INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Haemophilus influenzae, a small, gram-negative bacterium, is a causative agent of invasive diseases, such as meningitis and epiglottitis, as well as mucosal infections, including otitis media and chronic bronchitis. The vast majority of H. influenzae strains, however, are harmless commensals of the human upper respiratory tract (25). Invasive disease (meningitis, epiglottitis, and septicemia) is associated with a minority of virulent strains, generally but not exclusively encapsulated and of serotype b (48). With the advent of an effective vaccine against the b capsular antigens, the incidence of H. influenzae type b (Hib)-associated meningitis has declined in industrialized countries; however, meningitis associated with serotype a strains and nonencapsulated varieties has occasionally been reported (28, 34). Genetic comparisons between virulent and nonpathogenic strains of H. influenzae are expected to shed light on mechanisms of host invasion (12).

In 1995, H. influenzae Rd became the first free-living organism to have its entire genome sequenced (12). The Rd strain, a nonencapsulated derivative of a serotype d isolate, is nonpathogenic; part of the impetus behind sequencing its genome was to provide a baseline comparison with pathogenic isolates. The virulent Eagan strain (serotype b), isolated from a patient with meningitis, has a genome approximately 270 kb larger than that of Rd (9), and comparisons of Hib and Rd genomes have uncovered virulence-associated genes exclusive to Hib (27, 49). The fimbrial gene cluster, which facilitates colonization and adhesion to host cells, is found on an 8-kb fragment of Hib DNA which is missing from Rd (49). In this study, we used genomic comparisons between Rd and Hib (Eagan) to identify a candidate virulence-associated marker, the tryptophanase operon of H. influenzae.

Humans are the only natural hosts for H. influenzae. Most H. influenzae isolates from healthy individuals are nonencapsulated, as are isolates from surface infections (40). Approximately 7% of all strains are encapsulated; these strains have been sorted into six serotypes (a through f) on the basis of capsular antigens (37, 40). Kilian (25) proposed subdividing H. influenzae into five biotypes based on metabolic criteria; according to this scheme, most Hib strains belong to biotype I, and most noninvasive encapsulated (d and e) strains belong to biotype III. Multilocus enzyme electrophoresis and outer membrane protein profiles have distinguished two major phylogenetic divisions among encapsulated strains, with invasive strains clustering in division I (31, 32).

As early as 1922, it was noted that H. influenzae strains differed in the production of indole, with nearly all pathogenic isolates testing positive (25). The indole test was one of the defining criteria of Kilian's biotypes (I, II, and V are indole positive, and III and IV are indole negative). Indole production results from the catabolism of tryptophan to indole, pyruvate, and ammonia, a pathway which allows tryptophan to be used as a carbon and nitrogen source (33). In Escherichia coli, the tryptophanase (tna) operon has been extensively studied as a regulatory model (13-15, 22, 26, 43-45). Located at 83.8 min on the E. coli genetic map, the tna operon consists of regulatory sequences and three genes: tnaA (the structural gene for tryptophanase), tnaB (encoding a low-affinity tryptophan permease), and tnaC, encoding a 24-residue leader peptide. The operon is inducible by tryptophan and catabolite repressed by glucose. Tryptophan induction requires the TnaC peptide, which acts in cis to inhibit Rho-mediated transcriptional termination downstream of tnaC, thus allowing transcription of tnaA and tnaB. A single Trp residue at position 12 is essential for induction. Current models propose that when tryptophan is abundant, TnaC inhibits ribosome release, thereby interfering with termination at a Rho utilization site downstream from the TnaC stop codon (26, 50).

Homologs of E. coli tna genes are found in other gram-negative bacteria and have been cloned from Proteus vulgaris (21), Enterobacter aerogenes (24), and Symbiobacterium thermophilum (18); the organization of genes within the operon and the location of presumed regulatory sites are conserved among the enteric bacteria (21).

Does the correlation between indole production and virulence identify tryptophanase as a virulence factor, or does it merely reflect a shared recent ancestry of virulent strains? What is the evolutionary origin of H. influenzae tryptophanase? To begin to address these questions, we have identified and sequenced the tryptophanase operon of Hib (Eagan).

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains and plasmids. Plasmids were maintained in E. coli DH5 and purified by standard methods (38) with Wizard or Qiagen kits. Bacterial strains are described in Table 1. Growth of H. influenzae was as previously described (41).

Indole test. H. influenzae strains were cultured overnight at 37°C in supplemented brain heart infusion (sBHI) agar. Strips of filter paper were saturated with p-dimethylaminobenzaldehyde (Kovács reagent), and isolated colonies were transferred to the filter paper. Development of a bright pink-red color within 30 s indicated a positive test (20).

Tryptophanase assays. Preparation of bacterial extracts and tryptophanase assays were as described by Gish and Yanofsky (13). E. coli strains were grown at 37°C to the mid-exponential phase in morpholinepropanesulfonic acid (MOPS)-based minimal medium [40 mM 3-(N-morpholino)propanesulfonic acid, 4 mM Tricine buffer (pH 7.2), 50 mM KCl, 10 mM NH₄Cl, 0.5 mM MgSO₄, 1.3 mM K₂HPO₄] (13) supplemented with 1% acid-hydrolyzed casein (ICN Pharmaceuticals Inc.) and 1 µg of thiamine per ml. Cultures were induced either with L-tryptophan (100 µg/ml) or with DL-1-methyltryptophan (10 µg/ml). H. influenzae (and E. coli in selected experiments) was grown in sBHI broth. Overnight cultures were diluted 200-fold into sBHI broth with or without inducer (100 µg of L-tryptophan per ml or 10 µg of DL-1-methyltryptophan per ml) and were shaken at 37°C until they reached the mid-exponential phase. Tryptophanase assays were performed with extracts of sonically disrupted cells and measured the conversion of S-o-nitrophenylcysteine (a gift from Robert Phillips, University of Georgia) to o-nitrothiophenolate as the A₄₇₀. All assays were performed in triplicate. Errors in mean specific activities in individual cultures were estimated from standard deviations of enzyme assays and of protein determinations by use of standard methods of error propagation. One unit of tryptophanase activity is defined as 1 µmol of o-nitrothiophenolate produced per min under standard assay conditions (13).

DNA and protein determinations. DNA concentrations were determined by A₂₆₀ measurements or by fluorometry with a Hoefer TKO fluorometer and Hoechst 33258 (29). Protein was determined by dye binding with Bio-Rad reagent and bovine serum albumin standards (7).

Cloning and sequencing of H. influenzae inserts. Long PCR to amplify a 3.6-kb fragment between nlpD and mutS was performed with the Expand Long Template PCR kit (Boehringer Mannheim Biochemicals) and primers nlpD-F (5'-GAAGTCAAAGCAGGTCAAGACATCGC, corresponding to nucleotides [nt] 750312 to 750337 in Rd) (12) and mutS-R (5'-GGGCTGTCCTGCAGATTGACCTCG, corresponding to nt 750760 to 750741 in Rd). The PCR fragment was purified with the Wizard PCR cleanup kit (Promega), digested with BamHI and PstI, and ligated into a Bluescript vector (pSK; Stratagene) by standard methods (38). The resulting plasmid (pE6) slowed the growth of its E. coli host; spontaneous insertions of IS1 and IS5 into pE6 arose during propagation and relieved growth inhibition. Subclones were prepared directly from the 3.6-kb PCR product by digestion with Sau3A or AluI restriction endonucleases and ligation of the fragments into the BamHI or SmaI sites of the vector. Both strands of the insert were sequenced with phage promoter primers from the vector and oligonucleotide primers for the internal sequence. Sequences near the junctions with nlpD and mutS were confirmed by direct sequencing of PCR fragments following gel purification with a Qiaquick kit (Qiagen). Sequencing reactions were performed with dye terminator chemistry (ABI Prism FS) and were run on an ABI 377 sequencer. Sequences were analyzed with BLASTX and BLASTN (2) and with DNAStar software.

Southern blots. Genomic DNA (0.5 to 1 µg/lane) was isolated (5), digested to completion with PstI or BamHI, electrophoresed in 1% agarose, and transferred to nylon membranes (Nytran; Schleicher & Schuell, Inc.) by capillary action (38). A ³²P-labelled probe was prepared by random oligonucleotide priming of fragments isolated by electrophoresis in low-melting-point agarose (FMC Corp.). Hybridization was done at 42°C in buffer containing 50% formamide, 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 50 mM sodium phosphate (pH 6.5), 1× Denhardt's solution, 100 µg of herring sperm DNA per ml, and 1% sodium dodecyl sulfate (SDS); filters were washed twice at 42°C with 2× SSC-0.1% SDS and twice at 56°C with 0.1× SSC-0.1% SDS. Filters were exposed at 80°C with Kodak XAR-5 film and intensifying screens.

PCR analysis of the nlpD-mutS region. Genomic DNA (60 to 120 ng) was amplified with the nlpD-F and mutS-R primers; for some strains, the nlpD-F primer was used in combination with the mutS-R2 primer (5'-GTGGCGGCTATTTTAGGTGCCG, corresponding to nt 759531 to 759510 in Rd) (12).

Nucleotide sequence accession number. The Hib (Eagan) tna insert has been assigned GenBank accession no. AF003252.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Organization of genes in the nlpD region. The tna cluster of Hib was identified in a region corresponding to nt ~740000 on the Rd genome. The organization of this region resembles that at 61 min on the E. coli genetic map, a conserved region containing a cluster of genes (pcm, surE, nlpD, and rpoS) involved in stationary-phase survival and oxidative-stress protection. In E. coli, this cluster is closely linked to mutS, and rpoS is immediately downstream from nlpD (19). In contrast, H. influenzae Rd has no rpoS homolog, and mutS is contiguous to and oriented in the same sense as nlpD (12). In order to investigate this region in Hib, PCR primers from a coding sequence near the 3' end of nlpD and the 5' end of mutS were used to amplify the intervening region (Fig. 1). Long PCR amplified the expected 0.45-kb product with an Rd template (R906) but a 3.6-kb fragment with the Hib (Eagan) template. A gene cluster homologous to the tna operon of E. coli was found on the Hib fragment, immediately downstream from nlpD, a position occupied by rpoS in enteric bacteria. A BamHI/PstI fragment containing the Hib insert region between nlpD and mutS was cloned into a Bluescript vector for sequencing.

View larger version (14K): [in this window] [in a new window]   FIG. 1.   Map of tna insert region showing positions on Rd genome at which tna DNA is inserted and of unique restriction sites. The scale of the insert (Hib) map is 50% that of Rd. The primers used to amplify the insert region are depicted with solid arrows; F, nlpD-F; R, mutS-R. Open arrows depict paired USSs. The position of a SacII/XbaI fragment (bp 705 to 2680) used as a tna-specific probe is indicated.

View larger version (61K): [in this window] [in a new window]   View larger version (21K): [in this window] [in a new window]   FIG. 2.   Nucleotide sequence of the tna insert region in Hib (Eagan), with open reading frames and deduced amino acid sequences. (a) Sequence of a 3,456-bp BamHI/PstI fragment with its 5' end in nlpD and its 3' end in mutS. All genes in this region are transcribed to the right. Underlined sequences (bp 1 to 84 and 3192 to 3456) are aligned with the matching sequence in Rd (12) (bp 1 to 84 and 85 to 350 on the corresponding BamHI/PstI fragment). Base sequence differences between Rd and Hib (Eagan) are indicated by nucleotides above the Eagan genetic sequence, which refer to the Rd sequence. They include (numbering Rd positions from the BamHI site) CT substitutions at Rd positions 65 and 191, a TC substitution at position 73, GA substitutions at positions 89, 118, 307, and 328, an AG substitution at 187, and deletion of a T (indicated as ^T) between 92 and 96. The paired USS repeats at either end of the insert are indicated by arrows. Putative regulatory elements (consensus CAP binding and promoter sites) are indicated with dotted lines, and potential ribosome binding sites (Shine-Dalgarno) are in boldface. The coding region for NlpD ends at 31; the open reading frame for putative TnaC extends from 232 to 312, that for TnaA extends from 400 to 1818, and that for TnaB extends from 1895 to 3112. The initiating ATG for MutS is at 3270. The locations of a conserved tryptophan residue in TnaC and of a conserved lysine residue at position 270 in TnaA are marked with asterisks; in E. coli tryptophanase, this lysine forms a Schiff base with the coenzyme pyridoxal 5'-phosphate. , insertion sites for IS1 (between 112 and 113) and IS5 (between 254 and 255) in subclones of pE6 which relieve growth inhibition in E. coli (see Materials and Methods). (b) Potential RNA stem-loop structures of the 43-bp repeat sequences corresponding to USS plus-minus pairs. G values were 18.2 kcal/mol for the Rd structure and 12.3 kcal/mol for each of the Hib structures.

The tna cluster has been inserted within a USS. H. influenzae exchanges genes by natural transformation, preferentially taking up DNA fragments containing a 29-bp consensus sequence termed the uptake signal sequence (USS). The genome of H. influenzae Rd contains 1,465 USSs, 5'-aAAGTGCGGT.rwwwww...rwwwww (plus orientation) or its complement, in which uppercase letters denote a conserved 9-bp core, lowercase letters denote the consensus, r is purine, w is A or T, and a dot is any base (42). Some USS sites are paired; most of these are inverted repeats (plus-minus or minus-plus pairs) and can form a stem-loop structure in mRNA. Paired USS sites typically occur in intragenic regions at the 3' ends of genes (42). One such paired USS site, in the plus-minus configuration (5'-aaAAGTGCGGTaaaaaatctcaacaaatttttACCGCACTTtt), occurs 11 bp downstream from nlpD in Rd (12) (Fig. 2a, positions 42 to 84). The conserved cores are separated by 21 bp, an arrangement which preserves most of the 29-bp consensus sequence for each USS; the two consensus sequences overlap in the middle (42). The tna insertion has occurred within this 43-bp stem-loop sequence. Nucleotides 44 to 84 and two additional 3'-terminal T's have been duplicated at the 3' end of the tna insert, which is accordingly flanked by 43-bp direct repeats of the USS. The second USS copy is located at the 3' end of the tna cluster, 37 bp from the tnaB stop codon. Each USS can assume a stem-loop conformation (Fig. 2b).

The tnaA gene of Hib is surprisingly similar to E. coli tnaA. The first long open reading frame in the insert encodes a predicted 472-amino-acid polypeptide with a molecular weight of 53,094 (pI, 6.4). This gene is strikingly similar to E. coli tnaA, encoding tryptophanase: 74% identical in the DNA sequence and 86% identical in the inferred amino acid sequence. In fact, it is much more similar to the E. coli homolog than the latter is to homologs from enteric bacteria that are close relatives of E. coli (Fig. 3a and b); for instance, P. vulgaris tryptophanase is only 50.3% identical to E. coli tryptophanase in amino acid sequence.

View larger version (84K): [in this window] [in a new window]   FIG. 3.   Amino acid sequence comparisons of the inferred TnaA polypeptide from Hib (Eagan) with tryptophanases from E. coli (8, 15), P. vulgaris (21), E. aerogenes (24), S. thermophilum (gene 1) (18), and the tyrosine phenol-lyase from Citrobacter freundii (3). (a) Alignment by the Clustal method; filled residues are identical to those of E. coli TnaA. (b) Phenogram of TnaA amino acid sequences.

tnaB gene of Hib. The second long open reading frame encodes an inferred 405-amino-acid polypeptide with a molecular weight of 43,852 (pI, 8.6). This polypeptide is homologous to low-affinity Trp permeases of other gram-negative bacteria, including E. coli TnaB. It is also homologous to the Mtr family of aromatic amino acid permeases, including the product of the mtr homolog at nt 320544 to 321797 in H. influenzae Rd (12, 39). Unlike the tnaA homolog, the tnaB homolog of H. influenzae is phylogenetically more distant from related genes in enteric bacteria than they are from each other (the tnaB product is 34.5% identical in amino acid sequence to E. coli TnaB) (Fig. 4a and b).

View larger version (92K): [in this window] [in a new window]   FIG. 4.   Amino acid sequence comparisons of the inferred TnaB homolog from Hib (Eagan) with TnaB from E. coli (8, 15) and P. vulgaris (21) and Mtr from E. coli (39) and H. influenzae (12). (a) Alignment by the Clustal method; filled residues match the consensus. (b) Phenogram of TnaB and Mtr amino acid sequences.

tnaC homolog of H. influenzae. E. coli TnaC is a 24-amino-acid leader peptide with a single Trp residue at position 12; the Trp residue is essential for the regulation of structural genes by antitermination (14, 15). Similarly, the putative tnaC gene of Hib (Eagan) encodes a 26-amino-acid peptide with a single Trp residue at position 11. The inferred TnaC peptide is highly hydrophobic (12 of 26 residues), as are other TnaC peptides. A comparison of TnaC peptides from E. coli, P. vulgaris, E. aerogenes, and Hib (Eagan) shows a conservation of amino acids surrounding the crucial Trp (Fig. 5). The inferred sequence from H. influenzae is phylogenetically more distant from those of the enteric bacteria than their sequences are from each other, particularly in the amino acids surrounding Trp. Unlike those of E. coli and P. vulgaris, H. influenzae tnaC does not have a boxA sequence near the terminal codon; however, the 87-bp intragenic region immediately downstream of tnaC is relatively rich in pyrimidines and poor in G's and may be a Rho-dependent termination site like that involved in antitermination regulation of the E. coli tna operon (1).

View larger version (11K): [in this window] [in a new window]   FIG. 5.   Amino acid sequence alignment (Clustal) of the putative TnaC peptide from Hib (Eagan) with TnaC peptides from E. coli (15), P. vulgaris (21), and E. aerogenes (24). Filled residues match the consensus.

Ancestry of the tna gene cluster. The location of the tna cluster between direct repeats of a USS suggests acquisition by lateral transfer from another organism. Assuming that the flanking USS repeats were identical at the time of acquisition and assuming a universal silent substitution rate of 1% per million years (35), the 2.3% divergence of the USSs suggests a recent acquisition time, probably within the last 5 million years. By way of comparison, the silent substitution rate between Hib (Eagan) and Rd in the neighboring nlpD-mutS region analyzed was 9 of 201 positions, or 4.5%.

Tryptophanase activity in H. influenzae. To establish that indole-positive H. influenzae has tryptophanase activity, we assayed extracts of Eagan, INT1 (an indole-positive, nonencapsulated strain), and Rd by using a standard assay for E. coli tryptophanase (14). H. influenzae is a fastidious organism which grows poorly in defined media; thus, initial assays were done with cultures grown in a complex medium (sBHI) with unknown levels of tryptophan and glucose. As a comparison, E. coli AB1450 (wild type for tryptophanase) was grown in sBHI as well as in minimal medium (Table 3). Extracts from Eagan cultures grown without added tryptophan or glucose had a tryptophanase specific activity higher than that of E. coli in the same medium. This activity was not inducible by 1-methyltryptophan (Table 3) or 100 µg of tryptophan per ml (data not shown). The addition of glucose depressed specific activity less than threefold; a similar modest effect was seen with E. coli grown in the same complex medium. INT1 specific activities (data not shown) were similar to those of Eagan, whereas extracts from the indole-negative Rd strain had no detectable tryptophanase activity. We conclude that the indole-positive strains Eagan and INT1 have tryptophanase activity, whereas the indole-negative Rd strain has no detectable tryptophanase activity. The lack of induction by tryptophan and the weak repression by glucose are likely explainable by the presence of tryptophan and glucose in the complex medium. An investigation of the regulation of tryptophanase in H. influenzae will require the formulation of appropriate defined growth media.

Phylogenetic distribution of the tna gene cluster in H. influenzae. The distribution of H. influenzae tna genes might be expected to coincide with the indole-positive phenotype. We isolated genomic DNAs from a diverse collection of H. influenzae strains, including reference strains of each of the six serotypes, several recent isolates of nonencapsulated strains, and other Haemophilus species (Table 1), and digested them with BamHI and PstI, neither of which cut within the tna cluster from Eagan. Insertion of the tna cluster within Rd DNA yields a predicted 5,785-bp BamHI fragment and a predicted 3,676-bp PstI fragment. Figure 6 is a blot of genomic DNAs digested with BamHI and probed with a sequence internal to the tna cluster (depicted in Fig. 1). Of the 15 strains of H. influenzae tested, all indole-positive strains yielded a band on Southern analysis with BamHI (Fig. 6) and PstI (data not shown), and all indole-negative strains were negative on DNA blots. Hence, the indole test is a reliable indicator of the presence of tna genes, and the DNA sequences identified by Southern analysis are likely to represent active genes. Genomes positive for tna sequences included those from reference strains of serotypes a, b, c, and f as well as R1965, the type strain of H. influenzae, which is nonencapsulated and classified in biotype II (25). Reference strains of serotypes d and e and strain Rd were negative on Southern analysis. Other strains that tested positive for tna DNA included Eagan, INT1 (a nonencapsulated strain from the blood of a meningitis patient), U11, C2859, and C2861 (nonencapsulated isolates from cerebrospinal fluid [CSF]), R3001 (a nonencapsulated strain isolated from the lower respiratory tract of a cystic fibrosis patient), and C2853 (isolated from the sputum of a cystic fibrosis patient). For every indole-positive strain, single BamHI and PstI fragments were identified, indicating tna integration at a unique site. Although restriction fragment length polymorphisms were found (for BamHI in the reference type f strain, R1965^T, and C2859 and for PstI in U11, R3001, and R1965^T), every strain except for R1965^T yielded either the predicted BamHI fragment or the predicted PstI fragment or both, indicating that the tna genes were located at the same site in each indole-positive strain. R1965^T displayed restriction fragment length polymorphisms with both enzymes; however, PCR analysis confirmed tna insertion at the unique site (see below).

View larger version (72K): [in this window] [in a new window]   FIG. 6.   Southern blot of genomic DNAs with a tna-specific probe. Genomic DNAs from various H. influenzae strains and from H. haemoglobinophilus (H. hb) were digested with BamHI and probed with a SacII/XbaI fragment (Fig. 1) internal to the Hib (Eagan) tna cluster, and blots were washed under stringent conditions. Strain designations are indicated above lanes; a to f indicate reference strains of serotypes a to f, respectively. The predicted 5.8-kb fragment from Hib (Eagan) is indicated with an arrow.

View larger version (76K): [in this window] [in a new window]   FIG. 7.   PCR amplification of genomic DNA between nlpD and mutS. (a) Primers nlpD-F and mutS-R were used to amplify the intervening sequences in genomic DNA from various H. influenzae strains by long PCR, and the products were resolved by agarose gel electrophoresis. Arrows indicate the positions of the 3.6-kb amplification product from Hib (Eagan) and the 0.45-bp product from Rd. Lane designations are as in Fig. 6; M,  DNA HindIII markers. (b) Southern blot of the same gel with a tna-specific probe.

Other Haemophilus species. Haemophilus species other than H. influenzae are indole negative, with two exceptions: H. haemolyticus (a human parasite) and H. haemoglobinophilus (isolated from dogs and occasionally from humans) (25). We tested H. haemoglobinophilus genomic DNA by Southern analysis and PCR with external primers. Under stringent hybridization conditions, the H. influenzae tna probe did not detect a target in H. haemoglobinophilus genomic DNA, even with long exposure times (Fig. 6). Long PCR with H. haemoglobinophilus under a variety of conditions yielded a single, ~3.2-kb fragment, significantly shorter than the H. influenzae products (Fig. 7). The H. haemoglobinophilus PCR fragment did not hybridize with the internal tna-specific probe, indicating that this DNA is unrelated to tna. Thus, if H. haemoglobinophilus has tna genes, they differ significantly in sequence and map location from those of H. influenzae.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In classical microbiology, the indole test provides a means of differentiating among H. influenzae strains, with the majority of isolates from invasive disease testing positive (25). Genes found in pathogenic bacteria but not in their nonpathogenic relatives are often virulence determinants (16). As a first step toward testing the hypothesis that tryptophanase is a virulence determinant in H. influenzae, we identified the genes responsible for tryptophanase activity in indole-positive strains of H. influenzae. Their organization and the inferred amino acid sequences of the encoded proteins are similar to those of the tna operon of E. coli. Evolutionary conservation of the regulatory tnaC peptide and the presence of putative CAP binding and promoter DNA elements at positions similar to those in the E. coli and P. vulgaris operons suggest that the H. influenzae tna gene cluster is an operon; however, we have not shown this directly.

Virulence-associated genes are often clustered in pathogenicity islands (PAIs), DNA sequences whose anomalous G+C content suggests lateral transfer and which are bordered by direct repeats of tRNA genes or insertion sequences (11, 17). In enteric bacteria, PAIs are transferred by phage or plasmids. The structure of the tna cluster resembles a PAI: the tna cluster has been inserted within a USS which, like tRNA genes, has a stem-loop secondary structure and exists in multiple copies dispersed throughout the genome. The tna insert is flanked by 43-bp direct repeats of the USS. Two known virulence determinants in H. influenzae also resemble PAIs. The fimbrial gene cluster of H. influenzae, found in serotype b strains but not in Rd, is flanked by 52-bp direct repeats (49), suggesting a transfer mechanism similar to that of tna. The capsular genes of H. influenzae are located between direct repeats of an insertion sequence and are known to be horizontally transferred among H. influenzae subpopulations (27).

Single USS sites facilitate genetic exchange by transformation in Haemophilus and related genera. The function of paired USSs is unknown, but they are frequently found at the 3' ends of transcription units and may act as transcriptional terminators or to stabilize mRNA (42). As is typical, the paired USS at the 3' end of nlpD in Rd and Hib (Eagan) occurs 11 bp after the translational stop codon. Duplication of this region has placed a second paired USS at the 3' end of the tna gene cluster, 37 bp after the termination codon of tnaB, an arrangement which may have regulatory significance for the presumed operon.

When and from where did H. influenzae acquire the tna cluster? Sequence comparisons suggest that acquisition occurred relatively recently, within the past 5 million years. Most Haemophilus species are indole negative, arguing for an extrageneric source. Possible donor species from which H. influenzae may have acquired the tna cluster are indole-positive members of the related genus Pasteurella (e.g., Pasteurella multocida), which have A+T-rich genomes.

The tna gene cluster is found at the same map location in a diverse collection of H. influenzae strains, representing several different serotypes (a, b, c, and f) and nonencapsulated isolates. There are several possible explanations for the phylogenetic distribution of tna. (i) During the divergence of H. influenzae strains, a unique integration event might have occurred in an ancestral clade; all indole-positive H. influenzae would then have descended from this relatively recent common ancestor. This phylogenetic clustering is not supported by available population genetics data (2, 31, 32). (ii) A more likely possibility is that the tna cluster was acquired by a progenitor of H. influenzae and is an ancestral trait of the species. Indole-negative strains would have lost tna secondarily, perhaps by recombination between the flanking direct repeats. This scenario is more consistent with population genetics data and with the high prevalence (>70%) of the trait among natural isolates. (iii) Following the acquisition by one H. influenzae strain of the tna cluster, other, independently derived lineages might have acquired it by horizontal transfer and homologous recombination within outside markers. A similar scenario has been proposed to account for the emergence of serotype a strains having a virulence-associated bexA deletion previously found only among Hib strains (28). This explanation would imply strong selective pressure favoring tna insertion, especially in pathogenic strains. (iv) Finally, the tna cluster might have been repeatedly inserted at the same location in different strains by site-specific integration.

Intergeneric comparisons of inferred amino acid sequences show that, as expected, H. influenzae tnaB and tnaC are evolutionarily more distant from homologous genes in enteric bacteria than their enteric homologs are from each other. In contrast, the H. influenzae tnaA gene is surprisingly similar to that of E. coli, suggesting the possibility that E. coli acquired its tnaA gene by lateral transfer from the same source as H. influenzae, perhaps incorporating it within a preexisting operon. Such mosaic patterns of acquisition and exchange are common among enteric bacteria (47). For both organisms, genetic drift and characteristic mutation rates have established G+C ratios for tnaA conforming to the genomic and coding sequence averages. If lateral transfer from a common source occurred (e.g., from an A+T-rich organism to E. coli), the event must have been relatively ancient. Application of a method for estimating time elapsed since lateral transfer from base composition at neutral sites (4, 46) (details not shown) yields a rough minimum estimate of 80 million years, significantly preceding the acquisition of tnaA by H. influenzae. At the opposite extreme, high sequence homology between E. coli and H. influenzae tnaA genes relative to other enteric bacteria indicates that any lateral transfer into E. coli must have occurred more recently than the divergence of enteric bacteria, which occurred ~500 million years ago (36).

Is H. influenzae tryptophanase likely to play a role in virulence? Figure 8 summarizes data from the literature on strains isolated from healthy throat tissue and individuals with various illnesses. Among harmless isolates, 70 to 75% are indole positive, suggesting that tryptophanase activity is more common than not in H. influenzae. However, among isolates from diseases other than conjunctivitis, 94 to 100% are indole positive. The correlation between virulence and indole production extends beyond serotype or capsule formation; for instance, most strains associated with otitis media are nonencapsulated, and nearly all are indole positive. The correlation may merely represent linkage disequilibrium among clonally derived strains. Alternatively, it may be causal, indicating that tryptophanase is necessary but not sufficient for virulence. A number of virulence determinants in the genus Salmonella are genes encoding metabolic enzymes; their role in pathogenesis (and their selective advantage) is thought to be the scavenging of nutrients in unusual host microenvironments (16). Most H. influenzae disease requires the colonization of tissues beyond the upper respiratory tract and may require special nutritional adaptations. We are currently using gene disruption to test the hypothesis that H. influenzae tnaA is a virulence determinant.

View larger version (20K): [in this window] [in a new window]   FIG. 8.   Prevalence of indole-positive () strains among H. influenzae isolates from various sources. Data were compiled from Kilian (25; Tables 1 and 18; 175 strains, including 45 from normal oral and respiratory tract flora), from Kawakami et al. (23; 27 strains from normal throat flora), and from the present study (6 strains, described in Table 1). Numbers in parentheses indicate the total number of strains from each source. Respiratory infections included chronic bronchitis, sinusitis, pneumonia, and unspecified respiratory infections; severe disease included meningitis or CSF infection, blood infections, epiglottitis, and pneumonia. Open bars show indole-negative strains.

The location of the tna insert within the H. influenzae genome may also be significant for the evolution of virulence. The tna gene cluster has been inserted within the intergenic region between the 3' end of nlpD and the 5' end of mutS. Pathogenicity in strains of E. coli and Salmonella enterica is associated with insertions of novel DNA near mutS, a gene which encodes a methyl-directed mismatch repair enzyme (30). Hypermutable subpopulations arise at a high frequency among such strains by means of deletions that extend from insert DNA into mutS. Defects in mismatch repair also facilitate the promiscuous uptake of laterally transferred DNA. LeClerc et al. (30) hypothesized that inserts near mutS may provoke a regulated "mutator phenotype": under adverse conditions, mutS expression is downregulated and bacteria undergo a burst of mutagenesis or acquire new genes. In Salmonella, a 40-kb PAI occurs just 5' to mutS, whereas in E. coli, a rearrangement occurs at the 3' end of mutS, potentially placing mutS under antisense control by rpoS. The tna cluster of H. influenzae is oriented in the same sense as mutS, with the initiating ATG codon of mutS 78 bp downstream from the 3' border of the tna insert. It will be interesting to see whether the presence of the tna island influences mutS expression or bacterial mutability.

Genomes of pathogenic bacteria are often larger than those of their nonpathogenic conspecific bacteria, having evolved through the acquisition of PAIs by lateral transfer. The genome of Hib (Eagan) is larger than that of Rd (9) and contains three known inserts (the fimbrial operon [49]), tna, and the cap locus [27]), each flanked by long direct repeats. The presence of the tna cluster as well as the other two islands in virulent strains of H. influenzae suggests that there may be divergent evolutionary trends within H. influenzae, one toward strains with small genomes, which are unlikely to evolve virulence, and one toward strains with larger genomes, scattered loops of inserted DNA, and virulence potential.