluorescens 1 o-cluster and fluorescens 2 o-cluster by Bodilis et al. (2004) (Fig. 1a). In addition, two putida o-clusters, also previously described, were found. The separation of the two fluorescens o-clusters by two nodes is not strongly supported (<90 %), especially in nucleotide analysis.
Test of incongruence
To evaluate the incongruence between oprF and species phylogeny, we conducted a maximum-likelihood (ML) analysis using the GTR (general time reverse) model of evolution with an among-site rate heterogeneity approximated by a discrete gamma distribution and a proportion of invariant sites (see Methods). Different ML searches for the best tree were performed with or without constraint topologies (Fig. 2, Table 2). For each set of sequences, the resulting trees were compared, using the Shimodaira–Hasegawa test, to determine whether constrained trees were significantly different from the unconstrained tree (Shimodaira & Hasegawa, 1999). If a constrained tree had the same likelihood as the unconstrained tree, we considered that the monophylies enforced by the topology constraint were present in the unconstrained tree (see Methods).
For concatenated ribosomal genes, since the unconstrained tree showed monophylies of P. syringae and P. syringae plus P. fluorescens (topology constraints 2 and 3), the OprF topology (topology constraint 8) was very significantly rejected (P<0·001), as were its intermediate topologies (5, 6 and 7). Although not significantly rejected (P=0·376), the monophyly of P. fluorescens (topology constraint 1) was not found on the unconstrained tree (Table 2).
For the 16S rDNA gene (7- and 20-sequence sets), the unconstrained tree also showed monophylies of P. syringae and P. syringae plus P. fluorescens, in addition to being significantly inconsistent with OprF topology (P<0·05). For the 20-sequence set, since constrained topologies assume a monophyly for all o-clusters and lineages (defined in Fig. 1), the topology VII (monophyly of P. syringae plus P. fluorescens) is not the best topology. For the two sets of sequences, the monophyly of P. fluorescens was not significantly rejected (P=0·213 and P=0·109 for the 7- and 20-sequence sets, respectively).
As expected, through similar ML analyses performed on the oprF gene with 7- or 20-sequence sets, we obtained the OprF topology (constrained topology 8 or III, respectively). However, for the two sets of sequences, the species topology (4 or VIII) is not significantly rejected (P=0·071 and P=0·163 for the set of 7 and 20 sequences, respectively). Therefore, there is no significant incongruence between the OprF and species phylogenies.
Synonymous codon usage
Alternative synonymous codons are generally not used with the same frequency. In E. coli, synonymous codons that are recognized more efficiently and accurately by the most abundant tRNA are preferred, especially in highly expressed genes (Sharp & Li, 1987). The strength of preference (i.e. bias of codon usage) is expected to be correlated with the level of gene expression. Moreover, a weak bias of codon usage may result from either a low expression of the gene, or a recent horizontal transfer from an organism that has a different synonymous codon usage. Since the OprF protein is the major surface protein of Pseudomonas, the oprF gene should be highly expressed. For this gene, the bias of codon usage is expected to be high, unless the gene was recently transferred.
To estimate deviations in codon usage, the codon adaptation index (CAI) was calculated for oprF gene. CAI is a measure of the relative adaptiveness of the codon usage of a gene towards the codon usage of highly expressed genes for that organism: the higher the index value, the greater the codon usage bias (Sharp & Li, 1987). As a reference for highly expressed genes, we used the 46 concatenated ribosomal genes for each organism (see Methods).
The genes of ribosomal proteins had high a CAI value (from 0·74 in strain PAO1, to 0·76 in strain KT2440) but CAI values for the oprF gene were even higher (from 0·82 in strain Pf-5, to 0·85 in strain KT2440) (Table 2). Therefore, there is a strong codon usage bias in the oprF gene, as expected for a highly expressed and not recently transferred gene. However, the absence of a recent horizontal transfer event cannot be deduced from this result: when the CAI value was calculated for the oprF gene of each strain, using the highly expressed genes of the other six Pseudomonas species as reference, the index value remained high (from 0·79 to 0·85), except when the reference genes from the PAO1 strain were used (from 0·61 to 0·74). These results suggest that, for the oprF gene, the CAI index is not sensitive enough to highlight any differences between synonymous codon usage in P. putida, P. syringae and P. fluorescens.
GC3 composition (GC content at the third codon position)
Bacteria vary widely in their genomic G+C content, especially at third positions of codons, suggesting variable mutational biases among organisms (Muto & Osawa, 1987). In the same way as with codon usage bias, this feature can be used to detect a recent horizontal transfer from an organism that has a different GC3 content.
The GC3 contents were variable between sequences of concatenated ribosomal genes (from 58·3 mol% in strain SBW25, to 71·9 mol % in strain PAO1) (Table 2). Interestingly, evolutionary distance is strongly linked to difference in GC3 content in Pseudomonas (Fig. 3a). Moreover, while some GC3 contents were higher in oprF genes (from 59·9 mol% in strain MFY143, to 79 mol% in PAO1), the correlation between difference in GC3 content between concatenated ribosomal genes and difference in CG3 content between oprF genes is quite good (r2=0·82) (Fig. 3b), suggesting an absence of recent horizontal transfer between distant Pseudomonas organisms. In fact, the GC3 values shown in Table 3 were quite informative. The GC3 value ranged from 59·9 to 71·0 mol% in the fluorescens 1 and 2 o-clusters, and from 72·5 to 78·3 mol% in the putida 1 and 2 o-clusters.
|
|
.
|
). Analyses of incongruence were also performed with the genes upstream and downstream from oprF (Table 2). The unconstrained tree showed monophylies of P. fluorescens or P. syringae plus P. fluorescens (topology constraints 1 or 3), for cobA and for five out of the seven genes upstream from oprF. There is even a significant difference from OprF topology (topology constraint 8) with three genes upstream from oprF (ppsA, cmaX and cmpX). Moreover, with the concatenation of the seven upstream genes, the OprF topology (topology constraint 8) was very significantly rejected (P<0·001), as well as intermediate topologies (5, 6 and 7), while the species topology (4) was the best.
Search for an adaptive selection
In comparison with upstream and downstream sequences, pairwise distances between protein sequences showed quite high divergence in the OprF protein, e.g. only 56·9 % identity between the protein sequences of strains PAO1 and SBW25 (Table 2). In fact, the real divergence is even higher, since the most variable positions of the alignment were removed for analyses (see Methods). Even so, two genes (estX and cmaX) have deduced protein sequences with a similar divergence to that of oprF. Since such a divergence in the oprF gene suggests an unusual evolutionary history, we attempted to detect positive selection through comparison of synonymous (Ks) and nonsynonymous (Ka) substitution rates.
Since saturation of synonymous sites is an important problem in the analysis of selective constraints when nucleotide sequences are compared, we initially analysed the correlation between the evolutionary distance and the synonymous distance on the concatenated ribosomal genes. If synonymous sites are saturated, we might expect a curve with a plateau. In fact, since we obtained a very strong linear correlation (r2=0·99), no saturation exists at synonymous positions between the ribosomal genes of Pseudomonas (Fig. 3c). Moreover, although the regression coefficient is quite weak (r2=0·68), the correlation between synonymous distances of oprF and ribosomal genes clearly showed that there was no saturation. We therefore conclude that Ks and Ka can be compared to infer modification of evolutionary constraints in the oprF gene.
While Ks is quite variable, even in the same genomic context as the oprF gene (from 0·45 in the oprF gene, to 0·85 in the sigX gene), the Ka/Ks ratio is even more variable (from 0·06 in the sigX gene to 0·54 in the oprF gene), suggesting variable evolutionary constraints between genes (Table 2). As expected, the two genes that have similar divergence to oprF in their deduced protein sequences (the estX and cmaX genes), also have a relatively high Ka/Ks ratio (0·40 and 0·27, respectively). Although the Ks value of oprF is slightly less than the Ks value of those two genes, the Ka value of the three genes is almost the same (but that does not suggest identical evolutionary constraints).
Since a Ka/Ks ratio greater than 1 is considered to be strong evidence of positive selection for amino acid replacements, a ratio less than 1 is generally taken as evidence that the proteins have evolved under negative or purifying selection. Significantly elevated Ka/Ks ratios have been shown for very few whole proteins (Endo et al., 1996). However, Ks and Ka values are computed as an average over both amino acid sites and time. This approach is very conservative, since it does not detect either episodic positive selections or positive selections that only occur for some codons (Messier & Stewart, 1997; Yang et al., 2000; Yang & Nielsen, 2002).
In order to check whether the oprF gene had caused episodes of positive selection on some lineages, we estimated the Ka/Ks ratio in the different o-clusters. Interestingly, while all Ka/Ks values calculated from within-cluster comparisons are lower than those calculated from overall comparisons, between-cluster comparisons show higher ratios, some of them being greater than 1 (Table 3). Because the higher ratio is too near to 1 (1·29; for the comparison between fluorescens o-clusters 1 and 2), we cannot conclude that positive selection occurred; relaxed evolutionary constraints would probably lead to a similar result. Even so, our results clearly show that selective constraints between o-clusters, especially between fluorescens o-clusters, have varied over time.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
). The differences corresponded essentially to a separation of the cluster containing P. fluorescens strains and the one containing P. putida strains, while no horizontal transfer was observed between clusters containing P. fluorescens and P. putida strains (termed o-clusters). To explain this phylogeny, we proposed three hypotheses: (i) an ancient horizontal transfer (after the P. fluorescens–P. putida speciation), (ii) an ancient duplication (before the speciation) followed by several losses, or (iii) an artifact in phylogeny reconstruction. These different possibilities have been investigated in this study.
The gene encoding 16S rDNA has been widely used as a molecular sequence to reconstruct species phylogenies because it was assumed that intraspecific variation and horizontal transfer was low for this gene (Woese, 1987). However, in some groups of bacteria, particularly in the genus Pseudomonas, the inferred phylogenies based on 16S rDNA lack resolution at the intrageneric level due a slow rate of evolution (Moore et al., 1996; Anzai et al., 2000). Concatenation of 46 genes encoding ribosomal proteins provided a very robust systematic analysis of the relationships between P. aeruginosa, P. putida, P. syringae and P. fluorescens (Fig. 1). Monophylies of P. syringae and P. syringae plus P. fluorescens, which are supported by maximum bootstrap values with both protein and nucleotide sequences, were also supported by high bootstrap values in studies based on other genes: gyrB-rpoD concatenation, atpD-carA-recA-16S rDNA concatenation and the rpoB gene (Yamamoto et al., 2000; Hilario et al., 2004; Ait Tayeb et al., 2005). Robust tests of incongruence have confirmed this topology on concatenated ribosomal genes and on the 16S rDNA gene. On the other hand, the monophyly of P. fluorescens is only supported with nucleotide sequences (but with a maximum value) and not found by ML analysis, although it was not significantly rejected. By contrast, since the distance tree of the oprF gene (or OprF protein) shows a separation of P. fluorescens by a strongly supported node (99 %), ML analysis does not significantly support the incongruence between this topology and the species topology. Therefore, an artifact in the reconstruction of oprF phylogeny cannot be rejected.
To search for clues about horizontal transfer or duplication, we investigated the synonymous codon usage, GC3 content and genomic context of the oprF gene. Altogether, these features clearly showed an absence of ancient duplication or recent horizontal transfer. Ancient horizontal transfer followed by modification of the GC3 content towards the new genomic features cannot be formally excluded but it is unlikely because the chromosome context around the oprF gene is conserved and only a double crossing-over could explain a horizontal transfer of the oprF gene. In strain Pf-5, an interesting deletion, just downstream of oprF, could correspond to a replacement of the gene, although this feature is not present in the second strain of the fluorescens 2 o-cluster (SBW25 strain). This conservation of the genomic context has even already been shown in additional Pseudomonas strains, including a third strain of the fluorescens 2 o-cluster, strain OE28.3 (Brinkman et al., 1999). In this study, Brinkman et al. (1999) showed an influence of the sigX gene (a putative ECF sigma factor gene) on expression regulation of the oprF gene. Moreover, it is unlikely that the GC3 of oprF gene has greatly changed after horizontal transfer because there is no saturation at synonymous positions in the oprF gene and there is similar synonymous usage in Pseudomonas (except in strain PAO1). Finally, the Ks values between fluorescens o-clusters were closer or identical (for the 7- and 20-sequence sets, respectively) than between fluorescens and putida o-clusters (Table 3). This feature is well illustrated by trees built from synonymous distances (Fig. 5). Although the species topology was not found with the oprF gene, there was a monophyly of P. fluorescens plus P. syringae, well supported with the 20-sequence set (Fig. 5a). For the concatenated ribosomal genes, a phylogeny with synonymous distance shows a monophyly of P. fluorescens supported by a maximum bootstrap value (Fig. 5c); the same is true for the phylogeny with evolutionary distance (with nucleotide sequences) (Fig. 1c). The P. fluorescens intracluster resolution was even slightly better with synonymous distance than with evolutionary distance.
|
). Because the higher ratio is too near to 1, we cannot exclude that only relaxed evolutionary constraints had occurred. There now exist more sensitive ML tests for adaptive evolution that investigate individual codons in a gene and/or individual branches in a phylogeny (Yang et al., 2000; Yang & Nielsen, 2002). However, those tests require a known topology of the phylogenetic tree, while the observed oprF phylogeny could result from an artifact.
In fact, modifications in evolutionary constraints were evident because the central region of the protein contains a much-conserved surface-exposed motif, in all Pseudomonas species except in the fluorescens 2 o-cluster and in two species of the aeruginosa o-cluster (P. mendocina and P. stuzeri) (Vermeiren et al., 1999; Bodilis et al., 2004). This motif has four cysteine residues probably implicated in the formation of two disulfide bonds. According to the tree topology (OprF or species topologies), we think that the absence of this cysteine-rich motif results from at least two independent deletions. The role of this motif is unknown, although similarities with the calcium-binding repeats of the thrombospondin protein (a eukaryotic extracellular matrix protein) have previously been shown (de Mot & Vanderleyden, 1994). Interestingly, there is a proline-rich variable region situated next to the conserved motif, on the N-terminal side; this region is slightly longer when the conserved motif is absent. Since all the central region of oprF was removed in this study, modifications in evolutionary constraints must also have occurred in other regions of the gene.
Altogether, these results support the hypothesis of trans-species polymorphism (i.e. presence of alleles in different species that are more similar to each other than they are to alleles in the same species) in the oprF gene.
In a previous study, we suggested that the polymorphism of OprF could be correlated with the ecological niche, since the fluorescens 1 o-cluster contained most of the bulk soil (non-rhizospheric) isolates and the fluorescens 2 o-cluster contained all the clinical isolates and most of the rhizospheric isolates (Bodilis et al., 2004). In order to estimate the proportion of oprF type 2 (in the cultivable and uncultivable fractions) from environmental DNA, we have developed a ratio PCR (J. Bodilis and others, unpublished). By using this molecular tool, we found that the proportion of oprF type 2 (without the cysteine-rich motif) from cultivated fields (wheat or flax) was significantly different between the rhizosphere samples and the adjacent bulk soil samples (J. Bodilis and others, unpublished). The physiological properties that are linked with oprF type 2 fitness are still unknown, but could concern adhesion (Bodilis et al., 2004) or resistance to a variation of osmotic pressure, although the cysteine-rich motif is not essential for channel size modulations and pore features, as shown by Jaouen et al. (2004). Since OprF is known to be a pleiotropic protein, modifications in evolutionary constraints could result from variations in cryptic functions. More ecological studies and experiments with mutants in gnobiotic systems should further improve our knowledge of the multifunction of OprF.
Finally, it was interesting to find OprF without the cysteine-rich motif in the two more divergent o-clusters (aeruginosa and fluorescens 2); this divergence cannot be explained by the simultaneous presence of the two OprF types in the fluorescens 2 o-cluster (Table 3). Nonsynonymous mutations must therefore have accelerated in these two evolutionary lineages though relaxed constraints and/or though episodic positive evolution. Either of these phenomena could have led to a phylogeny reconstruction artifact, by attracting the two longest branches of the tree towards each other.
|
ACKNOWLEDGEMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Ait Tayeb, L., Ageron, E., Grimont, F. & Grimont, P. A. D. (2005). Molecular phylogeny of the genus Pseudomonas based on rpoB sequences and application for the identification of isolates. Res Microbiol 156, 763–773.[Medline]
Anzai, Y., Kim, H., Park, J., Wakabayashi, H. & Oyaizu, H. (2000). Phylogenetic affiliation of the pseudomonads based on 16S rRNA sequence. Int J Syst Evol Microbiol 50, 1563–1589.[Abstract]
Bellido, F., Martin, N., Siehnel, R. & Hancock, R. (1992). Reevaluation, using intact cells, of the exclusion limit and role of porin OprF in Pseudomonas aeruginosa outer membrane permeability. J Bacteriol 174, 5196–5203.
Bodilis, J., Calbrix, R., Guerillon, J., Merieau, A., Pawlak, B., Orange, N. & Barray, S. (2004). Phylogenetic relationships between environmental and clinical isolates of Pseudomonas fluorescens and related species deduced from 16S rRNA gene and OprF protein sequences. Syst Appl Microbiol 27, 93–108.[CrossRef][Medline]
Brinkman, F. S., Schoofs, G., Hancock, R. E. & de Mot, R. (1999). Influence of a putative ECF sigma factor on expression of the major outer membrane protein, OprF, in Pseudomonas aeruginosa and Pseudomonas fluorescens. J Bacteriol 181, 4746–4754.
Brinkman, F. S., Bains, M. & Hancock, R. E. (2000). The amino terminus of Pseudomonas aeruginosa outer membrane protein OprF forms channels in lipid bilayer membranes: correlation with a three-dimensional model. J Bacteriol 182, 5251–5255.
Dé, E., de Mot, R., Orange, N., Saint, N. & Molle, G. (1995). Channel-forming properties and structural homology of major outer membrane proteins from Pseudomonas fluorescens MFO and OE 28.3. FEMS Microbiol Let 127, 267–272.[CrossRef][Medline]
Dé, E., Orange, N., Saint, N., Guerillon, J., de Mot, R. & Molle, G. (1997). Growth temperature dependence of channel size of the major outer- membrane protein (OprF) in psychrotrophic Pseudomonas fluorescens strains. Microbiology 143, 1029–1035.[Abstract]
de Mot, R. & Vanderleyden, J. (1994). A conserved surface-exposed domain in major outer membrane proteins of pathogenic Pseudomonas and Branhamella species shares sequence homology with the calcium-binding repeats of the eukaryotic extracellular matrix protein thrombospondin. Mol Microbiol 13, 379–380.[CrossRef][Medline]
de Mot, R., Proost, P., van Damme, J. & Vanderleyden, J. (1992). Homology of the root adhesin of Pseudomonas fluorescens OE 28.3 with porin F of P. aeruginosa and P. syringae. Mol Gen Genet 231, 489–493.[CrossRef][Medline]
de Mot, R., Schoofs, G., Roelandt, A., Declerck, P., Proost, P., Van Damme, J. & Vanderleyden, J. (1994). Molecular characterization of the major outer-membrane protein OprF from plant root-colonizing Pseudomonas fluorescens. Microbiology 140, 1377–1387.[Abstract]
Endo, T., Ikeo, K. & Gojobori, T. (1996). Large-scale search for genes on which positive selection may operate. Mol Biol Evol 13, 685–690.[Abstract]
Herbeck, J. T., Wall, D. P. & Wernegreen, J. J. (2003). Gene expression level influences amino acid usage, but not codon usage, in the tsetse fly endosymbiont Wigglesworthia. Microbiology 149, 2585–2596.
Hilario, E., Buckley, T. & Young, J. (2004). Improved resolution on the phylogenetic relationships among Pseudomonas by the combined analysis of atpD, carA, recA and 16S rDNA. Antonie van Leeuwenhoek 86, 51–64.[CrossRef][Medline]
Jaouen, T., De, E., Chevalier, S. & Orange, N. (2004). Pore size dependence on growth temperature is a common characteristic of the major outer membrane protein OprF in psychrotrophic and mesophilic Pseudomonas species. Appl Environ Microbiol 70, 6665–6669.
Jukes, T. & Cantor, C. (1969). Evolution of Protein Molecules. Mammalian Protein Metabolism. New York: Academic Press.
Kersters, K., Ludwig, W., Vancanneyt, M., de Vos, P., Gillis, M. & Schleifer, K.-H. (1996). Recent changes in the classification of the pseudomonads: an overview. Syst Appl Microbiol 19, 465–477.
Kimura, M. (1980). A simple method for estimation of evolutionary rate of base substitutions through comparative studies of nucleotide sequences. J Mol Evol 16, 111–120.[CrossRef][Medline]
Kumar, S., Tamura, K., Jakobsen, I. B. & Nei, M. (2001). MEGA2: molecular evolutionary genetics analysis software. Bioinformatics 17, 1244–1245.
Lafay, B., Atherton, J. C. & Sharp, P. M. (2000). Absence of translationally selected synonymous codon usage bias in Helicobacter pylori. Microbiology 146, 851–860.
Messier, W. & Stewart, C.-B. (1997). Episodic adaptative evolution of primate lysozymes. Nature 385, 151–154.[CrossRef][Medline]
Moore, E. R. B., Mau, M., Arnscheidt, A., Böttger, E., Hutson, R., Collins, M., Van de Peer, Y., Wachter, R. & Timmis, K. N. (1996). The determination and comparison of the 16S rRNA gene sequences of species of the genus Pseudomonas (sensu stricto) and estimation of the natural intrageneric relationships. Syst Appl Microbiol 19, 478–492.
Muto, A. & Osawa, S. (1987). The guanine and cytosine content of genomic DNA and bacterial evolution. Proc Natl Acad Sci U S A 84, 166–169.
Nikaido, H. (2003). Molecular basis of bacterial outer membrane permeability revisited. Microbiol Mol Biol Rev 67, 593–656.
Orange, N. (1994). Growth temperature regulates the induction of beta-lactamase in Pseudomonas fluorescens through modulation of the outer membrane permeation of a beta-lactam-inducing antibiotic. Microbiology 140, 3125–3130.[Abstract]
Palacios, C. & Wernegreen, J. J. (2002). A strong effect of AT mutational bias on amino acid usage in Buchnera is mitigated at high-expression genes. Mol Biol Evol 19, 1575–1584.
Pautsch, A. & Schlulz, G. (1998). Structure of the outer membrane protein A transmembrane domain. Nat Struct Biol 5, 1013–1017.[CrossRef][Medline]
Ramette, A., Moenne-Loccoz, Y. & Defago, G. (2001). Polymorphism of the polyketide synthase gene phlD in biocontrol fluorescent pseudomonads producing 2,4-diacetylphloroglucinol and comparison of PhID with plant polyketide synthases. Mol Plant Microbe Interact 14, 639–652.[Medline]
Rawling, E., Martin, N. & Hancock, R. (1995). Epitope mapping of the Pseudomonas aeruginosa major outer membrane porin protein OprF. Infect Immun 63, 38–42.[Abstract]
Rawling, E. G., Brinkman, F. S. & Hancock, R. E. (1998). Roles of the carboxy-terminal half of Pseudomonas aeruginosa major outer membrane protein OprF in cell shape, growth in low-osmolarity medium, and peptidoglycan association. J Bacteriol 180, 3556–3562.
Rebière-Huët, J., Di Martino, P., Gallet, O. & Hulen, C. (1999). Interaction of Pseudomonas aeruginosa outer membrane proteins with plasmatic fibronectin. A root for new bacterial adhesins. C R Acad Sci Paris 322, 1071–1080.
Rediers, H., Vanderleyden, J. & de Mot, R. (2004). Azotobacter vinelandii: a Pseudomonas in disguise? Microbiology 150, 1117–1119.
Rice, P., Longden, I. & Bleasby, A. (2000). EMBOSS: the European Molecular Biology Open Software Suite. Trends Genet 16, 276–277.[CrossRef][Medline]
Saitou, N. & Nei, M. (1987). The neighbor-joining method: a new method for reconstructing phylogenetic tree. Mol Biol Evol 4, 406–425.[Abstract]
Sharp, P. & Li, W. (1987). The codon adaptation index – a measure of directional synonymous codon usage bias, and its potential applications. Nucleic Acids Res 15, 1281–1295.
Shimodaira, H. & Hasegawa, M. (1999). Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol Biol Evol 16, 1114–1116.
Tamber, S. & Hancock, R. (2004). The outer membranes of pseudomonads. In Pseudomonas, pp. 575–601. Edited by J. L. Ramos. New York: Plenum.
Thompson, J. D., Gibson, T. J., Plewniak, F., Jeanmougin, F. & Higgins, D. G. (1997). The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res 25, 4876–4882.
Vermeiren, H., Willems, A., Schoofs, G., De Mot, R., Keijers, V., Hai, W. & Vanderleyden, J. (1999). The rice inoculant strain Alcaligenes faecalis A15 is a nitrogen-fixing Pseudomonas stutzeri. Syst Appl Microbiol 22, 215–224.[Medline]
von Specht, B., Knapp, B., Muth, G., Broker, M., Hungerer, K., Diehl, K., Massarrat, K., Seemann, A. & Domdey, H. (1995). Protection of immunocompromised mice against lethal infection with Pseudomonas aeruginosa by active or passive immunization with recombinant P. aeruginosa outer membrane protein F and outer membrane protein I fusion proteins. Infect Immun 63, 1855–1862.[Abstract]
Woese, C. R. (1987). Bacterial evolution. Microbiol Rev 51, 221–271.
Woodruff, W. & Hancock, R. (1989). Pseudomonas aeruginosa outer membrane protein F: structural role and relationship to the Escherichia coli OmpA protein. J Bacteriol 171, 3304–3309.
Worgall, S., Krause, A., Rivara, M. & 8 other authors (2005). Protection against P. aeruginosa with an adenovirus vector containing an OprF epitope in the capsid. J Clin Invest 115, 1281–1289.[CrossRef][Medline]
Wu, L., Estrada, O., Zaborina, O. & 12 other authors (2005). Recognition of host immune activation by Pseudomonas aeruginosa. Science 309, 774–777.
Yamamoto, S., Kasai, H., Arnold, D. L., Jackson, R. W., Vivian, A. & Harayama, S. (2000). Phylogeny of the genus Pseudomonas: intrageneric structure reconstructed from the nucleotide sequences of gyrB and rpoD genes. Microbiology 146, 2385–2394.
Yang, Z. & Nielsen, R. (2002). Codon-substitution models for detecting molecular adaptation at individual sites along specific lineages. Mol Biol Evol 19, 908–917.
Yang, Z., Nielsen, R., Goldman, N. & Pedersen, A.-M. K. (2000). Codon-substitution models for heterogeneous selection pressure at amino acid sites. Genetics 155, 431–449.
Zhang, J., Rosenberg, H. F. & Nei, M. (1998). Positive Darwinian selection after gene duplication in primate ribonuclease genes. Proc Natl Acad Sci U S A 95, 3708–3713.
Received 2 November 2005;
revised 9 December 2005;
accepted 15 December 2005.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
50 % are reported. The horizontal length of branches is proportional to the estimated number of substitutions.
