High transcriptomic differentiation between parental isolates with admixture in the genetic cross

We generated RNA-seq data from male and female worms from both parental isolates (MHco3(ISE); drug susceptible reference isolate, referred to from here on as ‘MHco3’ and MHco18(UGA2004); multi-drug resistant and referred to from here on as ‘MHco18’) and from the F2 generation of the genetic cross after ivermectin treatment (F2IVM) and without treatment (F2CTL) (S1 Fig). We also collected samples from the F2 after treatment with benzimidazole (F2BZ) for comparison. An average of 72 million (SD = 27.5 million) 150 bp paired-end reads were generated per sample (S2A Fig). Between 75 and 82% of MHco18 and MHco3 reads mapped to the reference (MHco3) genome, respectively, with the F2 samples showing intermediate mapping percentages (S2B Fig). This is consistent with the expected levels of polymorphism in each population relative to MHco3, with lower mapping percentages as populations diverge from the reference isolate [18]. Transcriptomic differentiation between isolates, treatment groups, and replicates was assessed using principal component analysis (PCA) and Poisson distance (Fig 1). Samples clustered well by isolate or treatment group for all male samples (Fig 1A and 1C); as expected, the parental isolates were highly differentiated (1A, PC1: 38% variance) and the genetic cross F2 samples were an admixture of the two. For the female samples (Fig 1B and 1D), the parental isolates were again well differentiated, but there was greater variance among all samples (1B, PC1: 53%) with less defined clustering of replicates and treatment groups. Differences between replicates broadly corresponded to the sample processing batch (highlighted by datapoint symbol in 1B and sample name in 1D); this batch effect was included as a secondary factor in the DESeq2 analysis. The female F2CTL group was an outlier in the dataset (S3 Fig), so was excluded from analysis. The specific cause of this variance is unknown: the cluster of genes (n = 129) that differentiated female F2CTL samples and the first replicate of all female samples were mostly uncharacterised; 16 had C. elegans homologues and in this subset there was no enrichment for expression in a particular tissue, or for any GO term.

Download:

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

Fig 1.

PCA plots (A. males, B. females) with regularised log transformed data, based on the 500 most variable genes, and heatmaps (C. males, D. females) showing Poisson distance between samples. MHco3 = drug-susceptible parent, MHco18 = drug-resistant parent, F2IVM = ivermectin-treated genetic cross F2, F2BZ = benzimidazole-treated F2, F2CTL = untreated control F2. The datapoint symbol in the PCA and the sample number in heatmap relate to the order that the samples were processed (circle = 1st, triangle = 2nd, square = 3rd).

https://doi.org/10.1371/journal.ppat.1010545.g001

Separating transcriptomic differences associated with drug resistance from other between-isolate variation

To investigate whether ivermectin resistance was associated with differential gene expression, we undertook multiple pairwise comparisons (S1 and S4 Figs). We predicted that the MHco3 and MHco18 isolates would have many constitutive differences in gene expression reflecting genetic diversity resulting from their distinct evolutionary histories, including differences in their sensitivity to the benzimidazoles, levamisole and ivermectin. However, we predicted that strain-specific and drug-specific differences could be separated using the genetic cross: the subset of genes specifically associated with resistance to ivermectin would be differentially expressed in the ivermectin selected genetic cross relative to the unselected genetic cross, but strain-specific differences would not. Consistent with these expectations, the largest number of differentially expressed genes that were unique to a pairwise comparison was between parental isolates (n = 1481 and 1290, males and females respectively, S4 Fig). Genes that were only differentially expressed in the parental isolates (Fig 2, grey points) were essentially ruled out as playing a role in ivermectin resistance mediated by gene expression because they were not also differentially expressed in the ivermectin selected genetic cross (Fig 2, blue points), with the caveat that our analysis was by choice conservative being based on a stringent cut off (adj P < 0.01) for every pairwise comparison. Using this approach, we identified differentially expressed genes that were specifically associated with ivermectin resistance.

Download:

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

Fig 2.

Scatter plot showing differentially expressed genes in comparisons of resistant and susceptible populations (adj P < 0.01), with log2 fold change of males shown on the x-axis and females on the y-axis. Grey points represent genes that are differentially expressed only in parental (MHco18 vs MHco3) comparisons, i.e. likely to represent between-strain differences not associated with ivermectin resistance. If blue they are also differentially expressed in the F2IVM vs MHco3 and F2IVM vs F2CTL comparisons, i.e. likely to be specifically associated with ivermectin resistance. The most highly upregulated gene in resistant isolates is the H. contortus homologue of C. elegans cky-1 (HCON_00155390). Equivalent scatterplots for all pairwise comparisons are shown in S5 Fig.

https://doi.org/10.1371/journal.ppat.1010545.g002

Differentially expressed genes are enriched within the ivermectin QTL on chromosome V after ivermectin selection in the genetic cross

We have previously identified a major effect QTL on chromosome V in response to ivermectin selection in laboratory strains and in field samples of H. contortus [17,18]. We inspected the distribution of differentially expressed genes at this locus and throughout the genome for each pairwise comparison (S6 Fig). These plots highlighted the large differences in gene expression between parental isolates (S6, A and B) and a reduction in the degree of transcriptomic differentiation in comparisons using the ivermectin treated genetic cross (S6, C, D and E). In comparisons between parental isolates, differentially expressed genes were present throughout the genome with no enrichment for differential expression in the ivermectin resistance locus centred around 37.5 Mb on chromosome V (X² = 2.62, P = 0.11). However, the subset of these genes that were also differentially expressed in all comparisons with the genetic cross post ivermectin selection (F2IVM) showed significant enrichment for genes located in the ivermectin resistance locus (X² = 37.76, P = 7.23E-10).

Upregulation of a single gene within the ivermectin QTL in all resistant isolates

Seventy-six genes were up- or downregulated in all male and female ivermectin resistant populations (Figs 2 and S1B and S1 Table). Sixty of the differentially expressed genes had homologues in C. elegans (S1 Table) but this gene set showed no tissue, phenotype or GO term enrichment from C. elegans data. The gene with highest upregulation, HCON_00155390, a homologue of C. elegans cky-1, lies within the major chromosome V locus under ivermectin selection in H. contortus (Fig 3). As shown in S1 Table, HCON_00155390:cky-1 is expressed between 62 and 132-fold higher (log2 fold changes from 5.95 to 7.04, adj P ≤ 0.00024) in males and females of all resistant populations relative to the susceptible MHco3 isolate and 2.7-fold higher (1.44 log2 fold change, adj P = 0.0019) in males of the genetic cross after ivermectin selection (F2IVM = resistant genotypes only) relative to the unselected population (F2CTL = mix of resistant and susceptible genotypes). In the latter comparison the control population is an admixture of resistant and susceptible genotypes so the relative difference is expected to be lower than in the former comparisons with the fully susceptible MHco3 control. Constitutive upregulation of HCON_00155390:cky-1 was confirmed in MHco18 males relative to MHco3 males with RT-qPCR [18]. Constitutive upregulation in MHco18 females was also detected relative to MHco3 females, but the fold-change in expression could not be calculated due to a lack of measurable expression in MHco3 females. In C. elegans, cky-1 encodes a transcription factor expressed in the embryo and in non-neuronal pharyngeal cells [19].

Download:

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

Fig 3.

Chromosome V karyoplots showing genomic position of genes with significant upregulation (yellow) or downregulation (blue). Each point represents a differentially expressed gene (adj P < 0.01) and point size corresponds to significance. Panels A-D compare ivermectin-resistant versus ivermectin-susceptible populations (A. MHco18 vs MHco3 males, B. MHco18 vs MHco3 females, C. F2IVM vs MHco3 males, D. F2IVM vs MHco3 females). Panel E compares an ivermectin-resistant versus a mixed (resistant and susceptible) population (F2IVM vs F2CTL males). Panel F shows genetic differentiation (F_ST) between the F3 generation of the genetic cross pre- and post- ivermectin selection to highlight the major locus of ivermectin selection [18]. Panel G shows relative absence of differential expression at the ivermectin QTL in a comparison with a benzimidazole selected population (F2BZ vs F2CTL males). Upregulation of the H. contortus homologue of C. elegans cky-1 (HCON_00155390) is identified in ivermectin resistant populations only.

https://doi.org/10.1371/journal.ppat.1010545.g003

One of the most highly downregulated genes also lies within the major locus under ivermectin selection: HCON_00155240, a homologue of C. elegans F59B1.10. In C. elegans, this gene encodes a checkpoint (CHK) kinase-like protein with an uncharacterised oxidoreductase Dhs-27 domain, expressed in GABAnergic neurons, dopaminergic neurons and muscle cells. One caveat to our measurement of expression for this gene is its high level of polymorphism in the MHco18 isolate, which could give the appearance of downregulation due to a greater proportion of unmapped reads relative to MHco3 (discussed in more detail below). Outside of the ivermectin QTL, downregulated genes included seven homologues of C. elegans vap-1, a cysteine rich secretory protein expressed in the amphid sheath cell.

Notably, no putative ivermectin target or resistance genes from the literature [7,8,11,20] were identified as differentially expressed in all resistant populations; normalised read counts for candidate genes are shown in S7 (males) and S8 (females) Figs. Genes that are differentially expressed in the parental isolates only are unlikely to be related to ivermectin resistance and none of these candidate genes lie within genomic loci under ivermectin selection [17,18].

Male and female worms show shared and sex-specific differences in gene expression associated with ivermectin resistance

We were concerned that the higher between-replicate variability of the female data could limit the detection of truly differentially expressed genes in the male analysis when results from both datasets were combined. It is also possible that ivermectin affects male and female worms differently [21,22]. For these reasons, we also analysed results from the male and female datasets separately (Figs 4 and S1B). None of the genes identified in these sex-specific analyses were within the major QTL. In the male worms, there were an additional 67 upregulated and 107 downregulated genes that were not differentially expressed in all pairwise comparisons of ivermectin resistant female worms (S2 Table). There was low but significant upregulation of two putative multi-drug resistance proteins associated with ivermectin resistance in the literature: HCON_00162780 and HCON_00175410, the homologues of C. elegans pgp-11 and mrp-4, respectively. PGP-11 is a P-glycoprotein known to modulate ivermectin sensitivity in other nematode species [23], and the pgp-11 locus has previously been associated with ivermectin resistance in H. contortus [16, 17, 20]. In this study, the expression of HCON_00162780:pgp-11 was markedly higher in males relative to females (Fig 5A and 5B), and upregulation in MHco18 versus MHco3 males was confirmed with RT-qPCR (Fig 5C). In C. elegans, mrp-4 is involved in lipid transport and lipid storage, but has also been shown to have higher constitutive expression in an ivermectin resistant strain of C. elegans relative to wild-type [24]. One putative ligand gated ion channel, HCON_00137455, a homologue of lgc-25, was downregulated, as was HCON_00078660, a homologue of the sodium leak ion channel nca-2. In C. elegans, nca-2 is expressed in acetylcholine and GABA motor neurons [25], and is involved in the propagation of neuronal activity from cell bodies to synapses [26] and in synaptic vesicle recycling [25]. There was upregulation of one chloride channel, HCON_00108990, the homologue of best-19. In C. elegans, this is a bestrophin chloride channel expressed in the NSM, a pharyngeal neurosecretory motor neuron.

Download:

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

Fig 4.

Volcano plots showing differentially expressed genes in male (A) and female (B) samples (adj P < 0.01). Blue points represent genes that are only differentially expressed in the parental (MHco18 vs MHco3) comparisons. If purple, they are also differentially expressed in the F2IVM vs MHco3 comparisons. If yellow (males only), they are also differentially expressed in the F2IVM vs F2CTL comparison.

https://doi.org/10.1371/journal.ppat.1010545.g004

Download:

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

Fig 5. Expression analysis of HCON_00162780:pgp-11.

Normalised read counts for male (A) and female (B) RNA-seq samples highlighting higher constitutive expression in all male populations relative to female populations. C. Comparison of RNA-seq and RT-qPCR data (mean of three biological replicates) for MHco18 males relative to MHco3 males, showing constitutive upregulation in the ivermectin resistant isolate. For RNA-seq data, error bars show the log fold change standard error (lfcSE) and for RT-qPCRs, error bars show the standard deviation.

https://doi.org/10.1371/journal.ppat.1010545.g005

Of the differentially expressed genes in ivermectin resistant males, 125 had C. elegans homologues (S2 Table). These homologues were enriched for expression in the ALA neuron (E = 0.24, O = 4, Q = 0.0011) (S9 Fig), a mechanosensory interneuron that projects into the nerve ring then joins the lateral cords, extending as far as the tail, adjacent to the excretory canals [27]. The ALA neuron inhibits locomotion and pharyngeal pumping during lethargus and ‘stress induced sleep’, a characteristic behaviour in C. elegans during recovery from exposure to damaging conditions [28, 29]. There was GO enrichment for a single term ‘neuropeptide signalling pathway’ (E = 0.92, O = 6, Q = 0.013), associated with nlp-9, nlp-14, flp-6, flp-7, flp-16 and flp-24, all of which were downregulated. Notably, there was also phenotype enrichment for 13 terms associated with movement (S9 Fig).

For female worms, the lack of an F2CTL sample meant the equivalent analysis (excluding differentially expressed genes in all resistant male comparisons) would be very likely to be confounded by differences between the female parental isolates that were not associated with ivermectin resistance (Fig 4B). To mitigate this, an alternative analysis including previously published RNA-seq data for females from two additional ivermectin resistant populations (MHco4(WRS) and MHco10(CAVR) versus MHco3 [30]) was undertaken (S1B Fig). This identified 85 upregulated and 97 downregulated genes in all ivermectin resistant female samples that were not differentially expressed in ivermectin resistant male samples (S3 Table). Downregulation of HCON_00076370, a homologue of the C. elegans sodium:potassium:chloride symporter nkcc-1, was identified in this analysis. In C. elegans, nkcc-1 is a chloride transporter and functions as a developmental switch that changes the response of GABA_A receptors in body wall muscles from depolarising in the early L1 to hyperpolarising from the mid-L1 stage onwards by controlling the cellular chloride gradient [31]. In female worms only, there was also downregulation of one putative ligand gated ion channel, HCON_00097440, a homologue of acr-24. Of the differentially expressed genes in ivermectin resistant females, 158 had C. elegans homologues. These showed no tissue or GO term enrichment but there was enrichment for a single phenotype ‘uptake by intestinal cell defective’ (E = 0.59, O = 4, Q = 0.075).

Transcriptomic response to ivermectin exposure implicates neuronal regeneration, neuropeptide signalling and chloride homeostasis

Having identified genes with constitutive differential expression in ivermectin resistant worms, we were also interested in genes that were only up or downregulated after exposure to ivermectin treatment. This subset of genes would be expected to be differentially expressed in the F2IVM versus F2CTL comparison and the F2IVM versus MHco3 comparisons, but not the MHco18 versus MHco3 comparisons, because the MHco18 isolate had not been exposed to ivermectin. As ivermectin has a short half-life in sheep (plasma t_1/2 = 2.7 days [32]) and the F2IVM worms were isolated 21 days after ivermectin treatment, this gene set will not reflect acute ivermectin exposure but rather the longer-term influence of drug treatment.

Of 120 genes that were differentially expressed in response to ivermectin treatment, 95 had C. elegans homologues (S4 Table); none of these genes were within the major QTL. The most highly upregulated gene with a C. elegans homologue was HCON_00068595, which was 5 to 32-fold more highly expressed in ivermectin treated populations (2.31–4.99 log2 fold change, adj P ≤ 2.52E-12 (S4 Table)). This gene is a homologue of emb-9, a collagen type IV alpha 5 chain, involved in many processes such as pharyngeal pumping and egg laying, but with a critical role in positive regulation of neuron regeneration following axon damage [33]. emb-9 is most strongly expressed in the basement membrane of the C. elegans pharynx [34].

There was no differential expression of any ligand gated ion channels after ivermectin exposure, but HCON_00068600, the homologue of a neuronal CLC-2 chloride channel, clh-3, was upregulated between 2 and 12-fold (1.28–3.62 log2 fold change, P ≤ 2.95E-10 (S4 Table)) in ivermectin treated populations. CLC-2 is activated by membrane hyperpolarisation and is thought to regulate neuronal excitability by providing an efflux pathway for chloride. In C. elegans, clh-3 is only expressed in the (spontaneously active) hermaphrodite-specific motor neuron (HSN), where clh-3 inhibits excitability by mediating chloride influx, resulting in suppression of egg laying [35]. In H. contortus, HCON_00068600:clh-3 is also constitutively expressed in both male parental samples, suggesting expression is not limited to the HSN in this species. Interestingly, HCON_00068595:emb-9 and HCON_00068600:clh-3 are neighbouring genes on chromosome III, although encoded on opposite strands. In C. elegans, emb-9 and clh-3 are encoded on different chromosomes (chromosome III and II, respectively) and have no known interaction.

For the C. elegans homologues, the most significant GO term was ‘neuropeptide signalling pathway’ (E = 0.71, O = 8, Q = 6e-06) (S9 Fig). The eight genes associated with this term encoded five neuropeptide-like proteins (nlp-7, nlp-11, nlp-13, nlp-49 and nlp-50) and three FMRF-like peptides (flp-1, flp-14 and flp-22); all were downregulated in F2IVM samples. In common with the constitutive datasets, multiple homologues of C. elegans vap-1 (15 in total) were downregulated.

Differential expression of genes associated with neuronal plasticity and chloride homeostasis are specific to ivermectin selection

To investigate the specificity of the transcriptomic differences we identified in the ivermectin-selected genetic cross and to explore the possibility of a common response to drug exposure (for example, a stress response or multi-drug resistance pathway), we also undertook RNA-seq analysis of the same F2 population after benzimidazole selection. There were fewer differentially expressed genes in the benzimidazole resistant populations relative to the equivalent comparisons for ivermectin (59 associated with benzimidazole resistance (S5 Table) versus 76 for ivermectin, and 55 in response to benzimidazole treatment (S6 Table) versus 120 for ivermectin). The benzimidazole resistant populations showed no differential expression of genes known to confer resistance through non-synonymous mutations in the target protein, β-tubulin (i.e., β-tubulin isotype 1 (HCON_00005260) or β-tubulin isotype 2 (HCON_00043670)) or genes associated with drug metabolism (UDP-glycotransferases or cytochrome P450 enzymes).

Notably, both the ivermectin and benzimidazole-selected populations appeared to be transcriptionally very similar (Fig 1), despite different regions of the genome being under selection by each drug, with no evidence of a shared genetic component to resistance [18]. In the benzimidazole resistant and response to treatment groups, 35 genes in each (59% and 65% of differentially expressed genes, respectively) were shared with the equivalent ivermectin selected populations. In the benzimidazole resistant populations, as for the ivermectin resistant populations, there was downregulation of multiple homologues of C. elegans vap-1 (five downregulated constitutively and 11 in response to benzimidazole treatment) and four of the same neuropeptides were downregulated in response to benzimidazole and ivermectin treatment (homologues of nlp-13, flp-1, flp-14 and flp-22). In male worms only, upregulation of one putative multi-drug resistant gene, HCON_00162780:pgp-11, was identified in all pairwise comparisons for both ivermectin and benzimidazole resistant populations.

However, in contrast with the ivermectin resistant populations, there was no enrichment for differential expression at the chromosome V locus (Fig 3, bottom panel) and no differential expression of HCON_00155390:cky-1 or any of the genes associated with neuronal regeneration and chloride homeostasis. This suggests the differential expression of these genes, identified using the same genetic cross, are unique to ivermectin resistance and response to treatment.

Use of a genetic cross minimises the impact of between-isolate polymorphism on differential expression analysis

Previous work highlighted the potential impact of high levels of polymorphism on between-isolate RNA-seq analysis in H. contortus [30], showing that polymorphic isolates (relative to the reference MHco3 isolate) had significantly fewer reads mapping to the genome, potentially biasing differential expression analysis. In this study, the difference in ‘SNP rate’ (the number of SNPs in a gene model; see Materials and Methods) between parental isolates had a small but significant negative correlation with gene expression (R² = 0.074 [P < 0.001] and R² = 0.099 [P < 0.001] for males and females, respectively). This correlation suggests that highly polymorphic genes were more likely to be identified as downregulated than upregulated in the parental isolates (S10 Fig). This effect was more striking for the MHco18 isolate (upper left quadrant) but could be seen to a lesser extent for polymorphic genes in the MHco3 isolate (lower right quadrant). Admixture of the parental genotypes in the genetic cross largely mitigated this bias in male analyses (R² = 0.0026 [P = 0.322]) (S10 Fig, dark blue points), although the lack of a female F2CTL limited our ability to control for polymorphism in the female-only analyses (R² = 0.074 [P < 0.001]).

The positive impact of the cross in the male analyses was further supported by a reduction in the proportion of highly polymorphic genes (defined as having a SNP rate of greater than 2% in MHco18) in the downregulated genes identified in the F2IVM versus F2CTL analysis compared to the MHco18 versus MHco3 analysis. Notably, 21% (469/2196) of genes identified as downregulated in MHco18 males were highly polymorphic, relative to 2.8% (58/2069) of upregulated genes, whereas in the genetic cross 6.2% (60/973) of genes identified as downregulated in F2IVM males were highly polymorphic, relative to 5.8% (71/1219) of upregulated genes. Finally, the relevance of controlling for polymorphism was highlighted by the subset of genes that were identified as the most highly polymorphic in the parental isolates: C. elegans homologues of the 1684 genes with a greater than 2% difference in polymorphism rate in the MHco18 isolate relative to the MHco3, were enriched for the phenotype ‘anthelmintic response variant’ (E = 3.2, O = 10, Q = 0.058). This was associated with highly polymorphic homologues of avr-15, acr-16, lev-1, osm-1, che-1, che-3, che-11, che-14, daf-10 and dyf-3. Although not identified in the enrichment analysis, a homologue of ben-1, HCON_00043670 (β-tubulin isotype 2), was also highly polymorphic. Overall, these findings highlight the need to control for polymorphism in transcriptomic analyses of genetically divergent populations and show that genes associated with drug response are among the most polymorphic genes in a resistant population.

Ethics statement

All experimental procedures were examined and approved by the Moredun Research Institute Experiments and Ethics Committee and were conducted under approved UK Home Office licenses following the Animals (Scientific Procedures) Act of 1986. The Home Office licence number is PPL 60/03899 and the experimental code identifier is E46/11.

Generation of parasite material

Fifteen donor sheep were treated with 2.5 mg/kg bodyweight monepantel (Zolvix Oral Solution for Sheep, Elanco AH) 14 days prior to infection to ensure they were parasite free. Each donor was orally infected with 5000 L₃ of a single H. contortus isolate; three donors were infected with the drug susceptible MHco3(ISE) isolate [59], three donors were infected with the triple resistant MHco18(UGA2004) isolate (benzimidazole, levamisole and ivermectin resistant) [7] and nine donors were infected with the F2 generation of a genetic cross between MHco3 females and MHco18 males [15]. For the genetic cross F2 infections, on day 35 post-infection (pi), three donor sheep were treated with 0.2 mg/kg body weight ivermectin (Oramec Drench, Boehringer Ingelheim Animal Health UK Ltd) (F2IVM), three were treated with 7.5 mg/kg fenbendazole (Panacur 10% Oral Suspension, MSD Animal Health) (F2BZ), and three were left untreated (F2CTL). Adult worms were harvested at post mortem, sexed and snap frozen in liquid nitrogen for storage at -80°C in batches of 20 worms. Post mortem dates differed for each isolate: day 62 pi for MHco3, day 28 pi for MHco18 and day 56 pi for the F2 infections.

RNA isolation and library preparation for RNA-seq

Adult worms from one donor each for the MHco3, MHco18, F2IVM, F2BZ and F2CTL populations were used for RNA-seq. Total RNA was isolated from three batches of 20 male or female worms per donor using a standard Trizol (Thermo Fisher Scientific, 15596026) extraction. Each RNA sample underwent DNaseI (Qiagen, 79254) treatment in solution followed by purification on a RNeasy Mini Spin Column (Qiagen, 74104), before freezing at -80°C. Total RNA (1 μg) from each sample was used for NEBNext PolyA selection and Ultra Directional RNA Library preparation (Illumina, E7490, E7420). Prepared libraries were sequenced using 2×150 bp paired-end read chemistry on an Illumina HiSeq 4000 platform.

Messenger RNA data processing

Raw Fastq files were trimmed for the presence of adapters using Cutadapt v1.2.1 [60]. Option -O 3 was used to trim the 3’ end of any reads matching adapter sequence for 3 bp or more. The reads were further trimmed using Sickle v1.200 [61] with a minimum window quality score of 20. Reads shorter than 20 bp after trimming were removed. Downstream analyses were undertaken in Galaxy [62] and R Studio v1.1.456 [63]. Sequence quality was assessed with FastQC [64] and MultiQC [65]. Reads were aligned to the H. contortus V4 genome (haemonchus_contortus_MHCO3ISE_4.0 [14]) with HISAT2 v2.1.0 using default settings [38]. Counts of reads mapping to gene models (haemonchus_contortus.PREJEB506.WBPS15.annotations.gff3) were determined with featureCounts in Subread v1.6.2 [66] with the -O option to allow reads to contribute to multiple features. Differential expression analysis was undertaken with DESeq2 v1.22.2 [67] with alpha set to 0.01.

Gene enrichment analysis

Putative C. elegans homologues were identified with BLASTP (C. elegans WS264 proteins, e-value < 0.001) and tissue, phenotype and gene ontology enrichment analysis was performed with WormBase Enrichment Analysis [73], which uses a hypergeometric test and Benjamini-Hochberg correction of false discovery rate (Q < 0.1 for significance).

Data visualisation

Upset plots were generated with UpSetR v1.4.0 [74] and genome-wide expression and F_ST plots were generated with KaryoploteR v1.8.8 [75]. All other plots were generated with ggplot2 v3.2.1 [76].

Real-time quantitative PCR (RT-qPCR)

Total RNA was extracted from 20 male or female worms from each of three different donor sheep per isolate (MHco3 and MHco18). Total RNA (1 μg) was used for oligo(dT) primed cDNA synthesis (SuperScript III First-Strand Synthesis System, ThermoFisher, 18080051), with no-reverse transcriptase controls included for each sample. cDNA was diluted 1:10 for RT-qPCR (other than for HCON_00155390:cky-1 in female samples, where neat cDNA was used due to the low abundance of this transcript) and 1 μl added to each reaction. RT-qPCR was undertaken following the Brilliant III Ultra Fast SYBR QPCR Master Mix protocol (Agilent Technologies, 600882) and results were analysed with MxPro v4.10. Gene expression was normalised to β-actin (HCON_00135080). Primer sequences are listed in S7 Table.