The circular mt genomes of Ancylostoma caninum and Bunostomum phlebotomum, sequenced from single adult worms, were 13717 and 13790 bp in size, respectively (Figure 1). Each genome contained 36 genes, including 12 protein coding genes (adenosine triphosphatase subunit 6 [atp6], the cytochrome c oxidase subunits [cox1-3], cytochrome b [cytb], and the nicotinamide dehydrogenase subunits [nad1-6 and nad4l]), 22 transfer RNA (tRNA) genes and 2 ribosomal RNA genes (small [rrnS] and large [rrnL] subunits), and was consistent with gene arrangement 2 (GA2) 49. This arrangement is characteristic for the mt genomes of all members of the Strongylida and Ascaridida, as well as the free-living nematode Caenorhabditis elegans (Rhabditida), but not for Strongyloides stercoralis (Rhabditida) 414243444546474849. In accordance with other species of Strongylida for which complete mt genome sequences are available 4142, the AT-rich regions for both Ancylostoma caninum and Bunostomum phlebotomum were located between the genes nad5 and nad6, flanked at the 5'-end by the tRNA gene for alanine, and at the 3'-end by the tRNA genes for proline and alanine.

Each protein-coding gene for each of the two species had an open reading frame (ORF), and all genes were located on the same strand and transcribed in the same direction (5' to 3'), consistent with the known mt genomes of secernentean nematodes 37. The nucleotide usages (coding strand) of A, C, G and T in each mt genome were 29.0%, 6.5%, 16.1% and 48.5%, respectively, for Ancylostoma caninum (Table 1) and 26.9%, 6.2%, 16.7% and 50.1%, respectively, for Bunostomum phlebotomum (Table 1), with overall A+T contents of 77.5% and 77.0%, respectively. The A+T content of protein coding genes ranged from 70.9% (cox1) to 81.3% (nad6) for Ancylostoma caninum, and from 70.4% (cox1) to 82.6% (nad3) for Bunostomum phlebotomum. The A+T content for rrnS, rrnL (= ribosomal RNA genes) and the AT-rich region were 78.1%, 80.9% and 90.1%, respectively, for Ancylostoma caninum, and 75.2%, 82.4% and 88.0%, respectively, for Bunostomum phlebotomum. For the mt genome of Ancylostoma caninum, codon usage in individual protein coding genes (n = 12) ranged from 0% for CGC (arginine) and CCC (proline) to 15.7% for TTT (phenylalanine). For the mt genome of Bunostomum phlebotomum, codon usage ranged from 0% for CGC (arginine), CAC (histidine), CTC (leucine), CCC (proline), TCC (serine) and GTC (valine) to 15.0% for TTT (phenylalanine). For both species, individual tRNA structures were consistent with those predicted previously for hookworms and other secernentean nematodes 3742455051. All tRNA genes, except trnS(AGN) and trnS(UCN), had a predicted secondary structure containing a TV-replacement loop instead of the TψC arm and loop (not shown). The predicted secondary structure of each of the two serine tRNAs contained the TψC arm and loop but lacked the DHU loop. The genes rrnS and rrnL were 694 bp and 935 bp in length, respectively; the predicted secondary structures for the ribosomal RNA gene subunits for Ancylostoma caninum and Bunostomum phlebotomum (not shown) were similar to those of Necator americanus and Ancylostoma duodenale 42, which is also supported by the high nucleotide sequence similarity in the mt genes among these four hookworms (see Tables 2 and 3).

The AT-rich regions for Ancylostoma caninum and Bunostomum phlebotomum were 272 bp and 234 bp, respectively, and both exhibited complex secondary structures (not shown), as predicted previously for the AT-rich regions of nematodes 4142454749. Four AT-repeat regions of variable length were identified in the AT-rich region of the mt genome of Ancylostoma caninum: two were 6 nucleotides (nt) (3 AT-repeats), one was 14 nt (7 AT-repeats) and the longest was 16 nt (8 AT-repeats). Similar dinucleotide repeats have been described in the AT-rich region of the mt genomes of other nematode species (e.g., 414244). Other repetitive elements have been identified within this region in the free-living nematode Caenorhabditis elegans, the largest and most conspicuous of which are the repetitive sequence motifs CR1-CR6 45. However, no such elements were identified in the AT-rich region of the mt genome of either Ancylostoma caninum, Bunostomum phlebotomum or any other species of animal-parasitic nematode sequenced to date 41444749.

The identities (%) in inferred amino acid sequences of each protein-coding mt gene were calculated based upon pairwise comparisons between Ancylostoma caninum and Bunostomum phlebotomum (Tables 2 and 3). Based on these comparisons, the sequence identities (in decreasing order) were COX1 (97.3%), COX3 (96.0%), NAD4L (93.5%), COX2 (89.6%), CYTB (87.0%), ATP6 (85.9%), NAD1 (84.4%), NAD3 (81.9%), NAD4 (79.2%), NAD5 (78.5%), NAD2 (72.5%) and NAD6 (67.3%). In addition, the amino acid sequences inferred from each coding mt gene of Ancylostoma caninum and Bunostomum phlebotomum were compared, again in a pairwise manner, with those inferred from published mt genomes of Anisakis simplex 43, Ascaris suum 45 and Toxocara canis 44 (Ascaridida), Ancylostoma duodenale, Necator americanus 42 and Haemonchus contortus 41 (Strongylida), and Brugia malayi 46, Dirofilaria immitis 47 and Onchocerca volvulus 48 (Spirurida). The most conserved protein sequences among all species, assessed relative to Ancylostoma caninum and Bunostomum phlebotomum, were inferred to be COX1, COX3 and NAD4L, and the least conserved were NAD2 and NAD6 (see Tables 2 and 3).

Because of the high degree of intraspecific variation in nucleotide sequence in the mt genes of nematodes 373852 and the limited availability or lack of multiple mt genome sequences for each species, previous work has suggested that phylogenetic analyses for nematodes be conducted using concatenated amino acid sequence datasets, utilizing sequences inferred from individual mt protein coding genes 47. In order to further assess systematic relationships within and among members of the Ascaridida, Spirurida and Strongylida, a phylogenetic analysis was carried out using Bayesian inference (BI) and maximum likelihood (ML) (Figure 2). Almost all clades in the consensus tree were supported by maximum BI posterior probability (pp) values (pp = 1.00; expressed as a percentage in Figure 2) and/or ML bootstrap support (100). The phylogenetic analysis conducted herein clearly supports the distinct classification of the orders Ascaridida, Spirurida and Strongylida, each as monophyletic clades with maximum statistical support. The order Rhabditida appears to be paraphyletic, with Caenorhabditis elegans grouping closely with the Strongylida, and Steinernema carpocapsae and Strongyloides stercoralis placed externally to a clade comprising the Ascaridida, Strongylida and C. elegans. This relationship is consistent with the proposed molecular phylogeny for the Nematoda based on small subunit (18S) nuclear ribosomal DNA data 53. In addition, the hookworms were represented as a monophyletic clade within the Strongylida.

For hookworms, the phylogenetic analysis using BI indicated a closer relationship between Ancylostoma spp. and Necator americanus than between either of them and Bunostomum phlebotomum. This finding conflicts with the current classification of the Strongylida 16, wherein both Necator and Bunostomum are placed within the subfamily Bunostominae, whereas Ancylostoma is placed within the subfamily Ancylostominae (poorly supported by the ML analysis; bootstrap support = 47). A larger analysis, including mt data for more hookworm species, is needed to test further this hypothesis.

The present phylogenetic analysis did not support the grouping of the Ascaridida and Spirurida to the exclusion of the Strongylida, which contrasts markedly the results of a previous study based on nuclear ribosomal gene data (e.g., clade III versus clade V; ref. 53). The "common heritage" hypothesized herein for the Ascaridida and Strongylida to the exclusion of the Spirurida has been supported by previous studies using mt gene order data 49 and using concatenated amino acid sequence data inferred from protein-coding mt genes 38. These findings stimulate further study of the evolutionary relationships among taxa within this phylum using mt datasets. The high-throughput sequencing potential of 454 technology 54 and the recent validation of this technique for the sequencing of mt genomes 41 should provide a platform for an in-depth analysis of the phylogeny of the Nematoda.

Although some nuclear genetic regions (e.g., ITS-1 and ITS-2 of nuclear rDNA 22232425 or the cAMP-dependent protein kinase gene 26) have been shown to be suitable for the specific identification and differentiation of hookworms, the nuclear loci examined to date do not usually display sufficient levels of intraspecific sequence variability for the investigation of the genetic structures of hookworm populations (or the identification of population variants or "strains"). The ability to estimate genetic variability within and among hookworm populations is central to studying their epidemiology and population genetics, and can have important practical implications in relation to control.

Sequence-based analyses (including mutation scanning) of protein-coding mt genes, such as cox1 and nad1, have been particularly useful or population genetic studies 21272930315556575859. For example, Hu et al. 21 employed a single-strand conformation polymorphism (SSCP)-coupled sequencing approach to explore haplotypic variability within a limited number of Ancylostoma caninum specimens from Australia and each of the human hookworms (Ancylostoma duodenale and Necator americanus). Significant population sub-structuring was recorded within each of these three species, and two genetically distinct subpopulations were detected within Ancylostoma caninum from dogs from Townsville, Australia. Previous morphological and clinical studies had shown that Ancylostoma caninum in Townsville (Australia) is not specific to dogs and can also infect humans (but not complete its life-cycle), causing eosinophilic enteritis 19. It has been speculated 21 that particular, genetically distinct subpopulations within Ancylostoma caninum can selectively infect the non-canine host. The pattern of haplotypic variability within Ancylostoma caninum might be due to secondary contact between populations or subpopulations, which could have arisen due to host movement from other geographical areas where this hookworm has been recorded and where ecological conditions are distinct; for example, Ancylostoma caninum is endemic in tropical north-east Queensland, Australia 60, but also occurs in the north-west area of Western Australia 61. It is also possible that feral dogs or dingoes (in different geographical or climatic regions) might harbour one or more genetic variants which might "spill-over" into domestic dogs and/or humans 60. Future study of the genetic variation among Ancylostoma caninum specimens from domestic and feral dogs, cats and humans as well as between populations from other geographical and climatic regions in Australia and South-East Asia would allow such questions to be addressed. A comparison of the genetic make-up of Ancylostoma caninum from humans affected by eosinophilic enteritis with those from domestic dogs in the Townsville area would be particularly interesting.

In contrast to Ancylostoma caninum, no studies have yet explored the genetics or molecular epidemiology of Bunostomum phlebotomum. Mitochondrial markers might be used to examine sub-structuring in Bunostomum phlebotomum populations in endemic regions of South Africa. In addition, although there has been anecdotal evidence suggesting that Bunostomum phlebotomum may cause cutaneous larval migrans in humans (20 and unpublished observations [JVW]), the zoonotic potential of this species of hookworm has not yet been tested molecularly. In view of the lack of distinguishing morphological characters allowing the identification of individual larvae, the provision of molecular markers for Bunostomum phlebotomum might allow the extent of the zoonotic potential of this species to be assessed for the first time.

The two mt genomes characterized herein provide a solid foundation for studies of the epidemiology, ecology and population genetics of both Ancylostoma caninum and Bunostomum phlebotomum, which could have important implications for the control of infections by these parasites. Given the lack of morphological characters for specific identification and differentiation of hookworm larvae, there is a clear need for species and population genetic markers for in-depth exploration of the epidemiology of hookworms 59. Combined with the use of specific markers in the internal transcribed spacers (ITS-1 and ITS-2) of nuclear rDNA 232425, investigating the mt haplotypic variability in populations of Ancylostoma caninum and Bunostomum phlebotomum (irrespective of developmental stage) could provide important insights into host affiliations, gene flow and transmission patterns (cf. 6263) and thus assist in the control of these hookworms. Furthermore, the direct sequencing of the mt genome of Ancylostoma caninum by 454 technology is the second example of the use of this approach for the sequencing of mt genomes of nematodes 41 and re-enforces the exciting potential of emerging technologies for the high-throughput sequencing of relatively small organellar genomes.

An adult male of Ancylostoma caninum (designated Ac1) was collected (by IB) at necropsy from the duodenum of a dog from Townsville, Australia 23. An adult male of Bunostomum phlebotomum (Bp1) was collected at autopsy from the same site from a calf monospecifically infected with an isolate of Bunostomum phlebotomum, originally derived from a Jersey cow in Pretoria North suburb, South Africa (by JvW). Nematodes were washed in physiological saline, identified morphologically to species 16, fixed in 50% (v/v) ethanol and stored at -20°C until use. Total genomic DNA was isolated from individual worms using sodium dodecyl-sulphate/proteinase K treatment 64, followed by spin-column purification (Wizard Clean-Up, Promega). The specific identity of each nematode was verified using the sequence of the second internal transcribed spacer (ITS-2) of nuclear ribosomal DNA, which provides species-specific genetic markers for hookworms 25. The ITS-2 sequence derived from sample Ac1 was identical to that reported previously for Ancylostoma caninum (accession number AJ001591) 25 and that obtained from Bp1 (accession number FJ616999) was 82.3% identical to the closely related species Bunostomum trigonocephalum (accession number AJ001595) 25.

The complete mt genome of each Ancylostoma caninum and Bunostomum phlebotomum was amplified as two overlapping amplicons (~10 kb and ~5 kb, respectively) from ~20–40 ng of the genomic DNA from each specimen by long-PCR (BD Advantage 2, BD Biosciences) using each of the primer pairs 39F-42R and 5F-40R 394042, as described by Hu et al. 39, with minor modifications. The cycling conditions (2720 thermal cycler, Applied Biosystems) were: 92°C, 2 min (initial denaturation); then 92°C, 10 s (denaturation); 50°C, 30 s (annealing); 68°C (for the ~10 kb region) or 60°C (for the ~5 kb region), 10 min (extension) for 10 cycles, followed by 92°C, 10 s; 50°C, 30 s; 68°C or 60°C, 10 min for 20 cycles, with an elongation period of 10 s for each cycle, and a final extension at 68°C or 60°C for 7 min. Following the PCR, individual amplicons were resolved in ethidium bromide-stained agarose (1%) gels and shown to represent single bands. Amplicon size was estimated based on comparison with a 1 kb DNA size ladder (Promega). Amplicons of ~10 kb or ~5 kb were purified over a mini-column (Wizard PCR Preps, Promega). Subsequently, the amount of DNA in each purified amplicon was estimated spectrophotometrically (ND-1000 UV-VIS spectrophotometer, v.3.2.1, NanoDrop Technologies). The purified amplicons were then subjected to sequencing.

The mt genome of Ancylostoma caninum (designated AcMG-454; GenBank accession no. FJ483518) was sequenced (454 technology by AJ/RBG) using a Genome Sequencer 20 (Roche), according to an established protocol 54. The complete mt genome of Bunostomum phlebotomum (designated BpMG-PW; GenBank accession no. FJ483517) was sequenced (AW/DTL) by primer walking, as described previously 44. The AcMG-454 sequence was assembled automatically, whereas that of BpMG-PW was assembled manually using Sequencher v.4.8 (Gene Codes Corporation). Both mt genome sequences were annotated and subjected to analysis using standard approaches 4144, and their structures were compared with each other and with those of the two human hookworms, Ancylostoma duodenale (GenBank accession number AJ417718; ref. 42 and Necator americanus (AJ417719; ref. 42, Haemonchus contortus (EU346694; ref. 41) (Strongylida); Anisakis simplex (AY994157; ref. 43, Ascaris suum (X53453; ref. 45), and Toxocara canis (EU730761; ref. 44) (Ascaridida); Brugia malayi (AF538716; ref. 46), Dirofilaria immitis (AJ537512; ref. 47) and Onchocerca volvulus (AF015193; ref. 48) (Spirurida).

The analysis of amino acid sequence data was conducted via Bayesian inference (BI) using the software package MrBayes v.3.1.2 http://mrbayes.csit.fsu.edu/index.php and maximum likelihood (ML) using GARLI (65; http://www.bio.utexas.edu/faculty/antisense/garli/Garli.html), each running on a four dual-core Opteron-based Unix cluster. For individual species, the amino acid sequences inferred from all protein coding mt genes were concatenated. A selection of published mermithid mt genomes were used as outgroups (Strelkovimermis spiculatus, accession NC_008047; Agamermis sp., NC_008231; Thaumermis cosgrovei, NC_008046) and included a range of ingroup taxa (Xiphinema americanum, NC_005928; Brugia malayi, NC004298; Onchocerca volvulus, NC_001861; Dirofilaria immitis, NC_005305; Strongyloides stercoralis, NC_005143; Caenorhabditis elegans, NC_001328; Steinernema carpocapsae, NC_005941; Necator americanus, NC_003416; Ancylostoma duodenale, NC_003415; Cooperia oncophora, NC_004806; Haemonchus contortus, NC_010383; Anisakis simplex, NC_007934; Ascaris suum, NC_003127; Toxocara canis, NC_010690). Amino acid sequences were aligned using MUSCLE 66. Ambiguous sites were excluded using G-Blocks (6768; see Additional File 1 for alignment).

For BI of amino acid data, tree construction and posterior probabilities (pp) were calculated via 2000000 generations (ngen = 2000000) using the Metropolis-coupled Monte Carlo Markov Chain (MCMCMC) method and four simultaneous tree-building chains (nchains = 4), with every 10^th tree being saved (samplefreq = 10). A suitable burnin (burnin = 1000) was chosen using 'Trace' in the program Tracer v1.4 http://beast.bio.ed.ac.uk/. Evolutionary distance was estimated using the most appropriate amino acid model and calculated employing the MrBayes program (aamodelpr = mixed), allowing for a gamma-shaped variation in mutation rates with a proportion of invariable sites (rates = invgamma). Upon completion of the analysis, a 50% majority rule = consensus tree was constructed in TreeviewX v.0.5.0 http://darwin.zoology.gla.ac.uk/~rpage/treeviewx/. For the ML analysis using GARLI, tree construction was estimated with the model GTR+I+g using the mtRev amino acid substitution matrix, for two replicate runs, and termination criteria with setting genthresholdfortopoterm = 20000 (no new significantly better scoring topology found in > 20000 generations). Nodal support in the ML analysis was estimated by bootstrap re-sampling (n = 100) using GARLI and the same model settings.