Sex-biased gene expression in L. salmonis

Sex-biased transcripts were identified in three populations of pre-adult II L. salmonis using a 38 K oligonucleotide microarray. Two of the populations were from the Atlantic subspecies L. salmonis salmonis and were collected from separate bay management areas (BMA-2a and BMA-2b) in the Bay of Fundy, New Brunswick [37], and the third was from the Pacific subspecies L. salmonis oncorhynchi [38] collected from the Broughton Archipelago, British Columbia (BC). Eighteen to 21 F1 generation preadult males and females from each population were analyzed in individual microarray hybridizations (total n = 117 individuals and hybridizations). A total of 34.7 – 42.7 % of all unique contigs passing quality control (QC) filters were significantly sex-biased (Benjamini-Hochberg multiple test correction; p < 0.01; fold change (FC) ≥ 1.5) in Atlantic and Pacific L. salmonis (Table 1). Including only the transcripts expressed in both sexes, a Principal Component Analysis (PCA) separated male and female samples along the first principal component (PC1; explaining the most variation) in all three populations, representing 50.2, 39.5 and 53.4 % of the transcriptional variation in BMA-2a, BMA-2b, and Pacific lice, respectively (Fig. 1). No consistent differences were observed between the proportions of transcripts overexpressed in males relative to females in each population (Table 1). Sex-biased transcripts for each population, including p-values, fold changes, annotations, and accession identifiers, can be found in Additional file 1.

L. salmonis populations	Unique contigs passing QC filter	Proportion sex-biased (%)	Male-biased contigs	Female-biased contigs	Proportion (%) of orphans
					Male-biased	Female-biased	Unbiased
Atlantic (BMA-2a)	11859	34.7	1955	2157	45.1	37.1	33.1
Atlantic (BMA-2b)	8527	40.0	1729	1682	48.9	32.8	32.4
Pacific	14923	42.7	3068	3303	51.8	28.3	34.0
Consensus	7889	N/A	368	461	50.7	20.1	N/A

Orphans are contigs without annotation (BLASTx; e < 10^-10). The numbers listed in this table are representative of unique contigs (i.e. duplicate probes removed). Only unique contigs were considered for each category

Sequence conservation in sex-biased transcripts

The L. salmonis contigs used for microarray construction [39] were annotated using BLASTx and RPS-BLAST against SwissProt and Conserved Domain Database (CDD; [40]), respectively [39]. Contigs without annotation (e > 10^-10) are marked as unknown in the additional files. The proportion of orphans relative to annotated transcripts in male-biased, female-biased, and unbiased categories was assessed for each population. This approach has been used in other model organisms, for example in flies and nematodes [11, 41]. Male-biased transcripts from all three L. salmonis populations had a higher proportion of orphans compared with female-biased and unbiased transcripts (Table 1). Female-biased and unbiased transcripts did not show consistent differences in the proportion of orphans (Table 1). These data suggest lower sequence conservation of male-biased transcripts in L. salmonis.

Consensus of sex-biased transcripts in three populations of L. salmonis

To assess the functional impacts of sex-biased expression in L. salmonis, differentially expressed transcripts between sexes from each population were used to generate a consensus list (Fig. 2). A total of 1470 unique transcripts, out of a total of 7889 were shown to be significantly sex-biased in all three populations with 829 of these showing concordant expression profiles (Fig. 2; Additional file 2). Using this consensus list (i.e. requiring concordant differential expression being identified in all three populations), 368 transcripts showed male-bias and 461 were female-biased. As expected from the individual population analyses, consensus male-biased transcripts showed a 2.5-fold higher proportion of orphans compared with those showing female-bias (Table 1). Differences in the degree of sex-biased expression, as measured by fold change (FC) also varied between male- and female-biased transcripts. On average, 84.8 % of the transcripts overexpressed in females had low sex-bias (FC ≥ 1.5 and ≤ 4), whereas transcripts overexpressed in males had equal proportions of high and low sex-bias (Table 2). Interestingly, 127 male-biased transcripts were not expressed above background levels in any of the 58 females assayed and therefore are referred to as male-specific. In contrast, only 20 transcripts were female-specific in the consensus list. Here, fold changes are reported as the range of differential expression between males and females across all populations, unless the transcript was sex-specific, and then it is denoted as such. Fold changes specific to each population for consensus sex-biased transcripts can be found in Additional file 2.

Sex-bias	Population	Proportion (%)
		Low fold change (≥1.5 but < 4)	High fold change (≥4 fold)
MALE	BMA2a	51.5	48.5
	BMA2b	55.3	44.7
	Pacific	37.7	62.3
	Mean FC	49.9	50.1
FEMALE	BMA2a	85.5	14.5
	BMA2b	98.5	1.5
	Pacific	77.3	22.7
	Mean FC	84.8	15.2

Mean fold change (FC) was calculated for each transcript using the average fold change across the three populations

Male-biased transcripts in three populations of L. salmonis

The majority of annotated male-biased transcripts had roles in reproduction, for example being accessory gland proteins (Acps) and seminal fluid proteins (SFPs; reviewed by [42]). Transcripts known to regulate proteolysis for reproduction-related functions were highly male-biased in L. salmonis and included 16 proteases and 13 protease inhibitors, 10 of each being male-specific. However, a high degree of variance was observed in the expression of proteolytic transcripts among Pacific males (Fig. 3). To better understand this expression pattern, a transcript similarity assessment using kunitz/BPTI-like toxin (probe ID: C259R052) showed that 110 transcripts were strongly co-expressed (Pearson’s correlation, 0.95 < r < 1.0; Fig. 3). Although Atlantic males showed constitutive expression of these transcripts, Pacific males showed a characteristic “on/off” expression profile, with 10 of 19 individuals showing low, or absence of expression (Fig. 3). This co-expressed transcript list contains numerous representatives from known functional categories of male reproduction including peroxidases, pH regulators, kinases, and transporters, among others (Table 3). As seminal fluid proteins (SFPs) are only expressed in males [43], transcripts exhibiting male-specific expression are putatively assigned as candidate SFPs in L. salmonis. Many of these transcripts also enriched the Swiss-Prot (SP) and Protein Information Resource (PIR) Keyword (SP_PIR_Keyword) secreted (19 transcripts; p < 0.0001, Additional file 3), further supporting the involvement of these transcripts as SFPs or accessory gland proteins (reviewed by [42]).

Functional category	Transcript description	Degree of Male-bias (FC)	Probes	SwissProt Accession	CDD Accession	Ref
Protease inhibitors	Keratin-associated protein 6-2	30.8 - 277.1	C088R043	O08884	smart00131	[95, 96]
	Kunitz_BPTI	MS	C084R101	NA	smart00131	[69, 95–100]
	Kunitz_BPTI	MS	C057R056	NA	smart00131
	Kunitz_BPTI	MS	C259R083	NA	pfam00014
	Kunitz/BPTI-like toxina	MS	C259R052	B5L5M7	pfam00014
	Papilina	MS	C063R028	Q868Z9	pfam00014	[101]
	Papilina	MS	C183R014
	Papilina	MS	C066R049
	Papilina	MS	C142R005
	Tissue factor pathway inhibitor 2a	MS	C213R048	O35536	pfam00014	[100, 102]
	Antichymotrypsin-2a	2.9 - 4.2	C077R006	P80034	cd00172	[59, 95, 96, 101, 103]
	SERine Proteinase INhibitors (serpins)	MS	C215R048	NA	cd00172
	Serpin-Z10	1.6 - 2.4	C182R015	Q9SIR9	cd00172
Kinases	Adenylate kinase isoenzyme 1	2.1 - 10.4	C104R155	P05081	TIGR01360	[96, 104]
	Adenylate kinase isoenzyme 1	1.9 – 4.8	C153R147	P00571	TIGR01360
	Adenylate kinase isoenzyme 1	2.1 - 7.1	C031R080
	Casein kinase I isoform alpha	MS	C070R100	P97633	cd14016	[96, 105]
	Casein kinase I isoform epsilon	1.8 - 4.8	C244R145	Q9JMK2	cd00180
	Hexokinase type 2	2.5 - 7.9	C066R139	Q9NFT7	COG5026	[95, 106]
	Probable adenylate kinase isoenzyme F38B2.4	1.9 - 5.7	C212R032	Q20140	TIGR01360	[96, 104]
	Probable adenylate kinase isoenzyme F38B2.4	2.0 - 5.7	C028R063
	Pyruvate kinase	2.3 – 5.0	C020R004	O62619	pfam00224	[95, 96]
	Pyruvate kinase	2.0 – 4.6	C015R041		cd00288
	Pyruvate kinase	1.9 - 5.0	C155R159
Proteases	Calpain-A catalytic subunit	MS	C197R005	Q11002	smart00720	[95, 102]
	Calpain-A catalytic subunit	MS	C018R134		smart00230
	Carboxypeptidase Ba	MS	C161R058	P04069	cd03860	[102, 106]
	Cytosolic non-specific dipeptidase	MS	C261R120	Q3ZC84	COG0624	[102, 106, 107]
	Cytosolic non-specific dipeptidase	1.9 - 2.9	C145R086	Q9D1A2	COG0624
	Proprotein convertase subtilisin/kexin type 5a	2.5 - 4.2	C118R013	Q9NJ15	cd00064	[96]
	Serine protease persephone	MS	C118R020	Q9VWU1	smart00020	[61]
	Testisin	2.2 – 9.8	C007R130	Q9JHJ7	smart00020	[108, 109]
	Tryp_SPc, Trypsin-like serine protease	5.5 – 322.5	C158R134	NA	smart00020	[95, 96, 103]
	Tryp_SPc, Trypsin-like serine protease	MS	C009R051
	Tryp_SPc, Trypsin-like serine protease	MS	C008R159
	Zinc metalloproteinase nas-15a	2.1 - 3.7	C134R018	P55115	cd04280	[95, 103]
	Prostasinab	MS	C135R082	Q16651	cd00190	[102, 110]
	ZnMc_adamalysin_II_like	MS	C083R024	NA	cd04269	[96, 102]
	Proclotting enzyme heavy chain	MS	C006R078	P21902	smart00020	[111]
	Gamma-glutamyltranspeptidase 1b	1.7 – 3.4	C120R152	P20735	cl19223	[112]
pH regulation	Carbonic anhydrase 1	MS	C183R004	P83299	cd00326	[96, 113–115]
	Carbonic anhydrase 9	MS	C196R116	Q8VHB5	cd00326
	Carbonic anhydrase 9	MS	C131R016
	Carbonic anhydrase 9	MS	C161R087
Structural	Actin	MS	C223R146	Q92192	PTZ00004	[106, 116]
	Lamin Dm0	1.5 - 6.6	C220R106	P08928	pfam00038	[117]
	Outer dense fiber protein 2-like	2.5 - 5.2	C121R150	Q08B20	pfam02463	[51, 106]
	Kelch-like protein 20	MS	C022R130	Q5R7B8	NA	[96, 118]
	Tubulin alpha-2 chain	MS	C160R074	P06604	cd02186	[95, 106]
Transport	Solute carrier family 15 member 1	4.1 - 9.3	C170R033	P46059	TIGR00926	[96, 106]
	Solute carrier family 2, facilitated glucose transporter member 1	1.5 - 2.8	C192R047	P11166	pfam00083
	Sodium/glucose cotransporter 4	1.6 – 3.1	C170R069	Q2M3M2	pfam00474
	Solute carrier family 22 member 6-B	2.5 - 4.3	C072R016	Q66J52	TIGR00898
	Sodium-dependent nutrient amino acid transporter 1	MS	C167R125	B4JMC1	pfam00209
	Aquaporin-12Ab (LsGlp1_v1; [81])	1.7 - 5.3	C096R035	Q8IXF9	NA	[96, 119]
	Aquaporin-3 (Lsaqp12L2; [81])	3.7 - 10.4	C030R103	Q8R2N1	cd00333
Other	Mucin-like glycoprotein	3.3 - 9.9	C218R155	NA	pfam01456	[102, 120]
	Major royal jelly protein 3a	MS	C089R070	Q17060	NA	[121–124]
	Chorion peroxidase heavy chaina	MS	C176R138	Q9VEG6	pfam03098	[125, 126]
	Chorion peroxidase heavy chaina	MS	C154R094
	Peroxidasea	MS	C026R132	Q01603	pfam03098
Energy	Fructose-bisphosphate aldolase	2.0 – 3.4	C230R040	O52402	PRK09197	[59, 80, 104]
	Fructose-bisphosphate aldolase	1.8 – 3.0	C085R145	P14540	PRK09197
	Fructose-bisphosphate aldolase	1.9 – 3.3	C069R104	P14540	cd00946
	Glycogen phosphorylase	1.7 – 2.5	C107R029	Q9XTL9	cd04300
	Glycogen phosphorylase	1.6 – 3.2	C085R148	Q9XTL9	cd04300
	Glycogen phosphorylase, brain form	1.6 – 3.5	C171R004	Q3B7M9	cd04300
Male Fertility	Protein ref(2)Pb	1.6 - 3.9	C036R126	Q24629	cd14320	[127]

References are provided based on the identification of similarly annotated sequences or proteins involved in male reproduction (e.g. spermatogenesis, seminal fluid proteins, testis expression) in related organisms. The degree of male-bias is indicated by a range of fold change (FC) across populations while MS indicates the listed probes showed male-specific expression. Transcripts denoted with ^a have a signal peptide for secretion (SwissProt) while those with ^b were annotated using e < 10^-5 due to the absence of homology at higher stringency. Transcripts without annotation are represented by NA (no annotation). Each transcript has a unique contig ID which can be found in Additional file 2 using the probes listed here

Several other male-biased transcripts had putative roles in morphogenesis and the nervous system. Male-biased transcripts were enriched for cellular component assembly involved in morphogenesis (4 transcripts, p = 0.03), ossification (here probably calcification; 4 transcripts, p = 0.02), and Z disc (5 transcripts, p < 0.0001; Additional file 3). Additionally, male-biased transcripts were enriched for potassium ion binding (4 transcripts, p = 0.02), calcium ion binding (12 transcripts, p = 0.03), ion channel activity (5 transcripts, p = 0.04), and solute:cation symporter activity (5 transcripts, p = 0.01; Additional file 3), showing differences in sensory-system related functions. Other sex-biased transcripts involved in the nervous system, including their sex-biased expression profiles, are reported in Fig. 4. As salmon lice display sexually dimorphic patterns of mobility (i.e. mate location [32] and frequency of host switching [31]) and responses to neurotoxic drugs [37, 44], these transcripts will serve as important markers to better understand sex-related differences in the L. salmonis sensory system.

Female-biased transcripts in three populations of L. salmonis

Female-biased transcripts were enriched for basic molecular processes including RNA processing (60 transcripts, p < 0.0001), ribosome biogenesis (38 transcripts, p < 0.0001), and transcription (41 transcripts, p < 0.0001) (Additional file 3). Within these categories, some multi-subunit protein complexes were completely female-biased including chaperonin-containing T-complex (CCT-complex; 5 transcripts, p = 0.0002), Nup107-160 complex (4 transcripts, p = 0.02), spliceosome (11 transcripts, p = 0.02), histone deacetylase complex (4 transcripts, p = 0.04) and ribonucleoprotein complex (30 transcripts, p = 0.0009). Protein complex formation was also enriched in the female-biased list, for example the GO category macromolecular complex subunit organization (28 transcripts, p < 0.0001).

Several transcripts related to cell division and organization were overexpressed in females and some significantly enriched GO categories included cell cycle (35 transcripts, p < 0.0001), DNA replication (17 transcripts, p < 0.0001), and cell proliferation (13 transcripts, p = 0.004) (Additional file 3). Furthermore, GO enrichment of ATP binding (60 transcripts, p < 0.0001), ATP-dependent helicase activity (10 transcripts, p = 0.001), and ATPase activity (15 transcripts, p = 0.005; Additional file 3) indicated female-biased energy generation.

As observed with male-biased transcripts, female-biased transcripts were also enriched for reproductive functions. The GO category in utero embryonic development (8 transcripts, p = 0.005; Additional file 3) was significantly enriched despite the absence of mating across all experiments (female lice are not sexually mature at the pre-adult stage; [34]). These included nuclear autoantigenic sperm protein (FC = 2.7 – 10.0), pre-mRNA processing factor 19 (FC = 1.6 – 2.6), and protein arginine N-methyltransferase 1, among others (FC = 1.7 – 2.5; Additional file 3). Female-biased transcripts were also enriched for nuclear hormone receptor binding (4 transcripts, p = 0.02) with an additional female-specific transcript containing the c4 zinc finger common to nuclear hormone receptors (Zn_C4; CDD: smart00390; e = 2.14^-15). Several other female-biased transcripts not included in GO categories also have potential roles in reproduction including piwi-like protein 1 (FC = 2.6 – 8.3) and peroxiredoxin 1 (FC = 1.6 – 3.1). Lastly, transcripts involved in sex determination were female-biased in all populations, including prohibitin-2 ([45]; FC = 1.5 – 9.6) and pre-mRNA-splicing regulator female-lethal(2)D ([46]; FC = 1.7 – 2.4).

Some transcripts related to morphology, feeding, and detoxification were also female-biased across all populations. Transcripts such as serine proteinase stubble catalytic chain (female-specific), la protein homolog (FC = 1.5 – 2.5), and digestive organ expansion factor homolog (FC = 1.7 – 2.9), are involved in development, while trypsin-1 (FC = 3.1 – 12.0) and quinone oxidoreductase (FC = 2.1 – 116.5) have putative roles in feeding and detoxification, respectively. Lastly, immune-related transcripts, like rhotekin-2 (female-specific), ras-related protein Rab-32 (female-specific), and complement component 1 Q subcomponent-binding protein, mitochondrial (FC = 1.6 – 4.0), were female-biased in all L. salmonis populations. Female-biased transcripts therefore represent several candidates responsible for controlling sexual dimorphism at the molecular level in L. salmonis. Based on these findings, immunity, energy expenditure, and organogenesis are examples of previously unknown sexual dimorphism in salmon lice.

Annotation of L. salmonis sex-biased orphan contigs

Sex-biased orphan contigs were compared to known sequences using UniProt (e ≤ 10^-5) to augment novel transcript discovery in the non-model L. salmonis. Although this threshold is less conservative than that used for the original microarray annotation [39], it has been frequently used in other sea lice transcriptomic studies [47–49], being generally acceptable for gene annotation [50]. This method was used for novel transcript discovery only and these newly annotated transcripts were not included in GO analyses. A total of 16 female-biased and 12 male-biased transcripts were subsequently annotated (Additional file 2) using this method. This additional annotation did not substantially change observations made on sequence divergence differences in male and female-biased transcripts, and many of the newly annotated female-biased transcripts had similar functions to those identified in the enrichment analysis described above, including transcription, translation, and cell cycle (Additional file 3).

Potential links to reproduction were observed among several of the newly annotated male-biased orphans, including prostasin (male-specific), gamma-glutamyltranspeptidase (FC = 1.7 – 3.4), and protein ref(2)P (FC = 1.6 – 2.3) (Table 3). A male-specific transcript, c-factor, was also identified. However, the function of this transcript in L. salmonis remains unknown. Additionally, male-biased transcripts involved in neuromuscular development and function were discovered, including excitatory amino acid transporter 3 (SLC1A1; FC = 2.3 - 3), twitchin (FC = 1.6 – 5.7), and sarcoplasmic/endoplasmic reticulum calcium ATPase 1 (FC = 1.5 – 7.6). Probe identifiers, e-values, bitscores, SwissProt accessions, and descriptions for newly annotated transcripts can be found in Additional file 2.

Discovery of sex-biased transcripts related to reproduction in L. salmonis

Male-biased transcripts displayed greater sex-bias compared with female-biased transcripts in L. salmonis, similar to findings in D. melanogaster [6]. Among the highly male-biased transcripts, 127 were not expressed above background detection in any of the 58 females assayed in this study. These transcripts represent putative seminal fluid proteins (SFPs) based on their expression profiles (SFPs are only expressed in males; [43]) and annotations (Table 3 and references therein). SFPs are transferred to females during mating, impacting a variety of physiological processes such as sperm storage, egg production, feeding, behavior, and receptivity to re-mating (reviewed in [42]). Although genes involved in reproduction tend to evolve rapidly [57], many of the functional constituents of seminal fluid are conserved from arthropods to mammals [43, 58]. For example, genes involved in proteolysis are essential for sperm transfer, storage, and activity, among other functions [59, 60]. A total of 10 proteases and 10 protease-inhibitors were identified as male-specific in this study with several others showing male-bias (Table 3). All protease inhibitors putatively involved in reproduction contained at least one kunitz or serpin domain (CDD accessions: pfam00014 and cd00172, respectively), which are the predominant classes of protease inhibitors in seminal fluid [59]. In turn, proteases such as serine protease persephone and calpain-A catalytic subunit also had relevant annotations to SFPs based on their ability to modulate toll signaling in D. melanogaster [61, 62]. The transfer of immunomodulatory and antimicrobial SFPs may aid females in defending against infections that are introduced during mating [60, 63]. Many of the L. salmonis proteases assigned as SFPs also contained trypsin-like domains (CDD: cd00190 and smart00020) including prostasin, proclotting enzyme heavy chain, and three additional transcripts without SwissProt annotation (i.e. CDD only). These transcripts represent important targets for understanding the proteolytic events controlling reproduction and mating in L. salmonis.

Female-biased transcripts also had putative reproductive functions, including five with nuclear hormone receptor activity, one of which was female-specific. Recently, a sex-linked SNP in the L. salmonis prohibitin-2 gene with a pattern of female-biased expression was identified in three strains of L. salmonis from Scotland [45]. Similarly, a transcript annotated as prohibitin-2 was female-biased in all three populations of L. salmonis assayed in this study. The prohibitin-2 gene is likely involved in sex determination [45] along with pre-mRNA-splicing regulator female-lethal(2)D [46], which was also female-biased in L. salmonis.

Overall, functional enrichment for reproduction was less clear in female-biased transcripts compared with male-biased transcripts. Many of the putative female-biased transcripts potentially involved with reproduction had GO annotation with many other functional categories and only a small number were female-specific. For example, nuclear autoantigenic sperm protein (NASP) and histone deacetylase 1 (HD1), were female-biased in L. salmonis and have reproductive roles in similar species [54, 64, 65]. However, these transcripts were in the enriched GO category in utero embryonic development (p = 0.005), which is taxonomically constrained [66] and inconsistent with the virgin status of the lice used here [34]. Furthermore, the functions of NASP and HD1 differ in other species such as C. elegans in which they are important for female development and male-specific gene repression [67]. These discrepancies make it difficult to identify putative female-biased reproduction genes in the present work. In an earlier study by Eichner and colleagues [47], L. salmonis genes involved in reproduction were overexpressed in adult females compared to preadult II females. As the present study used preadult II females, it may have missed the identification of some female reproductive genes that are induced later in development than the stages evaluated here. Future studies assessing the expression and localization of these transcripts will offer insight on their exact functions. Much work remains to identify female-biased reproductive genes in L. salmonis, including the extensive changes likely to occur in female mating-responsive genes post transfer of male spermatophores (e.g. see [68, 71, 72]).

In general, sex-biased genes evolve faster than unbiased genes, with those exhibiting male-bias showing the highest rates of evolution (reviewed by [4]). This trend is heavily influenced by sex-biased transcripts involved in reproduction as these typically exhibit higher than normal rates of positive selection (reviewed by [4, 57]). Only male-biased transcripts showed a consistently higher number of orphans in each L. salmonis population (45.1 – 51.8 %) when compared to female-biased (28.3 – 37.1 %) and unbiased (32.4 – 34.0 %) transcripts. This trend was largely driven by the putative SFPs in this study (i.e. male-specific transcripts), of which 65.4 % were orphans. As higher rates of nucleotide substitutions are often caused by a relaxed functional constraint or greater positive selection on certain transcripts (reviewed in [4, 57]), higher-resolution analyses including non-synonymous/synonymous mutation ratios (dN/dS; [70, 71] and codon-usage bias [72, 73] will be important in clarifying the effects of selection on these L. salmonis genes. These analyses will yield important information regarding evolutionary processes affecting reproduction, population dynamics, and drug resistance. In particular, it will be important to identify whether genes involved in drug resistance are similar to those involved in sex-biased expression as this will provide insight on molecular mechanisms behind the higher rate of drug resistance in male L. salmonis. Specifically this will inform on whether the increased resistance is due to inherent physiological factors that differ between the sexes, or to the evolutionary rate of resistance mechanisms.

Sex-biased transcription related to sexually dimorphic phenotypes of L. salmonis

L. salmonis exhibits extensive sexual dimorphism in morphology at the pre-adult and adult stages [27]. Several sex-biased transcripts discovered here have related functions including the male-biased transcripts enriching the GO categories cellular components involved in morphogenesis (4 transcripts, p = 0.03) and Z-disc (sarcomere; 5 transcripts, p < 0.0001). Additionally, serine proteinase stubble catalytic chain was female-specific in L. salmonis and is required for proper formation of appendages in Drosophila spp. [74]. Therefore, these transcripts represent targets for understanding sexually dimorphic morphology including features that are important in mating (described in [27, 31, 34]). The sex-biased orphans described here are also ideal candidates for understanding mate guarding, a common mating behavior in crustaceans, which is known to be sexually antagonistic [75]. Therefore, transcripts involved in mate guarding are unlikely to be annotated due to rapid sexual selection [75] and taxonomic constraint [50].

Other molecular links to sexual dimorphism in L. salmonis were observed in transcripts such as longitudinal lacking protein, trypsin-1, and digestive organ expansion factor, which were female-biased in this study and have roles in salivary gland, trachea, and digestive organ development and function [76, 77]. Based on female L. salmonis having a greater requirement for blood in the meal [33], these transcripts will serve as important markers for understanding sexually dimorphic feeding patterns and related host-parasite interactions.

Male-biased L. salmonis transcripts also enriched several GO categories related to the sensory system, which is generally known to be more refined in male copepods [23]. These included potassium ion binding (4 transcripts, p = 0.02), solute:cation symporter activity (5 transcripts, p = 0.01), and calcium ion binding (12 transcripts, p = 0.03), among others (Additional file 3). Males are known to transfer between salmonid hosts more frequently than females [30, 31] with mate location primarily being the responsibility of the male [32]. Thus, transcripts involved in the sensory system and muscle development (e.g. Z disc; Additional file 3) represent putative targets for understanding neuromuscular differences related to increased mobility in males. Furthermore, chemical cues are essential components of host and mate location in L. salmonis and serve to optimize the probability of mating through balancing the proportion of males relative to females on each host [29]. Chemosensory signaling and behavioral responses to non-host semiochemical treatments known to interfere with host recognition are sexually dimorphic in L. salmonis [32, 78]. Based on the high evolutionary rates of genes involved in chemoreception [79], sex-biased orphans should be considered candidate targets for understanding the chemical ecology of L. salmonis.

The identification of sex-biased transcripts in L. salmonis also provided preliminary evidence for sexually dimorphic characteristics previously unknown in salmon lice. For example, immune-related transcripts such as rhotekin-2 and ras-related protein Rab-32 were female-specific in this study while other related transcripts were female-biased (Additional file 2). Gene Ontology analysis also indicated a higher energy expenditure in female L. salmonis based on the female-biased expression of 15 transcripts enriching ATPase activity coupled and 60 transcripts enriching ATP binding (Additional file 3). This is a similar finding to the GO analysis of sex-biased expression in D. pulex [11], and may serve as an explanation for the lower frequency of inter-host transfer in female L. salmonis [30, 31]. Although male-biased transcripts such as glycogen phosphorylase, fructose-bisphosphate aldolase, and hexokinase-2 enriched the GO category glucose metabolic process (7 transcripts; p =0.02), these are likely acting as an energy source for spermatogenesis [80]. These data indicate an increased basal energy demand in female L. salmonis that is important for understanding observed sex differences in drug tolerance and resistance [35, 37]. Collectively, the candidate sex-biased transcripts described here represent putative markers controlling energy expenditure, morphology, and the immune and sensory systems of L. salmonis.

Strengths and limitations of using a consensus-based approach to identify sex-bias

Recently, seven aquaporin paralogs were characterized in L. salmonis, each with different expression profiles across stage and sex [81]. Two of these aquaporins (LsGlp1_v1 (KR005660.1) and Lsaqp12L2 (KR005666.1)) showed male-specific and male-biased expression, respectively. In the present work, two male-biased contigs annotated as aquaporin 3 (BT121448.1) and aquaporin-12A (BT121051.1; see Table 3) showed more than 99 % sequence similarity to LsGlp1_v1 and Lsaqp12L2, respectively (Additional file 4). Therefore, aquaporin 3 and LsGlp1_v1 appear to be the same transcript based on sequence alignment, as are aquaporin-12A and Lsaqp12L2. Here, aquaporin 3 was only expressed above background fluorescence in one of the 58 females assayed, supporting the expression profile previously reported [81]. Additionally, aquaporin-3 was shown to be highly male-biased in the closely related C. rogercresseyi [54], suggesting this gene is important for a male-specific function in salmon lice. This type of transcriptomic consensus will be important for functional categorizations in future sea lice studies.

Within the 1407 transcripts shown to be sex-biased in Atlantic and Pacific populations, only 829 showed concordant expression profiles. The majority of transcripts in the discordant list (Additional file 2) were involved in cuticle formation and molting and, therefore, transcripts that potentially oscillate in expression levels at different molt intervals [47, 82] were eliminated from this interpretation. However, the consensus identification of sex-biased transcripts in L. salmonis did prove to be over-conservative in some cases, causing particular transcripts with putative sex-bias to be overlooked. For instance, trypsin-4 (probe: C054R168), annotated from Anopheles gambiae and involved in host seeking behavior and blood feeding [83], was female-specific in BMA2a and in Pacific L. salmonis. However, this probe did not pass quality filters in the BMA2b population and was eliminated from the consensus list. Female L. salmonis are known to feed more heavily on blood than males [33], with certain trypsins known to be involved in digestion and immune evasion on salmonid hosts [84–86]. Additionally, a similar transcript annotated to trypsin-1 was female-biased in this study (Additional file 2). Therefore, monitoring individual population analyses from this work is also important for identifying potential sex-biased markers in L. salmonis. Nonetheless, the consensus set of sex-biased transcripts identified here supports the characterization of L. salmonis transcript as markers for reproduction, morphogenesis, behavior, and other sexually dimorphic traits for targeted approaches (i.e. knock-out/knock-down, recombinant production, in vitro characterization, etc.) in future studies. This improved understanding of sex-biased gene expression in L. salmonis will inform future studies examining host-parasite interactions, drug resistance, reproduction, and novel drug discovery.

Lepeophtheirus salmonis populations and collections

Adult L. salmonis were collected from Atlantic salmon aquaculture farms on the Atlantic and Pacific coasts of Canada. Two populations of Atlantic L. salmonis were collected in the spring of 2013 from Bay Management Area 2a (BMA-2a; Back Bay) and 2b (BMA-2b; Grand Manan) in the Bay of Fundy, New Brunswick (NB), as described in full previously [35, 37]. A third population representing Pacific L. salmonis was collected from the Broughton Archipelago, British Columbia (BC) in 2010. Atlantic and Pacific L. salmonis are considered allopatric subspecies [38]. For all collections, egg strings were removed from adult females and larvae reared to the infective copepodid stage in static seawater hatch systems as previously described [35]. Copepodids (F1 generation) were then used to infect Atlantic salmon (Salmo salar) and allowed to develop to the pre-adult II stage. Pre-adult (F1) lice from all populations were used in 24 h in vitro EMB bioassays (described below) before collection and storage at -80 °C for RNA extraction.

Atlantic and Pacific L. salmonis Microarray Datasets

Two microarray datasets were used to compare sex differences in L. salmonis from the Atlantic (2013 collection) and Pacific (2010 collection) coasts of Canada. The Atlantic dataset was accessed from NCBI through Gene Expression Omnibus (GEO) accession GSE56024 [37]. In this study a total of 77 pre-adult Atlantic L. salmonis, 38 females and 39 males from BMA-2a and BMA-2b were exposed to four concentrations of EMB (0.1, 25, 300, and 1000 ppb) and a seawater control, as previously reported. The bioassay protocol was identical for both populations. This study compared the effects of EMB on L. salmonis including the interactions between population (BMA-2a is more EMB-resistant than BMA-2b; [37, 87]) and sex. However, baseline differences between males and females were not reported. The Pacific dataset was provided by the same laboratory group (B. Koop and S. Jones, unpublished data), which exposed 39 pre-adult Pacific L. salmonis, 21 females and 19 males, to low doses of EMB (0.01, and 0.1 ppb) or a seawater control. Lower doses of EMB were selected based on the high EMB-sensitivity of this population [88]. For all F1 generation cultures, lice were maintained in filtered sea water at 10 ± 2 °C and 32 ± 2 ppt. The Pacific dataset has been uploaded to GEO under the accession GSE73734.

A 38 K oligonucleotide microarray (eArray, Agilent) designed with expressed sequence tags (ESTs) from Atlantic and Pacific L. salmonis [39] was used to analyze all lice in this study. Annotation of each contig was completed using BLASTx and RPS-BLAST against SwissProt and Conserved Domain Database (e < 10^-10), respectively. A total of 18 – 21 hybridizations for each sex and population combination (117 total hybridizations) were completed. Hybridizations were completed using methods for sample preparation, microarray hybridization, and scanning as previously reported [37, 89]. Briefly, all slides were scanned using a Perkin Elmer ScanArray® at 5 μm resolution and optimized PMT intensities (1-2 % of array spots saturated). Filtering and quantification was completed using Imagene 8.1 (Biodiscovery) before completing statistical analyses in GeneSpring GX v12.6 (Agilent). A quality control (QC) filtered probe list was created for each population (Table 1) with probes included for statistical analysis only if at least 65 % of the samples in any one condition had raw fluorescent intensities ≥ 500 and showed no poor quality spots.

Sex-biased Gene Expression in L. salmonis

Microarray data was used to characterize baseline expression differences between male and female L. salmonis. Sex-biased probes were identified using a two-way ANOVA with sex and EMB as explanatory variables (Benjamini-Hochberg multiple test correction; p < 0.01; fold change (FC) ≥ 1.5). The effects of EMB on Atlantic L. salmonis transcriptomes was minor [37] and a significant transcriptomic effect was not detected in Pacific L. salmonis used here (i.e. no probes showed differential expression by EMB). The effects of EMB exposures were also controlled for in the statistical model by only using transcripts affected by a main effect of sex and through consensus-based analyses (described below). All probes with a main effect of sex for each individual population are compiled into Additional file 1.

Sex-biased probes from each population were used to create a consensus list for functional analyses. This list was limited to probes exhibiting significant and concordant sex-bias in all three populations of L. salmonis described here (Fig. 2). Sex-biased probes from individual population analysis can be found in Additional file 1, while sex-biased probes identified using the consensus sex-biased method can be found in Additional file 2. As duplicate probes represent unique contigs (i.e. transcripts) on the array [39], contig IDs are also included in the additional files. The variation between duplicate probes of consensus sex-biased transcripts is quantified in Additional file 2. Only unique transcripts are used to calculate the proportion of sex-biased expression in Atlantic and Pacific L. salmonis.

Differences in the degree of sex-bias were assessed by binning transcripts based on their degree of differential expression between male and female L. salmonis [6]. Transcripts with low sex-bias were those overexpressed by a fold change (FC) of ≥ 1.5 but < 4, while highly sex-biased transcripts had a FC ≥ 4. A mean FC value representing all three populations was also included (Additional file 2). Any transcript that did not pass the background QC filter in 100 % of the individuals within one sex was considered to be sex-specific in this study. However, based on lower limits of detection for microarrays, these transcripts may not be biologically sex-specific. A transcript similarity assessment was also completed using all QC-filtered probes against kunitz/BPTI-like toxin (probe ID: C259R052). Transcripts with similar expression patterns to kunitz/BPTI-like toxin were determined using a Pearson’s correlation (0.95 < r < 1.0). These transcripts are described in Additional file 2.

Functional enrichment of the consensus sex-biased transcript list was done using Gene Ontology (GO), InterPro, and SwissProt (SP) and Protein Information Resource (PIR) Keywords (SP_PIR_Keywords) with DAVID bioinformatics [90–92] using a modified Fisher’s exact test (p < 0.05; genes/enrichment category ≥ 4). Unique SwissProt accession ID’s were compared against a QC filter background list designed to include transcripts passing QC filters in all populations (Table 1). To reduce redundancy of Gene Ontology categories, GO Trimming was used with an 80 % soft trim threshold [93]. All GO analyses can be found in Additional file 3.

Sequence conservation in sex-biased and unbiased transcripts

To investigate rates of sequence divergence in sex-biased transcripts, unique contigs passing QC filter for each population were binned into one of three categories: Male-biased (overexpressed by males), female-biased (overexpressed by females) or unbiased (no expression difference between sexes). All contigs on the microarray were annotated using UniProt [94] and Conserved Domain Database (NCBI, [40]) with the best match being the alignment with the lowest Expect value followed by the highest bitscore [39]. Transcripts with no significant match (e > 1.0^-10) were considered orphans (labeled as “unknown” in additional files). To assess the degree of sequence conservation between male-biased, female-biased and unbiased transcripts, the proportion of orphans relative to annotated transcripts in each of these categories was assessed based on similar methodologies [11, 41]. Sequence conservation in male-biased and female-biased transcripts from the consensus list was also analyzed (Table 1).

Novel transcript discovery using sex-biased orphans

Based on the robust approach using a consensus sex-biased transcript list, orphans showing concordant sex-bias across all populations were re-annotated using a less conservative threshold of e > 1.0^-5. This annotation threshold is common amongst other sea lice transcriptomic studies and is generally considered an acceptable cut-off for annotation [50]. This annotation was used for exploratory transcript prediction but not for the main analysis or the Gene Ontology enrichment analysis. Sex-biased orphan annotation results can also be found in Additional file 2.