Structural and functional insights into the Diabrotica virgifera virgifera ATP-binding cassette transporter gene family

Background The western corn rootworm, Diabrotica virgifera virgifera, is a pervasive pest of maize in North America and Europe, which has adapted to current pest management strategies. In advance of an assembled and annotated D. v. virgifera genome, we developed transcriptomic resources to use in identifying candidate genes likely to be involved in the evolution of resistance, starting with members of the ATP-binding cassette (ABC) transporter family. Results In this study, 65 putative D. v. virgifera ABC (DvvABC) transporters were identified within a combined transcriptome assembly generated from embryonic, larval, adult male, and adult female RNA-sequence libraries. Phylogenetic analysis placed the deduced amino-acid sequences of the DvvABC transporters into eight subfamilies (A to H). To supplement our sequence data with functional analysis, we identified orthologs of Tribolium castaneum ABC genes which had previously been shown to exhibit overt RNA interference (RNAi) phenotypes. We identified eight such D. v. virgifera genes, and found that they were functionally similar to their T. castaneum counterparts. Interestingly, depletion of DvvABCB_39715 and DvvABCG_3712 transcripts in adult females produced detrimental reproductive and developmental phenotypes, demonstrating the potential of these genes as targets for RNAi-mediated insect control tactics. Conclusions By combining sequence data from four libraries covering three distinct life stages, we have produced a relatively comprehensive de novo transcriptome assembly for D. v. virgifera. Moreover, we have identified 65 members of the ABC transporter family and provided the first insights into the developmental and physiological roles of ABC transporters in this pest species.

as a novel insect pest control technology [18], especially in instances where target species are sensitive to oral RNAi [19]. D. v. virgifera is highly sensitive to oral RNAi [20][21][22], suggesting that it could suppress feeding damage caused by this pest [23].
ATP-binding cassette (ABC) proteins comprise one of the largest gene families, and are found across prokaryotic and eukaryotic domains [24]. Most of these proteins function as transmembrane transporters, which actively move a myriad of molecules across cellular membranes [25]. ABC transporter proteins have a two-domain structure: a highly conserved nucleotide-binding domain (NBD) and a variable transmembrane domain (TMD) [26]. The NBD binds and hydrolyzes ATP to provide the energy required for translocating a substrate across cell membranes, while the TMD forms a channel through which the substrate is transported [27]. Each NBD possesses several highly conserved, characteristic motifs, including Walker A, Walker B, Q-loop, D-loop, H-loop, and ABC signature motifs, while each TMD is made up of five to six transmembrane α-helices that dictate substrate specificities [27]. ABC transporter proteins require two NBDs and two TMDs for functionality. Some ABC transporters are full-transporters (FT) in that two TMDs and two NBDs are encoded in a single protein, whereas most are half-transporters (HT; one TMD and one NBD) and form functional units following homodimerization or heterodimerization [24,27,28]. Due to the relatively conserved sequence of the NBD, it has been used for the phylogenetic classification of the ABC transporter superfamily into eight subfamilies designated A to H (ABCA to ABCH) [29].
Among insect species, ABC transporters are implicated in diverse functions, including transportation of eye pigments [30][31][32][33][34], and resistance to chemical insecticides [35,36]. Within the model species for Coleoptera, the red flour beetle, Tribolium castaneum, Broehan et al. [30] reported that RNAi-mediated knockdown of some ABC transporters resulted in mortality, or phenotypes characterized by arrested growth, abnormal cuticle formation, defective eye pigmentation, or abnormal egg-laying or -hatching. Changes in the expression level or structure of some ABCA, ABCC and ABCG subfamily members have been associated with Bt toxin resistance in species of Lepidoptera [37], while paralogs of an ABCB transporter were linked to Bt Cry3Aa resistance in the coleopteran species, Chrysomela tremula [38], and were found to be in proximity to a quantitative trait locus (QTL) for Cry3Bb1 resistance in D. v. virgifera [39].
Similar investigations of ABC transporters in D. v. virgifera are arguably limited due to the dearth of genomic resources available for this species, which are currently comprised of Sanger and Roche 454 read-based transcriptome assemblies [40][41][42][43]. Complicating the development of genomic tools is the 2.58 GB size and complex repetitive structure of the D. v. virgifera genome [4,44]. Regardless, RNA sequencing (RNA-seq) has become an expeditious and cost-effective method for obtaining a wealth of transcriptome sequence data in non-model insects [45]. In the following, a de novo transcriptome assembly approach was used for the first prediction, annotation, and functional analysis of the ABC transporter gene family in D. v. virgifera. Specifically, eight ABC transporters were identified as putative orthologs to those previously reported to have a defining RNAi phenotype in the model coleopteran species, T. castaneum [30] (DvvABCA_ 50718, DvvABCB_39715, DvvABCE_2830, DvvABCF_ 2701, DvvABCG_3712, DvvABCG_14042, Dvvw and DvvABCH_5118). Subsequent RNAi-mediated knockdown demonstrated conservation of function with T. castaneum, as well as established potential new insecticidal targets for the control of this devastating agricultural pest.

Results
Transcriptome sequencing, assembly, and annotation Over 22 million raw Illumina (MiSeq) sequencing reads were generated across four libraries (  Table S1). Analogously, assemblies from Trinity and SOAPdenovo-Trans respectively produced 162,897 and 133,180 contigs, each with an N50 ≤ 439 bp (Additional file 1: Table S1). Clustering by CD-HIT-EST reduced complexity 3.5 to 34.9% across assemblies, and the number of predicted open reading frames (ORFs) within clustered transcripts ranged from 18,305 to 40,087 (Additional file 1: Table S1). BLASTx query of transcripts by Blast2GO against the arthropod-specific section of NCBI's non-redundant (nr) protein database generated annotations for 18,343 DNASTAR contigs (Evalue cutoff of 10 − 6 ), with a subset of these receiving gene ontology (GO) mapping and additional annotation terms (Additional file 2: Figure S1). Sequences lacking identity to known arthropod proteins above E-value thresholds were attributed to poor sequence conservation and/or novel sequences, as well as non-coding RNAs. The distribution of top BLASTx hits by species showed that T. castaneum was the most frequent, representing 65% of the matches (Additional file 3: Figure S2). Among ontologies assigned via mapping at GO level 2, a majority of the associated terms were assigned to cell structural component, metabolic process, and catalytic activity respectively for GO Cellular Component, Biological Process and Molecular Function (Fig. 1).
The DNASTAR assembly showed a high degree of completeness based on a BUSCO score of 928, or 89.6%, of the 1066 genes in the arthropod reference set (v. 9.0) being represented, with analogous levels of representation in both SOAPdenovo-Trans and Trinity assemblies (Additional file 1: Table S1).    Figure S3).

Gene expression across developmental stages
Since prior research in T. castaneum revealed that only 10 ABC transporters had obvious phenotypic consequences following RNAi-mediated knockdown [30], our study focused on functional analysis of their predicted D. v. virgifera orthologs. From this list, our initial predictions from the D. v. virgifera transcriptome (DNASTAR assembly) identified eight orthologs (Table 3). Differences resided in that T. castaneum has two closely related ABCA genes (TcABCA-9A and TcABCA-9B) which appear to represent a T. castaneum-specific duplication (Additional file 5: Figure S3). We were unable to identify a direct ortholog for these genes, but the closest homolog we found in the D. v. virgifera transcriptome appeared to be DvvABCA_50718. Analogously, we were unable to identify a direct ortholog to TcABCG-8A, so we targeted DvvABCG_14042, the closest identifiable homolog according to BLASTp results. Finally, while the ABCG genes TcABCG-9A and TcABCG-9B represent the orthologs of the T. castaneum eye-color genes scarlet and white, respectively [32], results of BLASTx searches of the DNASTAR assembly resulted only in the identification of an ortholog of white, Dvvw [46]. Semi-quantitative PCR of these eight D. v. virgifera ABC transcripts showed that all are expressed across all of the developmental stages examined (Fig. 3a).

RNAi knockdown phenotypes
Different growth stages of D. v. virgifera were microinjected with dsRNAs (Table 3), after which the level of each corresponding transcript was below or nearly below semiquantitative PCR detection limits. Specifically, the level of each targeted D. v. virgifera transcript was reduced at 5days post-injection as compared to buffer-injected controls (Fig. 3b). Moreover, injection of each of the eight dsRNAs resulted in defined phenotypes among dsRNA treated cohorts (Table 3; Fig. 4). The knockdown of DvvABCA_50718 led to approximately 60% mortality among treated prepupae, compared to 5% for the buffer-treated control group. In addition, the adults that survived pre-pupal injection and successfully eclosed had defects in their wings and elytra (Fig. 4a), while no phenotypic effects were observed among buffer-injected controls. Injection of DvvABCB_ 39715 dsRNA into larvae resulted in 100% mortality, and injections into pre-pupae led to defects in their development, which caused individuals to be unable to complete the pupal-adult molt and ultimately resulted in 100% mortality (Fig. 4b). Knockdown of DvvABCB_39715 in newly-eclosed adult female D. v. virgifera resulted in significant reduction in egg laying compared to untreated females (Fig. 5a). Upon further investigation, we discovered that injection of this dsRNA also affected ovary development, causing underdeveloped ovaries, hence the failure to produce eggs (Table 3; Fig. 4j).  RNAi-mediated knockdown of DvvABCE_2830 and DvvABCF_2701 in larvae resulted in 100% mortality. Prior to death, it was noted that the body mass of treated individuals was less than that of similarly-aged larvae treated with buffer alone (Fig. 4g, h). Analogously, injection of DvvABCE_2830 and DvvABCF_2701 dsRNA separately into pre-pupae both caused 100% mortality with no adult eclosion (results not shown). Injection of dsRNA specific for DvvABCH_5118 into early-instar D. v. virgifera larvae and pre-pupae caused development to arrest as individuals prepared to molt, thus resulting in 100% mortality (Fig. 4f). Affected individuals appeared to desiccate prior to death (personal observation).
Injection of dsRNA targeting DvvABCG_3712, DvvABCG_14042, and Dvvw resulted in phenotypes similar to those seen with RNAi knockdown of the corresponding T. castaneum orthologs [30]. Specifically, injection of dsRNA targeting Dvvw, gave the expected white-eye phenotype (Fig. 4e); indeed, we had identified this white ortholog previously [46]. Injection of DvvABCG_3712 dsRNA into pre-pupae caused developmental defects that resulted in 80% mortality (Table 3; Fig. 4c). Interestingly, adult females treated with DvvABCG_3712 dsRNA produced fewer eggs compared to females injected with buffer alone (Fig. 5b), and the eggs that were laid lacked obvious signs of embryonic development (Fig. 4i) and ultimately failed to hatch (Additional file 6: Figure S4). Injection of DvvABCG_14042 dsRNA into larvae and pre-pupae resulted in molting defects; about 80% of these died during their next molt (Table 3), while the 20% that survived through subsequent larval molts died following pupation (Fig. 4d).

Discussion
In recent years, ABC transporters have become a major focus for research in arthropods. This is in part due to their overall role in xenobiotic transport and insecticide resistance [25,[47][48][49][50], but more specifically, due to their suspected role in susceptibility to Bt toxins [38,51,52]. For example, Gahan et al. [53] reported genetic linkage of Heliothis virescens HvABCC2 with resistance to Cry1Ac, while changes in the structure, splicing, or expression level of ABCC2 orthologs were later associated with Cry1Ac resistance in Helicoverpa armigera [54], Bombyx mori [55], and Spodoptera exigua [56]. Indeed, expression of the P. xylostella ABCC2 ortholog in Drosophila melanogaster conferred susceptibility to this lepidopteran-specific toxin [57]. An ABCC2 ortholog is also linked to Cry1F resistance in Ostrinia nubilalis [58] and S. frugiperda [59]. Additionally, structural mutations in a member of subfamily A, HaABCA2, were implicated in Cry2Ab resistance in H. armigera [60], and, more recently, researchers were able to recapitulate an ABCA2 resistance allele in a susceptible population of H. armigera [61], providing further evidence for the importance of normal ABCA2 function in Cry2Ab toxicity. Reduced expression of ABCG members have been associated with Cry1Ac resistance in P. xylostella [62], as well as Cry1Ac and Cry1Ab resistance in O. furnacalis [63]. More recent studies in species of Coleoptera have implicated ABCB subfamily members in Cry3Aa resistance in C. tremula [38] and in Cry3Ab1 resistance in D. v. virgifera [39].
The study of ABC transporters in several arthropod species have relied on genomic data, including T. castaneum [30], Aethina tumida [64], B. mori [33], D. melanogaster [28], Bemisia tabaci [50], Daphnia pulex [65], and Tetranychus urticae [66]. Due to the status of D. v. virgifera as a major pest of cultivated maize (see Introduction) and current fragmented state of the unpublished draft genome assembly of this species (GenBank accession PXMJ00000000.2), the Illumina-based transcriptome assemblies reported here represent a particularly valuable genetic tool for gene discovery, characterization, and genome annotation. In particular, the 65 ABC transporter genes we identified are expected to be useful in downstream studies on insecticide resistance traits in D. v. virgifera.
Broehan et al. [30] previously identified 73 ABC transporters in T. castaneum, and a 74th ABC transporter was more recently reported by Grubbs et al. [32]. There are several possible reasons for why the 65 DvvABC transporters we identified are comparatively fewer than in T. castaneum. Firstly, our transcriptome was derived from lowerthroughput sequencing data (Illumina MiSeq), therefore genes expressed at very low levels may not have been represented within our raw Illumina data. Secondly, our RNAseq libraries were not comprehensive of all possible life/ growth stages or conditions, such that transcripts not expressed during growth states or under conditions used in Fig. 4 Effects of DvvABC transporter-specific RNAi on D. v. virgifera development. Buffer-injected controls (right) are shown next to dsRNA-injected individuals. a Injection of DvvABCA_50718-specific dsRNA into pre-pupae (PP) caused defects in adult wing development, while b PP injection of dsRNA for DvvABCB_39715 caused molting defects during eclosion. c, d and f Injection of dsRNAs for DvvABCG_3712, DvvABCG_14042 or DvvABCH_5118 into PP each resulted in severe molting defects during eclosion. e PP injection of Dvvw-specific dsRNA caused loss of eye pigmentation (arrow), while (g-h) larval injection of DvvABCE_2830 or DvvABCF_2701 resulted in a reduction in body mass and death prior to molting. i Injection of DvvABCG_3712-specific dsRNA into adult females interfered with embryonic development (arrow indicates location of head capsule in a control embryo). j DvvABCB_39715-specific dsRNA injected into adult females disrupted ovary development this study would have been missed. Regardless, BLASTx analyses of the 65 putative D. v. virgifera ABC transporters identified in this study demonstrate their greatest sequence similarity to T. castaneum and A. glabripennis orthologs. This is probably a consequence of the extensive publicly available genomic data for both T. castaneum and A. glabripennis, as well as their close phylogenetic relationships to D. v. virgifera. Furthermore, the putative one-to-one relationship among orthologs from D. v. virgifera and T. castaneum may suggest the retention of copy number without extensive gene loss or gain across evolutionary time.
Despite the relatively large amount of genomic and transcriptomic data available for model and some nonmodel coleopteran species, there is a comparative overall dearth of functional data available to support automated computational annotations. To partially address this shortfall, we generated functional information based on RNAi knockdown of eight D. v. virgifera ABC transporters, each of which demonstrated fairly conserved roles relative to their T. castaneum orthologs [30]. While some D. v. virgifera RNAi-mediated loss-of-function phenotypes include visible developmental defects, such as loss of eye pigmentation, others cause growth arrest and/or death. For example, knockdown of DvvABCA_ 50718 led to death during the pupal-to-adult molt and also caused deformation of wings and elytra in surviving adult beetles, which was the same as previously seen in T. castaneum [30] (Table 3; Fig. 4a). Since subfamily A transporter members are implicated in mammals with lipid transport, which can impact cell physiology [67], it is conceivable that the effects of the knockdown of DvvABCA_50718, and of its homologs, TcABCA-9A/9B, in T. castaneum [30], could be the result of disrupting critical lipid transport. DvvABCB_39715 RNAi also recapitulated the lethal effects of its T. castaneum ortholog; the effects on female fecundity could make this gene a particularly interesting target for RNAi-based pest control. It is worth noting that D. v. virgifera is predicted to have one more ABCB HT subfamily member compared to other insects [25], especially other beetles [30,64]. While ABCB FTs have been implicated in chemical insecticide resistance among insects [47], HTs are known to be mitochondrial transporters in humans, with roles in iron metabolism and transportation of Fe/S protein precursors [68,69]. These possibilities were outside the scope of our research, but future investigations into the DvvABCG_3712. Females injected with DvvABCB_39715 dsRNA failed to lay eggs, and those injected with DvvABCG_3712 dsRNA laid fewer eggs and those that were laid failed to develop. In both cases, buffer-injected females lay near-normal numbers. The eggs laid within a period of 2 weeks were counted every other day function of DvvABCBs could be beneficial for deciphering mechanisms of resistance evolution in D. v. virgifera.
RNAi knockdown of DvvABCE_2830 and DvvABCF_ 2701 resulted in 100% larval mortality. ABCE and ABCF subfamilies are highly conserved across all phyla, and due to their lack of TMDs are considered non-transporters. Instead, they appear to play roles in regulating translation [70,71], indicating that ABCE and ABCF proteins are essential. Thus, given that these genes are highly conserved across taxa in sequence, function, and RNAi phenotype [30], it may not be surprising that lethal RNAi knockdown phenotypes were obtained in D. v. virgifera.
The phenotypes observed following independent RNAi knockdown of DvvABCG_14042 and DvvABCH_5118 involved molting defects that resulted in near complete mortality. While these results are consistent with functional analysis of their T. castaneum orthologs, RNAi knockdown of the DvvABCG_14042 homolog TcABCG-8A in T. castaneum produced an additional phenotype of premature development of compound eyes [30]. In contrast, we did not observe any analogous eye phenotypes in D. v. virgifera following RNAi knockdown. It is likely that since the injected D. v. virgifera larvae died prior to reaching the next stage of development, there was no opportunity for compound eyes to form. In other species, orthologs of DvvABCH_5118 are known to transport cuticular lipids that are deposited in the outer epicuticle layer to form a waterproof barrier [30,62]. Therefore, it could be that cuticular lipid deposition may be reduced following RNAi knockdown of this ABCH transporter, which could promote desiccation and subsequent mortality of affected individuals.
The ABCG proteins are HTs, and, with 12 predicted members, form the second largest subfamily of ABC transporters identified in D. v. virgifera (Table 2). Among insects, some of the first ABCGs to be characterized were the pigment transporters (white, scarlet and brown) in D. melanogaster [34,72]. Mutants of white are characterized by white eyes (i.e. complete loss of eye pigmentation), scarlet mutants by bright red eyes (i.e. loss of brown pigments), and brown mutants by dark brown eyes (i.e. loss of red pigments) [34]. Studies have revealed that some ABCG proteins perform other crucial physiological roles in the transport of lipids, sterols, and drugs [73]. In the current study, RNAi-mediated knockdown of Dvvw resulted in a white-eyed phenotype consistent with prior observations in T. castaneum [30], and with our own previous findings in D. v. virgifera [46]. Our findings support a prediction that Dvvw is part of the ommochrome pathway, where it is likely acting within a heterodimeric complex to import ommochrome pigments into the pigment granules of the compound eye. As mentioned above, loss of white function in D. melanogaster, results in white-eyed flies, while mutations in scarlet lead to red-eyed flies.
However, RNAi-mediated knockdown of the corresponding gene, ABCG-9A (scarlet), in T. castaneum produces white-eyed beetles [30,32]. This finding was not surprising, since a previous report of RNAi targeting vermilion, a pivotal gene in the ommochrome pathway, also generates a white-eyed phenotype in T. castaneum [74], leading the authors to conclude that the T. castaneum eye is pigmented by ommochromes alone, and that the ommochrome biosynthetic pathway in T. castaneum produces red pigments as end products, rather than brown pigments as in D. melanogaster. Unfortunately, our initial survey of the D. v. virgifera transcriptome failed to identify a scarlet ortholog in our DNASTAR assembly, thus its function was not assessed. We did identify a scarlet ortholog from the Trinity assembly (See Table 2 and Additional file 5: Figure S3) after we had completed our functional analyses, but we were still unable to find any evidence of a brown ortholog. So, it will be interesting to investigate in future studies if pigmentation of the D. v. virgifera eye is more similar to that of T. castaneum or D. melanogaster. Specifically, in D. melanogaster a third ABCG transporter, brown, is required for wild-type pigmentation of the eye. In flies, Brown heterodimerizes with White and transports pteridine-based pigments into the eye. Although an ortholog of brown has been identified in the T. castaneum genome, no function has been identified [32].

Conclusion
This study provides a relatively large transcriptomic resource comprising genes expressed across several life stages of the arthropod pest species, D. v. virgifera. Due to potential omission of orthologs from our assembly, undoubtedly additional research will need to be performed in order to identify the full compliment of ABC transporters encoded by D. v. virgifera, and further functional assays will be needed to validate putative biochemical roles. Regardless, our work represents the initial description of the ABC transporter gene family in D. v. virgifera. Furthermore, the knockdown of ABC transporters DvvABCB_39715 and DvvABCG_3712, each of which reduced egg production and/or prevented embryonic development, could provide novel targets for D. v. virgifera population suppression and use as an insecticidal control agent. This research is a contribution to a growing set of genomic resources for arthropods, and provides information that may facilitate the development of methods to enhance the control of a devastating agricultural pest species.

Insect rearing
All D. v. virgifera used in this study are nondiapausing, from a colony previously established at North Carolina State University using beetles obtained from both Dr.
Wade French (USDA-ARS-NGIRL, Brookings, SD) and Crop Characteristics, Inc. (Farmington, MN, USA) (see [75]). Eggs deposited in an oviposition chamber (agar plate with cheese cloth) were collected weekly, pipetted into soil-filled containers, and held at 26°C for 1 week. Larvae were reared on roots of germinated corn seed in 16-oz containers, while adults were maintained in a 30cm 3 BugDorm (MegaView Science, Taiwan) at 26°C, 70% relative humidity with an L14:D10 photoperiod and fed an artificial diet (Western Corn Rootworm w/o Pollen Substitute, Frontier Insect Diet, Newark, DE, USA). Injected individuals were reared in small containers with corn seedlings to allow downstream observation.

Transcriptome sequencing, assembly, and annotation
Total RNA was extracted from mixed-staged D. v. virgifera embryos (n = 500 from an overnight egg lay aged up to 14 days), mixed-stage larvae (first-instar larvae (n = 20); second-instar larvae (n = 10); and third-instar larvae (n = 2), as well as an adult male, and an adult female (n = 1 each) using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and treated with DNase I (Qiagen) according to the manufacturer's instructions. The isolated total RNA was submitted to the Genomic Sciences Laboratory (North Carolina State University, NC, USA) for quality assessment, poly(A) selection, fragmentation, selection of 650 bp fragment sizes, Illumina TruSeq® library preparation, and 300 bp paired-end sequencing on an Illumina MiSeq sequencer (Illumina, San Diego, CA, USA).
The complexity of SOAPdenovo-Trans and Trinity assemblies were reduced by clustering allelic variants using CD-HIT-EST [80] with default parameters, except for change of sequence identity (−c 0.95), word length (−n 10), and length of throw-away sequence (−l 11). The relative completeness of each clustered D. v. virgifera transcriptome assembly was evaluated by comparison with the universal single-copy orthologs from Arthropoda obtained from OrthoDB v 9 [81] using BUSCO v 3 [82] (Evalue cutoff 0.001). Full-and partial-length open reading frames and corresponding derived amino-acid sequences were predicted from the resulting SOAPdenovo-Trans clusters with TransDecoder v3.0.0 [83] using a minimum length of 100 amino acids.
The transcript sequences assembled by SeqMan NGen® (DNASTAR, Madison, WI) were imported into Blast2GO v4.0 [84,85] and annotations acquired via BLASTx [86] comparison to the non-redundant (nr) arthropod-specific protein database at the National Center of Biotechnology Information (NCBI). The combined graphs were created at level 2 for Biological Process (P), Cellular Component (C), and Molecular Function (F) categories from Blast2GO.

Bioinformatic analysis of the D. v. virgifera ABC transporter family
A searchable database was created from the combined DNASTAR D. v. virgifera transcript assembly, and subsequently searched with the set of deduced T. castaneum ABC transporter amino-acid sequences [30,32] as queries using the tBLASTn algorithm in BlastStation software (TM Software Inc., Arcadia, CA, USA). Homologous sequences were selected based on sequence identity and Evalue (< 10 − 6 ). Putative D. v. virgifera ABC sequences were then used as BLASTx queries of the non-redundant NCBI protein database using the web blast interface (https://blast.ncbi.nlm.nih.gov/Blast.cgi) to confirm their identity as insect ABC genes; those that appeared to not be of non-insect origin, or were otherwise not ABC genes, were discarded. The number and positions of transmembrane domains were assessed via query of the NCBI Conserved Domain Database [87]. Finally, each D. v. virgifera ABC gene was putatively assigned to a subfamily (A-H) based on greatest similarity assigned to orthologs within BLASTx results. This BLAST search procedure was analogously repeated for SOAPdenovo-Trans and Trinity assemblies. The complexity of each ABC gene set was reduced by clustering allelic variants (sequence) across assemblies, and a comprehensive non-redundant set of putative D. v. virgifera ABC transporter contigs were generated (Additional file 7). Assembly of origin is denoted in sequence names as follows: DNASTAR (D), Trinity (T), and SOAPdenovo-Trans (S = "scaffold" and C = "contig") within the FASTA files. The full translation product of each contig can be found in Additional file 8.
Phylogenetic relationships among derived D. v. virgifera ABC transporter protein sequences were reconstructed from the conserved NBD. A multiple sequence alignment was performed with MUSCLE using MEGAX [88] (default parameters) and used within a subsequent phylogenetic analysis. The unrooted Maximum Likelihood phylogenetic trees were constructed in the MEGAX program using default parameters in all categories except: LG model of amino-acid substitution with Gamma distributed substitution rates (based on Best Model determination within the MEGA program), Partial Deletion treatment of gaps/missing data, and 1000 bootstrap replicates [89]. ABC transporter subfamilies were assigned to D. v. virgifera sequences and clades within this phylogenetic analysis by comparison to similarities from our BLASTx search results and tree topologies among nearest orthologous gene family members in T. castaneum [30,32], and D. melanogaster. Multiple sequence alignments were generated as described above, wherein the deduced D. v. virgifera amino-acid sequences included full-length sequences when possible, but some were incomplete partial-protein sequences. All phylogenetic reconstruction methods were performed as described above.

Gene expression across developmental stages
Preliminary analysis to estimate the relative expression levels for eight transcripts (DvvABCA_50718, DvvABCB_ 39715, DvvABCE_2830, DvvABCF_2701, DvvABCG_ 3712, DvvABCG_14042, Dvvw and DvvABCH_5118) across growth stages was made via semi-quantitative PCR in order to ensure dsRNA injections would be performed prior to the time of corresponding peak expression. Total RNA was extracted from each developmental stage [embryo (E), larval (L), pupal (P), and adult male (M) and female (F)], from which cDNA was reverse transcribed using the Superscript™ III First-Strand Synthesis System (Invitrogen, Carlsbad, CA, USA) using an anchored poly(T) primer. These cDNA pools were then used individually as template in eight separate PCR reactions each using D. v. virgifera ABC transporter transcript-specific primer pairs (Additional file 9: Table  S3). Primers for the D. v. virgifera ribosomal protein S6, DvvRPS6, were used as an external control. PCR reactions were set up using MyTaq™ DNA polymerase according to manufacturer instructions (Bioline, Memphis, TN, USA), and subsequent amplification reactions were performed in a C1000 Thermal Cycler (Bio-Rad Laboratories Inc., Hercules, CA, USA) with the following cycling conditions: (95°C for 3 min), 25× (95°C for 30s, 58°C for 30s, 72°C for 10s), (4 min incubation at 72°C). Amplification products were then visualized and compared using 1.5% agarose gel electrophoresis.

RNAi knockdown phenotypes
Primers were designed for the generation of dsRNA using Vector NTI Advance (VNTI) software (Invitrogen), for all ABC genes whose orthologs are known to produce obvious RNAi phenotypes in T. castaneum [30]. These primer sets targeted regions that encoded transcriptspecific TMD domains; this was done in order to potentially reduce unintended off-target effects by avoiding the more conserved NBD domains. Partial cDNAs were amplified for the 8 genes (DvvABCA_50718, DvvABCB_39715, DvvABCE_2830, DvvABCF_2701, DvvABCG_3712, DvvABCG_14042, Dvvw and DvvABCH_5118), as described above for developmental stage expression. Nested PCR was performed with an initial denaturation of 95°C for 3 min, 35 cycles at 95°C for 30s, 58°C for 30s, and 72°C for 10s, and then a 4 min incubation at 72°C on a C1000 Thermal Cycler (Bio-Rad). PCR products were purified using the QIAquick PCR Purification Kit (Qiagen) according to the manufacturer's instructions, ligated into the pGEM-T vector (Promega, Madison, WI, USA), and the resulting plasmids were used to transform TOP10 competent E. coli (Invitrogen). All positive clones were cultured in a selective LB medium containing 100 mg ampicillin L − 1 . The recombinant plasmid DNAs were isolated using the QIAprep® Spin Miniprep Kit (Qiagen), and the inserts were Sanger sequenced and confirmed by use as BLASTn queries (https://blast.ncbi.nlm.nih.gov/Blast. cgi). Purified plasmids with each cloned ABC transporter were used as template in separate PCR reactions primed with the following primers: T7 as a forward primer (due to location in pGEM), and a pGEM-specific reverse primer that was tailed with T7. This enabled all amplification reactions to be performed using the same set of primers under conditions described above. PCR products were analyzed by 1.5% agarose gel electrophoresis, purified using the QIAquick PCR Purification Kit (Qiagen), and then1 μg of each was used as template for dsRNA synthesis using the MEGAscript T7 in vitro Transcription Kit (Ambion, Austin, TX, USA). Each of the synthesized dsRNAs were purified using the MEGAclear Kit (Ambion) and concentration determined using a Nanodrop 1000 (Thermo Scientific, Waltham, MA, USA) using the singlestranded RNA setting.
RNAi assays were conducted by injecting dsRNA corresponding to each of the 8 specific D. v. virgifera ABC genes individually into the hemocoel of third-instar larvae, pre-pupae and/or newly-eclosed female adults. Before microinjection, experimental insects were anesthetized on ice for 30 min, then injected with~0.2 μl of a genespecific dsRNA at a concentration of 1-2 μg/μl. Each treatment was replicated three times, with ≥20 individuals in each replicate. Following injection, larvae and pre-pupae were allowed to recover at room temperature for 1 hour, and then moved to germinated corn for further monitoring and phenotypic analysis. Phenotypes were observed daily using a stereomicroscope, and transcript levels assessed at 5 days post-injection by semi-quantitative PCR using RNA isolated from pools of injected individuals (one individual per replicate, for a total of three individuals per PCR reaction).
Treated females were kept in an oviposition chamber (agar plate with cheese cloth) and maintained on an artificial diet. At 2 days post-injection, females were mated to untreated males, and generally started to lay eggs1 0 days later. To determine egg viability, eggs were harvested from the oviposition chamber and placed on moistened filter paper in Petri dishes and held at 26°C, 70% relative humidity with an L14:D10 photoperiod. Females were allowed to lay eggs over a two-week period, and eggs were counted every other day to assess the rate of egg laying. Hatch rate counts were made every other day, beginning 10 days after the first egg lay (22-days post-injection) and continuing for 4 weeks until no further hatching was observed.