Bordetella holmesii: initial genomic analysis of an emerging opportunist

Abstract

Bordetella holmesii is an emerging opportunistic pathogen that causes respiratory disease in healthy individuals and invasive infections among patients lacking splenic function. We used 16S rRNA analysis to confirm B. holmesii as the cause of bacteremia in a child with sickle cell disease. Semiconductor-based draft genome sequencing provided insight into B. holmesii phylogeny and potential virulence mechanisms and also identified a toluene-4-monoxygenase locus unique among bordetellae.

Bordetella holmesii is a brown pigment-producing Gram-negative coccobacillus that is increasingly recognized as a cause of invasive disease in immunocompromised patients (Weyant, et al. 1995, Shepard, et al., 2004). Reports have linked B. holmesii to bacteremia, meningitis, pneumonia, pericarditis, endocarditis, and other clinical syndromes including pertussis-like illnesses (Mazengia, et al., 2000, Shepard, et al., 2004, Mooi, et al., 2012). Though most cases are sporadic, a cluster of B. holmesii bacteremia among children was reported in 2010 (Layton & Weiss, 2010). Definitive identification of B. holmesii can be difficult using routine phenotypic methods, and sequence-based approaches are frequently required (Russell, et al., 2001). Since only limited genetic information is available to help understand the unusual predisposition of B. holmesii to cause invasive disease in asplenic hosts including those with sickle cell disease (McCavit, et al., 2008, Gross, et al., 2010), we constructed a draft whole genome sequence of a B. holmesii clinical isolate and used this information to analyze putative virulence determinants.

A seven year-old girl with sickle cell (hemoglobin SS) disease presented with fever and tachycardia. She had been receiving routine oral penicillin prophylaxis. Physical examination did not reveal a focal infection. Laboratory studies revealed a peripheral leukocytosis with a neutrophil predominance and an elevated platelet count. Blood samples were taken for culture, and empiric therapy was initiated with intravenous ceftriaxone. Fevers continued during the first day of hospitalization and then resolved. Two blood cultures were positive for a slow-growing (> 48 hours) Gram-negative microorganism on chocolate and sheep blood agar with poor growth on MacConkey agar. The isolate was oxidase-negative and produced a diffusible brown pigment. Automated identification systems failed to conclusively identify the pathogen due to its asaccharolytic nature. Based upon these data, B. holmesii was suspected. Antimicrobial susceptibility testing by E-test revealed minimum-inhibitory concentrations (in μg/ml) as follows: penicillin, > 32; cefoxitin, >256; ceftazidime, 1; levofloxacin, 0.032; erythromycin, 0.125; azithromycin, 0.047; trimethoprim-sulfamethoxazole, 0.125. Using 16S rRNA sequencing as described (Schuetz, et al., 2012), we confirmed the isolate as B. holmesii (1525-bp, 100% match to B. holmesii type strain ATCC51541^T). Subsequent blood cultures were negative, as was a cerebrospinal fluid culture. Echocardiography revealed neither vegetations nor valvular abnormalities. The patient improved and was discharged on oral levofloxacin.

In order to better understand the biology of this unusual organism, we undertook a sequencing and assembly project to generate a draft genome of the strain from this case (B. holmesii 44057). Bacterial genomic DNA was purified using a commercial kit (DNeasy, Qiagen) and quantified by spectrophotometry. Library preparation and sequencing using the Ion Torrent Personal Genome Machine 316 chip were performed by the Columbia Genome Center using manufacturer-specified protocols. A single run generated 2.8 × 10⁶ reads (mean length 118 bp; 3.3 × 10⁸ total bases). Sequence assembly was performed with CLC Genomics Workbench (version 5.0.1), generating 299 total contigs (>250 bp each.) The total size of the contigs was 3.43 Mbp, which may be less than the actual genome size, as the assembler was unable to match all bases into contigs >250 bp (95% of reads, 94% of bases assembled.) The GC% for all contigs was 62.9%, consistent with the value of 61.9% reported for the B. holmesii type strain (Weyant, et al., 1995).

After removal of a single contig with greater than 5% ambiguous bases, the draft B. holmesii genome was annotated (3742 predicted coding sequences, 44 RNAs; see Supplemental Information associated with this manuscript) and compared to known Bordetella sequences using the Rapid Annotation using Subsystem Technology (RAST) server (Aziz, et al., 2008). We focused our initial analysis on known Bordetella virulence factors and gene clusters that might be unique to B. holmesii with the intention of performing more detailed genomic analysis as more complete and widely representative sequence data become available.

Early phylogenies based on 16S rDNA indicated that B. holmesii is most closely related to B. pertussis (Gerlach, et al., 2001). However, subsequent analysis based on multiple housekeeping genes demonstrated that B. holmesii is a closer relative of B. avium and that it had likely acquired both its 16S rRNA and iron uptake island by lateral transfer from B. pertussis (Diavatopoulos, et al., 2006). In order to help resolve these discrepancies, we employed the program INSID, which utilizes raw, unprocessed sequencing reads to create a whole genome phenogram. The resulting tree is congruent with taxonomies that place B. holmesii close to B. avium. (Fig. 1) To survey the similarity of B. holmesii to other species on a gene-by-gene basis we used a BLAST-based approach to assign a best hit for each open reading frame (ORF) in the draft genome. At the nucleotide level 42% (1576/3742) of genes had a best hit in B. avium while only 11% (408/3742) had a best hit in B. pertussis. Amino acid comparisons showed even stronger similarity with 61% B. avium best hits and 4% B. pertussis best hits. These data are consistent with a closer overall relationship between B. avium and B. holmesii.

Figure 1.

Phenogram of representative whole genomes from the genus Bordetella. Whole genome similarity between strains was calculated using the program INSID. Publically available whole genomes were shredded into 200,000 segments consisting of 400 base pairs each, and compared to the unprocessed sequencing reads generated by the Ion Torrent platform for B. holmesii. The pairwise similarity matrix was based on comparisons using a 15 megabase (Mb) query and 5 Mb reference. For bootstrap values, each set of genomic fragments/reads was resampled 100 times with replacement, generating 100 new read sets and 100 new similarity matrices. All trees were generated using the UPGMA algorithm in PAUP. Note that the phenogram topology is remarkably robust to bootstrap resampling. Branch length is proportional to average pairwise similarity as calculated by INSID.

Many potential Bordetella virulence factors have been previously described, some of which are present in the non-classical bordetellae (i.e. not B. pertussis, B. parapertussis, or B. bronchiseptica) (Gross, et al., 2010). Consistent with prior findings (Gerlach, et al., 2004, Horvat & Gross, 2009), the draft genome of B. holmesii 44057 encodes open reading frames with significant similarity to the bvg virulence regulatory system (RAST 35814.9.peg.2521–2524). Likewise, the alcaligin siderophore biosynthesis operon, part of the laterally transferred iron uptake island (Diavatopoulos, et al., 2006), was present in this strain (35814.9.peg.84–96). B. holmesii filamentous hemaglutinin, encoded by fhaB, has been described previously (Link, et al., 2007). Following manual curation, the B. holmesii 44057 fhaB (35814.9.peg.2499) predicted amino acid sequence was found to be 99% identical to the published sequence. Numerous components of flagellar biosynthesis, types II and III secretion systems, and capsular polysaccharide synthesis and export were identified, although the existence of an intact locus for each of these systems has not yet been confirmed. No significant sequence similarity with pertussis toxin subunits or with dermonecrotic toxin from B. pertussis was found within the B. holmesii 44057 draft genome. A partial sequence (35814.9.peg.2138) has short areas of sequence similarity with B. pertussis pertactin but on directed database query appears to be more closely related to an autotransporter from outside Bordetella.

B. holmesii 44057 contains a region encoding a predicted toluene-4-monoxygenase (T4MO) system without detectable similarity to other sequenced Bordetella genes (35814.9.peg.146–150). By BLAST query of the non-redundant NCBI databases, the locus has significant sequence similarity to the T4MO systems from Gordonia polyisoprenivorans and Pseudonocardia dioxanivorans (Fig. 2) and was likely acquired by horizontal transfer. In order to determine whether this locus is present in another B. holmesii strain, we performed screening by PCR using primers targeting the 1439 bp tmoA gene (primer sequences: TmoA-F: GTGTACCCGGACTGGCTAACAGGC; TmoA-R: TCACGCTCCTGTCTGGATGATGCC). We detected a band corresponding to the predicted size of tmoA in B. holmesii 44057 and the B. holmesii type strain (ATCC 51541^T), but not in B. pertussis ATCC 9340 (Fig. 3). Neither the specific functional properties of the T4MO locus nor how widespread it is among B. holmesii isolates are known. Other members of the T4MO family may participate in generation of pigments from indole precursors (McClay, et al., 2005), and we hypothesize that the T4MO system may be involved in pigment production by B. holmesii.

Figure 2.

A novel toluene-4-monoxygenase locus in Bordetella holmesii 44057 compared to the homologous region from Gordonia polyisoprenovorans VH2. The area of significant nucleotide identity between the two loci is indicated by the bracketed region. There was no detectable nucleotide similarity in the flanking regions. Homologous genes are indicated using the same color, and percentages above gene arrows represent values for amino acid similarity between homologs. Note that there is low-level similarity at the amino acid level and in gene annotation in some flanking genes, suggesting insertion in a similar genomic location after the putative horizontal gene transfer event. Pseudonocardia dioxanivorans CB119 (not pictured) has at least 3 separate loci with strong nucleotide similarity to this locus. The largest of these regions has 65% percent nucleotide identity but does not span the entire locus.

Figure 3.

PCR-based detection of the tmoA gene from the toluene-4-monoxygenase locus. DNA from B. holmesii strain 44057, the ATCC type strain (51541), and B. pertussis (ATCC strain 9340) was amplified as described in the text. Detection of a 1.4 kb band was consistent with the presence of the tmoA gene. MW: molecular weight; L: 1 kb ladder (NEB).

B. holmesii causes bacteremia and other invasive diseases in humans with impaired splenic function, including the child with sickle cell disease presented here. To better understand the biology of B. holmesii, we generated a draft genome using a single run of semiconductor-based sequencing. Obtaining a finished, closed genome sequence would have allowed more detailed analysis of genomic rearrangements and architecture. However, this economical approach allowed us to evaluate the phylogenetic position of B. holmesii using an unbiased technique, improve our understanding of its complement of virulence factors, and to generate hypotheses regarding its unique characteristics. Sequences from this investigation will aid the development of new strategies to distinguish among Bordetella species. Determination of complete genome sequences of several clinical isolates of B. holmesii is underway and will provide additional data regarding diversity within this species and its relationship to other members of the bordetellae (Preston, et al., 2004).

This Whole Genome Shotgun project has been deposited at DDBJ/EMBL/GenBank under the accession ANGE00000000. The version described in this paper is the first version, ANGE01000000. Raw sequence reads are publicly available through the NCBI Sequence Read Archive (BioProject ID: 158371; accession number SRX201839), and the B. holmesii 44057 RAST annotation is included as supplemental data with this manuscript.