Introduction

The class Insecta is one of the most diverse metazoan groups, composing nearly 80 % of all known animal species (Grimaldi and Engel 2005). The success of this group can at least partly be credited to their diverged wing structures that provide benefits for their successful radiation, such as mobility, protection, and camouflage (Crowson 1981; Grimaldi and Engel 2005). Insect wings and their derivatives (i.e., dorsal appendages) vary greatly among different insect lineages, and also between the forewing and hindwing in individual species, providing a unique opportunity to study the molecular and genetic mechanisms underlying morphological evolution.

The processes guiding the formation of dorsal appendages are best understood in a dipteran insect, Drosophila melanogaster, and the knowledge acquired from these Drosophila studies has been utilized as a paradigm to study wing development in other insects (see Brook et al. 1996 to review Drosophila wing development. Some examples of studies in other insects are; Tomoyasu et al. 2009 and Tomoyasu et al. 2005 in beetles; Bowsher et al. 2007 and Abouheif and Wray 2002 in ants; Weatherbee et al. 1999, Martin et al. 2012, Macdonald et al. 2010, Shirai et al. 2012, and Stoehr et al. 2013 in butterflies). The genes identified as essential for wing development in Drosophila (i.e., wing genes) have been a focus of these studies, as developmental genes are often used in a similar developmental context in different organisms, with slight modifications corresponding to morphological evolution. This type of approach, the candidate gene approach, has been quite effective in elucidating the conserved as well as diverged genetic mechanisms underlying the diversification of insect wings.

Despite the effectiveness of the candidate gene approach, it is limited by the fact that the pool of genes to choose for analysis can only stem from those identified in Drosophila. While the conserved and diverged functions of the wing genes that have been evolutionarily maintained across species can be evaluated by this approach, genes uniquely important for wing development in each species cannot be identified through these candidate genes. In addition, this limitation has a risk of creating a fly-biased view of evolution, in which we treat the fly paradigm as a “starting point” of evolution. To go beyond the candidate gene approach and also escape this fly-centric viewpoint, we need to analyze insects other than Drosophila at a similar level of detail as has been achieved in the fly.

With a fully sequenced genome (Tribolium Genome Sequencing Consortium 2008), available mutants (Brown et al. 2003; Klingler 2004; Denell 2008; Trauner et al. 2009) and a robust and systemic RNA interference (RNAi) response (Brown et al. 1999; Bucher et al. 2002; Tomoyasu and Denell 2004; Tomoyasu et al. 2008), the red flour beetle, Tribolium castaneum, may offer such an opportunity. Coleopterans (beetles) contain vastly diverged wing structures compared to dipterans (flies). Flies, such as Drosophila, have typical flight wings on the second thoracic segment (T2) but have strongly reduced dorsal appendages (halteres) on the third thoracic segment (T3). In contrast, the hindwings of beetles are used for flight, while the forewings have been modified and now act as protective armoring structures, called elytra. The acquisition of elytra was an important step in the lineage leading to beetles, which has made Coleoptera one of the most successful animal groups on the planet (Lawrence and Britton 1991; Grimaldi and Engel 2005; Crowson 1981). The candidate gene approach has provided intriguing insights into beetle elytron evolution and development, which include the Hox-free nature of elytra (Tomoyasu et al. 2005) and the multiple co-options of exoskeletalization genes into the conserved wing gene network (Tomoyasu et al. 2009). However, it is also necessary to analyze genes unrelated to the fly candidate genes to gain a more comprehensive view of elytron evolution in beetles.

As a first step to escaping the limitations of the candidate gene approach, 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 not been implicated in wing development. We hypothesized that the highly pleiotropic nature of these developmental genes would increase the likelihood of finding novel wing genes in Tribolium. We call this a “semi-candidate gene” approach, as it is still relying on Drosophila studies but is independent of genes identified to be important for fly wing development.

We disrupted the function of these genes by RNAi and assessed the effects on the postembryonic development in Tribolium. Among the 25 genes we have tested (Table 1), RNAi for Tc-cactus (Tc-cact) suppressed the proliferation of wing and elytron discs in Tribolium, indicating a less characterized (but potentially evolutionarily conserved) role of Tc-cact during wing development. We also found that RNAi for the odd-skipped family genes disrupted the formation of the pleural plates on the first thoracic segment (T1), which have recently been discovered as wing serial homologs in Tribolium (Clark-Hachtel et al. 2013). In addition, we obtained several novel insights into the function of the developmental genes during postembryonic development in Tribolium. For instance, mille-pattes (mlpt) and Tc-odd-paired (Tc-opa) appear to be important for metamorphosis in Tribolium, and the odd-skipped family genes show how the redundancy of paralogs shifts over evolutionary time periods. However, despite these findings, no gene we examined was found to have novel wing-related roles in Tribolium. Although disappointing, this outcome provides us with two important insights. First, it suggests a relatively conserved nature of developmental toolkit genes and highlights the limited degree to which these genes are co-opted during insect wing evolution. Second, it reveals the difficulties of a semi-candidate gene approach and demonstrates the necessity of utilizing an approach that is truly independent of candidate genes, such as RNA sequencing and a genome-wide RNAi screen (Bucher and Klingler 2014; also mentioned in Knorr et al. 2013).

Table 1 List of semi-candidate genes examined in this study
Full size table

Materials and methods

RNA interference in Tribolium

Beetles were cultured on whole wheat flour (+5 % yeast) at 30 °C with 70 % humidity. pu11 nubbin (nub) enhancer trap line (Lorenzen et al. 2003; Tomoyasu et al. 2005; Clark-Hachtel et al. 2013) was used for all RNAi experiments. The plasmids and primers used to make double-stranded RNA (dsRNA) in this study are summarized in Table S1. dsRNA was synthesized using the MEGAscript T7 Transcription Kit (Life Technologies). Detailed dsRNA synthesis and injection procedures were described previously (Linz et al. 2014; Philip and Tomoyasu 2011). At least 20 last instar larvae were injected for each single gene tested (see Table 2 for numbers of larvae injected and their survival rates). The stage of larvae for injection was determined by the size of EYFP positive wing discs in the pu11 nub enhancer trap line. We selected larvae that had the initial stage of EYFP expression in the wing disc, which corresponds to 2.5 days after the final larval molt (see Figure S4 of Clark-Hachtel et al. 2013). dsRNA was injected into the dorsal side of the first abdominal segment at a concentration of 1 μg/μl unless otherwise stated. Each larva can hold up to 0.7 μl of dsRNA solution when injected at the stage described above. Detailed information including primer sequences and off-target effect assessment are included in Table S1.

Table 2 Summary of RNAi effects
Full size table

Phylogenetic analysis

Pfam (Finn et al. 2013) was used to identify conserved protein domains. Multiple alignments were created and curated in MEGA 5.2.1 (Tamura et al. 2011). Neighbor-joining and maximum likelihood analyses were performed in MEGA 5.2.1 with bootstrapping using 2000 replicates. Bayesian trees were generated using the GTR+I+ Γ model in MrBayes (Ronquist et al. 2012) and visualized with FigTree 1.4.2 (Rambaut 2014; Hall 2011). The conserved Zn finger domains (amino acid sequences or the corresponding nucleotide sequences) were used to create multiple alignments for the odd-skipped family members (see Document S1 for both amino acid and nucleotide alignments).

Image processing and documentation

The Tribolium adults were fixed in 95 % ethanol at least overnight. EYFP larval images were taken by submerging larvae in double-distilled water. The images were captured by Zeiss Discovery V12 with AxioCam MRc 5. 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 CS3.

Results

An RNAi survey for semi-candidate genes in Tribolium

The insect body is primarily patterned along two major axes (anterior/posterior (A/P) and dorsal/ventral (D/V)) during embryogenesis. The genes driving these processes are among the most thoroughly examined in Drosophila (for a general review, see Carroll et al. 2001; Gilbert 2014). For this study, we chose 25 Tribolium genes orthologous to these Drosophila embryonic patterning genes (Table 1) and assessed their functions at later postembryonic stages in Tribolium. From the four major classes of A/P patterning genes (maternal, gap, pair-rule, and segment polarity genes) (see Sanson 2001 for A/P patterning review), we focused mainly on the first three classes. The last class, the segment polarity genes, has been excluded because these genes are known to be involved in wing development in Drosophila. From the D/V patterning genes, we have selected dorsal and cact (see Lynch and Roth 2011 for D/V patterning review). The paralogs of some appendage genes with no known wing function (e.g., odd paralogs) and a unique Tribolium gap gene (mlpt) were also included (Table 1).

RNAi was performed at the standard 1 μg/μl concentration for each of the 25 genes (Table 2). Lower concentrations were used for genes with high lethality (mlpt and Tc-cact). A large subset of the 25 genes (15/25) failed to show any noticeable adult abnormalities (Table 2). A small group of genes (3/25) showed RNAi phenotypes previously described or parallel to Drosophila loss-of-function phenotypes (Fig. S1), while another set of genes (6/25) showed previously undescribed phenotypes outside of the wing tissues (Table 2, Figs. 1 and 3). Only one of the 25 genes (Tc-cact) had a clearly observable effect on wing and elytron development (Table 2 and Fig. 2a–d). The survival rate for genes that showed no phenotype stayed at high levels (>~70 % except Tc-slp, which resulted in a lower survival rate due to unknown pleiotropic effects) (Table 2). For the genes whose RNAi have resulted in previously undescribed phenotypes (mlpt, Tc-opa, and Tc-cact), RNAi was performed with two non-overlapping dsRNA fragments that target the same gene to confirm that the observed phenotypes were indeed caused by the reduction of the targeted genes and not by off-target effect (see Table S1 for the non-overlapping dsRNA fragments used in this study).

Fig. 1
figure 1

Metamorphosis defects induced by RNAi for Tc-cact, mlpt, and Tc-opa. a wild-type metamorphosis from a prepupa (left) to a newly eclosed adult (right). b Tc-cact RNAi. c mlpt RNAi. d Tc-opa RNAi. Tc-cact and mlpt RNAi beetles failed to pupate, while Tc-opa RNAi prevented adult eclosion

Full size image
Fig. 2
figure 2

Wing disc proliferation in Tc-cact and mlpt RNAi beetles. a, b pu11 nub enhancer trap line showing wing development in the last larval (a) and prepupal stages (b). Both elytron and hindwing discs are marked by EYFP shown in green (arrows in a–f). c, d Tc-cact RNAi larvae. Last instar larvae (c) and prepupae (d) with reduced wing disc proliferation at both stages. These larvae die just prior to pupation. e, f mlpt RNAi larvae. Last instar larvae (e) and prepupae (f), showing normal wing disc proliferation even with larval lethality similar to Tc-cact RNAi

Full size image

Evolutionarily conserved postembryonic roles of the embryonic patterning genes

Three genes, Sp8, Tc-bric-a-brac (Tc-bab), and Tc-aristaless (Tc-al), when knocked down, displayed previously described phenotypes without the indication of novel wing roles in Tribolium (Beermann and Schroder 2004; Beermann et al. 2004; Angelini et al. 2009; Angelini et al. 2012). RNAi for Sp8 (the Tribolium ortholog of Drosophila Sp1) showed disruptions in leg and antennal segmentation (Fig. S1a-e). Tc-bab RNAi induced fusion of tarsal segments in the leg (Fig. S1f-g), and Tc-al RNAi showed disruptions in the distal antennae (Fig. S1h-i). These RNAi phenotypes accentuated the evolutionarily conserved roles of these genes and also confirmed the effectiveness of our knockdowns.

Tc-cact has an uncharacterized role in wing imaginal disc development in Tribolium

cact codes for a IkB family protein that regulates the nuclear entry of Dorsal proteins during the D/V axis formation in the Drosophila embryo (Roth et al. 1989; Rushlow et al. 1989; Steward 1989; Moussian and Roth 2005). It has been reported that the D/V axis function of Tc-cact and Tc-dorsal is conserved in Tribolium (Nunes da Fonseca et al. 2008). Our screening has revealed that RNAi for Tc-cact caused pupation defects, inducing severe lethality at the last larval stage in Tribolium (Table 2 and Fig. 1a–b). The same lethality was observed even with 10 times diluted dsRNA (100 ng/μl) (Table 2). Further examination revealed that the hindwing and elytron imaginal tissues in the Tc-cact RNAi larvae failed to proliferate (Fig. 2a–d). In wild-type Tribolium, wing tissue proliferation, visualized by the EYFP expression of the pu11 nub enhancer trap line, begins at the early last larval stage (Fig. 2a) and continues to the prepupal stage (Fig. 2b) (also see Figure S4 of Clark-Hachtel et al. 2013). In Tc-cact RNAi, although we saw an initial sign of proliferation (Fig. 2c), wing imaginal tissues failed to proliferate further. Instead, we observed that the discs maintained their early size until just prior to larval lethality at the time of pupation (Fig. 2d). In contrast, larval RNAi for Tc-dorsal, whose protein products are regulated by Cact during the embryonic D/V axis formation, did not cause any noticeable morphological or developmental defects in Tribolium (Table 2). These results suggest that Tc-cact is essential, while Tc-dorsal is dispensable or redundant with other genes, for the development of wing-related tissues in Tribolium.

The odd-skipped family genes are critical for the formation of the wing serial homologs in Tribolium

The odd-skipped gene family was another class of genes included in our screening. In Drosophila, this family consists of four genes: odd, sister of odd and bowl (sob), brother of odd with entrails limited (bowl), and drumstick (drm) (Hart et al. 1996). Two members of this gene family, odd and bowl, are involved in embryonic segmentation in Drosophila (Nusslein-Volhard and Wieschaus 1980; Wang and Coulter 1996). The odd-skipped gene family in Drosophila is also important for appendage segmentation, with three members acting redundantly and the fourth member, bowl, having a unique function (Hao et al. 2003; de Celis Ibeas and Bray 2003). The Tribolium genome also contains four odd-skipped family genes, some of which have been implicated in embryogenesis (Choe et al. 2006), as well as in appendage segmentation (Angelini et al. 2012, 2009).

We first utilized dsRNA molecules for the odd-skipped family genes that contain the conserved Zn finger coding regions and are likely to knock down multiple odd-skipped paralogs in Tribolium (we hereafter refer these potential cross-reacting dsRNA as Tc-sob cross, Tc-drm cross, Tc-bowl cross, and Tc-odd cross; see Document S2 for the nucleotide sequence alignments). RNAi with these dsRNA molecules, except Tc-odd cross, caused defects in leg joint formation similar to the previously reported phenotype (arrow in Fig. 3a, d, g and Fig. S2a-e; also see Angelini et al. 2012 and Angelini et al. 2009). odd cross RNAi did not cause any noticeable abnormalities in adult morphology (Fig. S2e), presumably because Tc-odd cross dsRNA failed to knock down multiple odd-skipped paralogs despite the high nucleotide sequence conservation in the odd Zn finger coding region.

Fig. 3
figure 3

Defects in the T1 wing serial homologs induced by RNAi for odd-skipped family genes. a–c wild-type. The anterior trochantin (red arrowhead in b), posterior trochantin (white arrowhead in b), and epimeron (white arrow and black arrow in b and c, respectively) as well as T2 leg (arrow in a) are indicated. d–f Tc-sob cross RNAi. g–i Tc-drm cross RNAi. j–l Tc-sob unique RNAi. Both anterior and posterior trochantin (red and white arrowheads in e, h, k) and epimeron (white arrow in e, h, k and black arrow in f, i, l) are significantly reduced. Leg segmentation was also disrupted (arrows in d, g, j)

Full size image

None of these cross-reacting RNAi caused any noticeable abnormalities in elytra or hindwings in Tribolium. However, we noticed that these RNAi caused abnormalities in the formation of the pleural plates in the T1 segment in Tribolium. This is very intriguing, as we have recently discovered that some of these T1 pleural plates are serially homologous to elytra in T2 and hindwings in T3 in Tribolium (i.e., T1 wing serial homologs) (Clark-Hachtel et al. 2013). There are several pleural plates in the Tribolium T1 segment, including anterior and posterior trochantin, epimeron, and endopleuron (see Fig. 3 and Fig. S1 of Clark-Hachtel et al. 2013 for detailed annotation of the T1 body wall structure). RNAi with Tc-sob cross, Tc-drm cross, and Tc-bowl cross dsRNA resulted in the reduction of both anterior and posterior trochantin (red and white arrowheads in Fig. 3a–i, arrowheads in Fig. S2) as well as epimeron (arrows in Fig. 3a–i) in Tribolium. No pleural plates in other thoracic segments were affected (data not shown). These results indicate that, although odd-skipped family genes are not important for wing development in T2 and T3, these genes are critical for the formation of the wing serial homologs in T1 in Tribolium.

We next sought to investigate the individual function of the odd-skipped family members during Tribolium adult morphogenesis by utilizing dsRNA fragments unique to Tc-sob and Tc-bowl (Tc-sob unique and Tc-bowl unique). We continued to use Tc-odd cross fragment for odd RNAi, as this fragment did not appear to cross react with other odd-skipped paralogs (Fig. S2e). Among RNAi with these dsRNA fragments, only Tc-sob unique RNAi caused abnormality in the T1 pleural plates (Fig. 3j–l); both the anterior and posterior trochantin (red and white arrowhead in Fig. 3k) as well as epimeron were reduced (arrow in Fig. 3k–l). However, the abnormalities were less severe compared to those of Tc-sob cross RNAi (Fig. 3d–f), suggesting that either the Tc-sob unique dsRNA is less efficient to trigger RNAi than Tc-sob cross dsRNA or some of the paralogs function partially redundantly with Tc-sob in pleural plate formation. In contrast to Tc-sob unique RNAi, single RNAi for Tc-bowl with Tc-bowl unique dsRNA did not induce any noticeable abnormalities (Fig. S2g). Tc-odd cross + Tc-bowl unique double RNAi produced weak leg defects; however, the pleural plates were not affected (data not shown). Furthermore, the injection of Tc-odd cross and Tc-bowl unique dsRNA along with Tc-sob unique dsRNA did not enhance (and instead reduced) the Tc-sob unique single RNAi phenotypes (Fig. S2h), suggesting that the function of Tc-odd and Tc-bowl during pleural plate formation is minor (if any) compared to the Tc-sob function. The reduction of RNAi phenotypes in the triple RNAi could be due to the reduction of dsRNA concentrations (Table 2) and/or competition among three RNAi treatments (Miller et al. 2012). We could not determine whether Tc-drm has a unique role, since we failed to design dsRNA specific to Tc-drm due to the short length of the Tc-drm gene. Since Tc-sob single RNAi phenotype was not enhanced by Tc-odd and Tc-bowl RNAi, it is possible that Tc-drm is the paralog acting in a partially redundant manner with Tc-sob in pleural plate formation.

Taken together, our RNAi analysis revealed that the odd-skipped family genes are essential for the formation of the T1 pleural plates, with Tc-sob (and potentially also Tc-drm) having a more significant role than Tc-odd and Tc-bowl. Since some of these T1 pleural plates are serially homologous to wings in Tribolium, our findings indicate that the odd-skipped family genes (or at least Tc-sob) are novel “wing serial homolog” genes in Tribolium.

Two novel metamorphosis genes in Tribolium

In addition to the above findings related to wings, RNAi for two more genes, mlpt and Tc-opa, induced previously uncharacterized defects in Tribolium metamorphosis. mlpt is a unique polycistronic gene, whose messenger RNA (mRNA) contains multiple short open-reading frames within a single transcript (Savard et al. 2006; Galindo et al. 2007; Kondo et al. 2007). RNAi for mlpt at the last larval stage caused complete lethality just prior to pupation (Fig. 1c). Unlike Tc-cact RNAi, the proliferation of wing imaginal tissues appears to be unaffected in the mlpt RNAi larvae (Fig. 2e–f), suggesting that mlpt is involved in metamorphosis but not in the wing tissue proliferation in Tribolium.

opa is a pair-rule gene expressed in stripes along the A/P axis of the developing fly embryo (Benedyk et al. 1994). In Tribolium, however, Tc-opa does not appear to function as a pair-rule gene, as no segmentation defects were observed via RNAi (Choe et al. 2006). Tc-opa RNAi at the last larval stage resulted in a defect in adult eclosion in Tribolium (Table 2). Approximately 70 % of the Tc-opa RNAi beetles failed to eclose, displaying pharate adult lethality, while the remaining 30 % die prior to adult eclosion (Fig. 1d). This result suggests that Tc-opa also has an important role in metamorphosis in Tribolium.

Although deviating from the primary goal of our screening for novel wing genes, the unique metamorphosis roles of these two genes we identified will contribute to our understanding of the functional evolution of developmental genes, since neither of these genes has been implicated in metamorphosis in Drosophila.

Discussion

Drosophila has long been a classic model organism that has pioneered the study for the molecular and genetic mechanisms underlying insect wing development. The genes identified as important for wing development in Drosophila (wing genes) have been the main foci when studying wing development in insects other than Drosophila. Although this candidate gene approach has been quite successful in elucidating the diverged as well as conserved aspects of insect wing development, it is also limited since the pool of genes for analysis can only stem from those identified in Drosophila. As a first step to escaping this limitation, we took a semi-candidate gene approach, in which the selection of the genes still relies on Drosophila studies but is independent of Drosophila wing genes. Through our screening, we have obtained several novel insights into the function of the developmental toolkit genes in Tribolium.

Tc-cact in wing development

RNAi for Tc-cact at the last larval stage prevented the growth of wing imaginal discs, indicating that Tc-cact has a previously uncharacterized function in Tribolium wing development. cact encodes an evolutionarily conserved IkB protein that regulates the nuclear localization of NFkB proteins (Roth et al. 1989; Rushlow et al. 1989; Steward 1989; Moussian and Roth 2005). Dorsal and Dorsal-related immunity factor (DIF) are two NFkB factors that are known to be regulated by Cact in Drosophila during the embryonic D/V axis formation (Dorsal) (Roth et al. 1989; Rushlow et al. 1989; Steward 1989) or immune response (Dorsal and DIF) (Ip et al. 1993). Interestingly, RNAi for Tc-dorsal, which codes for the dominant binding target of Cact in the embryonic patterning event, did not cause any noticeable wing abnormalities in Tribolium. This suggests that either other NFkB factors act redundantly or Cact acts independent of NFkB factors during wing development in Tribolium. Recently, a potential role of cact in the development of the wing in Drosophila has been reported through an RNAi screen (Friedman et al. 2011). This implies that the wing-related function of cact might be conserved between flies and beetles. An alternative explanation to the wing imaginal tissue specific role of Tc-cact is that Tc-cact RNAi might have severely affected larval physiology and growth, causing overall cell proliferation including wing tissues to halt. We favor the former interpretation, as the Tc-cact RNAi larvae appear to reach the late prepupal stage with normal body size and shape (Fig. 2d). Nonetheless, further analyses will be required in order to characterize the molecular mechanisms behind the function of cact during adult morphogenesis in insects.

Function of odd-skipped family genes in the T1 wing serial homologs

We and another group have recently reported that there are tissues serially homologous to wings in some of the wingless segments in beetles (Clark-Hachtel et al. 2013; Ohde et al. 2013). One of the wing serial homologs identified in T1 belongs to a part of the dorsally originated body wall tissue (called carinated margin in Clark-Hachtel et al. 2013 and hypomeron in Ohde et al. 2013). In addition to this dorsal tissue, we found that some of the body wall plates of lateral origin (pleural plates) in T1, trochantin and epimeron, are also wing serial homologs in Tribolium (Clark-Hachtel et al. 2013). The merger of the dorsal and lateral wing serial homologs appears to be essential for the complete induction of ectopic wing structures in T1 through homeotic transformation in Tribolium.

Pleural plates in insects are thought to have evolved from the most proximal components of legs (such as epicoxa and subcoxa) in ancestral arthropods (Kukalova-Peck 1983, 2008). Since the odd-skipped family genes are critical for leg segmentation, the involvement of odd-skipped family in the formation of the T1 pleural plates may provide the first genetic evidence supporting the leg origin of insect pleural plates.

The recognition of the two classes of the T1 wing serial homologs being homologous to the two proposed insect wing origins (the paranotal lobe via paranotal hypothesis and the ancestral proximal leg segments via gill/exite hypothesis, respectively) have led us to promote a dual origin hypothesis of insect wings (Clark-Hachtel et al. 2013; Kukalova-Peck 1983, 2008; Niwa et al. 2010; Rasnitsyn 1981). The identification of genes involved in the formation of these T1 wing serial homologs is essential to further test the validity of the dual origin hypothesis. Therefore, although not directly related to the elucidation of the mechanisms involved in the diversification of wings on T2 and T3, our finding regarding the odd-skipped gene family in pleural plate formation provides valuable clues to understand the origin and evolution of insect wings.

Independent subfunctionalization among odd-skipped gene paralogs between Drosophila and Tribolium

In addition to our finding regarding the novel function of the odd-skipped family genes in the T1 wing serial homologs in Tribolium, our RNAi analysis of the odd-skipped gene family also provided an interesting look at the evolution of paralogous genes in separate organisms.

In Drosophila, bowl appears to be an outlier that possesses some non-redundant functions during postembryonic development (Hao et al. 2003). Through the usage of dsRNA fragments that uniquely target some of the odd-skipped paralogs, we found that Tc-sob (and possibly Tc-drm) has a more prominent role than odd and bowl during pleural plate formation in Tribolium. We saw the same tendency even for the evolutionarily conserved functions of this gene family, such as leg and antenna segmentation functions, in Tribolium (Fig. 3 and Fig. S2). This suggests that, although the functions of the odd-skipped paralogs as a gene family during appendage segmentation are conserved between Drosophila and Tribolium, the four paralogs have been subfunctionalized and/or neofunctionalized in a lineage-specific manner. Thus, the odd-skipped gene family provides us with a unique opportunity to investigate how the functional allocation among paralogous genes changes during evolution. Paralog-specific functional analyses for the odd-skipped family in other insects will help further investigate lineage-specific subfunctionalization among family genes.

odd-skipped gene family: lineage-specific duplications or an ancient duplication?

One caveat to the above interpretation regarding odd-skipped paralogs is that, although the Tribolium odd-skipped family genes have names corresponding to the four Drosophila paralogs, these genes might have emerged via linage-specific duplication and the functional assignments may thus be unique to each lineage. Previous phylogenetic analysis for the odd-skipped family genes indeed suggested lineage-specific duplication events (Angelini et al. 2009). This analysis was based on the nucleotide sequence of the genes, which maintains a unique degree of conservation among paralogs in each organism (Angelini et al. 2009).

Our RNAi analysis prompted us to further evaluate the orthology among the insect odd-skipped family genes. We first examined the domain architecture of the odd-skipped family proteins from three different insect orders (D. melanogaster, T. castaneum, and Apis mellifera). Pfam analysis indicated that odd-skipped family proteins contain several Zn fingers unique to this family (Fig. S3). Both the Bowl and Sob proteins from all three species contain four consecutive Zn finger domains, while Odd and Drm proteins contain fewer. Odd has three Zn fingers, except Tc-Odd, which retains a diverged (potentially degenerated) forth Zn finger. Drm is a very short protein, consisting of only two very conserved Zn fingers in all three species. Therefore, the Zn finger configuration in odd-skipped family protein is more similar among the same class of the protein than among the proteins in one species. Furthermore, in addition to the Zn finger domains, we also recognized several evolutionarily conserved motifs unique to each class of the proteins (Fig. S3). The unique Zn finger configuration along with the presence of additional motifs unique to each class are strong indications that the emergence of the four different classes of the odd-skipped family proteins preceded the radiation of the three holometabolous insect orders examined.

We next attempted to reevaluate the phylogenetic analysis for the odd-skipped gene family. The amino acid sequence multiple alignment composed of the two Zn finger domains that are conserved among all four classes of the odd-skipped family proteins contains minimal information regarding the divergence of these proteins (95 % identical, data not shown); therefore, we have decided to exclude Drm (the shortest class of the four) from the multiple alignment and focus on the Zn fingers that are conserved among the Odd, Sob, and Bowl classes of proteins. We used the Pfam-defined Zn fingers to compose the multiple alignment (Document S1) and built the phylogenetic trees by using three independent methods; neighbor-joining (Fig. 4), maximum likelihood (Fig. S4), and Bayesian (Fig. S5a). Unlike the previous tree based on the nucleotide sequence (Angelini et al. 2009), all three methods indicated that the duplication of the family preceded the split of Tribolium, Drosophila, and Apis. Interestingly (and rather puzzlingly), however, by using the nucleotide sequences that exactly correspond to the conserved Zn finger domains used for the amino acid-based tree constructions, we were able to recreate the tree obtained by Angelini et al. that suggests lineage-specific duplications (Angelini et al. 2009) (Fig. S5b). It is currently unknown why the nucleotide sequence-based phylogenetic analysis produces such a different outcome compared to that with amino acid sequences.

Fig. 4
figure 4

Phylogenetic analysis of odd-skipped family members from three holometabolous insect orders by neighbor joining. The neighbor-joining (NJ) tree is based on the alignment of the conserved Zn finger domains (see Document S1 for the alignment). The individual paralogous members cluster together, suggesting that the duplications that produced these paralogs preceded the emergence of the major holometabolous clades. Dm Drosophila melanogaster, Tc Tribolium castaneum, and Am Apis mellifera

Full size image

In addition to the domain architecture and phylogenetic analyses, we also examined the genomic structure to explore the origins of this gene family. We noticed that the synteny of the odd-skipped family genes is conserved between Drosophila and Tribolium (Fig. S6). In Tribolium, all four odd-skipped family genes are clustered within a 200-kb region. The odd-skipped family genes are also clustered in Drosophila, although bowl resides in a slightly more distant position from the other members with an inverted orientation. The conserved microsynteny between Drosophila and Tribolium further supports the idea that the duplication of the odd-skipped gene family preceded the divergence of beetles and flies.

Taken together, our domain architecture analysis, phylogenetic analysis with the amino acid sequences, and genomic architecture analysis are all in agreement that each odd-skipped paralog has the one-to-one ortholog in each organism examined, and the emergence of those paralogs preceded the evolution of three holometabolous orders. However, it will require further analysis to understand the discrepancy between the nucleotide-based and amino acid-based phylogenetic trees.

Novel genes in metamorphosis

Although unrelated to wing development, we found that mlpt and Tc-opa have previously uncharacterized roles in metamorphosis in Tribolium. The involvement of mlpt in metamorphosis is especially intriguing as the mlpt locus is transcribed into a polycistronic mRNA, which codes for a series of evolutionarily conserved small peptides (Savard et al. 2006). The orthologous gene in Drosophila has been identified (polished rice in Kondo et al. 2007 or tarsal-less in Galindo et al. 2007), which is involved in the formation of cuticular structures during embryogenesis (Kondo et al. 2010) and the formation of distal appendage structures (Pueyo and Couso 2008). However, the involvement of the Drosophila mlpt ortholog in metamorphosis has not been described. It would be interesting to investigate how the small Mlpt peptides interact with known metamorphosis pathways and if the metamorphosis function of mlpt is conserved in Drosophila and other insects.

The role of Tc-opa in metamorphosis is also noteworthy, as this is the first reported function for this gene in Tribolium. In Drosophila, opa is an important component of the pair-rule patterning mechanism during A/P axis specification (Benedyk et al. 1994). In Tribolium, however, Choe et al. did not find a patterning role for the beetle ortholog (Choe et al. 2006). Similar to mlpt, determining whether the metamorphic role of opa is conserved across other insects is an interesting question to explore in the future.

Evolution of beetle elytra

The evolution of the beetle unique wings, elytra, was an important step in beetles becoming one of the most successful animal groups on the planet (Lawrence and Britton 1991; Grimaldi and Engel 2005; Crowson 1981). Elytra therefore provide an interesting opportunity to study the diversification of insect wings in a context where the structure itself has been evolutionarily integral to the success of the insects bearing those structures. We sought to identify the genes that have been critical to the acquisition of these wing structures unique to beetles. Our semi-candidate gene approach, however, proved to be difficult. Although the Tribolium orthologs of Drosophila embryonic patterning genes are highly pleiotropic, and we identified several novel insights into the function of these genes, we could not identify any novel wing-related roles of these genes that are unique in Tribolium. These data indicate that, instead of screening limited subsets of genes, it will be more beneficial to utilize an unbiased approach to identify novel genes that have been critical to the evolution of beetle elytra. A genome-wide RNAi screening (such as the iBeetle project) (Bucher and Klingler 2014) or wing transcriptome analyses by RNA sequencing will give us a more comprehensive view of genes involved in the formation of beetle elytra and shed light on the evolutionary mechanisms underlying the vast diversity in wing structures seen among insects.