Transcriptional variation and divergence of host-finding behaviour in Steinernema carpocapsae infective juveniles

Background Steinernema carpocapsae is an entomopathogenic nematode that employs nictation and jumping behaviours to find potential insect hosts. Here we aimed to investigate the transcriptional basis of variant host-finding behaviours in the infective juvenile (IJ) stage of three S. carpocapsae strains (ALL, Breton and UK1), with a focus on neuronal genes known to influence behaviour in other nematode species. Identifying gene expression changes that correlate with variant host-finding behaviours will further our understanding of nematode biology. Results RNA-seq analysis revealed that whilst up to 28% of the S. carpocapsae transcriptome was differentially expressed (P < 0.0001) between strains, remarkably few of the most highly differentially expressed genes (> 2 log2 fold change, P < 0.0001) were from neuronal gene families. S. carpocapsae Breton displays increased chemotaxis toward the laboratory host Galleria mellonella, relative to the other strains. This correlates with the up-regulation of four srsx chemosensory GPCR genes, and a sodium transporter gene, asic-2, relative to both ALL and UK1 strains. The UK1 strain exhibits a decreased nictation phenotype relative to ALL and Breton strains, which correlates with co-ordinate up-regulation of neuropeptide like protein 36 (nlp-36), and down-regulation of an srt family GPCR gene, and a distinct asic-2-like sodium channel paralogue. To further investigate the link between transcriptional regulation and behavioural variation, we sequenced microRNAs across IJs of each strain. We have identified 283 high confidence microRNA genes, yielding 321 predicted mature microRNAs in S. carpocapsae, and find that up to 36% of microRNAs are differentially expressed (P < 0.0001) between strains. Many of the most highly differentially expressed microRNAs (> 2 log2 fold, P < 0.0001) are predicted to regulate a variety of neuronal genes that may contribute to variant host-finding behaviours. We have also found evidence for differential gene isoform usage between strains, which alters predicted microRNA interactions, and could contribute to the diversification of behaviour. Conclusions These data provide insight to the transcriptional basis of behavioural variation in S. carpocapsae, supporting efforts to understand the molecular basis of complex behaviours in nematodes.


Background
Parasitic nematodes employ a variety of behaviours that maximise opportunity for host contact and invasion. These behaviours vary across species, ranging from passive reliance on host ingestion, through pro-active host-finding by migration, nictation, and even jumping [1][2][3]. Each of these behavioural strategies rely on the incorporation of multiple sensory inputs, spanning chemosensory, olfactory, mechanosensory, thermosensory and hygrosensory circuits [2,4]. Nematode host-finding strategies are also remarkably plastic, varying in response to experience and environment [5]. Despite the obvious importance of parasite host-finding behaviour to medical, veterinary and agricultural interests, we know relatively little about the genes involved in regulating these behaviours. It may be possible to develop new approaches to parasite control by targeting components of the parasite host-finding apparatus.
The evident complexity of nematode behaviour belies their relative neuroanatomical simplicity. It is thought that the neurochemical complexity of nematodes is central to their diverse behavioural capacity and adaptability [6]; neuropeptides in particular are highly enriched and conserved between nematode species with diverse life history traits [7]. It has also been demonstrated that FMRFamide-like peptide (flp) genes are co-ordinately up-regulated in host-finding stages of diverse parasitic nematodes [8], and that they contribute to various host-finding behaviours [8,9], along with insulin-like peptides [10] and neuropeptide-like proteins [11]. Diverse neuronal gene families also contribute to the surprising behavioural enrichment of such simple organisms [12]. Variation in gene transcript abundance and sequence identity is central to the phenotypic plasticity of cells, tissues and organisms, underpinning behavioural variation.
Small non-coding RNAs also contribute to gene regulation, as a factor of developmental stage, and in response to the environment. Small RNAs have been implicated in driving phenotypic novelty and adaptation within and between species [13][14][15][16]. MicroRNAs are small non-coding RNAs that negatively regulate target gene expression across higher organisms [17], and have been shown to modulate neuronal connectivity, synaptic remodelling [18] and memory within the olfactory system of other invertebrates [19]. In vivo cross-linking and immunoprecipitation of microRNA-specific argonaute proteins coupled with sequencing of bound mRNA transcripts in Caenorhabditis elegans, demonstrates an enrichment of neuronal gene families. This confirms that many neuronal genes with known involvement in behaviour are biologically relevant microRNA targets [20]. Key to understanding the contribution of neuronal gene function and microRNA regulation to host-finding behaviour in parasitic nematodes is the development of a suitable model system through provision of foundational datasets.
Steinernema spp. nematodes are obligate entomopathogens that invade and kill insect hosts through coordinated action with commensal Xenorhabdus bacteria [21]. Steinernema infective juveniles (IJs) display qualitatively different host-finding strategies between species [22], representing a unique resource for the comparative analysis of behaviour. Steinernema carpocapsae is generally considered to employ an 'ambushing' strategy, characterised by nictation and jumping behaviours. Nictation is enacted by nematodes that stand upright, waving their anterior in the air [23]. During nictation, the nematode can respond to sensory stimuli in one of three ways: (i) it can cease nictation and transition to a migratory phase; (ii) it can engage in a torpid 'standing' phenotype that may enhance opportunity for host attachment; and (iii) it may jump directionally in response to volatile and mechanosensory input [8,23]. Whilst the jumping behaviour is thought to be unique to a small number of Steinernema spp., nictation is shared amongst many economically important animal parasitic nematodes, alongside the model nematode C. elegans, for which nictation represents a long-range phoretic dispersal behaviour that is regulated by IL2 neurons [24]. In this study our aim was to profile the host-finding behaviours of S. carpocapsae strains, and to identify protein-coding and noncoding RNAs that are differentially expressed in strains with variant behaviours. These data will underpin future efforts to understand nematode host-finding behaviour.

S. carpocapsae culture
Steinernema carpocapsae strains (ALL, Breton and UK1) were maintained in Galleria mellonella at 23°C. S. carposapsae strains were gifted by Prof Ali Mortazavi (University of California, Irvine), Prof Nelson Simões (University of the Azores, Portugal), and BASF UK, respectively. IJs were collected by White trap [25] in a solution of Phosphate Buffered Saline (PBS). Freshly emerged IJs were used for each experiment. Individual biological replicates and RNAseq libraries described below were derived from a mixed population of IJs that emerged from multiple G. mellonella cadavers on the same day.

Behavioural assays
Galleria mellonella host-finding assays and dispersal assays were conducted as published [9]; in both instances five biological replicates were assayed across three technical replicates each. A chemosensory index (CI) was calculated following host-finding assays, giving a measure of relative attraction for participating IJs [8]. For the nictation assays, micro-dirt chips were made from a PDMS mould [26], with 3.5% ddH 2 O agar. 20 IJs suspended in 1.5 μl PBS were pipetted onto the micro-dirt chip, under a binocular light microscope. Once the liquid had evaporated and the IJs could move freely, the number of nictating IJs was counted at 1, 2.5 and 5 min intervals. Nictation assays were conducted over five biological replicates, each with five technical replicates of 20 IJs each. Each dataset was assessed by Brown-Forsythe and Bartlett's tests to examine homogeneity of variance between groups. One way ANOVA and Tukey's multiple comparison tests were then used to assess statistically significant differences in mean across experimental groups. Tests were conducted in Graphpad Prism 7.02.

RNA-seq, differential expression and isoform variant analysis
Six biological replicates of each strain were prepared from approximately 10,000 individuals (80 μl packed volume after centrifugation at 2000 rpm for 2 min) each.
Total RNA was extracted from IJs using TRIzol Reagent (Invitrogen) and DNase treated using the Turbo DNase kit (Ambion) following manufacturer's instructions. RNA quantity were assessed by gel electrophoresis and quantified using Qubit RNA BR Assay Kit (Life Technologies) as per manufacturer's instructions. A total of six transcriptome libraries (150 bp, paired end) were prepared for each S. carpocapsae strain, from 1 μg of total RNA each, using the TruSeq RNA Library Prep Kit v2 (Illumina) following manufacturer's instructions. Sequencing was performed on the HiSeq2500 instrument. Fastq files were assessed for quality using the FastQC (v. 0.11.3) package [27]. Adapters, low quality bases, and reads shorter than 36 bp were removed using the Trimmomatic (v. 0.35) package [28]. The trimmed data quality was then re-assessed using the FastQC package. The S. carpocapsae genomic contigs (PRJNA202318.WBPS9) and associated GFF file (PRJNA202318.WBPS9) were downloaded from https:// parasite.wormbase.org/index.html. Annotations were converted to GTF format using Cufflinks (v. 2.2.2.20150701) [29]. High quality reads were then mapped to the S. carpocapsae genome [30] using the STAR (v. 2.5.3a) package [31]. Isoform expression levels were quantified using the RSEM (v. 1.2.19) package [32], and integrated EBseq package [33]. Raw read counts mapping to each gene in each sample were consolidated into a single count table. This process was repeated for each isoform. Downstream analyses was performed using R (v. 3.3.1) [34]. Differential expression of genes was quantified using the DESeq2 (v. 1.14.1) package [35]. Graphics were generated in R using RColorBrewer (v. 1.1-2) [36], gplots (v. 3.0.1) [37], geneplotter (v. 1.52) [38] and pheatmap (v. 1.0.8) [39] packages with custom R scripts. Additionally, we conducted a GO enrichment analysis of differentially expressed genes that correlated with behavioural variation between strains, using g:Profiler (https://biit.cs. ut.ee/gprofiler/gost), using standard parameters.

MicroRNA sequencing, discovery and quantification
Small RNA libraries were generated from the same RNA samples that were used for matched transcriptomes. A total of six small RNA libraries were prepared for each strain, from 1 μg of total RNA each, using the TruSeq Small RNA library Kit (Illumina) following manufacturer's instructions. 50 bp single-end libraries were sequenced on the HiSeq2500 instrument. Fastq files were assessed for quality using the FastQC (v. 0.11.3) package [27]. Adapters, low quality bases, and reads shorter than 13 bp were removed using the Cutadapt package (v. 1.8) [40]. Sequence reads without adapters were also discarded. Reads that passed QC were mapped to the genome sequence of S. carpocapsae, and microRNAs were identified by miRDeep2 (v. 2.0.0.8) [41], using a training set of mature and precursor microRNA sequences downloaded from miRBase (http://www.mirbase.org/). Naming of microRNAs was preferentially aligned with C. elegans, as indicated by miRDeep2 output. Novel S. carpocapsae microRNAs were named and numbered sequentially, taking care to avoid overlap with any C. elegans microRNA. Differentially expressed microRNAs were identified as above, using the DESeq2 package, and were presented using RColorBrewer, gplots, geneplotter and pheatmap packages.

MicroRNA target gene prediction
Three and five prime UnTranslated Regions (UTRs) of computationally predicted S. carpocapsae genes were exported from wormbase parasite [30,42] using the biomart function. Retrieved sequences were converted to fasta format and predicted microRNA binding sites were identified using miRanda [43]. Two separate miRanda analyses were performed, using i) default settings and ii) strict settings that require perfect conservation of seed site sequence complementarity between microRNA and target mRNA. Experimental verification of microRNA target predictions indicate that perfect complementarity between the microRNA seed region and target mRNA provides the highest degree of specificity and sensitivity [44]. However, Argonaute CLIP-seq analyses indicate that around 40% of all microRNA-mRNA interactions lack perfect seed region complementarity [45]; limiting analyses to perfect seed region requirements will lead to a substantial number of false negatives. MiRanda allows analyses that span canonical seed region complementarity, and non-canonical interactions, providing a robust overview of interactions that follow experimentally validated examples [44]. In each instance, we have included information of relative target site predictions using both strict and default target identification approaches.
Annotation of neuropeptide, neurotransmitter, GPCR, innexin and ion channel genes S. carpocapsae neuropeptide gene orthologues were identified via reciprocal BLAST analysis. A list of available C. elegans FLP, NLP and INS pre-propeptide sequences were obtained from Wormbase [46] and used as BLASTp and tBLASTn search strings via the Wormbase ParaSite BLAST server [42] under default settings. The protein sequences for overlapping genes associated with each high scoring pair were then employed as BLASTp search strings against the available C. elegans protein dataset via the Wormbase BLAST server [46]. Where no overlapping gene annotation was available, novel predicted proteins were generated by concatenating high-scoring return sequences to facilitate reciprocation [47]. The top reciprocal BLAST hit from C. elegans was used to assign putative neuropeptide gene names, and subsequent manual comparison of S. carpocapsae neuropeptide primary sequences to established nematode neuropeptide motifs [48] was used to confirm or reassign gene names where appropriate. GPCR, ion channel and innexin genes were exported from Wormbase parasite according to GO term, using the biomart function. All neurotransmitter genes were identified by reciprocal BLAST, as above. Additionally, the identity of each gene represented in heatmaps and tables of this manuscript was confirmed by reciprocal BLAST, as above. In a number of instances, different S. carpocapsae genes reciprocated to the same C. elegans gene. In any such case, a simple numbering system was applied to gene names in order to reflect clustered identity.
22 (0.08%) of the 1509 (5.3%) most highly differentially expressed genes across all pairwise strain comparisons (> 2 log2 fold, P < 0.0001****) were representative of neuronal gene families, based on GO term annotations (Additional file 1). Of these 22 neuronal genes, only eight were observed to share differential expression  Fig. 1b). Specifically, shared up or down-regulation in S. carpocapsae Breton relative to both ALL and UK1 strains would correlate gene differential expression with increased chemotaxis to G. mellonella (Fig. 1a). Conversely, shared up or down-regulation in S. carpocapsae UK1 relative to both Breton and ALL strains would correlate gene differential expression with reduced nictation behaviour (Fig. 1b). In order to explore the transcriptional regulation of neuronal gene families implicated in these behavioural differences, we assessed each gene family represented in the highly differentially expressed category, as defined above. This encompassed G-protein coupled receptors (GPCRs), ion channels, innexins, neuropeptides and neurotransmitter synthesis, degradation and transport genes (Additional file 1).
Although falling below our threshold of shared > 2 log2 fold change across both strain pairwise comparisons, Scinx-7ii demonstrated the largest expression change in the innexin / gap junction gene family, correlating with increased chemotaxis behaviour (down-regulated 1.03 and 2.65 log2 fold in S. carpocapsae Breton relative to UK1 and ALL strains, respectively; Fig. 4c-d; Additional file 2). Two paralogous asic-2 sodium channel genes were found to be differentially expressed, and inversely related to altered chemotaxis behaviour (Sc-asic-2ii; down-regulated 2.3 and 2.7 log2 fold, P < 0.0001**** in S. carpocapsae Breton, relative to ALL and UK1 strains, respectively; Fig. 4e, Additional file 2), and reduced nictation behaviour (Sc-asic-2i; up-regulated 2.9 and 3 log2 fold, P < 0.0001**** in S. carpocapsae UK1, relative to ALL and Breton strains, respectively; Fig. 4f, Additional file 2). carpocapsae UK1 relative to BR and ALL; assessing gene expression patterns that correlate with reduced nictation behaviour. a, b Differential expression analysis of neuropeptide-like protein genes; c, d differential expression of insulin-like peptide genes; adjusted P values are indicated for all log2 fold changes; padj P < 0.05*. P < 0.01**, P < 0.001***, P < 0.0001****. Flp and neurotransmitter gene maps are included in supplemental as there are no differentially expressed genes satisfying a > 2 log2 fold threshold MicroRNA prediction and quantification Following miRDeep2 identification, quality control and manual curation, a total of 283 high confidence micro-RNA genes and 321 predicted mature microRNAs (306 of which are unique within the genome) were identified across a deep analysis (n = 18 libraries) of S. carpocapsae strains (Additional file 3). One hundred fifty of the unique mature microRNA sequences were previously identified in S. carpocapsae Breton [49], with an additional 156 unique mature microRNAs identified within this study. 102 (36%) and 103 (36%) S. carpocapsae Breton microRNAs were differentially expressed (P < 0.0001****) relative to UK1 and ALL strains, respectively. 50 (17.6%) S. carpocapsae UK1 microRNAs were differentially expressed (P < 0.0001****) relative to S. carpocapsae ALL ( Fig. 5; Additional file 3).
Comparatively fewer microRNAs were differentially regulated and correlated with reduced nictation behaviour in S. carpocapsae UK1, with Sc-mir-772 (up-regulated 12.9 and 12.7 log2 fold, P < 0.0001**** relative to Breton and All strains, respectively) and Sc-mir-773 (up-regulated 8.7 and 8.5 log2 fold, P < 0.0001**** relative to Breton and All strains, respectively) representing notable exceptions. A further five microRNAs (Sc-mir-754, Sc-mir-756, Sc-mir-760, Sc-let-7 (See figure on previous page.) Fig. 4 Differential expression analysis of GPCR, innexin and ion channel gene families that correlate with S. carpocapsae strain behaviour. Figures a, c and e represent pairwise comparisons of S. carpocapsae Breton (BR) relative to UK1 and All; assessing shared gene expression patterns that correlate with increased chemotaxis to G. mellonella; Figures b, d and f represent pairwise comparisons of S. carpocapsae UK1 relative to BR and All; assessing gene expression patterns that correlate with reduced nictation behaviour. a, b putative GPCR genes with > 1 log2 fold difference between at least one of the pairwise comparisons; c and d putative innexin genes across all log2 fold changes; e and f putative ion channel genes with > 1 log2 fold difference between at least one of the pairwise comparisons. P values are indicated for all log2 fold changes; padj P < 0.05*. P < 0.01**, P < 0.001***, P < 0.0001**** and Sc-mir-84-5pi) were likewise differentially expressed and correlated with reduced nictation behaviour (> 2 log2 fold, P < 0.0001***) ( Fig. 6b; Additional files 3 and 4).

MicroRNA target analysis
In silico microRNA target prediction through miRanda suggests a substantial bias towards interactions with gene 5'UTRs. A total of 231,120 microRNA interactions, representing 247,237 binding sites are predicted for the 5'UTR of S. carpocapsae genes, relative to 127,796 microRNA interactions, representing 132,309 binding site predictions within the 3'UTR of S. carpocapsae genes. Through screening each of the most differentially expressed (> 2 log2 fold change, P < 0.0001****) microRNAs that correlate with behavioural variants across pairwise comparisons ( Fig. 6; Additional file 4), we identify a substantial number of predicted interactions with neuronal gene families considered in this study (Table 1). These datasets demonstrate potential cooperation between microRNAs that are predicted to interact with shared target genes. For example, the ion channel gene Sc-asic-2ii is a predicted target for Sc-mir-147, Sc-mir-301-3p, and Sc-mir-316, all highly differentially expressed, and correlated with increased chemotaxis behaviour (Table 1; Fig. 5a). Similarly, the let-7 family members, Sc-mir-84-5pi and Sc-let-7 co-ordinately target two ins-1 paralogues; both microRNAs are highly differentially expressed, and correlated with reduced nictation behaviour in S. carpocapsae UK1 (Fig. 5b). Our data also suggest that different microRNAs are predicted to interact with and converge on a number of shared neuronal gene targets considered here (Table 1). For example, Sc-mir-301-3p is predicted to simultaneously target the twk-12, egas-1 and asic-2 ion channel genes, alongside the nlp-39 neuropeptide-like protein gene, correlating with increased chemotaxis behaviour (Table 1; Figs. 5a and 1a). Mi-Randa microRNA target predictions across both strict and default settings for global 5′ and 3'UTRs are presented in Additional files 5, 6, 7 and 8.

Differential isoform usage between strains
Two genes implicated in the synthesis and transport of classical neurotransmitters were found to exhibit statistically significant differences in isoform preference across S. carpocapsae strains following RSEM analysis (Figs. 6, 7 and 8; Additional file 9). Interestingly, these isoform variants exhibit altered UTR sequences in Fig. 6 Differential expression analysis of microRNAs that correlate with S. carpocapsae strain behaviour. Heatmaps showing differential expression of microRNAs with at least one pairwise expression difference of > 2 log2 fold change, P < 0.0001****. a Differentially expressed microRNAs in S. carpocapsae Breton, relative to ALL and UK1 strains; assessing shared expression patterns that correlate with increased chemotaxis behaviour. b Differentially expressed microRNAs in S. carpocapsae UK1, relative to ALL and Breton strains; assessing shared expression patterns that correlate with reduced nictation behaviour (Additional file 4) Table 1 Predicted gene targets for differentially expressed microRNAs (> 2 log2 fold, P < 0.0001) that correlate with S. carpocapsae strain behaviour. Differentially expressed S. carpocapsae microRNAs were assessed for binding sites across 5′ and 3'UTRs of all ion channel, innexin, GPCR, neurotransmitter and neuropeptide genes. Ce target refers to direct 1 to 1 gene orthologues that are biochemically confirmed microRNA targets in C. elegans [20]. U refers to the number of predicted interacting microRNAs following default microRNA target prediction; S refers to the number of predicted interacting microRNAs following strict microRNA target prediction. Within the implicated behaviour column, "Both (inverse int.)" refers to involvement in both behaviours, through polarised regulation states (significantly upregulated for one, and significantly downregulated for the other)

Discussion
This study has generated high quality transcriptomes (n = 18) and small RNA libraries (n = 18) for the IJ stage of three S. carpocapsae strains, and has catalogued a number of protein-coding and non-coding RNAs that are differentially expressed between strains with distinct hostfinding behaviours. We have applied an arbitrary threshold of > 2 log2 fold change (P < 0.0001****) across both pairwise strain comparisons in order to link gene expression markers to behavioural differences. Many of these genes have already been shown to regulate behaviour in the model nematode C. elegans, however none have been implicated specifically in host-finding behaviours of parasitic nematodes. These data support the potential benefit of establishing and investing in natural diversity resources for parasitic nematodes, alongside more expansive studies in the model nematode C. elegans [50,51]. GO term enrichment revealed that a range of gene families and biological processes were enriched in differentially expressed gene sets that correspond with behavioural variation in S. carpocapsae strains (Additional file 1).
Transmembrane transport was the single most enriched biological process when considering either nictation or chemotaxis behaviours, following by a series of metabolic and biosynthetic processes. Of particular interest, several neuronal functions were also enriched. Given that we know much of how neuronal genes regulate behaviour in the model nematode C. elegans, neuronal gene families became the focus of our analysis. Only one neuropeptide gene, nlp-36, was found to exhibit differential expression (> 2 log2fold, P < 0.0001****) that correlates with distinct strain behaviours. NLP-36 is positively regulated by the cyclic nucleotide-gated channel subunit TAX-2 in C. elegans, which is implicated in the regulation of olfaction, chemosensation, thermosensation and axon guidance for a number of sensory neurons [52,53]. In S. carpocapsae UK1, significant up-regulation of nlp-36 correlates with reduced nictation behaviour (Figs. 1b and 3b). Down-regulation of insulin signalling during C. elegans development has been found to influence nictation behaviour in dauer juveniles, indicating a tuning of behaviour to environmental conditions [10]. Gruner et al. [54] point to a potential neuronal programming circuit for behavioural adjustment in C. elegans through the combined influence of insulin signalling, TRPV signalling and agonism of NPR-1, a known receptor of neuropeptides, FLP-21 and FLP-18. Although the differential expression of daf-28 does not meet the shared > 2 log2 fold change threshold we have applied to sequencing datasets, it is nonetheless down-regulated in S. carpocapsae Table 1 Predicted gene targets for differentially expressed microRNAs (> 2 log2 fold, P < 0.0001) that correlate with S. carpocapsae strain behaviour. Differentially expressed S. carpocapsae microRNAs were assessed for binding sites across 5′ and 3'UTRs of all ion channel, innexin, GPCR, neurotransmitter and neuropeptide genes. Ce target refers to direct 1 to 1 gene orthologues that are biochemically confirmed microRNA targets in C. elegans [20]. U refers to the number of predicted interacting microRNAs following default microRNA target prediction; S refers to the number of predicted interacting microRNAs following strict microRNA target prediction. Within the implicated behaviour column, "Both (inverse int.)" refers to involvement in both behaviours, through polarised regulation states (significantly upregulated for one, and significantly downregulated for the other) (Continued) Breton, relative to the other strains, correlating with enhanced chemotaxis toward G. mellonella. Conversely, daf-28 was up-regulated in UK1 relative to both Breton and All strains (Fig. 3a, b), correlating with reduced nictation (Fig. 1b). As was found for the two asic-2 sodium channel paralogues, daf-28 and other genes could function as part of a neurobiological switch, tuning behavioural strategy toward either migratory or stationary host-finding modes. It may be informative to track the dynamic abundance of such genes in IJs naturally enacting and transitioning between these behaviours, or in conditions that are known to enhance or suppress these behaviours to further strengthen the correlation of expression with behaviour [55,56]. The up-regulation of srsx GPCR genes in S. carpocapsae Breton correlates with enhanced chemotaxis to the lab host insect G. mellonella (Fig. 4a). One of these genes is orthologous to a cluster of srsx genes (srsx-22 and srsx-24) that are enriched in dauer stage C. elegans. It has been shown that dauer stage Caenorhabditis spp. are attracted to certain insect species, increasing the opportunity to engage in phoresis [57]. This cluster of shared srsx GPCRs may therefore mediate this attraction, in isolation, or in synergy with other such receptors. Two asic-2 sodium channel paralogues (denoted here as Sc-asic-2i and Scasic-2ii) are also notable as being differentially expressed and inversely correlated with both increased chemotaxis and reduced nictation behaviours (Fig. 4e-f). ASIC-2 is known to regulate aspects of nematode body posture and mechanosensation in C. elegans [58].
Small non-coding RNAs, including microRNAs, are increasingly implicated in complex aspects of biology and behaviour [51,[59][60][61]. We conducted a deep analysis of small RNA profiles across S. carpocapsae strains, and found a substantial degree of variation in relative abundance (Figs. 5 and 6, Additional file 3). Numerous micro-RNAs are differentially expressed, and correlate with behavioural differences across pairwise comparisons (Fig. 6, Additional file 4). Key to extrapolating biologically relevant information from microRNA networks is the identification of gene targets. Although there are many in silico tools that predict microRNA-mRNA transcript interactions, false positives are likely to be common [62]. Whilst many factors influence the reality and significance of predicted interactions, bioavailability of target One way ANOVA and Tukey's multiple comparison tests were conducted using Graphpad Prism 7.02. P < 0.05*, P < 0.001***. c Diagrammatic depiction of the relative position of VGluT in a glutamatergic neuron gene and microRNA in terms of spatial and temporal expression patterns will be key, along with the number of available microRNA binding sites, the relative enthalpy of binding interactions, and local competition for available microRNAs. In order to build on the basic knowledge presented in this study, it will be necessary to biochemically validate gene transcripts as microRNA targets through Argonaute ClIP-seq [45], and further, to demonstrate colocalisation of microRNA and target mRNA transcript to confirm interaction of discrete partners. This represents a substantial, but necessary task if we are to unravel the biological significance of microRNA regulation in the context of complex phenotypes and behaviours. Previous publications have employed a hierarchical and cooperative in silico target prediction approach using several programs simultaneously to arrive at an agreed set of targets [59]. Whilst this will certainly reduce the complexity of any target gene set, it will also constrain the output according   (27) to the most stringent program, leading inevitably to false negatives when perfect seed site complementarity is required by one or more programs. Here we have employed a dual analysis strategy within the miRanda discovery environment [43], using both default and strict discovery modes; the latter requires perfect seed site complementarity between microRNA and target mRNA transcript. Collectively, this strategy maps well to biochemicallyvalidated microRNA target interactions, including those that do not require perfect seed site binding [45]. However, as with any in silico prediction approach, biological validation is still required to corroborate interactions. Our datasets suggest a strong microRNA target site enrichment within predicted 5'UTRs of S. carpocapsae. This may be biologically significant, or could represent an artefact of computational UTR prediction. Ultimately, higher confidence interactions could be established by sequencing full length transcripts and confirming UTR identity on a transcriptome wide scale. In silico target prediction for differentially expressed (> 2 log2 fold, P < 0.0001****) and behaviourally correlated microRNAs points to instances of potential microRNA co-operation. For example, the ion channel genes Sc-asic-2ii and Sc-egas-1 are predicted targets for three and five individual differentially expressed microRNAs each ( Table 1). The neuropeptide GPCR gene, Sc-npr-23 is likewise targeted independently by two differentially expressed microRNAs. The differential expression pattern of each targeting microRNA correlates with altered chemotaxis behaviour, suggesting that microRNAs may also drive behavioural variation ( Table 1).
The predicted targeting of Sc-npr-11 by sc-miR-301-5pi reveals that UTR sequence variation between the two annotated npr-11 isoforms allows the dominant isoform (representing~75% of all npr-11 transcript across strains) to escape mir-301-5pi interaction (Table 1; Additional file 9). Likewise, variation in the 5'UTR and 3'UTRs of Sc-vglu-2 and Sc-cha-1 (Figs. 6 and 7, Table 2) reveal a potential quantitative difference in microRNA interaction events across different gene isoforms. In the case of Sc-cha-1, gene-level regulation of microRNA target site visibility could represent a strain specific mechanism to dampen global cholinergic signalling, or could allow selective dampening in a subset of cholinergic neurons. Ascertaining the relevance of isoform variance as it relates to microRNA interactions will require detailed gene localisation and functional studies. Our datasets highlight 5′ and 3'UTR variation as a factor in differential microRNA target site visibility, although it seems especially likely that alternative polyadenylation signals in the 3'UTR of protein-coding genes will exert substantial control over microRNA target site visibility, especially given the evident pervasiveness of 3'UTR variation in nematodes [63,64]. Any such regulation could modify the relative percentage of gene isoform variants accessible to microRNAs, which could allow for cell, or tissue-specific regulatory events that may be difficult to ascertain from organism-wide datasets. Collectively, our data point to a complex co-regulatory environment involving gene-specific isoform variation, and microRNA transcriptional regulation that is likely to influence various aspects of biology, including behaviour.
In the same way that "the best model for a cat is several cats" [65], the best model for a nematode parasite of vertebrates, is a sustainable population of the very same. However, this might not always be preferable in terms of ethics, available tools, sustainability or reproducibility. 'Model' status for any organism is contingent on high quality genomic and transcriptomic resources, easily amenable research tools, both in terms of genetic and molecular manipulation, alongside robust behavioural and phenotypic assays. Ease of culture, handling, and short generation times should also be major considerations. A model organism and the data generated from it must also be sufficiently relevant to trigger near-term impact on other species that have implications for health and economy. The datasets presented in this study build upon a growing catalogue of tools and resources for Steinernema spp. entomopathogenic nematodes [9,21,30,49,[65][66][67]. The close phylogenetic relationship between Steinernema spp. and economically important mammalian parasites [68], coupled with striking behavioural similarities, and a pathogenic life style that can be fully recapitulated in a Petri dish suggests that these entomopathogenic nematodes could represent attractive and convenient new surrogate models for parasite biology and behaviour. In addition to their potential worth as model organisms, Steinernema spp. have attracted considerable attention as bioinsecticidal agents [69]. The interest in Steinernema spp. as biological models, and as economically relevant end-point organisms in their own right represents a unique proposition for researchers interested in comparative biology, behaviour and host-parasite interactions.

Conclusions
This study is the first to identify differentially expressed genes that correlate with behavioural variation across three S. carpocapsae strains. We have identified four srsx GPCR genes and a sodium transporter gene that correlate with enhanced chemotaxis of S. carpocpsae IJs. A neuropeptide-like protein gene, srt GPCR gene and sodium transporter gene are found to correlate with reduced nictation behaviour. Numerous microRNAs are predicted to target neuronal genes, and display expression profiles that correlate with behavioural variation. These data provide important insight to the transcriptional basis of host-finding behaviours in this species, highlighting a potential role for microRNAs and gene isoform variation.