Horn primordial tissue shows sex-specific development in prepupae

In T. dichotomus, males develop exaggerated horns on both the head and thorax while female beetles are hornless (Fig 1A). In order to determine when male-specific horn morphogenesis begins, we first compared the development of male horn primordia and tissue from the same region in females. In T. dichotomus, sexually dimorphic horn development becomes apparent during the prepupal stage [9]. Approximately ten days prior to the end of the last (third) larval instar of males, cells of the dorsal head epidermis begin to evaginate to form a sac (Fig 1B). The surface of the sac continues to grow and fold during prepupal development, and forms four concentric circles at its distal tips that correspond to the branched tips of the adult head horn [3] (arrowheads in Fig 1C–1E). The development of the thoracic horn follows a similar progression, although the onset of evagination occurs later than in the head horn (Fig 1F–1I). In addition, the thoracic horn forms a different surface folding pattern than head horns, reflecting the differences in adult horn shape.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Transcription factors and signaling molecules are differentially expressed between male and female horns, and between horn types in T. dichotomus.

(A) Male (top) and female (bottom) adult of T. dichotomus. (B–I) Male head (B–E) and prothoracic (F–I) horn primordia during the prepupal stage. The dorsal aspect of the horn primordium is shown. Arrowheads indicate concentric circles corresponding to the distal tips of the adult head horn. (J–Q) Horn counterparts in the female head (J–M) and prothorax (N–Q). Scale bar is 1 mm. (R–U) MA-plots for each RNA-seq data set comparison. Panels R and S show the intersexual comparisons, whereas T and U show intrasexual comparisons. Male head horn versus female head horn (R), male thoracic horn versus female thoracic horn (S), male head horn versus male thoracic horn (T), and female head horn versus female thoracic horn (U). Grey circles indicate transcripts, tan circles represent genes DEG at an FDR of < 0.05, orange circles represent TF, and blue circles represent SM. HH, head horn; TH, thoracic horn; DEG, differentially expressed gene; FDR, false discovery rate; TF, transcription factor; SM, signaling molecule.

https://doi.org/10.1371/journal.pgen.1007651.g001

In T. dichotomus, the female pupa has a small protrusion on the head, and no visible horn on the thorax. This small pupal head horn in females disappears after eclosion, likely through programmed cell death during the adult molt [10]. The primordial tissue for the head structure in prepupal females displays less folding and lacks the four concentric circles typical of male horns (Fig 1J–1M). Prepupal female tissue located at the same region as male thoracic horn tissue, while still displaying a folded morphology, displays less folding than male tissue and lacks the morphology of male thoracic primordia (Fig 1N–1Q). Consistent with their different morphology at the pupal stage, females show no clear evagination in their horn primordia late in development (Fig 1M and 1Q).

To discover horn formation genes we sampled horn primordia at the onset of differentiation (Fig 1C, 1G, 1K and 1O) because (i) male-specific tissue folding is present during this stage, and (ii) the primordia are more clearly recognizable than in earlier stages, enabling consistent tissue collection among samples.

Discovering 11 transcription factors that affect horn formation in T. dichotomus

We then performed RNA-seq analysis to identify genes that contribute to elaborate horn morphology. To construct a T. dichotomus transcriptome for read mapping, we assembled four cDNA libraries, comprising male head horn tissue (111.2M reads, see Materials and methods for detail), male thoracic horn tissue (107.8 M reads), female head horn tissue (113.6 M reads) and female thoracic horn tissue (115.1 M reads). The summary of sequencing and de novo transcript assembly is shown in S1 Table. The total number of trinity transcripts is 82,108 and the contig N50 based on all transcript contigs is 3,158 (S1 Table). We evaluated the quality of the assembled transcript model by performing a BLAST search against both the NCBI nr database as well as the OrthoDB 5 database. The assembled transcripts showed the highest similarities to Coleopteran (beetle) genes, particularly to genes of the red flour beetle Tribolium castaneum in the NCBI nr database search (S1A–S1C Fig). In the OrthoDB search, 75.3% of the assembled transcripts had putatively orthologous genes in D. melanogaster and 89.3% of the transcripts were orthologous to T. castaneum (S1D and S1E Fig). As T. dichotomus and T. castaneum are both beetles, the increased percentage of transcripts orthologous to T. castaneum demonstrates the quality of the T. dichotomus transcriptome constructed in this study. We further evaluated our transcriptome with BUSCO [11]. Our transcriptome indicates 97.8 and 95.7% coverages over complete BUSCOs in Metazoa and Insecta, respectively (S1F Fig), reflecting the high quality of the T. dichotomus transcriptome.

We mapped short read sequences to the transcriptome and calculated mRNA abundance. We checked the distribution of count data with a multi-dimentional scaling plot, and found that gene expression between males and females was distributed distinctly, and that at least two biological replicates for each sample clustered together (S2 Fig). We then made two different types of comparisons to identify transcripts involved in horn formation. First, we made an intersexual comparison between the same horn types of male and female beetles (e.g. between head horns in male and female). This was performed in order to identify genes driving the development of different morphologies between male and female horns. Next, we compared tissue intrasexually, between different horn types in either males or females (e.g. between head horn and thoracic horn in males). Our goal was to identify genes that contribute to the unique horn morphologies present in each segment. Our intersexual comparison identified 739 differentially expressed genes (DEGs), and our intrasexual comparison identified 814 DEGs at a false discovery rate lower than 0.05 (S2 Table; S1–S4 Appendices).

To understand the developmental processes each tissue undergoes during the stage we investigated, we categorized genes enriched in each comparison by gene set analysis using ErmineJ software [12]. In our intersexual comparisons, Gene Ontology (GO) terms associated with muscle formation (e.g. myofibril assembly, striated muscle cell development) and metabolism (e.g. cellular amino acid catabolic process, lauric acid metabolic process) were overrepresented (S3 Table; S3 Fig). This suggests that sex differences in horn morphology (male versus female head horn, male versus female thorax horn), at least during the prepupal period (Fig 1C, 1G, 1K and 1O), arise primarily from differences in the amount of growth of each horn type. In contrast, differences between horn types (head horn versus thorax horn) are associated with differential expression of appendage patterning genes. GO terms associated with morphological differentiation, such as cell fate specification, leg disc pattern formation and head development, were clearly enriched in the comparisons between head and thoracic horns within each sex (S3 Table; S3 Fig).

RNAi-mediated knockdown affects horn formation: SP8 and pnr

We evaluated the function of DEGs during horn formation by using RNAi-mediated gene knockdown. Given the important functions of transcriptional regulation in animal development, we focused on DEGs annotated either as transcription factors (TFs) or as signaling molecules for our RNAi screening. In our annotation based on BLAST search against the NCBI nr database, we identified 49 candidate horn formation genes comprising 38 TFs and 11 signaling molecules (Fig 1R–1U; S4 Table). We first performed RNAi for all 49 candidate horn formation genes as an initial screening, and repeated experiments for 13 TFs and a signaling molecule in which we observed visually detectable effects on head and/or thoracic horn morphology (S4 Table). We consequently obtained 11 TFs with clear functional roles in horn development (S4 Table).

Among the 11 genes we identified, SP8 and pannier (pnr) RNAi most drastically affected horn phenotypes. SP8 is a member of the SP family of transcription factors, and has an orthologous amino acid sequence to D. melanogaster Sp1 (S4 Fig). In D. melanogaster, SP family genes Sp1 and buttonhead (btd) play partially redundant roles in development of ventral appendages and mechanosensory organs [13–15]. D. melanogaster Sp1 is involved in leg disc fate determination and postembryonic growth of ventral appendages [13, 14]. The role of a D. melanogaster Sp1 ortholog on growth of ventral appendages appears to be conserved in the beetle T. castaneum [16]. There are three sets of SP family genes in metazoans [17], and while members of all three families were present in our transcriptome, SP8 was the only SP gene identified as differentially expressed. SP8 knockdown induced an extra horn-like outgrowth from the anterior proximal region of the male head horn (Fig 2A and 2B). Head horns in T. dichotomus are unusual in having four tips, suggesting two successive bifurcation events (Fig 3A). This new RNAi-induced horn outgrowth—a horn on a horn—exhibited a bifurcated tip (Fig 2C), similar to the bifurcated tip of T. dichotomus thoracic horns (Fig 3A), as well as head horns of other, more typical, Dynastinae species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Extreme horn phenotypes after SP8 and pnr RNAi.

(A, B) Lateral view of adult heads from EGFP (A) and SP8 (B) RNAi beetles. The arrow in B indicates the ectopic horn. (C) A magnified SEM image of an ectopic horn in a SP8 RNAi beetle from ventral view. (D, E) Dorsal view of adult prothorax areas of EGFP (D) and pnr (E) RNAi beetles. Arrowhead indicates the furrow formed along dorsal midline instead of a thoracic horn. Scale bars are 5 mm in A and B, 1 mm in C, 3 mm in D and E.

https://doi.org/10.1371/journal.pgen.1007651.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. RNAi-mediated gene knockdown alters horn shapes and sizes in head and thorax.

(A–K) RNAi effects on horn shapes. (A) Dorsal view of head horn and prothorax in an EGFP RNAi beetle. (B) Schematics indicate measured regions. Gray shaded regions are a central groove (CG) and side grooves (SGs) in head horn, and a groove in thoracic horn (right). (C–K) Dorsal view of head horn and prothorax in RNAi beetles are displayed. Arrowheads indicate significantly decreased areas. Horn area and body length are plotted, and linear regression lines are drawn for each gene in red. Gray dots and gray regression lines are for EGFP RNAi beetles. Asterisks show significance of each RNAi treatment using a Wald test on the logistic regression. Asterisks for significantly affected areas are in red. Asterisks denote significance: * P < 0.05, **P < 0.01, ***P < 0.005. Scale bars are 5 mm. (L–Q) RNAi effects on horn size. (L) Lateral view of head and prothorax in an EGFP RNAi beetle. (M) Green lines indicate measured segments. (N–Q) Lateral view of head and prothorax in RNAi beetles are displayed. Horns that are statistically shorter than controls are denoted with arrowheads, whereas arrows indicated abnormally wider proximal parts in Sox14 (P) and ab (Q) RNAi beetles. Horn length and body length are plotted, and linear regression lines are drawn for each gene in red. Gray dots and gray regression lines are for EGFP RNAi beetles. P-values show significance of each RNAi treatment according using Wald test. Asterisks denote significance: * P < 0.05, **P < 0.01, ***P < 0.005.

https://doi.org/10.1371/journal.pgen.1007651.g003

In D. melanogaster pnr expression is localized to the dorsal midline, where it acts as a selector gene specifying dorsomedial identity within the head and thorax [18]. pnr also specifies dorsal regions of eye-antennal and wing imaginal discs and is an upstream regulator of both decapentaplegic (dpp) and wingless (wg), making it an ideal candidate for a master regulator of patterning and growth of head and thoracic horns [19]. Knockdown of pnr led to the development of a furrow along the dorsal midline of the T1 segment of the thorax, instead of a horn (Fig 2D and 2E). This complete loss-of-horn phenotype strongly suggests that pnr also functions as a dorsal-medial selector in T. dichotomus prepupal development. Interestingly, pnr did not affect development of the head horn, which suggests that cells giving rise to head horns may not be dorsal in origin—a possibility we discuss further below.

The remaining genes analyzed in our RNAi screening displayed only modest phenotypes, and thus we quantitatively assessed the effect of RNAi treatment by measuring both horn shape and length (S4 Table).

Nine additional transcription factors contribute to a characteristic horn shape and size

Common gene functions between female pre-segmental regions and male head horns suggest roles of clypeolabrum formation genes in horn shape, size and number

To investigate the regulatory and functional relationships among our 11 genes of interest, we searched the integrated D. melanogaster genomics database FlyMine using D. melanogaster orthologs as a query [23] (S7 Table). Although we found no pathway enrichment that suggests cooption of a specific developmental signal from this analysis, we found two terms from the Berkeley Drosophila Genome Project (BDGP) that were significantly enriched. These terms are assigned based on expression patterns of the genes in D. melanogaster embryonic development, and both the terms “clypeolabrum” (P = 2.022393e^-4, Holm-Bonferroni test) and “clypeo-labral primordium” (P = 0.020819, Holm-Bonferroni test) were significantly enriched (S7 Table), suggesting that our 11 candidate genes are involved in formation of this region. This finding agrees with data from Onthophagus beetles, where it has been reported that head horns are anatomically positioned around the pre-segmental ocular and clypeolabral regions, and genes involved in the patterning of embryonic pre-segmental regions are important for post-embryonic head horn differentiation [6].

The arthropod labrum originates as an ectodermal outgrowth arising just in front of the mouth (i.e., pre-segmental), in a domain of the head defined by expression of Optix [24]. Whether the labrum is a true appendage, or instead a non-appendicular projection, is debated [25]; but it forms in many insects (including T. castaneum beetles) from a pair of appendage-like outgrowths that later fuse medially into a single structure [24]. In T. castaneum, appendage patterning genes including dpp and wg are expressed in labral buds, but their domains of expression are reversed compared with other trunk appendages, leading Posnien et al. (2009) to propose that the labrum arose as an anterior outgrowth of the head from an ectopic redeployment of the appendage patterning network [24].

Our observation that genes functionally involved with head horn growth are associated with clypeolabral identity led us to hypothesize that the novel head horn in T. dichotomus is at least partially derived from pre-segmental regions, and that some of the RNAi induced changes in horn shapes we observed in this study were a consequence of perturbing development in the clypeolabral region. To test this hypothesis, we first analyzed the phenotype of the anterior head in RNAi treated females, as the morphology of the pre-segmental region is much clearer in hornless females than in males (Fig 4A). We analyzed female heads for genes which were assigned the BDGP term “clypeolabrum” in the FlyMine analysis including Rx, Optix, SP8 and Tbx20 (S7 Table). Rx and Optix are required to form the larval clypeolabral region in T. castaneum [26]. Sp1 loss-of-function mutants in D. melanogaster lack the mandibular head segment tissue (note that the D. melanogaster Sp1 ortholog in beetles is SP8, see S4 Fig for SP family gene phylogenetic tree) [17, 27]. Although midline, a Tbx20 homolog, is expressed in the clypeolabral region in D. melanogaster embryos, no head patterning function has yet been reported for any Tbx20 homologs in insects.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Depletion of “clypeolabral” genes affected formation of pre-segmental region in both females and males.

(A–E) Anterior region of adult female heads after RNAi treatment are displayed in a dorsal view, with the clypeolabral regions in red. Arrowheads in D indicate defects in the lateral section of the clypeolabral region in a Tbx20 RNAi beetle. Regions boxed with white dotted lines in A–E are magnified in A’–E’. (F–J) Corresponding head regions in male RNAi beetles. Arrowheads in I indicate defects at lateral parts. (K–M), EGFP (K and M) and SP8 (L and N) RNAi beetles from frontal (K and L) and ventral (M and N) views. Arrowhead in L indicates the loss of dorsal part of labrum in an SP8 RNAi beetle. The ventral side of labrum is in dotted line in M and N. Scale bar is 1 mm.

https://doi.org/10.1371/journal.pgen.1007651.g004

In female T. dichotomus, the anterior region of the clypeus has a characteristic curved shape, and has shorter hairs on the surface than the posterior region (Fig 4A and 4A’; clypeolabral region is in red). Rx RNAi in females created a furrow in the clypeolabral dorsal midline, and dramatically changed both the shape and the surface hair pattern (Fig 4B and 4B’). Optix RNAi females showed a wider anterior end compared to the control counterparts (Fig 4C and 4C’). Tbx20 RNAi resulted in a less pronounced anterior curved shape, and the posterior part of the clypeolabral plate was less extended laterally (Fig 4D and 4D’, arrowheads). We found that male beetles exhibited similar shape changes in the anterior head region compared to controls, indicating that these gene functions are shared between sexes, despite the corresponding body wall changes in males that ultimately form the horn (Fig 4F–4I).

We further discovered that SP8 RNAi, which forms a small ectopic horn in males, also affected the clypeolabrum shape in both males and females. In SP8 RNAi females, the anterior central part of the head has a u-shaped form, a rough surface, and long hair (Fig 4E and 4E’). Similar changes in anterior head shape were observed in male SP8 RNAi beetles (Fig 4J). The labrum and clypeus are separate in control beetles, but were instead fused in SP8 RNAi males (Fig 4K and 4L). The labrum segment lost nearly completely the normal dorsal structure, and the hairy ventral region was expanded in treated animals (Fig 4M and 4N).

These drastic changes of the clypeolabrum structure that coincide with ectopic horn formation suggest that the anterior-most tissue of the head has lost its identity, and that both dorsal and ventral characters are juxtaposed in atypical ways in SP8 RNAi animals. Artificial juxtapositions of dorsal-ventral signals in D. melanogaster imaginal discs can lead to new axes of outgrowth and ectopic miniature wings or legs that form at the base of existing structures [28, 29]. It is conceivable that altered dorsal-ventral patterning in the T. dichotomus clypeolabral region created a new center for horn growth, and thus induced the formation of a small horn following SP8 RNAi.

Combined, our analyses reveal that Rx, Optix, Tbx20 and SP8 function in the formation of both the male and female clypeolabral region and in the male head horn, and changes in the expression level of these genes alter horn shape and size.

Appendage patterning genes play roles in T. dichotomus horn formation

Previous work suggests that appendage patterning gene networks were coopted to form novel horn outgrowths in Onthophagus beetles, as both appendages and horns deploy similar developmental pathways [5]. We thus hypothesized that differentially expressed genes in our study would include appendage patterning genes. Four of our 11 genes are homologs of known appendage formation genes in D. melanogaster: BarH1, dac, SP8 and ab [13, 30–32]. RNAi knockdown of these four genes led to defects in both antenna and leg development, confirming their functional role in appendage growth in T. dichotomus (Fig 5A–5E, 5G–5K and 5G’–5K’). In addition, Sox14 RNAi led to fusion of appendage segments in the distal tip, a phenotype that has not been reported for this gene in any other insect (Fig 5F, 5L and 5L’). RNAi knockdown of dac, and ab affected both head and thoracic horn shape (Fig 3E, 3H and 3K), and ab affected the size of both horns (Fig 3Q). Our finding that dac plays a role in horn shape formation in T. dichotomus is noteworthy because RNAi knockdown of dac does not affect Onthophagus taurus horn formation [5]. While SP8 was assigned the BDGP term “clypeolabrum”, and functions in patterning this region, BarH1, dac, ab, and Sox14 were not assigned this term (S7 Table), suggesting independent recruitment of appendage patterning genes into the horn development program.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Five genes showed defects in appendage formation after RNAi.

(A–F) Antennae in RNAi beetles. (G–L) Metathoracic legs in RNAi beetles. Arrowhead indicates fusion of segments. Corresponding parts are indicated with red parentheses for antennae (A–F) and legs (G–L), respectively. Boxed regions in G–L are magnified in G’–L’. Scale bar is 1 mm.

https://doi.org/10.1371/journal.pgen.1007651.g005

Differential function of clypeolabral genes and conserved function of appendage genes between T. castaneum and T. dichotomus

To understand the ancestral function of genes involved in T. dichotomus head horn development, we performed RNAi analysis in the red flour beetle T. castaneum. Although T. castaneum is a member of the Tenebrionidae, and not a direct ancestor of the Scarabaeidae, this beetle is considered to represent the ancestral head shape [24]. In addition, we used this species for comparison because of the availability of both genome sequence and the ability to easily perform systemic RNAi [33, 34]. We analyzed orthologs of the T. dichotomus clypeolabrum patterning genes we identified, Rx, Optix, Tbx20 and SP8, and appendage-patterning genes, BarH1, dac, Sox14 and ab.

Contrary to the drastic change in horn and clypeolabral morphologies in T. dichotomus, Rx RNAi caused no detectable changes to adult head morphology compared to control injections in T. castaneum (S7A and S7B Fig; n = 22 and 20 for EGFP and Rx, respectively). We could find all three SP family gene orthologs in the T. castaneum genome, which enabled us to target the SP8 gene specifically [17] (S4 Fig). RNAi treatment for SP8 in T. castaneum affected appendages as previously reported [16]. SP8 RNAi resulted in fused appendage segments similar to results seen in T. dichotomus. (S7I Fig; n = 22). Notably, after SP8 RNAi, the clypeolabral region of the flour beetle was unaffected (S7F–S7H Fig). In fact, we detected no obvious morphological changes in the head surface of T. castaneum after RNAi for either Optix or Tbx20 (S7C and S7D Fig; n = 22 and 18 for Optix and Tbx20, respectively). We note that Optix RNAi affected compound eye formation, as has been reported (S7E Fig; t = 7.69, degree of freedom = 42, P = 1.53e^-09, student’s t-test) [6].

These results suggest that the clypeolabrum patterning genes we examined, although critical for formation of the embryonic head, no longer exert detectible effects when knocked down in late-stage larval T. castaneum. This constitutes an important difference between head patterning in Tribolium flour beetles and both Trypoxylus rhinoceros beetles and Onthophagus dung beetles, and suggests that the origin of head horns in beetles may involve heterochronic shifts in the timing of patterning of the clypeolabral region. In contrast to the non-conserved function of clypeolabral patterning genes, the function of appendage patterning genes was well conserved between T. dichotomus and T. castaneum, as we observed similar effects between both species for all five genes (Fig 5G–5L; S7I Fig).

Developmental origin of horns in T. dichotomus

Recent advances in developmental genetics have elevated the flour beetle T. castaneum to become a model system for studying development in insects generally [35, 36]. In particular, studies of embryonic expression of trunk and appendage patterning genes recently led to a new model for the formation of segments and sutures in the insect head [35]. This “bend and zipper” model proposes that the flat epithelial band containing head segments and ventral appendage primordia migrates anteriorly, folding upwards and backwards (Fig 6A and 6B) [37]. The anterior head lobes grow around the clypeolabral region, eventually fusing with each other along the anterior midline of the head. This now-inverted epithelium fuses with the maxillary and labial segment regions of the layer below, completing the head capsule (Fig 6C and 6D). The bend and zipper model accounts for the mysterious placement of frontal/ clypeolabral appendages in Paleozoic Euarthropoda, and for the coronal, frontal, and subgenal sutures demarcating head capsules of many insects [35, 38].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Developmental and evolutionary origin of head and thorax horns in Scarab beetles.

(A–C) “Bend and zipper” model of head development in T. castaneum from Posnien et al. (2010) [35]. The clypeolabral region (light blue) lies anterior to the mouth and is originally part of the “pre-segmental” region of the germ band (A); This region migrates anteriorally and vertically, inverting and folding backwards and becoming enframed by the lateral head lobes (gray) (B) which fuse along the midline of the head (C). The labrum (red) is formed from the fusion of a pair of non-appendicular ventral evaginations in the clypeolabral region. (D, E) Head and prothorax fate map in larval T. castaneum (D) and inferred fate map in adult T. dichotomus (E) based on gene expression and RNAi knockdown. We propose that the head horn originated from the clypeolabral region (light blue), explaining our finding that ventral appendage patterning genes and clypeolabral genes function in horn growth even though it extends from the top of the head. Thoracic horns appear to arise from the dorsal midline at the juncture of the head and thorax, a region defined by both Scr (light green) and pnr (dark green) expression. Coronal and frontal sutures are shown in dotted lines in C and D. (F) Partial phylogeny for the families and subfamilies of scarab beetles (Coleoptera: Scarabaeidae), showing the presence of head and/or thoracic horns. The majority of species with exaggerated horns are concentrated within three distantly related clades (Geotrupidae, Scarabaeinae, Dynastinae) that collectively represent only 20% of extant scarab species. For this reason, horns are thought to have arisen multiple times independently within the superfamily (blue shading), leading Darwin and others to speculate on the “special tendency” of the scarabs towards evolution of enlarged horns. (G) However, eleven of the included clades contain either rudimentary horns, at least a few genera or species with enlarged horns, or mutant individuals with horns (black shading), and our present results reveal striking similarities in the development of Scarabaeinae and Dynastinae horns. Thus, an alternative explanation of our results is that the ancestor to the scarabs had both horns, and mechanisms for suppressing horn growth (e.g., sexual dimorphism). If true, then the repeated, parallel evolution of horns within the scarabaidae would reflect taxic atavism of an earlier ancestral state. Tree topology is derived from Ahrens et al., (2014) [7], but see also [71,72]; species numbers and taxon descriptions are derived from Ratcliffe and Jameson (2006) [73].

https://doi.org/10.1371/journal.pgen.1007651.g006

Importantly, this model also provides a basis for proposing the developmental origins of horns in beetles, possibly explaining the paradoxical finding of our study as well as studies of dung beetles in the genus Onthophagus [6], that head horns, though appearing to lie on the top, or dorsal region of the beetle head, in fact express genes typical of anterior “pre-segmental” (clypeolabral) body regions and ventral imaginal discs. For example, Optix, Rx, and Tbx20, genes whose expression is confined to the pre-segmental clypeolabral region in diverse bilaterians including T. castaneum [26], are expressed in head horn tissues in Trypoxylus dichotomus and appear functionally involved with specifying horn size and shape (Fig 3). Similarly, the zinc-finger transcription factor Sp8 (D. melanogaster Sp1) is known to affect relative amounts of growth of ventral appendages in D. melanogaster [13], T. castaneum [16] and milkweed bugs [39], and we show that disruption of Sp8 through RNAi knockdown is sufficient to induce formation of an ectopic head horn in T. dichotomus (Fig 2B and 2C).

Together these results suggest that the head horns in both Onthophagus and Trypoxylus beetles form from appendage-like outgrowths in the clypeolabral region of the head, a pocket of anterior (pre-segmental) cells with ventrally-patterned outgrowths that folds upwards and backwards during embryogenesis such that the novel appendages growing from this region extend vertically from the top of the head in adult beetles (Fig 6E). Whether this means that beetle head horns are homologous with the various non-appendicular clypeo-labral evaginations of fossil and extant panarthropods (e.g., primary antennae of onychophorans) remains to be investigated [38].

In T. castaneum and other insects, including D. melanogaster, the homeotic gene Scr is localized to the dorsal ridge, the anterior-most region of the body capable of having a dorsal fate (note the consistency of this expression pattern with the “bend and zipper” model of head development, Fig 6) [40]. The dorsal ridge forms the boundary of the head and thorax, and is comprised of regions of the maxillary and labial segments, as well as parts of the first thoracic (T1) segment (prothorax). Although Scr retains this regional specification in insects pre-dating the origin of wings (e.g., firebrats), its best-studied function in pterygote species is to repress growth of wing primordia in the prothorax [41]. RNAi knockdown of Scr leads to vestigial T1 wings on the prothorax of D. melanogaster, milkweed bugs, cockroaches, mealworm beetles, T. castaneum as well as Onthophagus beetles [21, 41–45]. Wasik et al. (2010) showed that Scr RNAi knockdown affected development of prothoracic, but not head horns, in Onthophagus, and we demonstrate here that Scr RNAi leads to reduced growth of prothoracic, but not head horns in T. dichotomus [21] (Fig 3O).

Similarly, pnr, a gene involved in embryonic dorsal closure in D. melanogaster, is typically expressed along the dorsal midline of the thorax and abdomen, where it acts as a selector gene specifying dorsal-medial identity to tissues including the heart [18, 19]. Here we show that pnr expression is necessary for growth of the prothoracic horn, as RNAi knockdown resulted in a complete loss of the dorsal-medial region of the thorax, including loss of the entire horn (Fig 2E). Together, these results suggest the developmental locus of the thoracic horn in T. dichotomus is the anteriormost zone of the dorsal midline, a region specified by intersecting domains of expression of Scr and pnr (a region defined by both light green and dark green in Fig 6C–6E).

Evolutionary origin(s) of beetle horns

The Scarabaeidae contain approximately 35,000 species, the overwhelming majority of which are hornless. Yet, horns are thought to have arisen many times independently within this clade, such that today several thousand species bear elaborate weapons. The extreme sizes of these structures, and their concentration within a single family of beetles, led Darwin to conclude that sexual selection acted especially effectively in scarab beetles [46], and Arrow suggested they have a “special tendency” to the acquisition of horns [47]. Arrow went so far as to conclude “it is certain that these horns have had no common origin” [47]. Horns are assumed to have arisen multiple times for two reasons: most scarab species (80%) lack horns; and the sub-families with the majority of horned species (“dor” beetles [Geotrupinae]; dung beetles [Scarabaeinae]; rhinoceros beetles [Dynastinae]) are too widely dispersed within the Scarabaeidae (Fig 6F).

Our study provides the first detailed characterization of horn development from a rhinoceros beetle, T. dichotomus, permitting the parallel origins of rhinoceros and dung beetle horns to be contrasted at a mechanistic level. Although a few genes clearly show lineage specific differences in function (e.g., dac affects horns in T. dichotomus but not Onthophagus), the overwhelming pattern is one of similarity. Head horns in both lineages arise from ventral appendage-like outgrowths in the anteriormost, pre-segmental clypeolabral region of the head; in both lineages the expression and function of these clypeolabral patterning genes appear to involve a heterochronic shift markedly divergent from head development in T. castaneum (S7 Fig); and thoracic horn outgrowths in both lineages appear to extend from the anteriormost region of the dorsal midline, a zone specified by the homeotic gene Scr, and, we now show, the domain of expression of pnr.

In addition to the precise locations of horn outgrowth being similar, the formation of the outgrowths themselves appears similar, involving in both horn types and in both lineages the partial deployment of appendage patterning networks. Finally, recent studies of the mechanisms of sexual dimorphism in beetles suggest that sexually dimorphic growth of both types of horns is regulated by the same pathway. Alternative splice forms of doublesex in males and females regulate sex specific patterns of growth of enlarged mandibles in stag beetles (Lucanidae), as well as horns in Onthophagus and Trypoxylus, consistent with a shared capacity for female-specific repression of weapon growth across the scarabs [9, 48, 49]. Consequently, our results reveal many layers of mechanistic parallelism between the horns of rhinoceros and dung beetles (S8 Fig), and point to a surprisingly repeatable path to the evolution of these extreme, sexually selected structures.

An alternative explanation is that horns arose once, before the diversification of the scarabs, and that the repeated evolution of horns in diverse lineages represents “taxic atavism” [50, 51] as has been described recently for amphibian teeth [52] and supersoldier castes in Pheidole ants [53]. Indeed, several clues suggest the ancestral scarab beetles may have been horned. First, most of the primarily-hornless subfamilies contain at least a few species with either rudimentary horns (e.g., Pleocomidae, Passalidae, Ochodaidae, Orphninae) or with fully-developed horns (e.g., Melolonthinae, Cetoniinae, Rutelinae; Fig 6G). Second, the pupal stages of many scarabs have thoracic ‘horns’, and these are often present in individuals (e.g. females) or species that lack this horn as adults. Pupal ‘horns’ may serve a current function as support structures protecting animals during the vulnerable metamorphic molt [54, 55], but they may also represent developmental carry-overs from a horn that was present in the adult stages of an ancestor [56, 57]. Third, even within completely hornless species—in one case a species within a completely hornless subfamily, the Ceratocanthidae, which have been a distinct clade for at least 65 million years—mutant adult individuals occasionally appear with fully developed horns, and these horns also resemble the horns of other scarabs [56, 58] (Fig 6G). These observations led Emlen et al. (2006) to propose that perhaps the ancestral scarabs did have horns, as well as a developmental capacity to shut off horn growth (e.g., in females) [59]. If true, this would mean that the hornless state of most present-day scarab species reflects a derived condition entailing the repression of horn growth. Future studies will be needed to distinguish between these alternatives, including examining horn growth in additional horned scarab lineages such as the Cetoniidae and Geotrupidae, and testing whether the putative ancestral developmental potential to produce horns remains in currently hornless species.

Beetle rearing

We purchased larvae of T. dichotomus from Kuwagata Koubo Mushikichi (Fukuoka, Japan), and Roiene (Gunma, Japan). Larvae were sexed as previously described [9], individually fed on humus in plastic containers, and kept at 10 °C until use.

T. castaneum was reared on flour powder in plastic containers at 30 °C. After dsRNA injections, each larva was separated in 24-well plates until adulthood.

Sample collection, cDNA library construction and sequencing

Head and thoracic horn primordia were manually dissected out from T. dichotomus larvae in ice-cold 0.75% sodium chloride, snap-frozen in liquid nitrogen and stored at -80 °C until use. The developmental stage of each tissue was determined from the external morphology of dissected horn primordia. Total RNA was extracted using the RNeasy mini kit (Qiagen, Valencia, CA, USA) according to manufacturer’s instruction. On-column DNase I treatment was performed. RNA purity was assessed with Qubit RNA HS assay kit (Thermo Fischer Scientific, MA, USA). RNA integrity was analyzed on a Bioanalyzer 2100 (Agilent Technologies, CA, USA), and RNA quality is shown in S8 Table. One μg of total RNA from a single beetle was used for each paired-end cDNA library construction using the TruSeq RNA sample preparation kit (Illumina, San Diego, CA, USA). Three biological replicates were sequenced for each sample. The cDNA library was sequenced on an Illumina HiSeq 2000 (Illumina, San Diego, CA) generating 150 bp paired-end reads.

De novo transcript assembly and sequence analysis

We performed de novo assembly of short read sequences using Trinity (version r2012-06-08) with default configurations without any additional options [60]. Raw RNA-seq data and the assembled transcripts are deposited in DDBJ Sequence Read Archive under project accession number PRJDB6456, and in DDBJ Transcriptome Shotgun Assembly division under accession numbers IADJ01000001-IADJ01127986 (127986 entries), respectively. The BLASTnr version released on Oct 30, 2012 was used for annotation of the transcript model. The same version of BLASTnr, and OrthoDB 5 databases were used for the qualification of transcriptome shown in S1 Fig. Cutoff e-values for the BLAST searches against these databases were 1.0e-4. Both Metazoa ODB9 and Insecta ODB9 datasets were used for BUSCO analysis [11]. We performed read counting and differential expression analysis using RSEM (version 1.1.21) with default configurations [61]. Multidimensional scaling of raw count values was performed and visualized with R (version 3.3.3 (2017-03-06)) [62]. For identification of differentially expressed genes (DEGs) between samples, we used the TCC package with default options, and multi-step normalization and edgeR-based DEG analysis. In this strategy, normalization of count data and DEG estimation are iterated for avoiding false positives, and we repeated this cycle three times in this study [63–65]. DEGs between samples were defined as genes at false discovery rate lower than 0.05. For MA plots in Fig 1, M and A values were calculated using TCC package, and visualized with R (version 3.3.3 (2017-03-06)) [62, 65]. Raw numerical data for the DEG analysis are provided as S1–S4 Appendices. As several new methods have been developed after our initial analysis, we now include summaries of sequence data analyses with these alternative methods, which include analysis using a newer version of Trinity, and with tag quantification with Kallisto and Salomon, in S7 Appendix [60, 66, 67].

Sp family gene phylogenetic analysis

Alignment and phylogenetic neighbor-joining (NJ) tree of SP family genes was constructed by Clustal X using putative amino acid sequences [68].

Gene ontology (GO) enrichment analysis

We searched orthologous genes to the assembled transcripts from OrthoDB database. GO terms were assigned to assembled transcripts based on GO information on orthologous genes found in the FlyBase database. We then analyzed GO terms over-represented in each comparison between RNA-seq data set using ErmineJ (version 3.0.2) [12]. GO terms enriched at an FDR < 0.5 for each comparison in ErmineJ anlaysis were summarized with REVIGO and visualized with R (version 3.3.3 (2017-03-06)) [62, 69].

RNAi-mediated gene knockdown

953 bp, 482 bp and 841 bp cDNA fragments of esg, B-H1 and dac were first subcloned into the plasmid pCR4-TOPO (Invitrogen), respectively. We in vitro transcribed double-stranded RNAs (dsRNAs) for these genes from purified PCR products using primers 5´-TAATACGACTCACTATAGGGAGACCACGTCCTGCAGGTTTAAACG-3´ and 5´-TAATACGACTCACTATAGGGAGACCACCGAATTGAATTTAGCGGC-3´.

For the remaining genes, first-stranded cDNA (fs cDNA) was synthesized with the SuperScript III Reverse Transcriptase (Thermo Fischer Scientific, MA, USA) using one μg total RNA extracted from head and prothoracic horn primordia of three males and three females at the prepupal stage. Sample collection and total RNA extraction were performed as described above. Equal amounts of each fs cDNA were mixed and used as a template for PCR. The PCR was performed using gene specific primers with T7 sequence at 5´ end listed in S9 Table, and in vitro transcribed RNAs were made from purified PCR products using the AmpliScribe T7-Flash Transcription Kit (Epicentre, WI, USA). These gene-specific primers were designed in open reading frames to produce a product of 300–400 bp length, for which the target specificity was confirmed by BLAST search against the transcriptome. We ensured that the expected size of a PCR product was amplified on an agarose gel, cut a single band from the gel, and purified the product prior to in vitro transcription. A single dsRNA was used for each target gene, and a dsRNA targeting EGFP gene was used as a control.

We injected dsRNA into the hemocoel of last instar larvae through the intersegmental membrane at the anterior-lateral position of the prothoracic segment using a syringe (Terumo Corporation, Tokyo, Japan) with a 30-gauge needle (Becton, Dickinson and Company, NJ, USA). Last instar beetle larvae were moved from 10°C to room temperature (about 25 °C) at least 2 days before dsRNA injection, and reared at room temperature after the injection until they developed into adults. Our timing of dsRNA injection preceded the onset of horn formation, as larvae normally form pupal chambers several days after moving from 10°C to room temperature, and development of sexually dimorphic horns begins during prepupal period after pupal chamber formation [70]. We note that beetles need to be stored at 10 °C as the supply is seasonal, and that storage duration of larvae at low temperature might influence the survival rate of injected beetles. The data shown include all experiments performed on both short-stored and long-stored larvae.

For T. castaneum, dsRNA was injected into late larvae using glass needles with a Femtojet (Eppendolf, Hamburg, Germany). We injected approximately 0.83 μg of dsRNAs into each larva. We repeated the experiment for Rx, Tbx20 and SP8 because we did not find any change after the first injections for Rx and Tbx20, and because a sufficient number of beetles were unable to eclose in SP8 RNAi. For this second experiment, the amount of dsRNAs were changed to approximately 3.32 μg for Rx and Tbx20, and 0.083 μg for SP8.

Horn area and length measurement

Horn area was measured in pictures of each horn region (i.e. central and side grooves in head horn, and thoracic horn groove) on a flat surface by using ImageJ 64. For the side grooves on the head horn, two areas were separately measured and averaged. Horn length was analyzed from image data of either heads or prothoraxes separated from the other body segments and imaged from the lateral aspect. Both dorsal and ventral length were separately measured using the “SegmentMeasure” plug-in for ImageJ 64 developed by Hosei Wada. Note that we included non-horn body wall parts to obtain “horn length” data in order to measure corresponding segments among different dsRNA treated animals. Dorsal and ventral length were summed after values from right and left views were averaged. Body length, from the anterior tip of clypeus to the posterior most region of the body, was measured with a digital caliper model DN-100 (Niigata seiki, Co., Ltd., Niigata, Japan). We utilized logistic regression and a Wald test in R (version 3.3.3 (2017-03-06)) to test for significance [62]. We omitted some eclosed individuals from the measurement because we were unable to apply this scheme for measurement due to highly malformed horns. Although the number of available larvae from the supplier is limited, we analyzed at least five surviving individuals per gene. Raw numerical data for the area and length measurements are provided as S5 and S6 Appendices, respectively.

Microscopy

Photographs were taken with digital microscopes; either the VHX-900 or VHX-5000 (KEYENCE, Co., Osaka, Japan). Scanning electron micrograph were taken with a VHX-D500 (KEYENCE, Co., Osaka, Japan). Adobe Photoshop CS5.1 and Adobe Illustrator CS5.1 (Adobe Systems, Inc., San Jose, CA) were used for image processing and assembly.