INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

More than 30 microbial genomes have been completely sequenced, and sequencing of many more is in progress (20). The public availability of completed genomes provides researchers in life sciences with new possibilities to address important biological issues. However, the annotation of complete genome sequences frequently fails to classify a significant proportion (40 to 60%) of gene products (20). These new challenges may be approached in silico. For example, analyses of functionally or structurally related proteins allow classification into families and subfamilies (29, 42), and such efforts may result in the identification of sequence motifs (46). (The terms "motif" and "pattern" are used according to the guidelines in the PROSITE documentation, which can be found at the website of ExPASy [www.expasy.ch/tools].) Traditionally, a collection of patterns are applied to a single protein in order to identify the presence of motifs, which offers the possibility of making structural or functional predictions. Conversely, a well-defined pattern could facilitate classification of gene products at the genome level (8).

The bacterial surface is a crucial site of interaction between microbe and host. Bacterial surface proteins constitute a diverse group of molecules with important functions, such as adhesion, signaling, and defense mechanisms. Moreover, surface proteins are potential drug or vaccine targets. Surface proteins in gram-positive bacteria are typically lipoproteins (52) or cell wall-attached proteins (CWPs) (27, 35). CWPs of gram-positive bacteria have a conserved COOH-terminal region containing a hexapeptide sequence known as the LPXTGX motif (18). The LPXTGX motif is followed by a hydrophobic stretch of amino acids and a short charged tail, all three of which are necessary for efficient sorting of a protein to the cell wall (49). A membrane-associated enzyme called sortase catalyzes the transpeptidation of the threonine residue in the motif to the amino acid cross bridge of the peptidoglycan cell wall (15, 33, 34, 36, 47, 53-55). Homologues of the srtA gene, which encodes the sortase, have been found in many genomes of gram-positive bacteria, indicating that this mechanism of sorting is universal among gram-positive bacteria (33). The hydrophobic region next to the LPXTGX motif traverses the plasma membrane, while the charged tail probably acts as a stop-transfer signal, retaining the protein at the bacterial surface until further processing occurs (48, 49).

This work attempted to refine and develop the previously described motif that distinguishes CWPs. We suggest a tripartite motif, which takes into account the three principal requirements for efficient sorting. The pattern devised for this motif was used in a screening of the complete Streptococcus pyogenes genome, and a number of novel putative CWPs were identified. Compared to the classical pattern (LPXTGX), the tripartite pattern has increased sensitivity and specificity as shown by analysis of six bacterial genomes. The procedures employed do not require extensive bioinformatics resources and are thus readily accessible to the research community.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

The collection of 65 COOH-terminal sorting signals was from a recent review (Table 1 in reference 35). Translation of genome sequences and searches with the tripartite and LPXTGX patterns were performed using MacVector (Oxford Molecular Ltd., version 6.5.3). The website of the University of Oklahoma's Advanced Center for Genome Technology was used to access genomes of S. pyogenes, Staphylococcus aureus (NCTC 8325), and Streptococcus mutans (UA159). The genomes of Enterococcus faecalis (V583) and Streptococcus pneumoniae (type 4) were obtained from The Institute for Genomic Research website. The Escherichia coli (K-12) genome was obtained from the University of WisconsinMadison website. BLASTp (version 2.0) searches were performed at the National Center for Biotechnology Information website. The tBLASTn searches of the S. pyogenes genomes were done using either the BLAST server at Oklahoma University (strain SF370) or the server at the Sanger Centre, University of Newcastle (strain Manfredo). Signal sequence predictions were made using the SignalP (version 1.1) program at the website of the Center for Biological Sequence Analysis (37). Only proteins where the SignalP combined cleavage site score (Y) exceeded the threshold were considered positive. PROSITE scans for detection of motifs in novel CWPs were made at the website of Pole Bio-Informatique Lyonnais.

Partial or whole-genome screening was conducted by translating genomic DNA sequences in all six reading frames to generate amino acid files, to which the patterns were applied. The sequence analysis software used was incapable of excluding matches containing stop codons within the motif, and these matches were thus excluded manually. A match was considered false positive if not located within an open reading frame (ORF) encoding a putative protein of at least 100 amino acids. Moreover, if the match was located within an ORF but more than 50 amino acids from the COOH-terminus of the putative protein, it was also considered a false-positive hit. Remaining matches were regarded as true positives.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of a tripartite pattern by analysis of COOH-terminal sorting signals. A recently published list of 65 CWPs from 19 different gram-positive bacteria was used to obtain a set of COOH-terminal sorting signals (35). COOH-terminal amino acid sequences from each protein were tentatively divided into three parts for further analysis: a sortase target region, a membrane-spanning region, and a charged tail region. The sortase target region consists of six amino acids, where the first residue corresponds to leucine in the LPXTGX motif. The first and second positions of this region were completely conserved in all 65 CWPs and contained leucine and proline, respectively. In positions 3 to 6, a variety of amino acids were found (Fig. 1A). Amino acids that occurred more than once in a particular position were considered appropriate for inclusion in a new pattern (Fig. 1A). For each position, more than 90% of the variation was thus included. The membrane-spanning region contains at least 15 amino acids, which is sufficient to traverse the membrane (48). Most proteins contained a gap of variable length between the sortase target region and the first hydrophobic amino acid. We therefore chose to define the membrane-spanning region in relation to the charged tail. Thus, the 15 amino acids preceding a COOH-terminal cluster of charged residues (K or R) were considered membrane spanning for the purpose of this analysis. There are no theoretical or experimental indications that specific positions in the membrane-spanning region are important. Thus, the amino acid distribution was calculated for the membrane-spanning region as a whole (Fig. 1B). The nine most frequent amino acids together represented more than 95% of the 15 positions in the 65 CWPs. Finally, the charged tail region, starting (+1) with a lysine or arginine residue immediately following the membrane-spanning region, was analyzed. Since several proteins had a charged tail of four or five amino acids only, an amino acid distribution was calculated for the first five positions (Fig. 1C). The results showed that the first three positions were dominated by lysine or arginine residues. This is in accordance with experimental data showing that two of three consecutive residues in the charged tail must be lysine or arginine to allow efficient sorting (48).

View larger version (36K): [in this window] [in a new window]   FIG. 1.   Analysis of the cell wall sorting signal in 65 CWPs. (A) Amino acid distribution in positions 3 to 6 of the sortase target region. Amino acids included in the novel pattern are in underlined boldface. (B) Amino acid distribution in the membrane-spanning region. Amino acids included in the novel pattern are in bold. (C) Amino acid distribution in the first five positions of the charged tail.

Specificity and sensitivity of the tripartite pattern. When applied to the set of 65 previously described CWPs, the sensitivity of the tripartite pattern was 98%, compared to 80% for the LPXTGX pattern (the tripartite pattern fails to properly classify the S. mutans dextranase, due to its unusually long gap). The tripartite and the LPXTGX patterns were used to search for genes encoding CWPs in the entire genome of S. pyogenes. In addition, a partial screening was performed on the genomes of E. faecalis, S. aureus, S. mutans, S. pneumoniae, and E. coli (the gap length between the first and second part of the tripartite pattern was increased to 11 amino acids for the analysis of S. mutans, since at least one CWP [dextranase] from this species has a longer gap). In all the genomes of gram-positive bacteria, both patterns identified genes encoding putative CWPs, of which approximately 50% were previously described (Table 2). In total, the tripartite pattern identified 35 genes and the LPXTGX pattern identified 29 genes encoding putative or known CWPs. Thus, the sensitivity of the tripartite pattern was 21% higher than that of the LPXTGX pattern. For all gram-positive bacterial genomes analyzed, the tripartite pattern had a higher sensitivity than the classical pattern. Also, the specificity of the tripartite pattern was higher (8 false positives) than that of the LPXTGX pattern (24 false positives) (Table 2). In one particular case, a putative protein containing the LPXTGX motif in its COOH-terminal part was considered false positive. This protein, which was not identified by the tripartite pattern, lacked both a hydrophobic region and a charged tail; thus, it cannot be attached to the cell wall (48). For comparison, the genome of the gram-negative bacterium E. coli was included, and neither pattern identified proteins containing gram-positive sorting signals in their COOH-terminal part. We made several attempts to increase the sensitivity of the tripartite pattern by allowing more mismatches in the three different parts of the pattern. However, this invariably led to a decreased specificity without apparent gain of sensitivity (data not shown).

The CWPs encoded by an S. pyogenes genome. In the complete genome of the S. pyogenes M1 strain SF370, the tripartite pattern identified 14 genes encoding putative CWPs (Table 3). Twelve of these genes were identified using the LPXTGX pattern. The tripartite pattern successfully identified the six previously known CWPs (11, 21, 30, 31, 39, 44, 45, 59) expected to be present in the SF370 strain. In addition, the tripartite pattern identified a putative CWP containing a subtilase type of serine protease motif (3). This protein is similar to the C5a peptidase of Streptococcus agalactiae and S. pyogenes (11, 12). A fragment of this serine protease was recently described to be present in culture supernatant of S. pyogenes (28). Recently, an LPXTGX-containing hyaluronidase (HylA) in S. pyogenes was described (24). In S. pneumoniae, a similar hyaluronidase containing a typical cell wall sorting signal has been proposed to be cell wall attached (5, 25). The patterns used here failed to detect HylA in S. pyogenes, but a tBLASTn search with HylA showed that hylA is present in the SF370 strain. The gene encodes a HylA protein lacking the cell wall sorting region. However, immediately downstream from the stop codon of hylA, a sequence encoding a characteristic cell wall sorting signal was present. Also, in the S. pyogenes strain Manfredo, a single stop codon separated the hylA gene from a sequence encoding a cell wall attachment signal.

Identification of putative proteins containing atypical sortase target regions. Analogous to the screening performed on S. pyogenes, the partial screening of genomes from S. pneumoniae, E. faecalis, S. aureus, and S. mutans resulted in identification of CWPs not detected by the classical pattern. Among the 35 proteins found, seven proteins contained an atypical sortase target region (Table 4). All proteins contained a typical membrane-spanning region and charged tail. All but one (Cwp5) had putative signal sequences. Databases were searched for proteins similar to these atypical CWPs, and the most similar proteins were all bacterial surface proteins, or potentially surface attached (Bacillus amylopullulanase). Among the putative CWPs with an atypical sortase target region, the pullulanase of S. pneumoniae and the C5a peptidase of S. pyogenes were previously described as surface located (9, 57). The S. pneumoniae pullulanase is homologous to SpuA in S. pyogenes (see above). Deviations from the classical sortase target region were found in positions 4, 5, and 6. The putative CWP in S. aureus and the Cwp3 homologue from S. pyogenes strain Manfredo contained alanine in position 4, which is fully compatible with efficient sorting (49). Like the C5a peptidase, which is a known CWP, the putative CWP in E. faecalis has asparagine in position 5 (11, 57, 58). Cwp5 contains a serine instead of a threonine in position 4. Opacity factor, a CWP found in certain serotypes of S. pyogenes, also shows this variation (43). These examples show that use of the tripartite pattern increases the possibility of performing accurate and complete genome screenings.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of whole genomes, i.e., similarity searches or targeted gene identification, has become an integral part of research in the life sciences. In microbiology, genomics may help to identify virulence factors, as well as drug or vaccine targets (50, 51). However, the readiness with which sequence information can be utilized is often hampered by limited annotation. The number of unclassified genes is steadily increasing, and there is a need for new tools for functional and structural classification of genes and gene products (4). The use of specific and sensitive patterns is one way to classify putative proteins into groups or families. In this work we present a refined pattern that readily detects CWPs of gram-positive bacteria, and we have successfully applied this pattern in whole-genome screening procedures. CWPs comprise a diverse group of bacterial surface proteins, sharing a common theme of cell wall sorting (35). Several studies have shown that isogenic mutants lacking specific CWPs are attenuated in virulence in animal models (1, 13). A sortase mutant in S. aureus, which missorts all CWPs, was significantly attenuated in virulence compared to the corresponding wild-type strain (32). Also, surface proteins are potential vaccine targets, and a recent study convincingly showed the strength of genome screening as an initial step in the identification of vaccine candidates (38). Among the CWPs discussed in this work, the M protein has been used for the development of a multivalent group A streptococcal vaccine (14). Moreover, the pneumococcal pullulanase is immunogenic and highly conserved, making it a potential target for vaccine development (9). Cell wall sorting signals have also been used to create fusion proteins with heterologous antigens, which can thus be expressed in a gram-positive commensal for oral vaccination (17, 40). Moreover, CWPs from gram-positive bacteria are known to interact with human proteins and as such may be used as biotechnological tools. For example, protein A of S. aureus (19), protein G of group C and G streptococci (7), and protein L of Peptostreptococcus magnus (6) are widely used as tools for purification or detection of immunoglobulins. The pattern described herein presents an opportunity to identify novel CWPs, which may be virulence factors or vaccine targets or which may be used as biotechnological tools. Presumably, the complete set of CWPs in any gram-positive bacterium can be obtained by using the tripartite motif. Optimization of the pattern may be needed for certain species in order to approach completeness. For this purpose, the most evident modification is to increase the gap length, as selected CWPs exceed the maximum gap of eight residues. For S. mutans and S. aureus, there are such examples, and a marginal gain of sensitivity may be accomplished by increasing the gap. Such a change (maximum gap = 11) was made in the partial screening of S. mutants, which allowed identification of dextranase. In the complete screening of S. pyogenes, an increase in gap (data not shown) did not result in additional hits. Another strategy could be to exclude the membrane-spanning domain, which is the most loosely defined component, thus operating with a bipartite motif. This will greatly increase the number of false-positive hits but may possibly offer a somewhat higher sensitivity.

A common theme of previously described CWPs is the presence of repeated sequences, sometimes involved in protein-protein interactions (35). Only one of the seven novel putative CWPs identified in S. pyogenes contained such repeats. Instead, at least four CWPs were predicted to have enzymatic activity. From our data on the S. pyogenes repertoire of CWPs it seems that cell wall-attached enzymes are more common than previously recognized. Whole-genome analyses of other genomes from gram-positive bacteria are needed to support this speculation. From partial screenings of four other gram-positive genomes it is clear that many novel CWPs remain to be discovered. Less than 50% of CWPs identified here have been described previously. Our data imply that the number of CWPs varies considerably between different species of gram-positive bacteria. However, such extrapolations from data on partial screening procedures are difficult, since it became apparent from the screening of the S. pyogenes genome that CWP-encoding genes are not evenly distributed on the chromosome. On the contrary, we identified clusters of CWP-encoding genes in the genomes of S. pyogenes, S. aureus, and E. faecalis (data not shown).

Since the discovery that CWPs contain a conserved LPXTGX motif, it has become increasingly apparent that variations within the motif exist. An early report indicated that there is an absolute requirement for the proline residue, but not threonine, in the LPXTGX motif for efficient sorting to occur (49). The lack of absolute conservation within the motif is supported by the fact that several well-known CWPs, like the C5a peptidase and opacity factor of S. pyogenes and the -antigen of Streptococcus agalactiae, deviate from the classical motif in positions 4, 5, and 6, respectively (11, 23, 26, 43). If the tripartite pattern presented here is utilized in whole-genome analyses, the list of atypical sortase target sequences can be expanded. Such a list of atypical sequences can serve as a base for detailed studies on sortase specificity. Despite sequence conservation of the srtA gene, which encodes the sortase (33), small variations in sortase specificity might exist among different species of gram-positive bacteria. Also, the amino acid composition and length of the transmembrane part or the charged tail may vary between different gram-positive bacteria. Genome-based identification of additional CWPs by the use of the tripartite pattern should provide important information on these topics. In conclusion, the use of this pattern could be an easily accessible and important tool in the discovery of novel CWPs in gram-positive bacteria, which will help to address a number of important biological issues.