A reference gene set for sex pheromone biosynthesis and degradation genes from the diamondback moth, Plutella xylostella, based on genome and transcriptome digital gene expression analyses

Background Female moths synthesize species-specific sex pheromone components and release them to attract male moths, which depend on precise sex pheromone chemosensory system to locate females. Two types of genes involved in the sex pheromone biosynthesis and degradation pathways play essential roles in this important moth behavior. To understand the function of genes in the sex pheromone pathway, this study investigated the genome-wide and digital gene expression of sex pheromone biosynthesis and degradation genes in various adult tissues in the diamondback moth (DBM), Plutella xylostella, which is a notorious vegetable pest worldwide. Results A massive transcriptome data (at least 39.04 Gb) was generated by sequencing 6 adult tissues including male antennae, female antennae, heads, legs, abdomen and female pheromone glands from DBM by using Illumina 4000 next-generation sequencing and mapping to a published DBM genome. Bioinformatics analysis yielded a total of 89,332 unigenes among which 87 transcripts were putatively related to seven gene families in the sex pheromone biosynthesis pathway. Among these, seven [two desaturases (DES), three fatty acyl-CoA reductases (FAR) one acetyltransferase (ACT) and one alcohol dehydrogenase (AD)] were mainly expressed in the pheromone glands with likely function in the three essential sex pheromone biosynthesis steps: desaturation, reduction, and esterification. We also identified 210 odorant-degradation related genes (including sex pheromone-degradation related genes) from seven major enzyme groups. Among these genes, 100 genes are new identified and two aldehyde oxidases (AOXs), one aldehyde dehydrogenase (ALDH), five carboxyl/cholinesterases (CCEs), five UDP-glycosyltransferases (UGTs), eight cytochrome P450 (CYP) and three glutathione S-transferases (GSTs) displayed more robust expression in the antennae, and thus are proposed to participate in the degradation of sex pheromone components and plant volatiles. Conclusions To date, this is the most comprehensive gene data set of sex pheromone biosynthesis and degradation enzyme related genes in DBM created by genome- and transcriptome-wide identification, characterization and expression profiling. Our findings provide a basis to better understand the function of genes with tissue enriched expression. The results also provide information on the genes involved in sex pheromone biosynthesis and degradation, and may be useful to identify potential gene targets for pest control strategies by disrupting the insect-insect communication using pheromone-based behavioral antagonists. Electronic supplementary material The online version of this article (doi:10.1186/s12864-017-3592-y) contains supplementary material, which is available to authorized users.


Background
In moths, reproductive isolation relies heavily on mature female adults producing and releasing species-specific sex pheromone components to attract conspecific males. Adult male antennae on the other hand specifically perceive these sex pheromone components from a distance resulting in the successful location of the female to initiate and complete mating behavior [1]. In lepidopterans, type I sex pheromone components are fatty acid derivatives with 0-4 double bond containing carbon chains of varying lengths (C10-C18) and an oxygenated functional group that can be an aldehyde, alcohol or an acetate ester. Type II sex pheromone components are hydrocarbons or epoxides [2,3]. In female moths, the type I sex pheromones are synthesized de novo and the genes involved in the biosynthesis are regulated in a specialized manner in their sex pheromone glands (PGs). Sex pheromone biosynthesis employs a modified fatty-acid biosynthesis pathway that includes various processes such as acetylation, desaturation, chain shortening, reduction, and oxidation, either separately or in combination [2,4].
Males on the other hand, have evolved a sensitive antennal olfaction system to trace even very low amounts of sex pheromone components released from the female located at a long distance [24]. These sex pheromone components enter the male antennal pores and are transported by pheromone binding proteins (PBPs) [25][26][27] through the antennal sensillum, and activate membrane-bound pheromone receptors (PRs) [28][29][30]. After PR activation, the pheromone molecules must be rapidly degraded to release and refresh the PR. Depending on the functional group on a pheromone, whether an aldehyde, alcohol or ester, degradation of pheromones involves specific enzymes in the pheromone degradation pathway. Aldehyde oxidase (AOX) is an antennal enzyme that catalyzes the oxidation of aldehyde sex pheromones to carboxylic acids, which were biochemically characterized in the antennae of Manduca sexta [31], Antheraea polyphemus and Bombyx mori [32]. Thus far, only one insect AOX (AtraAOX2) from Amyelois transitella has been functionally characterized to hydrolyze aldehyde sex pheromone components and some plant volatiles in vitro [33]. However, a number of putative AOX genes have been reported in lepidopterans at the nucleic acid level [34][35][36]. Aldehyde-degradation involves the NAD(P) − dependent aldehyde dehydrogenase (ALDH) and AR enzymes that convert sex pheromones to their corresponding acids and alcohols, respectively [37][38][39]. Antennal carboxyl/cholinesterase (CCE) hydrolyzes ester sex pheromone components. In previous studies, we functionally characterized five carboxyl/cholinesterases (CXEs) belonging to the CCE family from S. exigua and Spodoptera litura; in vitro expressed SexiCXE10 specifically degraded ester plant volatiles and three other CCEs (SexiCXE4, SexiCXE13 and SlitCXE13) had dual roles, i.e., they hydrolyzed sex pheromone components as well as plant volatiles [40][41][42][43]. Other members of the multi-function enzyme families include CYPs [44,45], UGTs [46] and GSTs [47], which are also considered as odorant-degrading enzymes (ODEs) or pheromone-degrading enzymes (PDEs).  [49][50][51][52]. These studies also revealed that both female and male DBM need precise regulation of sex pheromone biosynthesis that facilitates their communication through the various structural components. In recent years, sex pheromone reception has been studied extensively because of its potential use as green targets for environmentally friendly pest control measures. Male DBMs have PBPs that are designed for sex pheromone reception. Thus far, three PBPs have been cloned and characterized in P. xylostella. These PBPs robustly bound all three sex pheromone components (Z11-16:Ald, Z11-16:OH and Z11-16) [53] thereby enhancing the electrophysiological responses of DBM to sex pheromone components in vitro [54]. In the recently sequenced DBM genome [54], two gene sequences likely to be odorant/pheromone-degrading enzymes such as CYPs [55] and GSTs [56] were identified. However, the tissue specific expression of these genes are not known. To identify additional genes involved in the DBM pheromone biosynthesis and degradation, in this study, we combined transcriptome based digital gene expression (DGE) and genome mapping. By using these methods, we identified a total of 299 genes belonging to 14 gene families potentially involved in the sex pheromone biosynthesis and degradation pathways. The tissue specific expression pattern of these genes was also confirmed qPCR.

Results and discussion
Overview of the transcriptomes Six transcriptomes from male antennae (m_Ant), female antennae (f_Ant), legs (L), heads (H), abdomens (AB) and female pheromone glands (PG) were sequenced using the Illumina HiSeq 4000 platform (Illumina, Tianjin, China) and assembled with Trinity (version r20140413p1). To obtain an accurate expression pattern, we removed as much of the ovipositor as possible from the females due to its adherence to the pheromone glands (Fig. 1). Illumina sequencing yielded 5.96-6.84 giga bases (Gb) for each of the six transcriptomes. We first mapped the transcriptome reads to the DBM reference genome (Genome assembly version 2, http://iae.fafu.edu.cn/DBM/) ( Table 1) and obtained the following mapping rates that were favourable for annotation: 42.29% (legs), 39.6% (male antennae), 38.65% (female antennae), 42.27% (heads), 43.7% (pheromone glands) and 41.65% (abdomens). Then, we supplemented the original genome annotation file (18,105 genes Genome assembly version 2, http://iae.fafu.edu.cn/DBM/) and found 3,839 novel genes (Additional file 1: Table S1). Further to obtain more unigenes, we also performed a de novo assembly of the six transcriptomes that yielded unigenes with a median length of 334 nt with the longest unigene being 28,680 nt in length. Finally, these reads were assembled into 125,385 transcripts and 89,332 unigenes, with N50 lengths of 1,831 and 1,499 nt, respectively (Table 2).
Then, we used Blast2GO to annotate unigenes into functional groups based on Gene ontology (GO). The GO annotations were used to classify the transcripts into functional groups according to specific GO categories. In the molecular function category, the genes expressed were mostly enriched for catalytic activity (e.g., hydrolase and oxidoreductase) and binding (e.g., nucleotide, ion, and odorant binding). In the biological process category, the most common were the cellular and metabolic processes. In the cellular component category, the most represented were the following two terms: cell (GO:0005623) and cell part (GO:0044464) (Additional file 2: Figure S1b).
The putative pathways of pheromone biosynthesis and degradation are shown in Fig. 2. We identified a number of genes related to the pathways by using the BLASTx search. Below, we describe the specific tissue expression patterns of these genes using qPCR. Fig. 1 Dissection of P. xylostella female sex pheromone glands. The pheromone glands from female P. xylostella were squeezed out from the abdomen using forceps (the gland is similarly inflated when the female calls). The abdomen of P. xylostella was cut at the sclerotized cuticle from the 8 th abdominal segment, and the sclerotized cuticle was removed before immersing the glands in liquid nitrogen. 1: Sclerotized ovipositor valves; 2: Pheromone gland; 3: Sclerotized cuticle that was removed Genes involved in the sex pheromone biosynthesis pathway Pheromone biosynthesis activating neuropeptide receptor (PBANr) Insect pheromone biosynthesis is regulated by PBANr that binds to PBAN released from the suboesophagal ganglion in the brain [5]. To date, no more than four alternative splice transcripts such as isoforms As, A, B and C with differences in their C-terminal sequences have been reported in one insect species [57]. B and C typically consist of the YXXΦ endosomal sorting motif and the complete seven transmembrane regions [58]. Although all four PBANrs are expressed dominantly in PGs, only PBANR-B and PBANR-C have been identified as intracellularly localized with some cell surface localization. PBANR-As and PBANR-A on the other hand, were completely intracellular, suggesting that PBANR-B and PBANR-C could be principal receptors participating in the PBAN signal reception [57]. In this study, we found two additional alternative splice transcripts (Px002138 and c63176_g2; see Additional file 3: Table S2) that possessed C-termini different from DBM PBANr genes previously described in the genome [59,60] (Additional file 4: Figure S2). We named them as PluxyPBANR-C, −B and -A. Multiple amino acid sequence alignment of these PBANr isoforms showed the presence of typical PBANr sequence features (Additional file 4: Figure S2). We further analyzed the tissue expression pattern of PluxyPBANR-C and -B by comparing their FPKM values (Additional file 3: Table S2 and Fig. 3) in the transcriptomes from the different tissues. The results showed at least 3-and 5-fold higher numbers for PluxyPBANR-C and -B transcripts, respectively in PG than in other tissues studied, and is consistent with reports from other moths such as Agrotis ipsilon [61], B. mori [62] and Heliothis virescens [63]. The first step in moth pheromone biosynthesis is the ATP-dependent carboxylation of acetyl-CoA to malonyl-CoA by ACC [4]. This is a rate-limiting step in fatty acid biosynthesis [51]. In our transcriptome data set, seven ACC transcripts were identified (Additional file 3: Table  S2) with all showing robust expression in the abdomen followed by PG (Additional file 3: Table S2). Among these, PxACC3, PxACC4 and PxACC5 showed higher expression than the others in PGs. In addition, homology analysis revealed that PxACC3 shared 73% amino acid identity with the ACC homolog from Papilio xuthus (Genbank XP_013176189.1), 70% with Helicoverpa assulta (Genbank (AKD01722.1) and 66% with D. plexippus (Genbank EHJ72299.1). PxACC4 is a full-length ACC, and the predicted protein from this transcript shared a high amino acid identity (83%) with the ACC from Agrotis segetum (AID66639.1) [64]. Similar to A. segetum ACC, PxACC4 also had dominant expression in the abdomen as well as in PG but without significant  In moths, saturated fatty acid is produced from malonyl-CoA and NADPH in a reaction catalyzed by FAS [4]. In this study, we found seven putative transcripts with high amino acid identities to known FAS genes (PxFAS1-PxFAS8) (Additional file 3: Table S2). Only one partial FAS transcript was found in the genome annotation, two novel transcripts were identified via genome mapping and four new FAS transcripts were found among the unigenes after de novo assembly. None of the seven PxFAS genes showed dominant expression in PGs (Additional file 3: Table S2). Five of the PxFASs displayed abdomen biased expression, unlike the two homologs, FAS from A. ipsilon (AGR49310.1) and A. segetum (AID66645.1) that shared 71 and 72% amino acid identity with PxFAS5 and PxFAS1, respectively and were expressed higher in PGs than in the abdomen [61,64]. The fatty acid transport protein (FATP) has been functionally characterized by in vitro expression and knock down assay in B. mori [65] and Eilema japonica [66]. These studies showed that FATP catalyzes the ATP-dependent esterification of extracellular long-chain fatty acids in sex PGs to produce the corresponding acyl-CoA derivatives. Four PxFATPs unigenes were found in this study; PxFATP1 shared high amino acid identity (74%) with FATP from E. japonica (BAJ33523.1); PxFATP2 and 4 shared 58 and 81% amino acid identity, respectively with FATP2 and FATP3 from Sesamia inferens [36]. All three PxFATPs displayed ubiquitous expression (Additional file 3: Table S2) indicating the lack of specialized function in the sex pheromone biosynthesis pathway. PxFATP3 was not detected in any tissue in our transcriptome data, and thus maybe a pseudogene (Additional file 3: Table S2).  [48,69]. Thus, Δ11 desaturase is expected to play a key role in this pheromone component biosynthesis. Investigation of the presence of desaturases in our transcriptome data set using genome mapping and de novo transcriptome assembly revealed 16 PxylDes transcripts (Additional file 3: Table S2). FPKM and qPCR analysis revealed that PxylDes12 and 16 were expressed abundantly in PG (Additional file 3: Table S2, Figs. 3 and  4). In addition, phylogenetic analysis showed that PxylDes12, 14, 15 and 16 clustered into the Δ11 clade, which also had the SinfDes4 homolog from S. inferens [36] ( Fig. 5). Thus, we propose that the two PxDes may be involved in the desaturation of Z11-16:CoA, the major sex pheromone component in DBM. Ten additional PxylDes were assigned to other desaturase groups and also did not have PG biased expression. These genes may also be involved in other physiological functions rather than sex pheromone component desaturation.
In DBM, Z11-16: OH is a main sex pheromone component [51,52], which is produced by the reduction of fatty acyl precursor, Z11-16:CoA, in an NADPH dependent reaction [22,[70][71][72]. In this study, we identified in total, seven putative FAR transcripts, named PxFAR1-7 (Additional file 3: Table S2). Among them, PxFAR6 and PxFAR7, clustered with the pgFAR clade (Fig. 6). All FARs in this group displayed PG dominant expression, including the first identified pgFAR from B. mori [70], SlitFAR3 from S. litura [73] and SinfFAR2 from S. inferens [36]. Together, the FPKM value and qPCR analysis showed that among the seven PxFARs, PxFAR6 had the highest expression (20-fold higher) in PGs when compared to the other tested tissues (Figs. 3 and 5). Therefore, PxFAR6 may likely play a major role in the conversion of Z11-16:CoA to Z11-16:OH in DBM. In addition, we found another FAR predominantly expressed in PG; PxFAR5 that did not cluster with the pgFAR group. It shared high amino acid  identity (71%) with YevoFARI from Yponomeuta evonymellus (ADD62438.1). A previous report on the PG specific expression of YevoFARI determined by qPCR analysis was also not clear [22]. The expression and functional analyses of this clade need further investigation.

Acetyltransferase (ACT)
ACT is known to convert fatty acid alcohol to the corresponding ester in plants [74][75][76]. However, no ACT related to sex pheromone component biosynthesis has been thus far identified at the molecular level. In this study, we identified 21 ACT transcripts, PxylACT1-22, by homolog searches using sequences (BLASTp) obtained from our previous studies as query; SlitACTs from S. litura [73], SinfACTs from S. inferens [36] and AsegACTs from A. segetum [64] (Additional file 3: Table S2). We found that only PxACT7 displayed PG abundant expression via FPKM and qPCR analysis (Additional file 3: Table S2, Figs. 3 and 4).
Further, BLASTx showed that PxACT7 shared a low amino acid identity (39%) with AsegACT (AIN34700.1) [64], which was shown to lack esterification activity by using in vitro yeast expression system and GC-MS analysis [65]. Not surprisingly, AsegACT was not abundant in PG, and its FPKM value in PG was almost equal to that in the abdomen [15]. Three ACTs from Agrotis ipsilon, ATF-KC007357, ATF-KC007358 and ATF-KC007360, were mainly expressed in PG than the rest of the body [61]. Their homologs in DBM are PxACT17, 19 and 14, which shared 65, 75 and 91% amino acid identities, respectively. However, none of these displayed PG dominant expression (Additional file 3: Table S2).

Alcohol dehydrogenase (AD)
Alcohol dehydrogenase converts alcohol compounds to corresponding aldehydes. However, no alcohol dehydrogenase has been identified at the molecular level from insect PGs although biochemical studies have shown the presence of alcohol dehydrogenses in the PGs of several species such as Heliothis virescens [77], Helicoverpa armigera [78] and H. assulta [78]. Z11-16:Ald was found to be a major sex pheromone in DBM, indicating that AD plays a crucial role in its biosynthesis. In this study, we conducted a BLASTp analysis by using previously identified SinfADs from S. inferens [36], AipsADs from A. ipsilon [61], CpomADs from Cydia pomonella [79], HassADs from H. assulta [78], HarmADs from H. armigera [78] and HvirAD from H. virescens [77]. We found in total, 22 ADs in DBM, named PxylAD1-22  Table S2). The FPKM values of these 22 PxylADs revealed that only one AD (PxylAD9) had PG dominant expression. qPCR assay also confirmed this (Figs. 3 and 4) and showed at least 2-fold higher expression in PG than in other tested tissues (Fig. 4). We also found several homologs of previously identified ADs from H. armigera (HarmAD5)  Aldehyde oxidase is considered as a sensillum enzyme that degrades redundant aldehyde odorants (including aldehyde sex pheromone components) to corresponding acids without the need for coenzymes [31][32][33]. We found three full-length AOXs in the DBM transcriptomes via genome mapping. These were named PxylAOX1, PxylAOX2 and PxylAOX3 (Additional file 3: Table S2), all containing the typical conserved motifs: a FAD-binding domain, two putative iron-sulfur (2Fe-2S) redox centers, and a molybdenum-cofactor (MoCo)binding site. One 2Fe-2S redox center contains four conserved cysteine residues that coordinate with iron. PxylAOX1 and PxylAOX3 likely originated from a gene duplication event because they were located in the same scaffold (scaffold_267). Based on FPKM values and qPCR analysis, PxylAOX1 and PxylAOX3 were identified to have antennae enriched expression without sex-bias; PxylAOX1 expression in the antennae was more than 60-fold higher and PxylAOX3 expression was 40-fold higher than in other tissues. In addition, PxylAOX1 expression was higher than PxylAOX3 (Additional file 3: Table S2, Figs. 3 and 4), while PxylAOX2 showed ubiquitous expression and therefore may perform a more generalized function. We also noticed a putative AOX identified in the DBM genome (ID: CCG014884.1) [54]. This sequence was identified as a xanthine dehydrogenase (XDH) member by BLASTx search. A phylogenetic tree constructed by using known insect AOXs and XDH as an out group (Fig. 7) showed the distribution of insect AOXs into three major groups; group 1 consisted of , all of which have antennae enriched expression pattern. However, like other AOXs, PxylAOX1 and 3 did not show sex biased expression suggesting that they may be specifically involved in aldehyde plant volatiles degradation rather than sex pheromone component pathways.

Aldehyde dehydrogenase (ALDH) and Aldehyde reductase (AR)
Unlike AOX, both ALDH and AR are thought to be intracellular and NAD(P) − dependent enzymes that catalyze the oxidation and reduction of a broad spectrum of aldehyde substrates to their corresponding acids and alcohols, respectively. Five PxylALDHs (named PxylALDH1-5) were identified from the DBM transcriptome data set. Among them, only PxylALDH4 was mainly expressed in the antennae as determined by the FPKM value and qPCR analysis (Additional file 3: Table S2). The expression level of PxylALDH4 in PG was at least 2-fold higher than in other tested tissues, indicating its likely function in odorantdegradation. Similar to AOX, this gene also showed no sexbiased expression suggesting that it is involved in plant volatiles degradation. A previous study showed that one ALDH, AaALDH3, oxidizes farnesal into farnesoic acid in Aedes aegypti [39]. But this gene only shared 11% amino acid identity with PxylALDH4, suggesting that they could act on different substrates. The ARs in general, include four reductase superfamilies; medium-chain dehydrogenases/reductases (MDR), short chain dehydrogenases/reductases (SDR), aldo-keto reductases (AKR) and NADB (NAD + binding) reductases. Through motif and BLAST analyses, we found in total, 15 ARs among which were six MDRs (PxylAR2, 3, 5, 6, 14 and 15), eight AKRs (1, 7, 8, 9, 10, 11, 12 and 13) and one NADB (PxylAD4). However, no SDR was found (Additional file 3: Table S2). In B. mori [82] and H. armigera [38], two AKRs have been characterized in vitro; these were reductases that converted the sex pheromone components, E10,Z12-16:Ald (B. mori), Z9-16:Ald (H. armigera) and Z11-16:Ald (H. armigera) to corresponding alcohols. FPKM values did not show AR expression specific to the antennae. Two AKRs shared the most amino acid identities with PxylARs; these were PxylAD9 (59%) and PxylAD7 (37%), but PxylAD9 and PxylAD7 displayed ubiquitous expression, indicating their involvement in other physiological functions instead of olfaction alone.

Carboxyl/cholinesterase (CCE)
Insect CCE is a superfamily, which participates in many physiological processes and contains various substrates [83]. Some of them are secretory enzymes that refresh ORs by clearing redundant esterase odorants around them. Besides Z11-16:Ald, Z11-16:Ac is another major sex pheromone component reported in P. xylostella [48]. We found 49 CCEs in the DBM transcriptome that was  Table S2). This total number is lower than 69 identified in B. mori [84,85], but higher than 39 found in H. armigera [86]. A phylogenetic tree of known insect CCEs is shown in Fig. 8. These CCEs were named from CCE001-30 sub-groups based on the B. mori [84,85] and H. armigera [86] 13 major clade nomenclature system. By FPKM values and qPCR analysis, five CCEs were found to have antennae enriched expression (Additional file 3: Table S2, Figs. 3 and 4); PxCCE005a, 016a, 016c, 016d and 029a. These were distributed into three clades. PxCCE016a, 016c and 016d were located in the same sub-clade containing most general intracellular enzymes with dietary detoxification functions or ester odorant degradation functions such as the two CCEs with antennae dominant expression, SlCXE10 [87] and SexiCXE10 [40], that degrade ester plant volatiles. Unlike these two CCEs, PxCCE016a, 016c and 016d had no obvious sex biased expression. Other CCEs involved in non-olfactory functions have been described in some previous studies [88][89][90], We found the homologs of these genes among the PxCCEs, and they did not show antennae dominant expression, thus they may not participate in olfaction. For instance, in Group E, the pheromone degrading   [92], SexiCXE13/SlitCXE13 from S. exigua and S. litura [43], and DmelEst6 from D. melanogaster [93] were able to hydrolyze ester sex pheromone components. These genes showed abundant expression in the antennae except SexiCXE13 and SlitCXE13. PxCCE026a was their closest homolog, but it showed ubiquitous expression by FPKM analysis. Moreover, PxCCE016b also showed ubiquitous expression. A recent study showed that PxCCE016b was involved in chlorantraniliprole resistance; its increased expression in resistant strains enabled better degradation of ester pesticides [94]. The four 016 sub-group CCEs are derived from the same ancestor because they clustered in the same clade in the phylogenetic tree and may have evolved to perform various physiological functions based on the difference in tissue expression patterns between PxCCE016b and three other PxCCE016 CCEs. PxCCE029a and PxCCE005a were in the uncharacterized functional clade. The 029 clade CCEs lacked one or more of the enzyme catalytic active site residues, but had another active site called carboxyl/cholinesteraselike adhesion molecules including transmembrane and binding domains [95]. Thus, PxCCE029a may be involved in neuro/developmental functions. We also investigated the intron-exon organization in PxCCEs (Fig. 9). The average intron size was 1290 nt, which was slightly lower than 1372 in B. mori [85]. The longest intron was between exons 2 and 3 in PxCCE027b and comprised 10,014 nt, similar to that observed in BmCCE027b [85]. The shortest intron was in PxCCE028b between exons 8 and 9, and contained only 33 nt. Further, the approximate distribution of introns is shown in Fig. 9, which shows the positions of introns in PxCCEs by multiple sequence alignment. We also collected and marked the splice sites into three phases according to their positions in the codons. We noticed the presence of two highly conserved intron insertion sites and the splice site phase was also conserved. The two most conserved introns were a phase 2 intron around position 1500 present in 42 CCEs (Fig. 9, red arrow) and a phase 0 intron around position 500 in 37 CCEs (Fig. 9, gray arrow). The two conserved intron insertion sites and splice site phases were identical to that reported in B. mori CCEs [85], indicating that these sites were likely conserved since the early stage of CCE evolution. In addition, several clade specific conserved sites were also observed and were similar to the clade-specific conservation of intron phase and position reported in P. xylostella GSTs [56]. All CCEs in clade 020 possessed two conserved introns, phase 2 introns at positions 700 and 800 (Fig. 9, blue and dark brown brackets, respectively). Notably, two intron insertions and phase splice sites were conserved in two distant clades, 024-026 and 030 (Fig. 9, light brown and gray brackets, respectively). These conserved introns were also observed in clades 020, 024, 026 and 030 in B. mori and the neuroligins of D. melanogaster and A. mellifera [85].

UDP-glycosyltransferases (UGT)
UGT is a superfamily of detoxifying enzymes that improve the excretion of certain molecules by transferring glycosyl residues from nucleotide sugars; for example, hydrophobic molecules are thus converted to hydrophilic molecules for easy excretion. Typical examples of such molecules in insects are odorant molecules [46,96,97]. In the present study, 26 UGTs were found via genome mapping and transcriptome assembly of all six transcriptomes from DBM (Additional file 3: Table S2). All PxUGTs presented the classic UGT signature motif (N-terminal substrate binding site and N-glycosylation sites) similar to that described for other insect UGTs [46,97,98]. These UGT sequences have been submitted to the UGT Nomenclature Committee (http://prime. vetmed.wsu.edu/resources/udp-glucuronsyltransferasehomepage) for nomenclature assignments [97] (the subgroup names are shown in Additional file 3: Table S2).
Similarly, most PxUGTs were clustered in UGT33 (n = 10) and UGT40 (n = 4) sub-families. In the UGT33 subfamily, BmUGT33D8 [99] and one UGT from M. sexta (AI234470.1) [100] displayed antennae enriched expression. By using FPKM values and qPCR analysis, we also identified three UGT33 family members among PxUGTs (PxUGT33AA1, PxUGT33W1 and PxUGT33Z1) that displayed antennae dominant expression (Additional file 3: Table S2, Figs. 3 and 5) indicating their olfactory function. Recently, two more UGTs from UGT40 and 46 sub-families, SlUGT40R3 and SlUGT46A6 from S. littoralis, showed abundant expression in the male antennae after exposure of moths to the sex pheromone component, Z9E11-14:Ac or an ester plant volatile, Z3-6:Ac. This suggested that the two UGTs may be odorantdegrading enzymes. We also found a UGT40 member in DBM, PxUGT40W1, which displayed the highest expression in antennae than in other tissues. However, there was no significant difference in the expression of PxUGT40W1 between male and female antennae, thus indicating that it may be involved in the removal of plant volatiles rather than sex pheromone components. In addition, PxUGT39B5 of the UGT39 sub-family also had an antennae enriched expression pattern indicating its olfactory function. However, the functions of its homologs have not been reported yet.

Cytochrome P450
Cytochrome P450 is a multi-functional enzyme in organisms, and is involved in detoxification, resistance, and odorant clearance [44,101]. A previous study annotated 85 P450 genes (PxCYPs) in DBM but the tissue specific expression pattern was investigated only in the larval stages but not in adult tissues [55]. In the present study, the 85 PxCYPs were analyzed in various adult tissue transcriptomes and validated via qPCR (Additional file 3: Table S2, Figs. 3 and 4). We found that eight PxCYPs had predominant expression in the antennae of both sexes, and belonged to three CYP sub-families, CYP3: PxCYP321E1, 366C1; CYP4: PxCYP 4G77, 4 M22, and CYP6: PxCYP6AN14, 6BG3, 6BG6 and 6CN1 (Additional file 3: Table S2, Figs. 3 and 4). One CYP3 member, CYP345E2, which is an antenna-specific CYP from the pine beetle, D. ponderosae, was shown to catalyze the oxidation of monoterpene volatiles from pine hosts [44].
In our study, we found 2 CYP345E2 homologs in DBM; PxCYP321E1 and PxCYP366C1, both with higher expression in the antennae. In addition, PxCYP366C1 had the highest expression in the antennae when compared to other candidate odorant-degrading CYPs. This gene was also reported as highly expressed in adult heads from different developmental stages [55], indicating that PxCYP366C1 plays a vital role in specific olfactory behaviors of DBM adults; an example of such behavior could be locating oviposition sites. Some members of the CYP4 sub-family were also identified to be olfactory related. CYP4AW1 in P. diversa was reported to be involved in pheromone inactivation by analyzing its antennae specificity and through electrophysiological studies [45]. In DBM, we identified 2 CYP4 genes; PxCYP4G77 and 4 M22 both with robust expression in the antennae. A previous study reported the high expression of PxCYP4M22 in an insecticide-resistance strain than in the susceptible strain [55]. Together with its tissue specific expression pattern, we propose that PxCYP4M22 may participate in insecticide detoxification in the antennae of the insecticide-resistance strain. In the CYP6 sub-family, no member has been reported to have odorant clearance function. However, several genes have been reported to have olfactory organ specific expression; for example CYP6W1 [102] and CYP6A20 [103] in D. melanogaster, and CYP6B48, CYP6B42 and CYP6AE49 expressed in the antennae of both sexes and male proboscis in S. littoralis [104]. We found four CYP6 members in DBM (PxCYP6AN14, 6BG3, 6BG6 and 6CN1) that were abundant in the antennae. PxCYP6BG3 and 6BG6 had high expression in the larvae than in adults [55], suggesting that the two CYP genes may be involved in odorant clearance in larvae.

Glutathione S-Transferase (GSTs)
The functions of GSTs in many insect physiological functions such as insecticide resistance [56] and detoxification [105,106] have been well-documented. In this study, we identified in total, 25 PxGSTs from the transcriptomes of DBM. In comparison with a previous study [56] we identified three new GSTs, PxGSTe6, PxGSTe7 and PxGSTu3 in DBM (Additional file 3: Table  S2). Since PxGSTu3 was not detected in the transcriptomes of all tested tissues we presume that this may be a pseudogene. One olfactory-specific GST from M. sexta named, GSTmsolf1, was reported to degrade the plant volatile, trans-2-hexenal [47]. When we used the deduced amino acid sequence of this protein as query in a BLASTp search, we found that it belonged to the GST sub-family Delta. However, all three antennae enriched PxGSTs we identified in this study belonged to the GST sub-family Epsilon (Additional file 3: Among these genes, two AOXs, one ALDH, five CCE, five UGTs, eight P450s and three GSTs displayed antennae dominant expression, and were proposed to be involved in sex pheromone components and plant volatiles degradation. Together, our data provides a strong reference for the genes involved in sex pheromone biosynthesis and degradation pathways, and will be a useful gene set for future gene function studies. This data set may also provide potential gene targets to design pest control strategies based on disrupting pheromone-based behaviors of insect pests.

Insect samples
The 5 th instar larvae of DBMs were purchased from an insect rearing facility in Henan province, China. The rearing conditions were 27°C, 14 L : 10D photoperiod and 65% ± 5% relative humidity (RH). We collected approximately 300 antennae, 300 heads, 300 legs, and 100 abdomens from 3-day old adults of both sexes (male/female = 1:1), and 100 female sex pheromone glands for transcriptome sequencing. Three biological replicates of these samples were also collected for qPCR. Ovipositors that were adhered to the abdomens were removed as much as possible to avoid tissue contamination. The tissue samples were stored in liquid nitrogen at −80°C until further use. and the library quality was assessed on an Agilent Bioanalyzer 2100 system.
De novo assembly of short reads and gene annotation Raw data (raw reads) in fastq format were first processed through in-house perl scripts. In this step, clean data (clean reads) were obtained by removing the adapter sequences, reads containing poly-N, and low quality reads from the raw data. Simultaneously, Q20, Q30, GCcontent, and sequence duplication level of the clean data were calculated. All downstream analyses were based on clean data with high quality. De novo transcriptome assembly was performed using the short reads assembly program Trinity (version: r20140413p1) with min_ kmer_cov set to 2 by default and all other parameters were set to default values. The overlap settings used for the assembly were 30 bp and 80% similarity, and all the other parameters were set to their default values. Unigenes >150 bp were aligned by BLASTx with protein databases, including Nr, Swiss-Prot, KEGG, and COG (e-value < 10-5), to identify proteins with high sequence identity and to assign putative functional annotations. Next, we used the Blast2GO program (version: b2g4pipe_v2.5, e-value = 1.0E-6) (https://www.blast2go. com/) to obtain GO annotations of the unigenes and we obtained the GO functional classifications using the WEGO software (http://wego.genomics.org.cn/cgi-bin/ wego/index.pl).

Reads mapping to the reference genome
Reference genome and gene model annotation files were downloaded directly from the DBM genome website: http://iae.fafu.edu.cn/DBM/. Index of the reference genome was built using Bowtie v2.2.3 and paired-end clean reads were aligned to the reference genome using TopHat v2.0.12. We selected TopHat as the mapping tool to generate a database of splice junctions based on the gene model annotation file to yield a better mapping result compared to other non-splice mapping tools.

Quantification of gene expression level
HTSeq v0.6.1 was used to count the read numbers mapped to each gene. Then, FPKM of each gene was calculated based on the length of the gene and reads count mapped to this gene. FPKM, which is the expected number of Fragments Per Kilobase of transcript sequence per Millions base pairs sequenced, considers the effect of sequencing depth and gene length for the reads count at the same time, and is currently the most commonly used method for estimating gene expression levels [107].

Differential expression analysis
Prior to differential gene expression analysis, for each sequenced library, the read counts were adjusted by the edgeR program package through one scaling normalized factor. Differential expression analysis of two samples was performed using the DEGSeq R package (1.20.0). The P values were adjusted using the Benjamini & Hochberg method. Corrected P-value of 0.005 and log2 (Fold change) of 1 (for de novo assembly: P value was adjusted using q value. qValue < 0.005 & |log2(foldchange)| > 1 was set as the threshold for significant differential expression).

Novel transcripts prediction
The Cufflinks v2.1.1 Reference Annotation Based Transcript (RABT) assembly method was used to construct and identify both known and novel transcripts from TopHat alignment results.

Phylogenetic tree
Amino acid sequences of selected AOX, CCE, FAR and DES were aligned with MAFFT (E-INS-I parameter). Thereafter, PhyML 3.1 with WAG substitution model was used to construct a maximum likelihood phylogenetic tree using Bayesian analysis. The deduced protein sequences used in the phylogenetic tree are listed in Additional file 5. Finally, the trees were viewed and group edited with FigTree v1.4.2 (http://tree.bio.ed. ac.uk/software/figtree/).

RNA extraction and cDNA synthesis
Total RNA was extracted using EasyPure RNA Kit (TransGen Biotech, Beijing, China) following the manufacturer's instructions, in which DNase digestion was included to avoid the genomic DNA contamination. RNA quality was checked with a spectrophotometer (NanoDrop 2000, Thermo Fisher Scientific, USA). The single-stranded cDNA templates were synthesized from 1 μg total RNA from various tissue samples using the PrimeScriptRT Master Mix (TaKaRa, Dalian, China) at 42°C for 1 h. The reaction was terminated by heating at 70°C for 15 min.
Quantitative real time PCR and data analysis qPCRs were performed for each sample using an iCycle iQ (Bio-Rad, CA, USA) according to the minimum information for publication of quantitative Real-Time PCR Experiments [108]. Gene-specific primers were designed by Beacon Designer 7.6 (PREMIER Biosoft International, CA, USA) and are listed in Additional file 6: Table S3. The mRNA levels were measured in three technical replicates from each of the three biological replicates by qPCR using TransStart Tip Green qPCR SuperMix, as described by the manufacturer (TransGen Biotech, Beijing, China). The mRNA levels were quantified using ribosomal protein L8 (RPL8) as the reference gene [56]. Means and standard errors were calculated based on at least two biological replicates. Relative expression level of the mRNAs for genes was calculated according to the 2 −ΔΔCq method. Relative fold-changes in the different tissues were calculated and normalized based on the transcript levels in the bodies.