Collection and DNA extraction

The collection of human patient samples was approved by the Human Research Ethics Committee of the Northern Territory Department of Health and Menzies School of Health Research (approval 13–2027) and informed consent was obtained in writing from each participant. Animal care and handling procedures used in this study followed the Animal Care and Protection Act, in compliance with the Australian code of practice for the care and use of animals for scientific purposes, outlined by the Australian National Health and Medical Research Council. The study was approved by the QASP and the QIMR Berghofer MRI Animal Ethics Committees (DEEDIAEC SA2012/02/381, QIMR A0306-621M).

Skin scrapings were collected from two unrelated patients with severe crusted scabies (patient A and patient B) from different regional areas of Northern Territory, Australia. The collections from individual patients were made 14 months apart and were likely independent. On both occasions, scabies mites (S. scabiei var. hominis) were individually picked from the skin. Each sample contained >1000 mites. Two samples of pig mites (S. scabiei var. suis) were collected from an inbred population of mites from a pig model [31]. The samples were taken from different pigs from consecutive cohorts (where infections are passed on to new piglets from a previous group) at different time points. The first sample consisted of >1000 whole adult mites (pig unwashed). The second sample also consisted of >1000 mites, but was split into three subsamples, which were washed using different protocols to reduce the amount of bacteria present on the surface of the mites due to the wound micro-environment (pig washed 1, 2 and 3 respectively). The protocol entailed: (1) 15 min wash at room temperature in 4% paraformaldehyde in water [32]; (2) 1 hour incubation at 37°C in 150 mM NaCl, 10 mM EDTA, pH8.0, 0.6% SDS, and 0.125 ug/ul lysozyme [adapted from 33] and; (3) 1 hour incubation at 37°C in 1% bleach (Sodium hypochlorite) in water. Mites were subsequently rinsed twice in water. Between wash steps mites were centrifuged at 10000 rpm for 2 min.

Whole genome sequencing

Whole mites were crushed and DNA was extracted from each sample using a Blood and Cell culture DNA Kit QIAGEN and a modified procedure, adapted from the manufacturer’s protocol. Washed mites were submerged in 1 ml of ice-cold lysis buffer (20 mM EDTA, 100 mM NaCL, 1% TritonX-100, 500 mM Guanidine-HCl, 10 mM Tris pH7.9) and homogenized with stainless steel beads of 2.8mm diameter at 6800rpm, 3 cycles, 30 sec per cycle, 30 sec between cycles. The suspension of lysed mites was supplemented with DNase free RNase A to 0.2 mg/ml and with Proteinase K to 0.8 mg/ml and incubated at 50°C for 1.5 h. After centrifugation at 4000g for 10 min to pellet insoluble debris the genomic DNA was isolated on the QIAGEN genomic tip as instructed in the manufacturer’s protocol. Six DNA libraries were constructed and 100 nucleotide (nt) long paired-end reads were generated on an Illumina HiSeq 2500. Additionally, 54 S. scabiei var. suis eggs were collected from the pig model. To reduce any surface bacteria, these were washed washed twice in 4% paraformaldyde and rinsed twice in water. DNA was extracted separately from 51 individual eggs, and from 3 pools of 5 eggs and 1 pool of 16 eggs. Sequencing libraries were constructed using the Nugen Ovation SP Ultralow DNA kit for 3 single eggs, and pools of 5 and 16 eggs, and sequenced using an Illumina HiSeq 2500 with 2x100 nt reads. Raw data is available via ENA accession PRJEB12428.

Initial de novo assembly

Sequence read quality was assessed using FASTQC [34]. Adapter and quality (Q≥20) trimming was performed using TrimGalore! (v0.3.1) [35]. For the human and unwashed pig samples, reads were de novo assembled using Velvet (v1.2.08) [36]. To establish the best k-mer size, we initially assembled all reads from patient B using k-mer values of k = 61, 63, 65, 67, 69, 71, 73, 75, 79, 85, 89 and 95. Based on the quality of these assemblies, we used k = 69, 75, 77, 79, 81, 83, 85, 89 and 95 for patient A and the unwashed pig samples.

Metagenomic profiling

Metagenomic profiling of the patient B mite assembly was carried out using PhymmBL (v4.0) [37]. We augmented the bundled PhymmBL model database with interpolated Markov models trained on the spider mite, Tetranychus urticae (strain London) (available at https://bioinformatics.psb.ugent.be/gdb/tetranychus/) [38] as a proxy for scabies mite. To estimate the species abundance, reads were aligned back to contigs using Bowtie2 (v2.2.3) in local alignment mode [39] and the number of reads aligning to each contig were counted.

De novo assembly of the mitochondrial genome

For the two human samples and the unwashed pig sample, the mitochondrial genomes were assembled individually using a bait-and-reassemble strategy (see Fig 1 for overview). First, contigs from the whole genome assemblies were aligned to European house dust mite (EHDM), Dermatophagoides pteronyssinus mitochondrial reference genome (NCBI GenBank: EU884425) [40] using LASTZ (v1.02.00) [41] with default settings. For patient B, contigs from the k = 65 assembly were used for alignment; for patient A, contigs from k = 77, 79 and 89 were used; and for the unwashed pig sample, contigs from the k = 69 and 81 assemblies were used. These different k-mer value assemblies varied in their N50 values, largest contigs sizes and median read coverage. The aligned contigs were then filtered against the National Center for Biotechnology Information (NCBI) NT database for scabies mite mitochondrial contigs using BLASTN (v2.2.29) [42] to remove likely false positives due to homology with other mitochondrial genomes (e.g. host mitochondrial genome). After the filtering step, contigs from patient B k = 65 assembly gave the best coverage of EHDM mitochondrial reference genome. Adapter and quality trimmed reads from the patient B sample were then aligned back to the filtered patient B k = 65 contigs using Bowtie2 (v2.2.3) in local alignment mode. Aligned reads were then used to de novo assemble the patient B scabies mite mitochondrial genome using Velvet (v1.2.08). Assemblies were run using k = 69, 79, 87, 89, 91 and 95. A k-mer size of 91 gave the longest single contig and was selected as the reference mitochondrial genome of scabies mite.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Bait-and-assembly workflow for metagenomics data.

Two iterations were performed. NT stands for nucleotide collection blast database (nt) from NCBI.

https://doi.org/10.1371/journal.pntd.0004384.g001

To generate the mitochondrial reference for patient A, adapter and quality trimmed reads from the patient A sample were aligned to the patient B mitochondrial reference genome, and matching reads were de novo assembled using Velvet. The final contigs were scaffolded and gaps filled with Ns. This scaffold was then used to realign the patient A reads and reassemble. This process was run iteratively. The second iteration gave the minimal number of Ns in the contig and this was retained as the patient A mite mitochondrial genome.

The same process was used for the unwashed pig sample, but in this case the first iteration gave the best contig with minimal Ns.

Annotation

Mitochondrial protein-coding genes were predicted on patient B mitochondrial reference genome using the MITOS pipeline (accessed online, August 2014) [43] using the invertebrate mitochondrial genetic code (with default settings). To verify the annotations, open reading frames (ORFs) between stop to stop codons were extracted from the assembled patient B mite genome using the GETORF program from the EMBOSS package (v6.6.0) [44] using the invertebrate mitochondrial codon table. The extracted 774 ORFs were then used to search the NCBI NR protein database using BLASTP (v2.2.29) (E-value threshold = 0.1). For further verification, 13 protein-coding genes from EHDM were aligned to the patient B mite reference genome using TBLASTN (v2.2.29).

12S and 16S ribosomal RNA were annotated using covariance models for those genes using Infernal (v1.1rc4) [45, 46]. The covariance models were built from multiple sequence alignments of 12S and 16S rRNA genes from three sarcoptiformes, Steganacarus magnus (NCBI RefSeq: NC_011574), Dermatophagoides pteronyssinus (NCBI GenBank: EU884425) and Dermatophagoides farinae (NCBI RefSeq: NC_013184) using Clustal-omega (run on the web version on 15/12/2013) [47]. MITFI (within the MITOS pipeline) was used to identify 21 of 22 standard mitochondrial tRNA genes.

Single nucleotide polymorphisms and diversity

For each of the six samples, reads from each sample were aligned to the patient B reference genome using the Bowtie2 aligner in local mode. Pileups were generated using SAMtools (v0.1.19-44428cd) [48] mpileup. Varscan (v2.3.6) [49] was then used to call SNPs with the mpileup2snp command (Min Coverage: 300; Min Variant Frequency: 0.01; otherwise default parameters). SNPs called within the 100 nt boundary at both ends of the genome (positions 1–100 and 13,820–13,919) were manually filtered out due to lack of reliability of alignments at those low complexity regions. For each sample, SNP frequency was estimated using the ratio of reads supporting the variant to total read depth at the SNP.

Annotation

Thirty-five predicted genes were identified in the scabies mite (var. hominis) mitochondrial genome (Fig 2). These include 13 protein coding genes, two ribosomal RNA genes, and 20 tRNA genes. All of the expected mitochondrial genes for a typical metazoan mitochondrial genome were identified, except for two tRNAs (Alanine and Tyrosine). Search with BLASTP of extracted stop-to-stop ORFs from the mitochondrial genome also verified 12 of the 13 protein-coding genes (except the ATP8 gene) and TBLASTN alignment of 13 protein coding genes from the EHDM mitochondrial genome to the scabies mite mitochondrial genome also confirmed 12 of the 13 protein coding genes (except the ND4L gene) in the scabies mite mitochondrial genome. Two tRNA genes identified by tRNAscan-SE were considered ambiguous due to their overlap with other genes (A is overlapping with C and Y is overlapping with NAD4L). However, the overlap is not on the same strand.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Annotation of the scabies mite, Sarcoptes scabiei (var. hominis), mitochondrial genome.

tRNA genes are abbreviated using the one letter amino acid code. The inner brown plot is plotted for GC content using window size of 500 and step 1. The plot shows deviations from the average GC content of the genome (0.1932). The plot is scaled based on the maximum (0.3034) and minimum (0.1078) GC content values.

https://doi.org/10.1371/journal.pntd.0004384.g002

The gene order and strand specification of the protein coding and ribosomal RNA genes are the same as that of EHDM. All identified tRNA genes, except tRNA-C and tRNA-V also maintain the same gene order and strand as EHDM. The tRNA genes that are not syntenic with EHDM (tRNA-C, tRNA-V) are also on the opposite strand to that of EHDM.

Single nucleotide polymorphisms and mitochondrial haplotypes

To identify genetic polymorphisms in the sequenced mite pools, reads were mapped back to the reference Mt genome (patient B). The average depth of coverage was 2914 across samples (average per sample ranging 1698–4299). A total of 665 single nucleotide polymorphisms (SNPs) were identified across all samples relative to the reference Mt genome (patient B assembly): 601 SNPs in the patient A sample, 598 SNPs in the patient B sample, and 102 SNPs in the unwashed pig sample, while the washed pig samples (w1, w2, w3) contained 102, 100, 102 SNPs respectively (S2 Table). The four pig mite samples are effectively replicates (biological and technical) and were highly concordant.

Within each sample, SNP allele frequencies, estimated from the ratio of reads supporting the variant to the total coverage, were tightly grouped into a small number of clusters (Fig 3A), suggestive of the presence of just a few mitochondrial haplotypes in each sample. To estimate the number and frequencies of haplotypes, and to infer their sequences, we used k-means clustering (S3 Table). SNPs common to all haplotypes within a sample have a frequency of 1. Additional haplotypes are defined by the presence of extra SNPs with mean frequency less than 1. We also examined reads supporting SNPs located close to each other (<100 nt) for evidence that SNPs were on the same haplotype (supported by the same reads) or distinct haplotypes (never co-occurring on one read). Only 4 pairs of SNP were located within 100 nucleotides of each other and in two different frequency clusters. In each sample, we gave the haplotype with the highest frequency the prefix H1. Haplotypes with successively lower frequencies are labeled H2 and H3, in that order.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Scabies haplotype groups.

(A) SNP allele frequencies in each sample show strong clustering and define haplotypes. (B) Phylogenetic tree of the inferred haplotypes from multiple human and pig samples group into 6 haplogroups and 2 broad clades (red and blue boxes).

https://doi.org/10.1371/journal.pntd.0004384.g003

In the patient B sample, a single clear cluster in the SNP frequencies was observed (Fig 3A). This implies the presence of two haplotypes. The dominant haplotype, H1_B_REF, has an estimated frequency of 0.98 and corresponds to the consensus obtained from de novo assembly. The second haplotype (H2_B) is defined by the presence of 598 SNPs and has an estimated frequency of 0.02.

The patient A sample contains a single clear cluster, containing 593 SNPs with a frequency of 1.0. An additional 8 SNPs do not cluster well with the dominant group. These fall into four other clusters identified by k-means. We chose to ignore the three smallest clusters, resulting in just two closely-related haplotypes: H1_A, which has a frequency of 0.98, and H2_A, which has a frequency of 0.02 and is separated from H1_A by just 4 SNPs. The lack of replication of these SNPs makes the significance of these closely related haplotypes unclear.

All four pig samples contain three clear clusters, forming three haplotypes in each sample. The haplotypes with the highest frequencies (H1_u, H1_w1, H1_w2, H1_w3) have estimated frequencies of 0.89, 0.91, 0.89 and 0.90 respectively. Additional haplotypes are composed of combinations of the clusters (S3 Table) and are concordant between samples. The second haplotypes have estimated frequencies of 0.10, 0.07, 0.09 and 0.09 for the unwashed and three washed samples respectively and we label these H2_u, H2_w1, H2_w2, and H2_w3. The third haplotypes, which we label H3_u, H3_w1, H3_w2 and H3_w3 for the unwashed and three washed samples respectively, have estimated frequencies 0.02, 0.01, 0.02 and 0.01 respectively. The H1_u, H1_w1, H1_w2, H1_w3 haplotypes have 82, 82, 81, 81 SNPs; H2_u, H2_w1, H2_w2, H2_w3 have 82, 82, 81, 81 SNPs and H3_u, H3_w1, H3_w2, H3_w3 have 87, 87, 85, 87 SNPs respectively.

Similar analyses of DNA sequencing data from a limited number of individual scabies mite eggs or small pools did not identify clusters in the SNP allele frequencies (S2 Fig), suggesting that multiple haplotypes provide evidence for genetic diversity rather than heteroplasmy.

To understand the relationship between the 16 inferred haplotypes, we constructed a phylogenetic tree based on the SNPs present in each haplotype sequence using MEGA5 (v5.2.2) [51] with a distance measure based on the number of differences between haplotype sequences (S4 Table). The tree shows that the haplotypes fall into two broad clades and six haplogroups (Fig 3B). Three haplogroups are found in clinical isolates (human mite haplogroups 1–3), while the pig mites comprise three haplogroups (pig mite haplogroups 1–3). The average difference between haplotypes within each of the pig mite haplogroups is 1–2 SNPs. The two haplotypes in human mite haplogroup 3 are almost identical and very similar to human mite haplogroup 2, forming one of the clades, while human mite haplogroup 1 is distinct from the other human mite haplogroups, and more similar to the pig mite haplogroups, forming the second clade.

Conclusion

Scabies is responsible for significant morbidity in Indigenous Australians in many remote communities of northern and central Australia. Effective control and prevention of scabies epidemics in those communities is of paramount importance as scabies has a long-term effect on the life expectancy and quality. We have sequenced, assembled and annotated the mitochondrial genome of the scabies mite using a bespoke bait-and-assembly approach; we identified SNPs in multiple isolates from patients and a laboratory pig model, and inferred the haplotype structure and diversity of individual infections. We used these tools to investigate the genetic diversity within individual infestations. Larger samples are now needed.

The development of genomics resources for studying the scabies mite will accelerate research into this parasite, just as genome sequences have for other neglected parasitic diseases. For example, in malaria genomic resources provided means to identify drug resistance causing mutations [59] or in schistosomiasis it helped suggesting new approaches to preventions and strategies for control [60]. The scabies mite mitochondrial genome sequence will facilitate further population genetics research in this area.