Transcriptomic exploration of the Coleopteran wings reveals insight into the evolution of novel structures associated with the beetle elytron

Abstract

The acquisition of novel traits is central to organismal evolution, yet the molecular mechanisms underlying this process are elusive. The beetle forewings (elytra) are evolutionarily modified to serve as a protective shield, providing a unique opportunity to study these mechanisms. In the past, the orthologs of genes within the wing gene network from Drosophila studies served as the starting point when studying the evolution of elytra (candidate genes). Although effective, candidate gene lists are finite and only explore genes conserved across species. To go beyond candidate genes, we used RNA sequencing and explored the wing transcriptomes of two Coleopteran species, the red flour beetle (Tribolium castaneum) and the Japanese stag beetle (Dorcus hopei). Our analysis revealed sets of genes enriched in Tribolium elytra (57 genes) and genes unique to the hindwings, which possess more “typical” insect wing morphologies (29 genes). Over a third of the hindwing‐enriched genes were “candidate genes” whose functions were previously analyzed in Tribolium, demonstrating the robustness of our sequencing. Although the overlap was limited, transcriptomic comparison between the beetle species found a common set of genes, including key wing genes, enriched in either elytra or hindwings. Our RNA interference analysis for elytron‐enriched genes in Tribolium uncovered novel genes with roles in forming various aspects of morphology that are unique to elytra, such as pigmentation, hardening, sensory development, and vein formation. Our analyses deepen our understanding of how gene network evolution facilitated the emergence of the elytron, a unique structure critical to the evolutionary success of beetles.

Transcriptomic comparisons between the forewings (elytra) and hindwings of beetles identified a set of differentially expressed genes. RNA interference (RNAi) for elytron‐enriched genes revealed novel genes with roles in forming various elytron‐unique features. These findings deepen our understanding of how changes in the gene regulatory network facilitate the evolution of novel traits.

Research Highlights

• Transcriptomic comparisons between the forewings (elytra) and hindwings of two beetle species identified a set of differentially expressed genes.

• The hindwing transcriptome shows enrichment of many previously identified wing genes, while the elytron transcriptome highlights gene regulatory networks patterning the morphological features shared between the elytron and body wall of beetles.

• RNA interference for some of the elytron‐enriched genes in Tribolium revealed novel genes with roles in forming various aspects of the morphological features unique to elytra, such as pigmentation, hardening, sensory development, and vein formation.

1. INTRODUCTION

Despite the fact that morphological innovation is a critical aspect of organismal evolution, the precise molecular mechanisms driving this process are poorly understood. Insect wings offer a powerful opportunity to study the mechanisms underlying the evolution of novel traits, as these dorsal appendages are highly diverged among the insects, varying in their shape, texture, color, and function. One distinct example of a novel wing structure is the highly modified forewings on the second thoracic segment (T2) of Coleopteran insects (beetles) called elytra. Elytra have evolved a number of unique morphological features including a modified shape; a heavily sclerotized, hardened, and pigmented surface (namely, exoskeletalized surface [Tomoyasu et al., ²⁰⁰⁹]); a distinct parallel vein pattern; an array of sensory structures covering the elytra; and a unique sealing mechanism that allows the left and right elytra to bind tightly together over the dorsal surface of the beetle (Crowson, 1981). Coleoptera is one of the most successful orders of organisms on the planet, and the morphologically unique elytra, which define this clade, were a key to this evolutionary success (Grimaldi & Engel, 2005; Hunt et al., ²⁰⁰⁷). Specifically, the elytra protect the dorsal surface of the beetle from a variety of harsh environmental factors by acting as an armored plate (Linz et al., 2016). This includes protecting their fully functional flight wings on the third thoracic segment (T3), which allowed beetles to explore a variety of niches all while maintaining their ability to fly (Crowson, 1981; Linz et al., ²⁰¹⁶). Despite the importance of elytra, the molecular route through which these novel wings evolved is not well understood.

A majority of the knowledge regarding insect wing formation stems from studies based on a dipteran insect (fly), Drosophila melanogaster (see [Beira & Paro, ²⁰¹⁶; Brook et al., ¹⁹⁹⁶; Ostalé et al., ²⁰¹⁸; Tripathi & Irvine, ²⁰²²] to review Drosophila wing development). These studies revealed a complex network of interactions among genes involved in wing development, namely the Wing Gene Network (WGN) (Figure 6a). This WGN has been used as a paradigm to identify changes in the gene regulatory network that facilitated the evolution of elytron‐specific morphological features in the lineage leading to beetles (i.e., the candidate gene approach). By using the red flour beetle (Tribolium) as a model, this approach showed that (i) elytra, despite their modifications, are the developmental default state of the Coleopteran wing and the Hox gene Ultrabithorax (Ubx) acts uniquely in T3 to cancel these modifications to form the hindwing (Tomoyasu et al., 2005) (Figure 6b), and (ii) how multiple conserved wing genes have diverged roles in the beetle lineage to form some of the elytron‐unique morphological features, such as the exoskeletalization and unique shape of elytra (Figure 6c) (see Tomoyasu et al., ²⁰⁰⁹ for exoskeletalization and Ravisankar et al., ²⁰¹⁶ for shape). Although effective, the candidate gene approach has a number of limitations. First, a candidate gene approach is limited to the genes in the WGN identified in Drosophila. Thus, only the conserved and diverged functions of genes evolutionarily maintained across species can be explored via this approach. The consequence is that genes uniquely important for wing development in the beetle lineage are barred from the candidate gene list. Second, a candidate gene approach can create a “Drosophila‐centric” view of evolution, where the fly WGN is viewed as the starting point of evolution. To escape the limitations of the candidate gene approach, we need to adopt unique, non‐fly‐oriented approaches to studying the mechanisms underlying the evolution of elytra.

Figure 6.

The gene network that governs the development of fly and beetle wings. (a) WGN established through decades of Drosophila studies (Beira & Paro, 2016; Tripathi & Irvine, ²⁰²²). Note that this is an abridged version of WGN, showing only some of the key components. (b) WGN for the Tribolium hindwing. The genes previously demonstrated to be involved in the Tribolium wing development (through expression and/or RNAi analyses) are highlighted in teal. A large part of the WGN is conserved between Drosophila and Tribolium, with one crucial difference in the involvement of Ubx. The WGN operates without Ubx in Drosophila, while Ubx has been integrated into the WGN and become indispensable for the formation of wings in Tribolium (some Ubx‐dependent interactions are highlighted in purple). The involvement of the gray‐colored genes in the Tribolium wing development is yet to be determined. The genes enriched in the hindwing transcriptome (kn, sal, omb, en, and Ubx), as well as the genes associated with the hindwing‐enriched genes (eg/kni, bx/ap, inv/en, and wnt2/wg), are boxed in red. (c) The WGN for the Tribolium elytron, showing the changes in the WGN that facilitated the evolution of elytra. Only the part that was demonstrated to be operational in the Tribolium elytron was included (teal genes in (b)). Previously identified changes are highlighted in orange, with whited‐out genes showing loss of expression as the consequence of these changes. The changes found in the current study that were functionally validated via RNAi are shown in red. RNAi, RNA interference; WGN, Wing Gene Network.

One option we previously explored was a “semi‐candidate gene” approach, where we examined the postembryonic function of a group of Tribolium genes whose orthologs in Drosophila are important for embryonic segmentation and axis establishment, but have no described role in fly wing development (Linz & Tomoyasu, 2015). We hypothesized that the highly pleiotropic nature of these genes would increase the likelihood of finding novel wing genes in Tribolium. Although some genes with novel functions in postembryonic development (such as metamorphosis and formation of wing serial homologs) were found, no gene we examined had novel wing‐related roles in Tribolium. These results demonstrate the importance of adopting a methodology that is fully independent of candidate genes: a true “non‐candidate gene” approach.

To examine the mechanisms driving the evolution of elytra using a non‐candidate gene approach, Tribolium offers a number of options, such as a fully sequenced genome (Tribolium Genome Sequencing Consortium, 2008), a variety of genetic tools and resources (Brown et al., 2003; Denell, ²⁰⁰⁸; Klingler, ²⁰⁰⁴; Trauner et al., ²⁰⁰⁹), and a highly effective systemic RNA interference (RNAi) response (Bucher et al., 2002; Tomoyasu & Denell, ²⁰⁰⁴; Tomoyasu et al., ²⁰⁰⁸) (see Klingler & Bucher, ²⁰²² for review). One such option is a genome‐wide RNAi screen. The iBeetle project implemented this approach and, with over 80% of the Tribolium genes examined and a publicly available database of results, created a powerful tool for the community (Hakeemi et al., 2022; Klingler & Bucher, ²⁰²²; Schmitt‐Engel et al., ²⁰¹⁵). For the study presented here, we adopted a second, more wing‐oriented approach by using RNA sequencing (RNAseq) to compare and contrast the elytron and hindwing transcriptome of Tribolium. In addition, we also integrated a second, less‐studied Coleopteran species: the iconic Japanese stag beetle, Dorcus hopei. By performing a similar RNAseq‐based comparison of elytron and hindwing transcriptomes in Dorcus, and then comparing the pools of fore and hindwing‐enriched genes across the two Coleopteran species, we sought to reveal the “core” genes critical to wing formation within the Coleopteran lineage.

Our Tribolium RNAseq experiments revealed: (i) relatively small groups of genes enriched in each respective wing tissue, (ii) a pool of genes enriched in hindwing tissue that have been previously implicated in insect wing formation via the candidate gene approach, and (iii) multiple genes differentially expressed in elytra with undescribed roles in forming the beetle‐unique wing. The Dorcus RNAseq experiments revealed similar patterns in elytron and hindwing transcriptomes; however, the overlap between Tribolium and Dorcus was limited to only a small subset of genes. Importantly, among the overlapping genes in hindwing tissue were two genes, Ubx and spalt (sal); key wing genes previously identified as expressed highly in the hindwings of Tribolium (Tomoyasu et al., 2005).

Although RNAseq is highly effective at revealing lists of differentially expressed genes, a caveat is that the functional significance of these genes is often unclear. The systemic RNAi response of Tribolium provides a unique opportunity to go beyond a simple RNAseq‐based approach and perform a functional analysis of groups of genes revealed by these sequencing experiments. Accordingly, we proceeded by removing well‐described and/or previously studied wing genes from the list of differentially expressed genes, and then performed an RNAi‐based functional screen of a subset of the genes uniquely expressed in beetle elytra (as well as hindwing). These analyses revealed genes with functions in forming a variety of elytron‐unique morphological features, including their (i) exoskeletalization, (ii) organized array of sensory structures, and (iii) vein structures. Importantly, these genes were identified and screened completely independent of Drosophila wing genes and represent one of the first sets of true non‐candidate Coleopteran wing genes, which allows elucidation of the beetle lineage‐specific changes in the WGN that drove the evolution of elytra. Ultimately, these findings deepen our understanding of how changes in gene regulatory networks facilitate the emergence and diversification of novel traits, a critical step during organismal evolution.

2. RESULTS

2.1. Tribolium elytron and hindwing differential gene expression analysis

Following sequencing, we first performed differential gene expression analysis to compare the Tribolium elytron and hindwing transcriptomes. At a given significance threshold (p ^adj value < 0.01) we detected relatively small pools of differentially expressed genes between the elytron and hindwing transcriptomes, with 57 (58) genes enriched in the elytron and 29 (30) genes enriched in the hindwing (values in parenthesis represent the number of genes before manual curation, as some genes computationally annotated as two separate transcripts appear to only represent one) (Figure 1a, Supporting Information: Tables S7 and S8). Within these groups of genes, differences in transcript abundance between tissues were relatively subtle (average change = ~two‐fold) (Figure 1a, Supporting Information: Tables S7, and S8). One clear outlier to this trend was the hindwing‐unique Hox gene Ubx (change = ~14‐fold) (Figure 1a). Strong enrichment of Ubx in hindwing tissue is expected (Tomoyasu et al., 2005), and provides the first evidence that our transcriptome sequencing and differential gene expression analysis were successful. In addition to Ubx, we also found that a considerable number of the genes (8/29) enriched in hindwing tissue have been (i) implicated in Drosophila wing formation, (ii) previously studied in Tribolium wing formation via candidate gene approaches, (iii) and/or are known targets of Ubx‐driven wing differentiation (marked genes in Figure 1a). Among these genes are knot (kn) and sal, which are known direct targets of Ubx in Drosophila (Galant et al., 2002; Hersh & Carroll, ²⁰⁰⁵) and, in Tribolium, unique hindwing expression of sal has also been previously shown (Tomoyasu et al., 2005). These data further confirm that our sequencing is revealing bona fide differentially expressed genes. In contrast to the hindwing‐enriched genes, most of the genes enriched in elytron tissue have not been implicated in wing development, suggesting that these genes might be involved in the unique features associated with elytra. We also performed a gene ontology (GO) analysis of the genes uniquely expressed in each respective wing tissue; however, this failed to reveal any striking trends (Supporting Information: Figure S2). In summary, our sequencing experiments appear to be successful and, although some genes are considered as candidate genes (many of which have been previously examined in Tribolium), a majority of genes differentially expressed in either the T2 or T3 wing transcriptome have not been implicated or studied in the context of beetle wing formation.

Figure 1.

Differential gene expression in the elytron and hindwing of Tribolium and Dorcus beetles. (a) DESeq2 generated MAplot showing Tribolium elytron and hindwing differential gene expression. Previously established wing genes unique to hindwing tissue are highlighted in purple. (b) Dorcus elytron and hindwing differential gene expression. Values in parenthesis indicate the number of genes with an identifiable Tribolium ortholog. (c) Overlap of Dorcus and Tribolium, elytron and hindwing unique genes. Gene names are indicated below Venn diagrams. Significance threshold (p ^adj value) is indicated in the respective panel.

Along with the differential expression analysis, we also discovered one striking pattern when we examined the clustering of our Tribolium sequencing replicates. Here we found that, despite isolating tissue for sequencing at similar developmental time points, samples are generally grouped by replicate rather than tissue type (Supporting Information: Figure S3a). This implies that the elytron and hindwing transcriptomes are highly similar at any given developmental time point, yet broadly dynamic over even short variations in developmental time.

2.2. Dorcus elytron and hindwing transcriptomes and a comparison across two Coleopteran species

In addition to Tribolium, we also performed RNAseq in Dorcus to reveal genes differentially expressed in the elytron and hindwing of a second Coleopteran species. These data allowed us to compare the elytron and hindwing transcriptomes across Coleopteran species and reveal the “core” genes shared within these tissues. RNAseq in Dorcus revealed similar trends to our Tribolium sequencing experiments. First, the pools of assembled genes differentially expressed in each respective wing tissue are relatively small at a given significance threshold (p ^adj value < 0.01), with 53 genes unique to elytron and 20 genes unique to hindwing (Figure 1b). Second, we again observed that, similar to Tribolium, the replicate clustering of our Dorcus RNAseq samples clustered by individuals rather than tissue type (Supporting Information: Figure S3b). This further supports that the elytron and hindwing transcriptomes are highly similar, but also dynamic throughout development. In addition to these findings, we also compared the groups of genes enriched in elytron and hindwing tissue that are shared in both Dorcus and Tribolium. Here, we observed an overlap of only two genes in the Coleopteran elytron tissue and four genes in the hindwing tissue (Figure 1c). Although minimal, the genes that overlap were biologically relevant. In hindwing tissue, the overlapping genes were Ubx and sal, as well as uncoordinated−5 (unc‐5), and invected (inv); all known wing genes except unc‐5, which was functionally analyzed in this study (see below). In elytron tissue, two genes were shared between the Coleopteran species: reversed polarity (repo) and scabrous (sca). These genes both function in peripheral nervous system (PNS) formation in Drosophila (for review see Jan & Jan, ¹⁹⁹⁴), and their shared unique expression in elytra is fitting given the array of sensory structures that are morphologically unique to elytra (Crowson, 1981; Tomoyasu & Denell, ²⁰⁰⁴). However, the function of these genes has not been tested in a Coleopteran, and accordingly, they were also included in our functional screen.

2.3. Functional analysis of differentially expressed genes

RNAseq and downstream differential gene expression analyses are highly effective at revealing genes unique to specific tissues; however, the caveat to these in silico‐based approaches is that the functional significance of these genes is unknown. The well‐established systemic RNAi response of Tribolium offers a powerful opportunity to functionally screen large pools of genes rapidly and with relative ease. Accordingly, we next sought to functionally examine a subset of the genes revealed by our RNAseq. To choose genes for our RNAi screen, we began with the raw output of our DESeq2 based differential gene expression analysis. From this list, we began by sorting based on p ^adj significance value and removing genes with greater than 0.01 p ^adj values. The remaining genes (57 for elytron and 29 for hindwing) were sorted by fold‐change. We then removed genes that have previously described roles in wing patterning and/or have been previously studied as candidate genes (1 in elytron and 11 in hindwing). From the remaining list, we chose the top 14 genes. We also included two additional genes for each tissue based on either, (i) cross‐referencing to the iBeetle screen (Schmitt‐Engel et al., 2015), (ii) interest based on gene family, or (iii) by overlap between Dorcus and Tribolium sequencing. In total, the list of genes for our RNAi screen contained 16 genes from elytron and 16 genes from hindwing.

One rate‐limiting and costly aspect of an RNAi screen is gene cloning. To facilitate a rapid RNAi screening, we synthesized plasmids containing 100 bp regions of each gene. For each of the genes within our chosen list, we began by examining the RNAseq read mapping in the Interactive Genomics Viewer (v2.3.68; Robinson et al., ²⁰¹¹) to choose 100 bp regions of each of the 32 genes in our list that (i) had visually identifiable differential read mapping between elytron and hindwing samples, and (ii) showed no tendency for off‐target effects (based on in silico analysis by the E‐RNAi tool; Horn & Boutros, ²⁰¹⁰) (see Supporting Information: Table S9 for additional details). We then used GenScript Gene Synthesis services to synthesize plasmids containing our sequences of interest. Each synthesized plasmid contained a 100 bp region of enhanced yellow fluorescent protein (EYFP) nucleotide sequence (see below), and on the 5′ and 3′ side of the EYFP sequence, each plasmid contained two 100 bp long gene sequences from our RNAi screen list (four 100 bp gene regions + one 100 bp EYFP sequence in each plasmid, see Supporting Information: Figure S1). In total, with four genes per plasmid, eight plasmids were synthesized.

2.4. Multigene RNAi screen

We began our RNAi screen by taking a multigene‐knockdown approach to rapidly screen the gene lists for each tissue. We generated 300 bp double‐stranded RNA (dsRNA) containing three contiguous 100 bp gene sequences: two genes from our RNAseq‐based gene list as well as EYFP as a positive control (see Supporting Information: Figure S1 for conceptual scheme and Miller et al., ²⁰¹² for an example of using chimeric dsRNA). Our initial multigene RNAi produced either (i) disruptions in metamorphosis preventing adult eclosion, (ii) defects in non‐wing related tissue, (iii) defects in elytra, or (iv) no observable abnormalities (see Supporting Information: Tables S4 and S5). Albeit positive initial results, multiple caveats exist for this initial multigene screen. First, although combining the three genes into a single contiguous dsRNA sequence may eliminate competition for cellular uptake of dsRNA, competition at the mRNA silencing level of the RNAi machinery can still create drastically hypomorphic knockdown of the individual genes within the single dsRNA (although EYFP knockdown was consistently strong in all multigene RNAi [data not shown]) (Miller et al., 2012). Second, three of the multigene knockdowns disrupted metamorphosis, preventing examination of the adult morphology (including wings). These caveats could produce a situation where we may miss disruptions in wing tissue. To avoid this issue, we proceeded by performing single RNAi for each of the 32 genes.

2.5. Single‐gene RNAi screen

Single gene RNAi was performed at the maximum possible concentration for each of the 32 genes to generate the strongest knockdown (Supporting Information: Tables S4 and S5). Lower concentrations were used for genes with high lethality and/or metamorphosis defects. A large subset of the genes (23/32) failed to show any observable abnormalities in wing tissue or other adult morphology (Supporting Information: Tables S4 and S5). Another group of genes showed previously undescribed RNAi phenotypes outside of wing tissue, including (3/32) with metamorphosis defects, and (2/32) with other morphological defects (Supporting Information: Figure S4). Surprisingly, no RNAi for genes from the hindwing list (16/32) displayed disruptions in wing formation. Although the contexts are different, the iBeetle genome‐wide RNAi screen reported that ~70% of genes examined after larval stage injections presented no phenotype (Schmitt‐Engel et al., 2015). Our lack of RNAi disruptions from hindwing‐enriched genes may be caused by excluding nearly 40% (11/29) as candidate genes with known functions in wing formation, and as such we have biased our list towards genes without observable function. Finally, four genes from the elytron list showed previously undescribed roles in forming various aspects of the elytron morphology (Figures 2, 3, 4, 5).

Figure 2.

Pigmentation and metamorphosis defects in Tc‐hr38 RNAi. (a–d) Un‐injected pu11. WT coloration of pu11 pupae (a), adult (b), and last larval stage larvae (c), as well as normal T2 and T3 disc formation as shown by EYFP nub expression at the end of the last larval stage (d). (e–h) Tc‐hr38 RNAi. RNAi at the last larval stage (e and f) caused discoloration of pupae (e) and adults, which die shortly after eclosion (f). RNAi at the penultimate larval stage (g and h) caused discoloration of the last larval stage larvae (g) and also disrupted the formation of T2 and T3 wing imaginal tissue (arrows in (h)). Scale bars in (a)–(D) (2 mm in (a)–(c), 0.5 mm in (d)) apply to all images below the respective panel. EYFP, enhanced yellow fluorescent protein; RNAi, RNA interference; WT, wild‐type.

Figure 3.

Disruptions of sensory structures in Tc‐sca RNAi. (a–c) Un‐injected pu11 elytron, hindwing, and body wall. WT sensory structures of the elytron (a), hindwing (b), and first thoracic segment body wall (c). (d–f) Tc‐sca RNAi. The sensory structures of the elytron (d), hindwing (e), and body wall (f) are duplicated and disorganized. Boxed regions in (b), (c), (e), and (f) are magnified in respective inboxes. All scale bars are 200 µm and apply to all images below the respective panel. RNAi, RNA interference; WT, wild‐type.

Figure 4.

Disruptions of sensory structures and veins in the elytron of Tc‐cs10 RNAi. (a–e) Un‐injected pu11 elytron and hindwing. WT sensory structures and veins of the elytron (a–c), and hindwing (d and e). (f–j) Tc‐cs10 RNAi. The formation of the sensory structures and veins of the elytron (f–h) is disrupted. The sensory structures of the hindwing are disrupted; however, the veins are intact (i and j). Boxed regions in (a), (d), (f), and (i) are the region magnified in (b), (e), (g), and (j), respectively. Arrows indicate absence of sensory structures. All scale bars are 200 µm and apply to all images below respective panel. RNAi, RNA interference; WT, wild‐type.

Figure 5.

Wing blistering caused by Tc‐msr RNAi. (a and b) Un‐injected pu11 adults. The WT elytra (a) and hindwing underneath a dissected elytron (b). (c–f) Tc‐msr RNAi. Elytra are blistered weakly (c), moderately (d), or severely (e). Hindwings occasionally produced a blistering effect (f). Arrows indicate elytron blisters and arrowhead indicates hindwing blisters. All scale bars are 2 mm. RNAi, RNA interference; WT, wild‐type.

The survival rate for genes that showed no phenotype stayed at relatively high levels (Supporting Information: Tables S4 and S5). In addition to the E‐RNAi tool used to check for off‐target effects (Horn & Boutros, 2010), two non‐overlapping dsRNA fragments that target the same gene were designed for genes with previously undescribed RNAi phenotypes in the elytra (Supporting Information: Table S6).

2.6. Genes with previously undescribed roles outside of wing formation

A group of genes we functionally examined (three genes enriched in elytron and two genes enriched in hindwing) did not produce wing defects, but did cause disruptions in either (i) adult tissues outside of wing or (ii) metamorphosis. In regard to adult tissues outside of the wing, one gene enriched in elytron tissue, Tc‐doublesex (TC008682, Tc‐dsx), caused disruptions in sex‐specific morphology after RNAi. Specifically, after Tc‐dsx RNAi, we failed to observe the presence of the male‐specific sex patch on the leg of the first thoracic segment (a structure similar to this, the sex combs, are known to be dependent on dsx function in Drosophila; Tanaka et al., ²⁰¹¹) and also observed that the adult reproductive organs were protruding from the posterior abdomen (Supporting Information: Figure S4a–e). In addition to Tc‐dsx, RNAi for one gene enriched in hindwing tissue (TC004474, Tc‐toll‐7) caused fusions in the club structure of adult antennae (Supporting Information: Figure S4f,g). A paralog of Tc‐toll‐7, Tc‐toll‐10 (TC004901) was also enriched in hindwing and was included in our RNAi screen. Single RNAi for Tc‐toll‐10 showed no observable abnormalities. To assess possible redundancy between these two paralogs, we also performed double RNAi for both genes simultaneously (Supporting Information: Table S5), but we did not observe any enhancement of the antennae disruptions nor in any other adult tissues. Interestingly, these two genes have been implicated in embryonic patterning and axis elongation in a variety of arthropod species (Benton et al., 2016). The role of Tc‐toll‐7 in Tribolium antennae formation may be the first described postembryonic role for this gene.

In addition to these two genes disrupting adult tissues, an additional two genes enriched in elytra (TC013309 and TC013527) and one gene enriched in hindwings (TC006510) showed disruptions in metamorphosis. TC013309 (Tc‐repo) caused lethality at the prepupal stage just before pupation, while TC013527 and TC006510 prevented successful adult eclosion (Supporting Information: Figure S4h,i). All three genes caused 100% lethality in injected individuals (Supporting Information: Tables S4 and S5). This effect was also observed after injecting a lower concentration of dsRNA (500 ng/µl). Combined, these data reveal a number of unique functions for genes that have not been previously examined in Tribolium postembryonic development.

2.7. Tc‐hr38 (TC013146) has a conserved role in cuticle pigmentation and hardening

The elytra of beetles are unique compared to other insect wings in that they are exoskeletalized. These characteristics are similar to the body wall and previous work has shown that the gene regulatory network for exoskeletalization is shared between these tissues (Tomoyasu et al., 2009). Although a complex process, the biochemical cascade begins with two important and well‐conserved enzymes, tyrosine hydroxylase (TH) which converts tyrosine to 3,4‐dihydroxyphenylalanine (DOPA) and DOPA decarboxylase (DDC) which converts DOPA to dopamine (see Noh et al., ²⁰¹⁶ for review). In Drosophila, which shares this pathway, the nuclear receptor hormone receptor 38 (Dm‐hr38) has been implicated in the regulation of these two enzymes and is thereby connected to the biochemical cascade controlling pigmentation and sclerotization (Kozlova et al., 2009; Sekine et al., ²⁰¹¹). In addition, Dm‐hr38 has also been implicated in regulating the hormonal signals (such as ecdysone) that drive developmental transitions (metamorphosis) in insects (Sutherland et al., 1995). The Tribolium ortholog of Dm‐hr38, Tc‐hr38 (TC013146), was differentially expressed in the T2 elytron tissue. RNAi for Tc‐hr38 at the last larval stage caused a subtle orange coloration to form throughout the entire Tribolium pupae (including elytra) (Figure 2e). Adult beetles failed to successfully eclose, although subtle changes in wing and body wall coloration appeared to persist (Figure 2f). We also observed that the cuticle of Tc‐hr38 RNAi individuals became brittle, as appendages would easily break from the adult body during dissection (data not shown). We also injected dsRNA for Tc‐hr38 at the penultimate larval stage and observed a more striking pigmentation defect when animals molted to the last larval stage (Figure 2g). Additionally, we also monitored wing imaginal disc proliferation and observed disruptions in wing tissue formation, which is likely a by‐product of disruptions in hormone signal regulation (Figure 2h). All individuals injected at the penultimate stage failed to pupate, preventing observation of adult morphology. Given these results, it appears that the role of Tc‐hr38 is conserved between beetles and flies (pigmentation and sclerotization, as well as hormone signal regulation). In addition, differential expression of this gene in elytra further supports that the gene regulatory network orchestrating the unique exoskeletalization of elytra is shared between the T2 wing and body wall. As such, disruptions caused by the knockdown of this gene occurred not only in the wing structures but were also shared across the entire body wall of the animal. It is also important to note that the regulatory role of hr38 is far upstream in the exoskeletalization biochemical cascade. Therefore, enrichment of this transcription factor in elytron tissue during earlier development (imaginal disc development) is fitting, compared to the large pools of exoskeletalization “realizator” genes, such as cuticle proteins, which are strongly enriched in elytra at later pupal stages (Dittmer et al., 2012).

2.8. Tc‐sca (TC003194) has a conserved role in formation of sensory structures

The dense array of sensory structures covering the elytra is another defining feature of the beetle unique wings (Figure 3a). However, these sensory structures are similar to the exoskeletalization of elytra, in that this morphological feature, although unique for a wing, is shared with the body wall of beetles (Figure 3c). Accordingly, we identified genes differentially expressed in Tribolium elytra whose orthologs in fly have roles in forming the PNS (examples include Tc‐pox‐neuro, Tc‐prospero, Tc‐repo and Tc‐sca). Tc‐sca was also uniquely expressed in Dorcus elytra. We included Tc‐sca (along with Tc‐repo—see above, and Tc‐pox‐neuro—no observed abnormality) in our RNAi screen (Supporting Information: Table S4). In Drosophila, Dm‐sca, acts to single out neuronal precursor cells from a cluster of proneural cells (Mlodzik et al., 1990). Disrupting the function of Dm‐sca causes duplication and disarray among final sensory bristles (Mlodzik et al., 1990). RNAi for Tc‐sca caused sensory structures of the elytra to duplicate and become disorganized (Figure 3d–f). The hindwing of Tribolium also contains sensory bristles (albeit far fewer than elytra) near the wing hinge (Figure 3b) and in the central region of the wing blade (data not shown). Tc‐sca RNAi also resulted in the duplication of hindwing sensory bristles (Figure 3e and data not shown) and drastic increase of sensory bristle‐associated small dent structures that cover the adult body wall (Figure 3f). Ultimately, the role of Tc‐sca and Dm‐sca appear to be conserved in patterning neuronal precursor cells. This likely means that the increased number of sensory structures covering the elytra has caused the enrichment of genes involved in PNS formation within the elytron transcriptome. In summary, the presence of differentially expressed genes involved in PNS formation suggests that the molecular mechanisms driving the formation of the elytra unique sensory structures are also shared with the gene regulatory network that forms sensory structures on the beetle body wall.

2.9. Tc‐chemosensory10 (TC008682) has a previously undescribed role in forming sensory structures and veins of elytra

The sensory structures covering beetle elytra are assumed to function at a molecular level in a similar manner to any insect sensory organ, be it tactile, olfactory, nociception, and so forth (areas of study that have received considerable attention in insects). Thus, along with the genes that pattern these sensory organs (see above), we also detected the enrichment of genes that function within these structures. One example was the differential expression of three chemosensory genes (Tc‐cs): Tc‐cs10, Tc‐cs11, and Tc‐cs12. Chemosensory proteins (CSPs) are a large family of soluble proteins, which play important roles in insect chemosensation similar to odorant‐binding proteins (for review see Pelosi et al., ²⁰¹⁴). Surprisingly, these small proteins also have a wide variety of roles outside of chemosensation including, among others, developmental processes (Maleszka et al., 2007). RNAi for one of the chemosensory genes, Tc‐cs10, caused gaps to form in the veins of elytra (Figure 4f–h). These gaps were marked by an absence of sensory structures, which are normally regularly positioned on either side of the vein (Figure 4h). In addition, RNAi also disrupted the sensory hairs on the hindwing; however, vein formation appeared to be unaffected (Figure 4i,j). These data suggest that Tc‐cs10 is involved in the formation of some sensory hairs on both elytron and hindwing structures, and in elytra, this gene may also have a role in forming the continuous veins that run parallel along the length of the wing structure.

2.10. Tc‐methionine sulphoxide reductase (TC005676) has an unknown role unique to wing formation

Both the elytron and hindwing of beetles are composed of a dorsal and ventral layer of epidermal cells. While the hindwing epidermal layers are bound together tightly, the situation in elytra is different. Here, the dorsal layer of the epidermis secretes the thick cuticle (discussed above) and the ventral layer has become markedly thin and membranous. In addition, there is also a space between the two layers in elytra, which are separated by pillar‐like structures called trabeculae (Noh et al., 2016). RNAi for the gene Tc‐methionine sulphoxide reductase (Tc‐msr) caused a dramatic blistering effect to take place in the elytra (Figure 5c–e). This effect ranged from weak (occurring partially in one elytron Figure 5c) to strong (completely “ballooning” the entire elytron structure Figure 5e). In all cases, the disruption separated (partially or completely) the two epithelial sheets of the elytron to create the blistered appearance. In addition, we also noticed that this effect occurred in the hindwing (Figure 5f), although this was far less frequent than in the elytron (1/14), potentially because the hindwing is trapped and held underneath the elytron structure through later pupal development, thus limiting the blistering effect. We did not detect disruptions in tissues outside of the wings in any Tc‐msr RNAi.

3. DISCUSSIONS

3.1. A non‐candidate gene approach

The evolution of elytra was a key driving force behind the massive radiation of the Coleopteran order of insects. Despite the evolutionary importance of this morphologically novel structure, the molecular mechanisms underlying its formation are still largely a mystery. In the past, the genes identified as important for wing development in Drosophila (fly wing genes) have served as the foundation on which the conserved and diverged genetic mechanisms promoting elytron formation have been based (Ravisankar et al., 2016; Tomoyasu et al., ²⁰⁰⁹) (Figure 6). And while effective, this candidate gene approach can only explore genes that have been previously identified in Drosophila and are also evolutionarily maintained across species. In the past, as a first step to moving beyond candidate genes, we explored less fly‐centric options such as a semi‐candidate gene approach, which still relied on fly genes, but not fly wing genes (Linz & Tomoyasu, 2015). In the present study, to become completely independent of Drosophila, we used RNAseq to explore the transcriptome of the elytron and hindwing of Tribolium, as well as another beetle species, Dorcus. Using RNAi, we then functionally screened a subset of the genes revealed by our sequencing (after removing known fly wing genes). Using this approach, we have revealed a number of interesting insights regarding the patterns of the Coleopteran fore and hindwing transcriptome and the function of elytron‐enriched genes in forming the novel structures found in elytra (Figure 6; also discussed in detail below).

3.2. Dynamic transcriptomic changes throughout the development of beetle wings

In an effort to explore the mechanisms underlying the formation of the beetle unique wings, we used RNAseq to explore the developmental transcriptome of the elytron and hindwing. The first striking pattern we observed in both Tribolium and Dorcus RNAseq was that sequencing replicates clustered by replicate, rather than tissue type (Supporting Information: Figure S3a,b). Some experimental caveats could explain this pattern. The non‐isogenic nature of our Dorcus individuals (obtained from a recreational supply store) could produce this effect; however, our Tribolium samples (pu11 line) are isogenic, and the similar pattern between the species suggests this is not the cause. Instead, this pattern is likely produced by a unique wing‐transcriptome landscape, which we were able to observe because of our specific method for isolation of individuals for replication.

For both beetle species, we visually determined the stage for dissection and isolated both fore and hindwing discs from each chosen individual as replicates. In Tribolium, we also pooled tissues for each replicate, but the fore and hindwing tissues of each replicate were still composed of discs from the same individuals. Each replicate series thus represents a unique developmental snapshot of the fore and hindwing discs. Combined, the patterns we observed in our samples, although not ideal, provide some biologically relevant insight into the patterns of the beetle fore and hindwing transcriptome landscape. These patterns tell us that the beetle fore and hindwing transcriptomes are each very similar at any single developmental time point, yet also highly dynamic over even short variations in developmental time. Similar patterns of large, stage‐specific, variations in transcriptome‐landscape have also been reported in the wings of Drosophila (Pavlopoulos & Akam, 2011). Ultimately, future work using alternative staging methods and/or more sensitive RNA isolation techniques (single disc per replicate) across larger developmental periods will help further resolve these variations in the wing transcriptomes over developmental time.

The dynamic aspect of wing transcriptomes likely limited our ability to detect differentially expressed genes in two ways. First, our differential gene expression analyses could be biased towards genes that maintain differential expression over broad developmental time periods. This may be why relatively few unique genes (with only modest fold changes) were detected in each fore and hindwing in both beetle species. Second, even a slight difference in the stage between the samples of two beetle species could significantly reduce the number of detectable differentially expressed genes that are shared between Dorcus and Tribolium. Although we dissected wing tissues from the larvae of both species at roughly the same stage (i.e., at the last larval stage just before the prepupal formation), the developmental time course is drastically different between Dorcus and Tribolium. Tribolium spends 7 days at the last larval stage with the last 2 of those days being immobile (i.e., prepupa), while Dorcus spends about 6–8 months at the last larval stage with the last 2 weeks as immobile prepupa. Therefore, the limited number of shared differentially expressed genes we detected between two species (less than five genes overlapping in each wing tissue) is likely caused by the dynamic nature of wing transcriptomes combined with the time course differences between the two species.

Despite these limitations, the genes that were ultimately enriched in each wing tissue confirmed that our sequencing approach was still effective at revealing unique genes. This was especially apparent in Tribolium hindwing tissue, which despite having only 29 unique genes, was composed of multiple established wing genes (from Drosophila and also studied in Tribolium and other insects) including (among others) Ubx, sal, and kn, which are known to be differentially expressed, and critical in promoting hindwing formation (in fly [Galant et al., ²⁰⁰²; Hersh & Carroll, ²⁰⁰⁵] and beetle [Tomoyasu et al., ²⁰⁰⁵]). In addition, Ubx and sal were two of the four genes uniquely expressed in both Tribolium and Dorcus hindwing tissue, suggesting that differential expression of these genes is maintained over broad developmental time periods across the Coleopteran order.

Conceptually similar Drosophila transcriptomic experiments (using microarray) have also been performed to compare fly fore and hindwing development (Hersh et al., 2007; Mohit et al., ²⁰⁰⁶; Pavlopoulos & Akam, ²⁰¹¹). These experiments reported greater numbers of differentially expressed genes; however, the overlap between the genes revealed in the separate studies was limited, again highlighting a dynamic wing transcriptome in insects (Tomoyasu, 2017). We, and others, have also previously performed formaldehyde‐assisted isolation of regulatory elements sequencing (FAIRE‐seq) in Tribolium (and also Drosophila) wing tissues to explore differences in open chromatin profiles (cis‐regulatory differences) of these two tissues (Lai et al., 2018; McKay & Lieb, ²⁰¹³). FAIRE‐seq revealed strikingly similar patterns to our RNAseq (but from the chromatin state point of view), with very few (<5) open‐chromatin differences occurring between elytron and hindwing tissue. In both fly and beetle, these peaks were nearly all associated with Ubx, which suggests that our observed gene expression differences are consistent with regulatory differences in these tissues. In summary, our RNAseq experiments appear to have revealed intriguing insights into the patterns of the developmental landscape of the beetle wing tissues. These patterns corroborate similar patterns observed previously in other insects at both the transcriptomic and cis‐regulatory level (in Drosophila [Nystrom et al., ²⁰²⁰] and more recently in the butterfly, Bicyclus anynana [Murugesan et al., ²⁰²²]). Additionally, we have successfully captured both known and unknown genes enriched in both elytron and hindwing tissue, genes that can serve as key non‐candidate genes for this, and future studies.

3.3. RNAseq and the formation of morphologically novel structures: Few unique genes with few unique functions

Our RNAseq experiments revealed a relatively small pool of differentially expressed genes in elytron tissue. We explored the function of a group (~25%) of these genes and found a subset with roles in forming elytron‐unique morphological features (exoskeletalization, sensory structures, and veins) (Figure 6c). Despite this, the precise role of a majority of these genes within their respective developmental pathway appeared to be conserved between beetles and flies. This highlights the conserved nature of a majority of genes during wing development, and also the difficulty of revealing uniquely expressed genes with unique developmental roles using RNAseq.

When we consider the intricate gene networks that pattern complex structures, there are a variety of stages during which genetic events can bring about morphological evolution. In a broad sense, these events/changes can occur during the early initial patterning of a tissue, during later intermediate patterning events, or even at the terminal branches of gene networks (the so‐called “realizator” genes). Our RNAseq analysis explored transcriptomic differences occurring during intermediate patterning events (inferred by the timing of our sequencing and the genes revealed in our differential expression analysis). During this period, we revealed that the fore and hindwing transcriptomes of Coleopterans are highly similar, despite the drastic morphological differences observed between these adult tissues. A few possibilities exist to explain these patterns; (i) similar genes are spatially differentially regulated at this time to produce downstream changes, yet their overall level of expression remains the same; thus, making them invisible to differential expression‐based analyses, or (ii) upstream (early to mid‐level) patterning events are similar and, instead, changes in morphology have been achieved through modifications at the terminal realizator genes. Deciphering these possibilities may require additional seq‐based approaches over larger developmental periods in combination with more intricate histological and functional dissections (such as single‐cell RNAseq; Deng et al., ²⁰¹⁹; Everetts et al., ²⁰²¹).

Another area of consideration revolves around a well‐described phenomenon in morphological evolution; that formation of novel structures relies on co‐option of entire gene networks (such as the beetle horn [Moczek & Rose, ²⁰⁰⁹], butterfly wing eyespot [Keys et al., ¹⁹⁹⁹], and elytron [Tomoyasu et al., ²⁰⁰⁹]). Our differential expression analysis appeared to be highly effective at revealing genes uniquely expressed within these networks (see below); however, the genes responsible for inducing the network itself remain more obscure. RNAseq is merely a single snapshot in the complex developmental cascade, and this likely makes the elucidation of the precise co‐option events difficult. It is also possible that differential expression of a gene is not necessary for a unique co‐option event (see point “i” above). For example, the Tc‐ap genes have co‐opted the exoskeletalization pathway, yet the ap genes maintain similar expression between elytron and hindwing (Tomoyasu et al., 2009) and thus would not be picked up by differential expression analyses. This may mean that elucidating the genes that control these co‐option events requires a more targeted experimental approach, such as unbiased loss‐of‐function studies for genes that are expressed early in the wing development regardless of their degree of elyton/hindwing differential expression.

3.4. RNAseq highlights gene regulatory networks patterning the morphological features shared between elytron and body wall

As a wing, a number of morphological features of elytra are highly unique, including their exoskeletalization as well as the array of sensory structures covering their surface (although the hindwing also possesses sensory structures, albeit far fewer than elytra). These features, however, are also shared with the Coleopteran body wall. The exoskeletalization of the elytra, specifically, has been previously shown to have occurred via multiple co‐options of the exoskeleton formation pathway that also functions in the body wall (Figure 6c). This co‐option event has occurred downstream of conserved wing genes (Tc‐ap genes, Tc‐acheate scute (Tc‐Ash), and the vein formation pathway) (Tomoyasu et al., 2009). The formation of sensory structures in elytra also appears to rely on the same gene network as in body wall tissue, since RNAi for the critical sensory formation gene, Tc‐Ash, removes sensory organs throughout the beetle (Tomoyasu et al., 2009; Tomoyasu & Denell, ²⁰⁰⁴; Wheeler et al., ²⁰⁰³). One observed (and expected) trend in our differential gene expression analysis was the enrichment of genes in elytron tissue whose orthologs in Drosophila are known to be critical to sensory structure formation, patterning, and function (14/57), as well as genes involved in pigmentation and cuticle formation, namely exoskeletalization (4/57). Additionally, our cross‐species comparison between Tribolium and Dorcus elytron unique genes also revealed two genes (sca and repo) known to function in sensory formation. A subset of these genes was included in our functional screen (sensory: Tc‐repo, Tc‐sca, Tc‐cs10, Tc‐poxn and cuticle: Tc‐hr38) allowing us to assess their function in Tribolium wing formation. Tc‐sca and Tc‐hr38 caused disruptions in sensory structures and pigmentation of the elytra, but also caused similar defects in tissues outside of the elytra including the body wall. These results further highlight that the genetic pathway involved in sensory formation and exoskeletalization of elytra is shared across adult tissues and was co‐opted, in its entirety, in elytra. Tc‐repo caused severe metamorphosis defects preventing pupation (even at reduced dsRNA concentrations) and as such, we were unable to assess its role in adult morphology. Tc‐poxn failed to produce any observable abnormalities, although subtle unnoticed changes in the size or shape of sensory bristles (reported in some Drosophila poxn mutants; Boll & Noll, ²⁰⁰²) are possible.

3.5. The unique role of a chemosensory gene in elytra

In contrast to other sensory‐related genes, Tc‐cs10 appears to have a unique role in elytron formation. Chemosensory genes are a large and diverged family of genes that code for soluble proteins that transport chemical signals from the environment to olfactory receptors within sensory organs. Three chemosensory genes (of the ~20 in beetle; Forêt et al., ²⁰⁰⁷) were uniquely expressed in Tribolium elytra (Tc‐cs10‐12). The reported chemosensation‐based role of CSPs in other insects initially suggested that enrichment in elytra was a product of sensory organ function; however, the unique expression of these genes during the early development of the elytra was puzzling. Surprisingly, RNAi for Tc‐cs10 did not simply disrupt the function of sensory organs, but rather sporadically eliminated the formation of both the sensory structure itself as well as the adjacent vein. These disruptions appeared to be dispersed randomly throughout elytron tissue and also varied between individual Tc‐cs10 RNAi adults (data not shown). Because the CSPs are such a large family of proteins, one possibility is functional redundancy among family members. We also tested triple RNAi for Tc‐cs10, Tc‐cs11, and Tc‐cs12 (Supporting Information: Table S6); however, we did not observe any enhancement of the vein disruptions. Nevertheless, there are ~20 CSPs in Tribolium, and excluding redundancy among large gene families is challenging.

A wide range of tissue and stage‐specific expression patterns have been reported for CSPs in other insects (Forêt et al., 2007; Wanner et al., ²⁰⁰⁵) and developmental roles of CSPs have also been reported in the embryo of the honeybee, Apis mellifera (Maleszka et al., 2007). To our knowledge, however, this is the first example where CSPs have been implicated in postembryonic development. The small soluble nature of these proteins may facilitate their roles in developmental processes, but further work will be needed to fully reveal how these genes function in beetle elytra.

3.6. Tc‐msr and epithelial sheet binding

A loss of adhesion between epithelial sheets of the wing can create a distinct blistering effect, which has been used in flies to identify components of integrin mediated adhesion (Prout et al., 1997; Walsh & Brown, ¹⁹⁹⁸). RNAi for Tc‐msr, which was enriched in elytron tissue, produced this blistering effect in Tribolium elytra (and in hindwings, although less frequently). The blistered elytron phenotype was a common disruption observed in the Tribolium wings throughout the iBeetle screen (49 genes), and in fact, Tc‐msr was one of the genes that caused this effect when knocked down (Schmitt‐Engel et al., 2015). Schmitt‐Engle et al. also examined 19 of the orthologs of the beetle blister genes in Drosophila wings by transgenic RNAi knockdowns. Disruption of Dm‐msr (or Dm‐Eip71CD) generated a curly wing phenotype (Supporting Information: tab. 4 in Schmitt‐Engel et al., 2015), suggesting that the role of these genes may be similar (curly wings can be caused by defects similar to blistering), but not fully conserved between flies and beetles. Our transcriptome study found that Tc‐msr is expressed higher in elytra than in hindwings, and RNAi for Tc‐msr resulted in the blistering effect more frequently in elytra than in hindwings. These outcomes suggest that this gene may have gained an additional role in mediating the unique trabeculae‐mediated binding of the epithelial sheets found in elytra (Noh et al., 2016).

3.7. D. hopei, the Japanese giant stag beetle

One unique aspect of the work presented here is the use of Dorcus as a second Coleopteran species. Although a number of beetle species with available sequencing data exist (e.g., see Du & Zeng, ²⁰¹⁶), we chose Dorcus for a number of reasons. First, Dorcus (and stag beetles in general) is ingrained as a part of popular culture in Japan, being utilized as a collector's item, a pet, and an object of intense fascination. Second, these beetles also have established culturing methods, well‐described life cycles, and are available through a multitude of commercial vendors. And third, their morphological features are also intensely fascinating. For example, their mandibles are variably enlarged in males, and the degree of the enlargement often depends on the environment. This provides a unique opportunity to study the evolution of both sexual dimorphisms and polyphenism. With their immense popularity, the study of stag beetles could mark one of the first pop‐culturally relevant model organisms ever utilized. The sequencing data produced in this study can serve as a foundation for future studies using Dorcus.

4. MATERIALS AND METHODS

4.1. Tribolium cultures

Beetles were cultured on whole‐wheat flour (+5% yeast) at 30°C with 70% humidity. We used the isogenic pu11 nubbin (nub) enhancer trap line, which expresses EYFP in nub‐positive tissue (including wings), in all experiments (Clark‐Hachtel et al., 2013; Lorenzen et al., ²⁰⁰³; Tomoyasu et al., ²⁰⁰⁵).

4.2. Tribolium tissue sample preparation and RNA extraction

Individual pu11 larva were staged and selected for dissection based on wing disc proliferation marked by EYFP expression. We chose larvae that had progressed for ~5 days in the last larval stage (see Supporting Information: figure S4 in Clark‐Hachtel et al., ²⁰¹³). Selected larvae were individually dissected in cold Phosphate‐buffered saline (PBS). Briefly, each larva was cut between the head and the first thoracic segment (T1) and also between the first and second abdominal segment (A1 and A2). This region was then cut at the dorsal and the ventral side of the wing discs on each (left and right) side of the larva, leaving two strips of tissue from T1 through A2 containing fore and hindwing discs. The discs were then each carefully removed entirely from the remaining body wall tissue using EYFP expression to confirm the specificity of dissection. The elytron and hindwing discs were then transferred separately to a tube containing homogenization buffer from the ARCTURUS PicoPure RNA Isolation Kit (Thermo Fisher Scientific). Six to seven elytron and hindwing discs were dissected for each replicate (3/4 individuals), with five replicates per tissue type. After dissection, tissues were briefly homogenized and RNA was extracted following the protocol provided by the kit. Samples were stored at −80°C until further processing.

4.3. Tribolium library construction and high‐throughput sequencing

Samples were transported to University of Cincinnati Bioinformatics Center for the following steps. Total RNA was quality‐checked and quantified using a BioAnalyzer 2100 (Agilent) and determined to be of good quality. A total of 10 (5 elytra and 5 hindwings) sequencing libraries were constructed using TruSeq RNA sample prep kit v2 (Illumina). Each library was indexed and sequenced on 1/6 of a lane on the HiSeq. 2500 platform (Illumina), producing 20–30 million 50 bp single‐end reads for each replicate (Supporting Information: Table S1).

Reads were quality‐checked using FastQC v.0.11.2 (Andrews, 2010) and cleaned using Trimmomatic v.0.32 (Bolger et al., 2014) to remove adapter sequences and perform quality trimming. Trimmomatic was run with the following parameters, “ILLUMINACLIP: TruSeq.3‐SE. faLEADING:2 TRAILING:2 SLIDINGWINDOW:4:2 MINLEN:25.” After trimming FastQC was run again. For all samples, 2% of reads or less were removed by trimming. The trimmed reads were then mapped against the Tribolium genome (Tcas 3.0) using Tophat2 (v.2.0.13) and allowing up to two mismatches per read (Kim et al., 2013; Langmead & Salzberg, ²⁰¹²). Approximately 90% of the reads for each sample were successfully mapped to the reference genome (see Supporting Information: Table S1 for detailed statistics). Read counting was done for each gene using the htseq‐count command from the HTSeq package v.0.6.1 (Anders et al., 2015). The Tribolium transcriptome data is available at NCBI Sequence Read Archive (accession number PRJNA744737).

4.4. Tribolium differential expression and Blast2Go analysis

To explore expression differences between Tribolium elytra and hindwings we used the R (v.3.1.2) package DESeq2 (v.1.6.3), which employs negative binomial modeling and adjusts for multiple testing using the Benjamini–Hochberg method (Benjamini & Hochberg, 1995; Love et al., ²⁰¹⁴). GO analysis was also performed using Blast2GO (v.4.1.2) (Conesa et al., 2005; Gotz et al., ²⁰⁰⁸).

4.5. Dorcus RNAseq, de novo transcriptome assembly, differential gene expression analysis, and annotation

Dorcus larvae were purchased from a stag beetle supply store (RTN & Mushi‐sha). When larvae were determined to be nearing the prepupal stage, the end of the last larval stage (by size, mobility, and puparium formation), two individuals were selected for dissection. In cold PBS, the left and right, hindwing and elytron disc were removed from the two individuals. Because of the size of the tissue, each individual disc was used as one replicate. RNA was isolated using Qiagen RNeasy kit following standard procedures. The following steps were performed by Laboratory for Phyloinformatics at RIKEN Center for Biosystems Dynamics Research. A total of eight (four elytra and four hindwings) sequencing libraries were constructed using TruSeq RNA sample prep kit v.2 (Illumina). Each library was indexed and sequenced on a quarter lane of the HiSeq. 1500 platform (Illumina) producing ~20 million 120 bp paired‐end reads for each replicate (Supporting Information: Table S2). The reads were quality‐checked (FastQC v.0.10.0) followed by trimming adapter sequences and filtering low‐quality bases as follows: trimming TruSeq adapters by cutadapt v.1.2.21, trimming 5′‐end 1 nt by seqtk v.1.0‐r31, trimming low‐quality nucleotides in 3′‐end by fastq_quality_trimmer implemented in fastx toolkit v.0.0.13 with the options “‐t 20 ‐l 60 ‐Q 33,” and filtering reads comprising of low‐quality nucleotides by fastq_quality_filter in FASTX‐Toolkit v.0.0.13 with the options “‐q 20 ‐p 80 ‐Q 33.” The resulting reads were pooled and de novo assembled by the Trinity assembler (v. trinityrnaseq‐r2013‐02‐25) using the “jaccard‐clip” option to remove chimeric reads (Bowtie [v.1.0.0] was also required for Trinity function) (Haas et al., 2013). The resulting assembly contained 55,541 components (referred to as “genes”), with an N50 length of 2851 bp (see Supporting Information: Table S2 for Dorcus sequencing and assembly statistics). We also assessed the completeness of the de novo assembled D. hopei transcriptome by Benchmarking Universal Single‐Copy Orthologs (BUSCO v.3.0.1) (Kriventseva et al., 2019; Simão et al., ²⁰¹⁵). We observed a high level of completeness (98.4%–97.0% complete with the “eukaryote,” “metazoan,” and “insecta” OrthoDB v.10 lineages) with low proportions of fragmented or missing BUSCOs (0.8%–1.2% and 0.8%–1.8%, respectively) and a moderate amount of duplication (35.7%–39.1%) likely due to inclusion of all putative splicing variants in this assessment. Combined, these data support a high‐quality de novo assembly (Supporting Information: Table S2). Reads for each sample were then mapped back to the assembly using Bowtie (v.1.0.0) (Li & Dewey, 2011), and expression of individual genes was quantified by RSEM (v.1.2.3). Differential expression analysis was then performed using DESeq2 (v.1.6.3) run in R (v.3.1.2) (Love et al., 2014). Assembled Dorcus genes were then assigned orthology to Tribolium proteins using BLASTx and reciprocal blasting using tBLASTn (evalue cutoff 1 × 10⁻²) (Altschul et al., 1990). Only genes with Tribolium orthologs were considered for comparisons between Dorcus and Tribolium differentially expressed genes. The Dorcus transcriptome data are available at NCBI (accession numbers PRJNA744737 and GKCD00000000).

4.6. Initial RNAi screen

To screen multiple genes simultaneously (see Section 2) we designed primers with T7 tails to generate template DNA containing two genes from our RNAi screening list as well as the EYFP sequence contained in each plasmid (a positive control for subsequent RNAi) (see Section 2, Supporting Information: Table S3 for primers, and Supporting Information: Figure S1 for the scheme). We designed additional primers with T7 tails to generate individual template DNA for each gene within our list (Supporting Information: Table S3).

4.7. Tribolium Gene cloning

Genes chosen for further analysis were amplified from pupal cDNA and cloned into PCR4‐TOPO vector (PCR4‐TOPO cloning kit; Thermo Fisher Scientific) using gene‐specific primers (Supporting Information: Table S3). Templates for dsRNA were synthesized by PCR using the TOPO_RNAi primer, or gene‐specific primers (Supporting Information: Table S3). Additionally, non‐overlapping off‐target regions for each cloned gene were also generated using gene‐specific primers with T7 tails attached to their 5′ ends (Supporting Information: Table S3). Additional gene fragments were also synthesized using GBlocks (Integrated DNA Technologies) and gene‐specific primers with T7 tails (Supporting Information: Table S3).

4.8. dsRNA synthesis

All dsDNA templates contained the T7 polymerase promoter sequence at the 5′ and 3′ end. In vitro transcription was used to generate dsRNA for each gene using MEGAscript T7 kit (Thermo Fisher Scientific). 1.5 μg of template was used for this step. The resulting dsRNA samples were then purified by MEGAclear kit (Thermo Fisher Scientific). We used agarose gel electrophoresis to confirm the specificity of the products. Concentration of dsRNA ranged from 3 to 10 μg/μl (Philip & Tomoyasu, 2011).

4.9. Tribolium injection and dissection

Injections were performed during the early last larval stage or penultimate larval stage. EYFP expression was used to select the correct stage. A Zeiss Discovery v12 or Leica M80 microscope were used for all injections. Number of larvae injected, stage of larvae, and dsRNA concentration for each injection can be found in Supporting Information: Tables [Link], [Link], [Link]. After injection, larvae were kept in flour until the adult stage when they were dissected and analyzed for phenotypic disruptions. Detailed protocols for injections can be found in (Linz et al., 2014; Philip & Tomoyasu, ²⁰¹¹). Resulting adults were allowed to age for 5 days, after which their elytra and hindwings were dissected and fixed overnight in 95% ethanol. Dissected wings were mounted on glass slides in Permount.

4.10. Image processing and documentation

Images of mounted wings were captured by Zeiss Axio Imager.M2 with Zeiss Axiocam 503 color camera. Whole adults were imaged using Zeiss AxioCam MRc5 connected to the Zeiss Discovery V12 microscope. Zeiss AxioVision Extended Focus module was used to obtain images with increased focus depth. Some pictures were enhanced only for brightness and contrast with Adobe Photoshop Creative Cloud.

5. CONCLUSIONS

As a morphologically novel structure, beetle elytra offer a unique opportunity to study the molecular mechanisms that underlie this critical evolutionary process. In the past, candidate genes or even semi‐candidate genes have been the framework for molecular‐based studies of elytron evolution. Here, using RNAseq, we have become independent of Drosophila and revealed sets of genes enriched in the developing transcriptome of the elytra (and hindwing). We then functionally screened a subset of these genes and found previously undescribed roles for a group of genes in forming some of the unique morphological features of this beetle‐specific wing. Future analyses will be needed to functionally examine the entirety of the genes enriched in elytron tissue. Also, future RNAseq‐based experiments using carefully choreographed staging of tissues can be used to resolve patterns in the complex developmental landscape of the beetle wings. Combined, these analyses will help provide a more comprehensive view of how gene network evolution facilitated the formation of novel structures, and help resolve the intricacies of this critical evolutionary process.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1002/jez.b.23188.

Supporting information

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Supporting information.

Linz, D. M. , Hara, Y. , Deem, K. D. , Kuraku, S. , Hayashi, S. , & Tomoyasu, Y. (2023). Transcriptomic exploration of the Coleopteran wings reveals insight into the evolution of novel structures associated with the beetle elytron. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, 340, 197–213. 10.1002/jez.b.23188

DATA AVAILABILITY STATEMENT

Supporting information and data files are available online in the Supporting Information: Material section of this article. Tribolium and Dorcus RNAseq reads are available in The Sequence Read Archive (SRA) at https://www.ncbi.nlm.nih.gov/sra/, reference number PRJNA744737. The Dorcus transcriptome assembly is available at Genbank (accession number GKCD00000000).