2.1.1. cys-loop ligand-gated ion-channel identification in P. ovis.

The Online Resource for Community Annotation of Eukaryotes (OrcAE; https://bioinformatics.psb.ugent.be/orcae/overview/Psovi) database was tBLASTn-searched for genes encoding for cysLGIC in the current P. ovis reference genome [39]. Known protein sequences of GluCls, GABA-Cls, pH-Cls, HisCls and nAchRs from the two-spotted spider mite, Tetranychus urticae, the fruit fly, Drosophila melanogaster, the scabies mite, Sarcoptes scabiei, the tick, Rhipicephalus microplus and the house dust mite, Dermatophagoides pteronyssinus were downloaded from the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov/) and used to screen the genome for homologues (e-value threshold of <1e-50). The most likely cysLGIC subunit encoding genes from P. ovis were identified based on their homology at the amino acid level.

Information on the transcription levels of the cysLGIC subunit encoding genes during the life cycle of P. ovis was extracted from OrcAE using lifecycle stage-specific gene expression data [40]. A heatmap was constructed with the transcription data of the genes across the different lifecycle stages, i.e., larvae, protonymphs, tritonymphs, adult females and adult males, from [40] in R [41]. CysLGIC subunits from P. ovis (based on the susceptible sheep strain), T. urticae, D. melanogaster, S. scabiei and D. pteronyssinus were aligned with the use of MUSCLE [42]. The Jones, Taylor and Thornton model was used for the phylogenetic analysis. A maximum likelihood analysis, bootstrapping 1000 pseudo-replicates, was performed with MEGA X [43] to construct a midpoint rooted tree.

2.1.2. Sample collection.

Pre-treatment skin scrapings from 8 farms corresponding to an ML field efficacy study [13] were used for the collection of 50–100 living P. ovis mites per farm (S1 Table). Mite isolates from the different farms differed in their susceptibility to ML treatment (ivermectin, doramectin and moxidectin), as determined by the calculation of the mite count reduction two weeks post-treatment. The mite population on a farm was considered sensitive when the average mite reduction was ≥ 95% and lower limit of the 95% confidence interval was ≥ 90%, suspected of resistance when the average mite reduction was < 95% OR lower limit of the 95% confidence interval of <90% and resistant when the average mite reduction <95% AND lower limit of the 95% confidence interval <90%. For deep amplicon sequencing only, mite containing skin samples from sheep were collected from infested donor animals held at the Moredun Research Institute, UK. These mites have never been exposed to acaricides throughout their maintenance as a reference population and are therefore completely susceptible to treatment with moxidectin (S. Burgess, Personal Communication). Before collection, skin scrapings and samples were incubated for 20 minutes at 37°C to increase mobility of the mites. Samples were screened with a stereomicroscope and viable mites from all lifecycle stages were collected with a needle. 25 mites of mixed life-stage and sex per sample were snap frozen in liquid nitrogen and stored at -80°C.

2.1.3. Deep amplicon sequencing.

Locus specific primers were designed to amplify a 219 bp fragment of interest from both GluCl-44 (PsoOvis1B009037 in the genome annotation described in Section 2.2) and GluCl-280 (PsoOvis1B011928). Details of DNA extractions, primers used, PCR conditions and MiSeq sequencing are given in S1 Methods and S2 Table. Sequencing was undertaken on the Illumina MiSeq platform and raw FASTQ files were analysed with the DADA2 v.1.11.5 [44] bioinformatic software package to ascertain the number of unique amplicon sequencing variants (ASV) contained in each sample. This workflow was adapted from the DADA2 analysis described at www.nemabiome.ca, with modifications to accommodate for the amplicons analysed in this paper. Briefly, FASTQ files were pre-filtered with the ‘FilterAndTrim’ function to remove any ‘Ns contained in the sequences. CutPrimers [45] was used to remove forward and reverse primer sequences in the amplicons. After primer removal, reads were filtered again with ‘FilterAndTrim’ with no N’s allowed, maxEE = 6, truncQ = 2, a minimum length of 200 bp for each forward and reverse read, and the removal of phiX if identified. A sequence table was constructed with ‘makeSequenceTable’ to display all ASVs present in the dataset. Chimeras were removed with ‘removeBimeraDenovo’. This provides a read count of each ASV present in each sample. A fasta file was also generated with the ‘getUniques’ and ‘uniquesToFasta’ commands to provide a list of all ASVs identified and their corresponding nucleotide sequence. Each ASV was blasted against a reference sequence to ensure that the ASV correctly matched the intended amplicon. Any off-target amplicons were subsequently removed from analysis. Furthermore, ASVs with extremely low reads were manually deleted from the dataset since these most likely represent PCR artifacts. All amplicon sequences were submitted to the European Nucleotide Archive (ENA) under PRJEB82994.

2.2.1. Sheep- and Cattle-derived P. ovis genome assemblies.

For the sheep-derived assembly, filtered long read (ONT) data was assembled with Flye v2.8.3 [46] with two iterations of polishing. To improve the base accuracy of the assembly, it was polished three times with HyPo [47] using both Illumina and ONT data. To remove potential contamination, we examined read coverage (ONT reads were aligned back to the assembly with minimap2 [48]) as well as taxonomic identification by searching against the NCBI nucleotide (‘nt’) database using megablast. Only those contigs with hits to mite sequences and with the expected read coverage were retained. The decontaminated assembly was scaffolded with 10X Linked-Reads using the Faircloth Lab pipeline (http://protocols.faircloth-lab.org/en/latest/protocols-computer/assembly/assembly-scaffolding-with-arks-and-links.html#). For the cattle-derived P. ovis data, the genome was assembled using the CLAWS v2.1 workflow [49] combining ONT long reads and 10X linked reads. As this assembly was of low contiguity, the assembly was then aligned to the sheep-derived assembly and reference-based scaffolding was carried out using RagTag [50]. We provide more in-depth methods of assembly, removal of contaminant sequences and assembly evaluation in the S2 Methods. The gene annotation for the sheep-derived P. ovis genome assembly was obtained by combining transcript alignments of available P. ovis RNASeq data, protein alignments of related species and ab initio gene predictions principally using PASA [51]. Finally, the sheep-derived P. ovis genome annotations were mapped to the cattle-derived P. ovis genome using Liftoff [52]. A more detailed explanation of genome annotation is given in S2 Methods.

Identification of cys-loop ligand-gated ion-channels genes (Section 2.1.1) and resulting qPCR analyses were undertaken using an earlier version of the P. ovis gene set as a basis [40]. A BLAST-based reciprocal best hits (RBH) of approach [53] of protein sequences was used to identify corresponding genes between our newly generated (CNAG) sheep-derived P. ovis gene set and the gene set of [40]. We have used gene names from our newly generated dataset throughout the text to avoid confusion and present corresponding gene names for both gene sets in tables where appropriate.

2.2.2. Population genomics of Belgian cattle-derived P. ovis.

Mites from seven Belgian farms of mixed susceptibility, suspected- and fully resistant phenotypes were whole genome sequenced using a PoolSeq approach (Table 1). Raw data generated by whole genome re-sequencing was deposited in the ENA under accession PRJEB84953. Reads for the seven samples of susceptible, resistant and suspected resistant cattle-derived P. ovis populations were quality and adapter trimmed with fastp (version 0.23.4) [54] default parameters and sequence quality checked with FastQC (v0.12.1) [55]. Trimmed reads for each population were aligned to the sheep- and cattle-derived P. ovis genomes separately with BWA (v0.7.17) [56]. PCR duplicates of aligned reads were marked and subsequently sorted by mapping coordinates using Picard (v2.18.29) [57]. Alignment statistics for each processed BAM file were generated using samtools “flagstats” and “stats” tools (v1.17) [58]. Variant predictions across populations were made in Varscan (v2.4.6) [59] as it is compatible with pooled template sequencing by estimating allele frequency per population. Samtools mpileup [58] was used to generate the Varscan input format and a p-value threshold of <0.05 set and a minimum variant frequency of 0.01 required for variant prediction. The gene body location of variants and potential effects were predicted in SnpEff (v5.2) [60] using a custom database for the sheep- and cattle-derived P. ovis genomes. Average coverage per gene was calculated with the bamstats05 tool of the Jvarkit package (version dbdbed3a9) [61]. Average read coverage for each gene across the sheep- and cattle-derived P. ovis genomes was divided by median coverage for all genes in order to identify outlier loci and coverage plots created in the R package ggplot2 (v3.4.4) [62]. Fixation indices estimated by the unbiased-Hudson method (F_st) were calculated from bam files per gene and in sliding windows of 10,000 bp with 1,000 bp steps using the Grenedalf package (v0.3.0) [63]. Grenedalf is a re-implementation of the popoolation2 package [64]. Poolfstat (v2.2.0) [65] and pcadapt [66] were used to generate global pairwise nucleotide distance and F_st estimates and perform principal components analysis (PCA) between populations respectively. Smoove (v.0.2.8) (https://github.com/brentp/smoove) was used to predict and annotate structural variants. We have used capitalised “Susceptible”, “Suspected Resistant” and “Resistant” when referring to each of the seven re-sequenced populations by phenotype in the Results and Discussion sections.

Download:

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

Table 1. Overview of the different samples used for DNA extraction together with the number of mites, region of origin, in vivo ML susceptibility and host species as established by [13].

https://doi.org/10.1371/journal.ppat.1012963.t001

Computationally identified regions of interest in the genome were visually checked using the Integrative Genomics Viewer (v2.16.2) (IGV) [67]. ONT reads sequenced for sheep- and cattle-derived P. ovis genome assemblies were revisited to identify the genomic context of genes of increased copy number among resistant and susceptible populations. Protein sequences of genes with elevated coverage were searched against ONT reads with Diamond (v2.1.8) [68] to determine if tandem duplications (cis) or insertions in other regions of the genome (trans) could explain observed increases in copy number. Diamond was run in ultra-sensitive mode with the number of matches for ‘query’ nanopore reads to ‘subject’ P. ovis protein sequences increased to 1000 to accommodate more than one copy of a gene occurring on a single read.

2.3.1. Sample collection.

A farm with a known ML susceptible P. ovis population, 99.3% mean reduction in mite counts post-treatment and one assessed as having a resistant mite population with a mean reduction of 36% in mite counts post-treatment from the field efficacy study [13] were revisited to collect susceptible and resistant mites before and after treatment. On both farms, a group of 12 animals with clinical psoroptic mange was selected and physically separated from the rest of the herd. From this group, 6 animals were selected for validation of the ML susceptibility at the time of sampling as described in [13], and 6 for the collection of mites for RNA extraction. Details of RNA extraction and sequencing protocols are given in S1 Methods.

The pre-treatment samples were taken from a maximum of half of the lesion surfaces. All animals were subsequently treated twice with a subcutaneous injection of Ivomec (Merial, FR) at a dose of 0.2 mg ivermectin per kg with a 7-day interval. On the suspected resistant and resistant farms, the post-treatment samples were collected 7 days after the second ivermectin treatment. All samples were processed within 8 hours after collection. After incubation at 37°C, adult female P. ovis mites were collected from the skin scrapings with a fine needle. Differentiation of life-cycle stages was based on [69]. Mites were snap-frozen in liquid nitrogen in batches of 25 and stored at -80°C in 2 ml tubes (VWR, USA).

2.3.2. Quantification and differential expression of RNASeq data.

Pseudo alignment of the RNA-seq reads to the CNAG sheep-derived P. ovis transcriptome annotation was performed using Kallisto (v1.15.0) [70] in unstranded mode as part of the rna-seq-pop package (v1.0.4) [71]. Kallisto generated read counts for all RNA-seq samples were used as input for DESeq2 (v1.30.1) [72] and Sleuth [73] to identify significantly differentially expressed genes and transcripts respectively between the three ML-resistant unexposed (RES_unexposed), three ML-resistant exposed (RES_exposed) and three ML-susceptible (SUS) replicates. Significantly differentially expressed genes and transcripts were classified as those having a fold change ≥ 2.0 between each of the pairwise comparisons and a False Discovery Rate (FDR) corrected p-value of ≤ 0.05.

2.3.3. Validation of RNA-seq data by Real Time qPCR.

qPCR was used to validate the transcription of genes associated with ML resistance in the tested P. ovis populations. qPCR was performed on cDNA samples from all nine RNA samples used for the RNA-seq analysis (SUS x3 replicates, RES_unexposed x3, RES_exposed x3) and six differentially expressed genes were selected along with three P. ovis housekeeping genes. cDNA libraries were constructed with the iScript cDNA Synthesis Kit (Bio-Rad Laboratories Inc., USA) following the recommended protocol. Forward and reverse primers were designed for six genes of interest from the RNASeq experiment using the gene sequences of [39]: two UGT genes, PsoOvis1B004414 and PsoOvis1B010992, an inositol oxygenase, PsoOvis1B000159, the CYP genes PsoOvis1B011549 and PsoOvis1B005763, and a glutathione S-transferase, PsoOvis1B008730, and for three housekeeping genes: a beta actin, PsoOvis1B008887/PsoOvis1B000861 [74], a mitochondrial ribosomal protein, PsoOvis1B008370 [21], and elongation factor 1-alpha, PsoOvis1B001219 [75]. A description of primers used for each gene is given in S3 Table. All primer sequences were BLAST searched against the P. ovis genome to rule out cross specificity. Specific qPCR conditions are given in the S1 Methods. Correlation between RNAseq and qPCR values was performed using the R package “rmcorr” Shiny application (https://lmarusich.shinyapps.io/shiny_rmcorr/) [76] following [77]. Rmcorr is able to account for non-independent measures, in our case each gene was assessed three times for each contrast between RES_unexposed, RES_exposed and SUS replicates. This allows us to include all contrasts in a single correlation analysis. Rmcorr uses analysis of covariance (ANCOVA) to statistically adjust for interindividual variability and fits a common slope and variable intercept to all data points, i.e., genes, included.

3.1.1. Identification of cys-loop ligand-gated ion-channels.

In P. ovis, 16 different cysLGICs were found, that are activated by either glutamate (n = 2), histamine (n = 2), GABA (n = 4), acetylcholine (n = 7) or environmental pH (n = 1). All cysLGICs showed homologies of 63% or more at amino acid level with at least one other cysLGIC from T. urticae, D. melanogaster, S. scabiei, D. pteronyssinus and R. microplus. Moreover, high levels of homology, between 82% and 99%, were observed with certain cysLGICs in other species, most often with D. pteronyssinus, probably due to the relatively close homology between these species [78]. An overview of the different cysLGICs in P. ovis is given in Table 2. together with the corresponding cysLGIC with the highest homology in T. urticae, D. melanogaster, S. scabiei, D. pteronyssinus and R. microplus.

Download:

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

Table 2. Summary of the cysLGIC encoding genes in P. ovis with the corresponding cysLGICs with the highest homology in T. urticae, D. melanogaster, S. scabiei, D. pteronyssinus and R. microplus.

https://doi.org/10.1371/journal.ppat.1012963.t002

Both P. ovis GluCls showed similar transcription patterns, with higher transcription levels in the larval stage (Fig 1, GluCl-44 218 reads per million and GluCl-280 192 reads per million) and, to a lesser degree, in protonymphs and adult males (83–119 reads per million). In the other two stages, the subunits had low levels of transcription, with between 10 and 33 reads per million. All other cysLGIC subunits had low transcription levels, between 1 and 50 reads per million. Fig 2 depicts phylogenetic relationship of the orthologues of the different cysLGIC subunits. Four of the five cysLGIC gene families included are monophyletic and have high bootstrap support although three nicotinic acetylcholine-receptor genes form their own poorly supported lineage (unshaded sequences, Fig 2). Only the Histamine-gated chlorine channels are not monophyletic consisting of three lineages within the purple shaded oval of Fig 2. DNA concentration after purification and Phusion high-fidelity PCR ranged from 33.6 ng/µl to 119.7 ng/µl.

Download:

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

Fig 1. Heatmap of the transcription levels (expressed in reads per millon) of the different P. ovis cysLGIC subunit encoding genes.

Blue indicates low levels of gene transcription and red relative high levels of gene transcription.

https://doi.org/10.1371/journal.ppat.1012963.g001

Download:

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

Fig 2. A maximum likelihood phylogenetic tree of cysLGIC subunits from P. ovis (PSOVI), T. urticae (Tu), D. melanogaster (Dm), S. scabiei (Ss) and D. pteronyssinus (XP_).

Bootstrap values are given at the nodes and gene families of included sequences are overlaid in coloured ovals. The three unshaded sequences are nicotinic acetylcholine-receptors which form a poorly supported (bootstrap support = 49) lineage separate from other nicotinic acetylcholine-receptors.

https://doi.org/10.1371/journal.ppat.1012963.g002

3.1.2. Deep amplicon sequencing.

No variation was observed between the 179 bp-long sequences of the GluCl-44 gene and the GluCl-280 genes between the analysed P. ovis mite populations. The amplified GluCl-44 and -280 sequences were identical to both gene sequences previously submitted to OrcAE (S1A and S1B Fig). S1C Fig illustrates how the location of these known mutations relates to the examined regions of GluCl-44 and GluCl-280.

3.2.1. Sample collection and RNA extraction.

The post-treatment reduction in mite counts on the susceptible farm, Farm 2 from [13], was 97.4% with a 95%CI of [91.8%:99.9%]. The reduction post-treatment on the resistant farm, Farm 5 in [13], was 66.1% with a 95%CI of [55.1-75.6%]. Based on the criteria described in [13] the mite population on the first farm was classified as sensitive and the latter as resistant. A detailed description of the samples can be found in S4 Table.

3.2.2. RNAseq analysis and identification of genes differentially expressed between P. ovis mite populations.

RNA sequencing generated between 30–39 million reads per sample. The pseudo-alignment rates varied between 84–90%. Detailed information per sample is given in S4 Table. A total of 10,516 genes were tested in each comparison and the majority of genes were assigned an adjusted p-value with the RES_unexposed versus RES_exposed comparison having the least at 8,877 genes. DESeq2 results for each of the three comparisons between the three treatments are given in S1–S3 Files.

3.2.3. Gene and isoform level differential expression.

A Venn diagram, volcano plots and heatmap of differentially expressed genes with greater than log₂-fold change of +1/-1 between the three pairwise contrasts are shown in Fig 3, parts A-C. Differential expression results implicated constitutively expressed detoxification genes in P. ovis ML resistance with 37 genes showing shared overexpression in RES_exposed and RES_unexposed versus SUS, 16 of which are doubled in expression versus SUS (> log-fold₂ change of 1, Fig 3A). In both RES_exposed and RES_unexposed versus SUS comparisons a cytochrome P450 3A4-like gene (PsoOvis1B011549) was significantly highly overexpressed with (Table 3) absolute fold-changes of 4.9 and 3.3 respectively. A blast search of the PsoOvis1B011549 protein sequence against a database of arthropod cytochrome P450s (https://arthropodp450.eu) indicates it is a member of the CYP3691A subfamily with greatest similarity to the European house dust mite, Dermatophagoides pteronyssinus. This gene was also significantly over-expressed in the RES_exposed versus RES_unexposed comparison with 1.5-fold higher expression in RES_exposed replicates. Two UGT genes (PsoOvis1B010992 and PsoOvis1B004414) were also highly overexpressed in RES_exposed and RES_unexposed versus SUS comparisons but not in the RES_exposed versus RES_unexposed comparison, as was an inositol oxygenase gene (PsoOvis1B000159) which may act in detoxification by UGTs. The two UGTs occur in tandem on chromosome 7 in the P. ovis sheep- and cattle-derived P. ovis genomes. Several CYPs, UGTs and other detoxification genes including carboxylesterases, ATP-binding cassette and glutathione-S-transferase genes exhibited patterns of induced expression both up- and down-regulated in RES_exposed (Table 3). Other potentially induced genes include a GRIN3B glutamate ionotropic receptor (PsoOvis1B009258). With respect to other possible resistance mechanisms two cuticular protein genes were overexpressed in RES_exposed versus SUS and RES_unexposed comparisons with 3.5-to-6-fold changes in abundance (Table 3 and S1–S3 Files). A voltage-dependent T-type calcium channel subunit (PsoOvis1B011113) was downregulated in both RES_exposed and RES_unexposed versus SUS comparisons. An acetylcholinesterase-like gene (PsoOvis1B007848) was downregulated in RES_exposed versus RES_unexposed and SUS.

Download:

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

Table 3. Overview of highlighted differentially expressed genes for each comparison.

https://doi.org/10.1371/journal.ppat.1012963.t003

Download:

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

Fig 3. Summary of RNASeq results.

A) Venn diagram showing concordantly up- and down-regulated genes between RES_exposed versus SUS, RES_unexposed versus SUS and RES_exposed versus RES_unexposed comparisons with greater than log-fold₂ 1 difference. Up refers to genes more highly expressed in the first treatment labelled for each comparison. Shared differentially expressed genes between two comparisons are found in the overlapping circles. B) Volcano plots of differentially expressed genes for the three contrasts with key genes labelled: P450 = Cytochrome P450, UGT = UDP-glucuronosyltransferases, pTE = putative transposable element. C) Heatmap of log-fold change for each of the three contrasts undertaken for all genes included in (A). fc = fold-change.

https://doi.org/10.1371/journal.ppat.1012963.g003

Further genes of interest included two argonaute genes (PsoOvis1B005563 and PsoOvis1B003877) which were significantly downregulated in RES_exposed versus SUS. A gene likely to encode a retrotransposon, PsoOvis1B002032, had striking expression dynamics with high relative expression in RES_exposed and RES_unexposed and zero expression in SUS replicates with associated very low adjusted p-value. As is often the case for organisms such as P. ovis, with poorly characterised genomes, functional domain annotations were unavailable for many of the differentially expressed genes identified between both groups. Further investigation of their potential function could identify additional detoxification mechanisms.

At the isoform level results are concordant with the gene results with the exception of one of the five isoforms of the Gamma-Aminobutyric (GABA) transporter (PsoOvis1B011106, interpro accession: IPR000175, CDD accession: cd11496) which is down-regulated in both RES_exposed and RES_unexposed treatments versus SUS but is not significant in any gene level comparison (S4–S6 Files).

3.2.4. Quantitative PCR results.

The results of the qPCR confirmed the results of the RNA seq analysis with a positive, significant rmcorr estimated r_rm value of 0.74 (95% confidence intervals 0.525-0.977, p < 0.004) as can be seen in Fig 4A. The expression of inositol oxygenase (PsoOvis1B000159), a UGT (PsoOvis1B010992) and the two CYPs (PsoOvis1B011549 and PsoOvis1B005763) followed the same trend in qPCR as for the RNAseq data. Only for the other UDP-glycosyltransferase (PsoOvis1B004414) was the transcription pattern of the RNAseq data not confirmed by the qPCR, indeed removing this gene from the rmcorr increased the r_rm value to 0.95 and narrowed the 95% CIs (Fig 4B, 0.729-0.992). This lack of concordance between qPCR and RNAseq data for PsoOvis1B004414 was unexpected, since the biggest differences in transcription after RNA seq were found for this gene.

Download:

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

Fig 4. Rmcorr correlation analysis between log₂ real-time quantitative PCR and RNASeq expression values for six selected genes.

Each gene has three points representing the three experimental contrasts between RES_exposed RES_unexposed and SUS conditions, and a common trend line as fit by rmcorr to all genes. FC = fold change. A) all six genes included correlation results: r_rm = 0.74, 95% CI [0.525, 0.977], p = 0.004. B) Gene PsoOvis1B004414 removed from the analysis given the difference between RNASeq and qPCR values resulting in an increased positive correlation: r_rm = 0.95, 95% CI [0.729, 0.992], p < 0.001, as shown by the trend line. Raw data for this figure is given in S13 Table.

https://doi.org/10.1371/journal.ppat.1012963.g004

3.3.1. Improved cattle- and sheep-derived P. ovis genome assemblies.

The genome assembly length of the sheep-derived P. ovis genome (62.7 Mb) and the number of large scaffolds (10) are close to the expected size and number of chromosomes (Genome assembly metrics: Table 4). The less contiguous sheep-derived P. ovis assembly of [38] GCA_002943765.1, also has a length of 63 Mb and the Genomes on a Tree (GoaT, https://goat.genomehubs.org) is nine based on ancestral genomes. Although not as completer, the cattle-derived P. ovis genome is of similar length (Table 4). Sequencing depths and alignment statistics are given in S5 Table. Of the 10,516 nuclear genes annotated in the sheep-derived P. ovis genome (Table 5), 300 were not transferred to the cattle-derived P. ovis genome assembly by liftoff. Of these, 206 genes lack informative annotations, i.e., have “protein_coding”, “N/A”, hypothetical or uncharacterised annotation only. A further 58 encode transposable elements and 36 encode annotated genes. Pairwise genome-wide genetic distances, F_st and PCA between the Susceptible and Resistant populations indicate separation between the two phenotypes and high-similarity among the Resistant populations (Fig 5A and 5B).

Download:

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

Table 4. Genome assembly metrics for sheep- and cattle-derived P. ovis assemblies.

https://doi.org/10.1371/journal.ppat.1012963.t004

Download:

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

Table 5. Genome annotation metrics for P. ovis used in RNASeq and population genomic analyses.

https://doi.org/10.1371/journal.ppat.1012963.t005

Download:

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

Fig 5. Relatedness of genome re-sequenced P. ovis populations.

A) Heatmap of global pairwise nucleotide divergence and F_st between all whole genome sequenced pooled template populations included here. D03 = Susceptible; D08 = Suspected Resistant; D05, D06, D07, D09 & D10 = Resistant. B) Principal components analysis for the first two components for distance between populations coloured by phenotype and labelled by population. X-axis explains 50% of variance and discriminates between the Susceptible population versus Suspected Resistant and Resistant populations.

https://doi.org/10.1371/journal.ppat.1012963.g005

3.3.2. Gene copy number variation.

Elevated coverage for each population was identified at two loci across the P. ovis cattle-derived P. ovis genome through comparison of average gene coverages and visual inspection of read data in IGV. On Chromosome 2 the Cytochrome P450 3A4-like gene PsoOvis1B011549 with high-expression and a neighbouring short-chain dehydrogenase gene PsoOvis1B003189 have elevated coverage in all Resistant populations but not the Susceptible population where coverage is not different from the genome-wide median (Fig 6B). The boundaries of the 3,347 bp elevated coverage region are clearly defined at positions 3,606,198–3,609,545 in the cattle-derived P. ovis genome (Fig 6D). For each of the six populations coverage is even across the elevated region including introns present in both PsoOvis1B011549 and PsoOvis1B003189 genes. Coverage varies widely as a multiple of the median across Resistant populations indicating possible variable copy numbers across sampled populations (Fig 6B). Gene PsoOvis1B003189 was also over-expressed in both RES treatments versus SUS but not in RES_exposed versus RES_unexposed. On Chromosome 7, the tandemly located UGTs PsoOvis1B010992 and PsoOvis1B004414 exhibit even more extreme coverage dynamics (Fig 6A). In contrast to the high-copy number CYP gene, the Susceptible population also exhibits increased copy number of the two UGT genes (Fig 6A and 6C) but without associated gene over-expression. Average coverage/median coverage is below 50 in these genes in the Susceptible population but above this level in all Resistant populations. The alignment profile of the elevated region also differs between Susceptible and other populations (Fig 6C).

Download:

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

Fig 6. Copy number variation of the over-expressed UGT (PsoOvis1B004414 and PsoOvis1B010992) and CYP genes (PsoOvis1B011549).

A) & B) Coverage of each gene divided by median genome-wide coverages per population sequenced for the two chromosomes encoding the two UGT and CYP genes. Resistant and Suspected Resistant populations are coloured in shades of blue and green and the Susceptible population in red, for B) gene labels were removed for several Resistant populations as they were overlapping. C) & D) Read alignment patterns across these loci showing insert size of paired end reads and overall read coverage (pink regions) with coding region locations for UGTs and CYP genes plotted beneath the X-axis.

https://doi.org/10.1371/journal.ppat.1012963.g006

Other loci with elevated coverage in Resistant and Suspected Resistant versus the Susceptible population include PsoOvis1B001111 also on Chromosome 2, a GTP-ase activating protein, although this gene is not differentially expressed in any comparison. For all populations, elevated average coverage of this gene is explained by a 19 bp spike in coverage against background genome coverage, possibly resulting from transposable element activity. A plexin A gene (PsoOvis1B001325) on Chromosome 2 is elevated in all populations but less so in the Susceptible population in the cattle-derived P. ovis genome aligned data but, surprisingly, not the sheep-derived genome (S6 and S7 Tables). This gene was not differentially expressed in any comparison. Two genes (PsoOvis1B007656 & PsoOvis1B000689) at the 3’ end of Chromosome 4 are also elevated in coverage, one of which encodes for a reverse transcriptase. Finally, genes on Chromosome 1 (PsoOvis1B004923) and Chromosome 10 (PsoOvis1B007634) were also elevated versus background coverage. Another gene with a distinct coverage profile is Skeletor (PsoOvis1B008827) which is elevated in alignments against the sheep-derived P. ovis genome but not the cattle-derived assembly.

We used ONT long reads to explore possible cis copy number variation through the occurrence of multiple copies of a gene on the same long read for both sheep- and cattle-derived P. ovis datasets. For sheep-derived P. ovis nanopore data, CYP gene PsoOvis1B011549 only occurs once in every read whereas 8 reads had evidence for more than one copy of PsoOvis1B011549 and PsoOvis1B003189 in cattle-derived ONT data (S8 Table). By contrast many reads had non-overlapping consecutive matches to the two UGT genes in the cattle-derived long-read data. The longest ONT read (eb6fdcaf-bc66–4353-a602-9c8fa9087462) at >70Kb encoded 40 copies of PsoOvis1B010992 with matches to three further genes including PsoOvis1B000689 (S2 Fig), 35 of the genes occurred consecutively along the read (S9 Table). The read (17f558b9-9b0f-4336-94f5-3de162674cb6) with most copies of UGT PsoOvis1B004414 encodes 20 copies of the gene alongside 20 copies of PsoOvis1B010992 and PsoOvis1B014440. In the sheep-derived P. ovis nanopore data the most copies of either UGT gene encoded by a single read was three copies (S10 Table).

Read alignments around PsoOvis1B011549 were inspected in IGV to identify the mode of copy number increase in the Resistant versus the Susceptible populations. The increase in coverage was clearly delineated in Resistant populations with no change versus surrounding sequence in the Susceptible (Fig 6). Discordant read mappings at the start and end of the sequence with elevated coverage were used to identify potential insertion positions in other regions of the P. ovis genome for both sheep- and cattle-derived P. ovis assemblies. Five sites were identified with elevated coverage of ~250 bp length in all Resistant populations but not the Susceptible population. Localised increases in read number concordant with putative insertion sites and confirmed by inspection in IGV (S3 Fig). Two of the potential insertion sites flank either side of a gene, PsoOvis1B003901, on chromosome 4. One further CYP gene on Chromosome 5, PsoOvis1B000586, had elevated coverage in one Resistant population (D06) in sheep-derived P. ovis genome data only, as this gene was not placed on the cattle-derived P. ovis genome assembly by liftoff (S6 and S7 Tables). Although significantly differentially expressed in two comparisons (RES_exposed versus RES_unexposed and RES_unexposed versus SUS) it is not a gene with high absolute expression. Furthermore, this gene and two 3’ loci (PsoOvis1B006329 & PsoOvis1B014010) were not placed by the liftoff annotation into the cattle-derived P. ovis assembly. Few other copy number variants correlate with resistance phenotype, including a deletion of 162 bp not found in the Susceptible population in an intergenic region of Chromosome 5 (position 1,943,240 cattle-derived P. ovis genome).

3.3.3. Genome wide F_st scans.

Many peaks of differentiation were evident in pairwise F_st scans between Resistant, Suspected Resistant (D05-D10) and Susceptible (D03) populations (Fig 7). Chromosomes 1 and 2 as the largest chromosomes contain the most signals of differentiation. Given the multiple pairwise comparisons, we decided to restrict analysis to genes with average F_st differences of 0.2 or greater between the average of all Susceptible (D03) versus Resistant/Suspected Resistant (D05, D06, D07, D08, D09, D10) pairwise comparisons and all Resistant/Suspected Resistant versus Resistant/Suspected Resistant comparisons. In this way we were able to focus on genes that had a consistent signal of differentiation between Susceptible and Resistant comparisons (x6) while avoiding differentiation shared with Suspected Resistant and Resistant-only population pairwise comparisons (x15).

Download:

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

Fig 7. Per gene absolute mean difference in F_st (which we term |F_st|) between all Resistant and Suspected Resistant versus Susceptible comparisons and Resistant and Suspected Resistant versus Resistant and Suspected Resistant comparisons.

Chromosomes are numbered along the X-axis with a change in colour indicating a break between chromosomes. A cut-off of 0.2 was applied shown by the horizontal red line, genes above this are labelled with font size changed to avoid overlap. The UGT PsoOvis1B004414 and P450 PsoOvis1B011549 are highlighted in bold and increased font size.

https://doi.org/10.1371/journal.ppat.1012963.g007

Only 48 loci passed the threshold for consideration (Fig 7 and S11 Table) using the cattle-derived P. ovis genome aligned dataset. This included the cytochrome P450, PsoOvis1B011549, which has an average of F_st of 0.4 across the six D0X versus D03 comparisons versus 0.0 for intra-Resistant/Suspected Resistant comparisons giving an absolute difference (|F_st|) of 0.4. One of the two UGTs, PsoOvis1B004414, was also in the 48 genes at |F_st| 0.23 versus |F_st| of only 0.06 for PsoOvis1B010992. The Gamma-Aminobutyric (GABA) transporter (PsoOvis1B011106) which exhibits differential transcript usage also had high |F_st| at 0.24. Only four genes overlap with differentially expressed genes for RES_exposed and RES_unexposed versus SUS comparisons. These included the cytochrome P450 PsoOvis1B011549 and adjacent amplified gene PsoOvis1B003189, the UGT PsoOvis1B004414 and PsoOvis1B013814 on Chromosome 3 which is highly overexpressed in RES_exposed and RES_unexposed versus SUS at 3.5 and 2.6-fold greater expression respectively. This gene is not annotated and does not encode a diagnostic protein domain (InterProScan search) and is most similar to clumping factor B-like in Dermatophagoides farinae by protein blast search (e-value 1 x 10^-8).

Extended regions of |F_st| > 0.2 occurs on Chromosome 2 at ~1 mb and Chromosome 3 at ~1.45 mb (Fig 7). Genes in these regions weren’t correlated to resistance phenotype by gene expression or copy number variation analyses. Some genes with high |F_st| encode nonsynonymous mutations with inverse frequencies between the Susceptible and Resistant populations (S12 Table).

3.3.4. Variants associated with ML-resistance candidate loci.

No non-synonymous mutations were detected in the high-copy number, over-expressed CYP genes in any populations. Three synonymous mutations occur at close to 100% frequency in the five Resistant and single Suspected Resistant populations which are not present (0% frequency) in the Susceptible population (S12 Table). A 5’ upstream variant in the amplified DNA region is fixed in the Susceptible population but not present in any other populations. The neighbouring short-chain dehydrogenase gene PsoOvis1B003189 that shares elevated coverage with PsoOvis1B011549 encodes a premature stop codon at amino acid 391 of 798. This stop codon is at 100% frequency in all six Resistant/Suspected Resistant populations but does not occur (0% frequency) in the Susceptible population. The UGT gene PsoOvis1B010992 encodes two frameshift mutations within three base pairs of each other. These mutations occur at intermediate levels in the Susceptible population at 34% and 37% respectively whilst occurring at low prevalence (<5%) in all other populations (S12 Table). As they involve a two base pair insertion and deletion respectively it is not clear if a frameshift results in the final protein. Three missense mutations follow a similar pattern of prevalence with intermediate levels in the Susceptible population and low frequencies in the six other populations. The Gamma-Aminobutyric (GABA) transporter (PsoOvis1B011106) encodes many mutations between the Susceptible and Resistant populations; however, none result in nonsynonymous changes, and we did not identify any non-synonymous mutations in the whole genome re-sequencing data at this gene.

4.2.1. Overexpression of metabolic resistance genes.

The most prominent genes emerging from the RNAseq experiment were two UDP-glycosyltransferases (UGTs) and a CYP gene which were constitutively up-regulated in all six RES_exposed and RES_unexposed P. ovis replicates. Although recognised as important components of metabolic resistance to pesticides in arthropods the role of UGTs is less well-characterised than for CYPs [96–99]. Our study is one of the first descriptions of UGT expression in an arthropod species of veterinary importance, as they are absent in some well-known non-insect arthropods, such as the Ixodes and Rhipicephalus ticks, the honeybee mite, Varroa destructor, and the predatory mite, Metaseiulus occidentalis [97,98]. Indeed, UGTs were lost early in chelicerate evolutionary history, but were re-acquired from bacteria in mites [98,100]. In the spider mites, T. urticae and Tetranychus cinnabarinus, up-regulation of a UGT was linked to resistance to the ML abamectin [96,98]. Multiple recombinant UGTs from T. urticae, expressed in E. coli, were capable of metabolising abamectin and milbemectin in vitro. Furthermore, inhibition of a UGT gene with RNAi resulted in increased mortality in abamectin resistant strains of T. cinnabarinus [96,99,101]. The constitutive up-regulation of an inositol oxygenase orthologue (PsoOvis1B000159) also indicates a possible role of UGTs in the detoxification of MLs. This enzyme catalyses the production of D-glucuronate from inositol. D-glucuronate is involved in mammalian Phase II detoxification processes and could potentially have a similar role in invertebrates [102]. Snoeck et al. (2019) [99] observed the use of UDP-glucose as a main donor in T. urticae, but UDP-glucuronic acid was also utilised by certain UGTs. To fully understand the role of this gene in P. ovis further studies are required to provide a functional characterisation. Results from qPCR broadly support the RNASeq results with only PsoOvis1B004414 having differing expression dynamics across the three treatments. In SUS P. ovis mites, this gene was also expressed highly, to a similar extent to the RES_exposed and RES_unexposed mites.

In T. urticae, CYPs have been linked to abamectin resistance [103]. Three different CYP genes were more highly expressed in different abamectin resistant strains [95]. A recombinant protein derived from one of these three genes was capable of detoxifying abamectin to a less toxic metabolite [95]. A nuclear receptor underlies trans-driven increases in detoxification genes including CYPs [19]. Sarcoptes scabiei utilises the CYP family for the detoxification of pyrethroids [104]. In other non-parasitic arthropods, CYPs have been linked to ML resistance [105]. Knockout of the CYP gene CYP9A186 gene reversed avermectin resistance in the army worm, Spodoptera exigua. Recombinant CYP9A186 was able to metabolise abamectin and emamectin benzoate in vitro [106]. CYP also play a role in ivermectin resistance in the human body louse Pediculus humanus humanus and the tick Rhipicephalus microplus, although ABC-transporters have a bigger impact on susceptibility in R. microplus than CYPs [92,107,108].

4.2.2. Combined overexpression of detoxification genes.

Combined over-expression of Class I (CYPs) and Class II (UGTs) detoxification genes observed here occurs in other arthropod pests and vectors resistant to a variety of pesticides. Examples include the Colorado potato beetle, Leptinotarsa decemlineata, in response to the neonicotinoid imidacloprid [91], the green peach aphid, Myzus persicae, against sulfoxaflor [94] and the beet armyworm, Spodoptera exigua, in response to multiple insecticides including abamectin [90]. Interestingly, for L. decemlineata silencing the CYP and UGT genes through RNAi did not improve the mortality of imidacloprid versus RNAi of each gene individually. UGTs specifically have been implicated in resistance to organophosphates in the house fly Musca domestica [109]. This is concerning as plunge dipping in the organophosphate, diazinon, is heavily relied upon for the effective control of sheep scab. This method of treatment has also been shown to be highly efficacious against ML-resistant sheep-derived P. ovis mites (S.T.G. Burgess, Personal Communication).

4.2.3. Other genes of interest.

We observed significant, but modest, differences in expression of further detoxification pathway genes including glutathione S-transferases, ATP binding cassette, carboxylesterase and further CYP genes between the SUS, RES_exposed, and RES_unexposed P. ovis isolates (S1–S3 Files). In S. scabiei and T. urticae, glutathione S-transferases of the µ- and ∂-classes have been linked to ML resistance, while a glutathione S-transferase of the ∂-class was capable of conjugation of abamectin [110,111]. The presence of environmentally induced genes offers phenotypic flexibility that makes it possible to adapt to new environmental challenges [112,113]. However, for the development of a potential diagnostic test for resistance, constitutively over-expressed genes would result in less-labour intensive diagnostics. As these have the possibility to be detected in acaricide RES_unexposed populations, while treatment induced over-expressed genes need acaricide exposure for their detection.

4.2.4. No differential expression of GluCls.

Finally, no cys-loop ligand gated ion channel component genes were found among the significant genes, nor were resistance-associated mutations discovered in the amplicon regions examined or whole genome re-sequencing data. This superfamily of arthropod ion-channels is regarded as the main target site of the MLs [31,114]. Although not of the same family of ion channels, in addition to differential expression, functional expression in Xenopus laevis showed that mutations can lower the susceptibility of GluCls for ivermectin [24]. The higher transcription of GluCls in the adult males and protonymphs (Fig 1) could be explained through the increased motility of these stages [40], as the main functions of GluCls in invertebrate nervous systems are control and modulation of locomotion, the regulation of feeding, and the mediation of sensory inputs [22]. However, an explanation for the higher levels in the larval stage is lacking. In lice, GluCls appear to play a role in feeding behaviour [115]. However, the low expression of GluCls in tritonymphs, a major feeding stage, is not in line with this finding [40].

We did, however, observe isoform level differential expression and elevated F_st of one isoform of a GABA transporter gene (PsoOvis1B011106) without associated differential gene expression of this highly expressed gene. This suggests a change in the proportion of isoforms of this gene versus whole gene expression in RES samples. GABA transporter genes belong to three families and help regulate the concentration of extracellular GABA. The ML abamectin is believed to activate GABA transporters which results in excessive GABA release and ultimately death of the target organism [116,117]. The role of differential transcript usage in acaricide, and wider pesticide resistance in arthropods, is little understood currently but results such as ours could form a basis for functional investigations. There was evidence for induced expression of an NMDA-selective glutamate receptor (PsoOvis1B009258) gene.

4.3.1. Gene amplification in subtelomeric regions.

Multiple rounds of gene amplification as we observed here at the tandem UGT gene locus, also explains resistance to the herbicide glyphosate in the parasitic weed, Amaranthus palmeri [118,119]. The 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS) locus is amplified 100-fold as part of a 400 kb long cassette of the EPSPS gene and 58 other genes. This EPSPS encoding cassette of genes occurs as extrachromosomal circular DNA and putatively originated from the activity of a transposase also encoded by the cassette. The EPSPS locus is also duplicated to a lesser degree in other weed species that have developed glyphosate resistance including Kochia scoparia [120,121]. In K. scoparia an initial unequal crossing-over is hypothesised to have led to duplication of EPSPS followed by selection for further unequal crossing-over events. Such a scenario could also explain our observation for the UGT locus as ONT long-reads indicate only the two UGT genes have undergone gene copy number amplification. The two P. ovis UGT genes are located close to the start of Chromosome 7 in the P. ovis genome (both sheep- and cattle-derived versions) at positions 159,563–161,298 kb (P. ovis cattle-derived genome), potentially in the subtelomeric region. Subtelomeric regions are enriched for repetitive elements, which result in genomic instability and increases the potential for unequal crossovers [122–124]. In the common bean, Phaseolus vulgaris, pathogen resistance (R) genes located in subtelomeric regions are highly-duplicated as a result of their location with associated resistance phenotypes [125]. Subtelomeric regions are variable in length from 10s to 100s of kilobases across organisms [122] and are currently not defined for P. ovis. The two P. ovis UGT genes may therefore occur within or nearby this region of instability which we hypothesise could, at least partially explain the elevated copy numbers of these genes through unequal crossing-over. This remains for confirmation in future studies alongside further investigation of inter-individual variability in UGT copy number and age of the expansion of UGT copy numbers in P. ovis.

4.3.2. Transposable element mediated increased copy number of a CYP gene.

The constitutively over-expressed CYP in our resistant RNASeq (RES_exposed and RES_unexposed) replicates was associated with increased copy number in the six Resistant/Suspected Resistant populations. The amplified region has clean break points and evidence for at least five potential insertion locations of this locus was provided by discordant read mappings (S3 Fig). It was not possible to confirm whether gene translocation events alone explain elevated coverage as several long-reads used to assemble the cattle-derived P. ovis genome indicated gene duplication in-situ may also have occurred at this locus. A relationship between transposable element (TE) activity and detoxification pathway gene overexpression has been observed in many arthropods [126,127]. The specific modes of TE action, especially as expression of duplicated genes is not simply doubled [128] remain to be elucidated but may involve loss or gain of promoter repressor and enhancement elements [127]. Several fixed synonymous positions in Resistant/Suspected Resistant and stop codon in adjacent PsoOvis1B003189 gene were identified populations which were not present in the Susceptible population. If a causal relationship between this gene and ML resistance is subsequently demonstrated in functional genomic experiments, these variants could form the basis of a simple PCR test for CYP-mediated ML resistance.

The putative transposable element encoding gene PsoOvis1B002032 which had zero expression in SUS RNASeq replicates but very high expression in all six RES replicates, had low genome sequencing coverage in the sheep-derived genome assembly and is not present in the cattle-derived genome gene set. This may reflect the difficulty in assembling active TEs correctly, even with the excellent genomic resources available to us. Its striking expression dynamics remain an interesting observation, albeit restricted to the RNASeq component of our analyses.