Transcriptome sequencing reveals maelstrom as a novel target gene of the terminal system in the red flour beetle Tribolium castaneum

Anterior-posterior patterning of the Drosophila embryo depends on spatial polarity cues provided by maternal coordinate systems (St Johnston and Nüsslein-Volhard, 1992). While anterior patterning is mediated by the Bicoid morphogen (Driever and Nüsslein-Volhard, 1988), posterior patterning largely depends on Nanos (Nos) and Pumilio (Pum) proteins (Barker et al., 1992; Macdonald, 1992; Nüsslein-Volhard et al., 1987; Wang and Lehmann, 1991). The terminal system acts to pattern the anterior and posterior non-segmented regions of the embryo through the torso-mediated receptor tyrosine kinase (Ras/MAPK) pathway (Furriols and Casanova, 2003; Li, 2005). Both Torso receptor and its putative ligand Trunk are ubiquitously expressed throughout the oocyte, and it appears that localized Torso activation results from activation of Trunk by Torso-like, the expression of which is restricted to ovarian follicle cells (Johnson et al., 2015; Mineo et al., 2015; Savant-Bhonsale and Montell, 1993; Stevens et al., 2003). Stimulation of Torso signalling results in de-repression of the terminal target genes tailless (tll) and huckebein (hkb). This activation is indirect and involves de-repression mechanisms. Terminal target genes are usually repressed and the activation of Torso relieves this repression, which allows tll and hkb expression at the poles of the embryo. Repression of terminal target genes requires the high-mobility group (HMG) protein Capicua (Cic) and the Groucho (Gro), which acts as a co-repressor. In absence of maternal Cic, tll and hkb are de-repressed and expression expands towards central regions of the embryo. This results in the loss of thoracic and abdominal anlagen (Ajuria et al., 2011; Astigarraga et al., 2007; de las Heras and Casanova, 2006; Jimenez et al., 2000; Liaw and Lengyel, 1993; Paroush et al., 1997).

The red flour beetle Tribolium develops as a short-germ embryo. The majority of segments get patterned in a secondary growth process from a so-called growth zone (Klingler, 2004; Richards et al., 2008). At the blastoderm stage, the anterior region of the Tribolium egg gives rise to extra-embryonic serosa and amnion, whereas embryonic anlagen are restricted towards the ventral-posterior region of the egg (Handel et al., 2000).

Despite these differences, Torso-signalling was shown to be active at both poles of the Tribolium egg (Schoppmeier and Schröder, 2005; Schroder et al., 2000). At the posterior, Torso is required for setting up or maintaining a functional growth zone. In the absence of Torso signalling, invagination fails and embryos are depleted of all growth zone-derived segments (Grillo et al., 2012; Schoppmeier and Schröder, 2005). At the anterior, terminal signalling is crucial for the formation of extra-embryonic membranes. In Torso or tsl RNAi, the serosa is reduced in size (Schoppmeier and Schröder, 2005; van der Zee et al., 2005).

Although the putative target genes of the Tribolium Torso pathway in the anterior region remain to be identified, posterior expression of Tc-wingless (Tc-wg), Tc-tailless (Tc-tll), Tc-caudal (Tc-cad) and Tc-forkhead (Tc-fkh) depends on Tor activation (Schoppmeier and Schröder, 2005). Yet Torso RNAi phenotypes cannot be entirely explained by the loss of the known target genes, indicating that crucial components for terminal patterning remain to be identified (Casanova, 2005). By comparing terminal-system loss-of-function (Tc-tsl RNAi) and gain-of-function (Tc-cic RNAi) transcriptomes with the wild-type transcriptome, we have identified the first comprehensive set of Tor-target genes in a short-germ insect.

To elucidate, whether Capicua fulfils a conserved function in repressing Torso target-genes, we characterized the Tribolium capicua ortholog (Tc-cic, TC004697). Tc-cic mRNA is expressed ubiquitously in unfertilized eggs and is also present in early blastoderm embryos (not shown), thus resembling the situation in Drosophila (Jimenez et al., 2000). Depletion of Tc-cic by parental RNAi results in severe patterning defects (Fig. 1; Fig. S1). Only about 7% of eggs collected in the first week after eclosion of injected animals were capable of secreting a larval cuticle, whereas 92.4% of the embryos die prior to completion of embryonic development (empty egg) (Schmitt-Engel et al., 2015). To determine whether or not the strong Tc-cic RNAi no-cuticle phenotype reflects a late function in maintaining embryonic anlagen, we performed time-lapse recordings of live embryos using a transgenic line expressing nuclear-localized GFP driven by the TriboliumEF1-α promoter (EFA-nGFP) (Sarrazin et al., 2012) (Fig. 2; Movies 1 and 2).

In wild-type embryos, at the differentiated blastoderm stage, a clear distinction arises between the wider-spaced serosal cells and the more densely spaced cells of the germ rudiment (Fig. 2A; Movie 1) (Handel et al., 2000). During gastrulation, embryonic anlagen condense and give rise to the germ rudiment that progressively becomes covered by the extending extra-embryonic membranes (i.e. serosa and amnion) (Fig. 2E,G; Movie 1) (Benton et al., 2013; Handel et al., 2000). Eventually, the head anlagen become morphologically visible, serosa window closure proceeds and germ-band elongation can be observed (Benton et al., 2013; Handel et al., 2000) (Movie 1).

Upon Tc-cic RNAi, extra-embryonic membranes are enlarged and expanded towards the posterior pole of the embryo (Fig. 2B,D; Movie 2). During gastrulation, the serosal cells spread in a posterior direction, while the residual embryonic tissue invaginates (Fig. 2F). Eventually, extra-embryonic membranes cover the entire egg, while embryonic tissue is restricted to the posterior pole of the egg and becomes internalized completely. This presumably embryonic tissue does not show any signs of morphological differentiation or axis elongation (Fig. 2H; Movie 2). At this point, embryonic development stops. Thus, the majority of Tc-cic depleted embryos die during or soon after gastrulation.

To analyse early patterning of the serosa anlage, we stained Tc-cic RNAi embryos for zerknüllt-1 (Tc-zen-1), which in wild type is already expressed in the presumptive serosa (Fig. 3G) (Falciani et al., 1996). As expected, in Tc-cic-depleted embryos the Tc-zen-1 expression expands towards the posterior, indicating that the serosa primordium is already enlarged (Fig. 3I). Interestingly, the serosa-embryo boundary in the differentiated blastoderm is still oblique in Tc-cic RNAi (Fig. 3), suggesting that Tribolium Cic may have no impact on embryonic DV patterning.

In order to reveal the function of Tc-cic in patterning gnathal and thoracic anlagen, we analysed the formation of segment primordia by Tc-even-skipped (eve) expression (Fig. 3A-F). In the blastoderm, the pair-rule gene Tc-eve is expressed in double-segmental stripes. Eventually, these primary domains will resolve into a segmental pattern (Patel et al., 1994; Schoppmeier and Schröder, 2005). Tc-eve stripe 1 correspond the maxillary primordia and Tc-eve stripe 2 covers the anlagen of the first thoracic segment. The third primary domain, however, resembles the anlagen of the posterior growth zone (Fig. 3B). In Tc-cic RNAi embryos, primary Tc-eve domains are still present, but are shifted towards the posterior pole of the embryo (Fig. 3C). In addition, the distance between Tc-eve stripe 1 and the serosa boundary is severely decreased in size, indicating the depletion of the pre-gnathal anlagen (Fig. 3F). This is supported by the loss of the ocular Tc-wg domain (Fig. 3L). In wild-type blastoderm, Tc-wg is expressed in a posterior domain as well as in the ocular anlagen of the anterior head (Fig. 3J) (Nagy and Carroll, 1994). Upon Tc-cic knockdown, this anterior domain is no longer present (Fig. 3L). Thus, gnathal and anterior thoracic segments are initially patterned, whereas pre-gnathal segments apparently are lost.

Although Tc-cic RNAi causes the expansion of extra-embryonic membranes at the expense of the anterior head anlagen, the depletion of Torso signalling results in opposing effects (Fig. 3). In Torso or tsl RNAi, the serosa is reduced in size, while the presumptive head region appears enlarged and extended towards the anterior (Fig. 3E). This finding is corroborated by the expansion of even-skipped and the reduced expression of the serosal marker zerknüllt-1 in Tc-tsl RNAi embryos (Schoppmeier and Schröder, 2005) (Fig. 3H). These phenotypes suggest that anterior Tc-cic RNAi reflects a de-repression phenotype of the terminal system.

Next, we tested whether posterior downstream gene activity is also affected in Tc-cic RNAi. To achieve this, we analysed the expression of Tc-tll, Tc-wg and Tc-wntD/8 in early Tc-cic and Tc-tsl RNAi embryos (Fig. 3). In early wild-type blastoderm, Tc-tll is expressed in a small posterior domain (Fig. 3P) (Schroder et al., 2000). Whereas the posterior tll domain is abolished in Tc-tsl RNAi (Fig. 3Q), we observed a massive expansion of Tc-tll expression in Tc-cic-depleted embryos (Fig. 3R).

A de-repression phenotype was also observed for Tribolium Wnt genes (Fig. 3J-O). At the blastoderm stage, wg/wnt1 and wntD/8 are expressed at the posterior pole (Bolognesi et al., 2008a,b). In Tc-tsl, posterior Tc-wg and Tc-wntD/8 domains are lost (Fig. 3K,N), indicating that posterior patterning by Torso may be realized, at least in part, through Wnt signalling. Again, the knockdown of Tribolium cic results in the expansion of posterior Tc-wg and Tc-wntD/8 domains towards central regions of the embryo (Fig. 3L,O). Thus, as in Drosophila (Jimenez et al., 2000; Paroush et al., 1997), Torso signalling in Tribolium induces gene expression by relieving Cic-mediated repression.

To uncover a comprehensive set of Torso target genes, we performed a differential gene-expression-screen, using next generation sequencing (NGS). To this end, we compared terminal-system loss-of-function (i.e. Tc-tsl RNAi) versus gain-of-function (i.e. Tc-cic RNAi) transcriptomes with the wild-type transcriptomes. Genes being differentially expressed in Tc-tsl versus Tc-cic RNAi (i.e. upregulated in Tc-tsl RNAi and downregulated in Tc-cic RNAi, or vice versa) likely depend on the activity of the terminal system.

Wild-type (T1), tsl RNAi (T2) and cic RNAi (T3) embryos were fixed at the differentiated blastoderm stage (7-11 h) and total RNA was isolated. Sequencing was carried out on a SOLiD 4 system (Life Technologies) and reads were mapped to the Tcas3 genome (http://bioinf.uni-greifswald.de/gb2/gbrowse/tribolium/). In total, 12,795 genes showed differences in mRNA expression levels (Fig. 4 and Table S1). We found 1801 genes to be upregulated in Tc-tsl RNAi and downregulated in Tc-cic RNAi, and 2790 genes that were upregulated in Tc-cic RNAi but downregulated in Tc-tsl RNAi (see Table S1 for details).

Candidate genes for further inspection were selected by different criteria. To account for biological background variations, genes were only considered to be differentially expressed if the fold change (T1 versus T2 or T1 versus T3) was at least 50% compared with the wild type (n=2241), yet as the data was taken from a single technical replicate, comparisons of fold change numbers have statistically no significance value.

Expression of all differentially expressed genes was monitored by assessment of maternal and different early zygotic wild-type transcriptomes (M.S., unpublished, see http://bioinf.uni-greifswald.de/gb2/gbrowse/tribolium/). Genes were only taken into account if they were expressed during early stages of development (i.e. maternal until differentiated blastoderm), reducing the number of putative candidates to 1142. iBeetle phenotypes of candidates were monitored by comparison with the iB-database (http://ibeetle-base.uni-goettingen.de). The iBeetle screen was a large-scale RNAi screen Tribolium, which identified functions in embryonic and postembryonic development for more than half the Tribolium genes (Schmitt-Engel et al., 2015). Of the differentially expressed genes, more than 300 were represented in the iB-database. This comparison allowed us to exclude genes without any obvious embryonic/larval phenotype, as well as adult lethality, defects in metamorphosis and egg laying/oogenesis upon pupal RNAi. However, genes with larval phenotypes related to anterior-posterior patterning were preferentially selected for closer inspection, reducing the number of putative candidates to 122. In addition, Drosophila orthologs of all candidates were determined and candidates were classified according to their molecular and biological function (Table S2). Genes with no Drosophila orthologs or with unknown function were preferentially selected.

Based on these criteria, we selected 50 candidate genes and proceeded to screen those for functions in early patterning (Table S2). Larvae were scored for RNAi phenotypes. Phalloidin (f-actin) and Hoechst (DNA) staining was used to monitor embryonic phenotypes. In addition, expression was monitored by whole-mount in-situ hybridization (Table S2).

We found two genes that did not result in any detectable phenotype; five genes affected oogenesis, as we observed the cessation of egg-laying upon RNAi; and in 13 cases all eggs displayed no-cuticle phenotypes. As deduced from Hoechst and β-Tubulin staining, these so-called empty-egg phenotypes are due to lethality prior to cellularization (Table S2). Depletion of 30 genes resulted in larval phenotypes, which in most cases were associated with various quantities of empty-egg phenotypes (Table S2 and Fig. 4).

A single gene did not show any obvious expression at all and six genes were expressed only during later stages of development (i.e. germ-rudiment and subsequently). Thirty candidates were expressed ubiquitously during early stages (Table S2). Yet we identified 13 genes with distinct early terminal expression domains (Figs 4 and 5, see Table 1 for details).

Of these 13 candidate genes, five genes were exclusively expressed at the posterior pole or in the segment addition zone (Fig. 4, Table 1). Eight of the 13 candidates were expressed at both poles at some point during embryonic development (Figs 4 and 5, Table 1). Interestingly, six out of these eight genes displayed an exclusive early anterior expression domain, either in freshly laid eggs (i.e. maternal expression, TC004989, TC006235, TC009922 and Tc-mael) or in undifferentiated blastoderm stages (i.e. early zygotic expression, TC004241 and TC006282), suggesting functions in patterning the anterior anlagen (Figs 4 and 5, Table 1).

To control for potential off-target effects, we tested these 13 genes by injecting dsRNA fragments not overlapping with the original dsRNA fragments (‘non-overlapping fragments’, NOFs). All phenotypes were identical to those observed for the original dsRNA fragments (not shown).

Larval phenotypes are correlated with the expression of the candidates: genes being expressed at the posterior pole predominantly display posterior truncations or aberrations (Fig. 4, Table 1). For example, larvae depleted for TC004989, TC005697 or TC003946 show the loss of abdominal segments, as seen after loss of Torso-signalling (Schoppmeier and Schröder, 2005) and RNAi with TC000868 and TC013474 results in the depletion of terminal structures (i.e. urogomphi and pygopods). Candidates that are also expressed anteriorly frequently show additional head defects. The knockdown of TC006181 or TC006282 not only affects axis elongation, but also results in deformation of the larval head (Fig. 4). For two genes (TC004241 and TC008122), we observed the loss of the entire head without obvious axis elongation phenotypes (Fig. 4). These head phenotypes resemble that of mild Tc-cic RNAi (Figs 1 and 4) and thus may represent gain-of-function phenotypes of the terminal system.

Although it remains to be elucidated the degree to which these candidate genes are indeed direct target genes of Torso signalling, our study establishes that transcriptome sequencing facilitates the identification of novel candidate genes for early anterior and posterior pattern formation in Tribolium.

We selected the Tribolium ortholog of the Drosophila maelstrom gene for closer inspection. Maelstrom (Mael) is a conserved PIWI-interacting RNA factor (piRNA), consisting of two domains: a HMG box and a MAEL domain (Sato and Siomi, 2015). In Drosophila, piRNA factors are required for transcriptional transposon silencing in both germ and ovarian somatic cells (Sato and Siomi, 2015). In addition, Mael coordinates axis specification through nucleating microtubule-organizing centre (MTOC) components (Sato et al., 2011).

The Tribolium genome contains a single maelstrom ortholog (Tc-mael, TC008172). Tc-mael is expressed maternally and transcripts accumulate at the anterior pole of unfertilized eggs and the polar body (Fig. 5A). During blastoderm stages, Tc-mael becomes expressed at the posterior pole. This domain persists until gastrulation (Fig. 5C,E). Later, Tc-mael expression is no longer detectable (not shown).

Knocking down Tc-mael by RNAi results in truncation phenotypes (Fig. 6 and Fig. S2). About 60% of the larvae lack growth-zone derived structures: abdominal segments and terminal structures are lost, resembling the loss of Torso signalling (Schoppmeier and Schröder, 2005). These posterior truncation phenotypes were sometimes (about 20%) accompanied by aberrations of the larval head, likely representing stronger phenotypes. Head phenotypes, however, are highly variable and may include deformations and/or deletions of pre-gnathal and gnathal segments. Although phenotypes likely reveal hat Tribolium mael functions in early embryonic patterning (see below), we observed additional effects of Tc-mael depletion. RNAi with Tc-mael results in incorrect dorsal closure, generating completely or partially everted (‘inside-out’) larvae (12.5%). These larvae may have escaped the primary truncation effect.

To analyse the impact of Tc-mael RNAi, we performed time-lapse recordings of EFA-nGFP embryos (Fig. 7 and Movie 3) (Sarrazin et al., 2012). Upon Tc-mael RNAi, anterior germ-rudiment formation is impaired: head anlagen of Tc-mael RNAi embryos are less condensed than wild type (Fig. 7E). During wild-type germ-band elongation, the head region extends to an anterior-dorsal position and the germ-band extends posteriorly (Fig. 7 and Movie 1). Upon Tc-mael RNAi, the head anlagen still moves in anterior ventral direction; posterior invagination, however, is delayed to some degree (Fig. 7D and Movie 3) and elongation fails (Fig. 7J,K,L; and Movie 3), supporting a function for Tc-mael in growth-zone formation and/or axis elongation (see below).

In Tc-mael RNAi embryos, the serosal window eventually closes and amniotic cavity formation seems normal (Fig. 7E,F; Movie 3). Unexpectedly, we observed the premature rupturing of the serosa at a posterior-dorsal position (Fig. 7K,L; Movie 3). In wild type, serosal withdrawal does not occur until the onset of dorsal closure at an anterior-ventral position (Hilbrant et al., 2016; Panfilio et al., 2013).

Rupture of extra-embryonic membranes is a coordinated process that involves opening both the serosa and the amnion in an anterior-ventral region, after germ-band retraction (Hilbrant et al., 2016; Panfilio et al., 2013). If the amnion fails to withdraw prior to embryonic dorsal closure, the embryo becomes everted (‘inside-out’) (Hilbrant et al., 2016; Panfilio et al., 2013). Proper serosal patterning (and subsequent withdrawal) involves the attachment of the serosa to the vitelline membrane at the posterior pole of the egg (Koelzer et al., 2014). In Tc-mael RNAi, however, the posterior serosa never seems to come into contact with the vitelline membrane at the posterior egg pole, but rather contracts anteriorly, loses tissue integrity and ultimately disintegrates (Fig. 7J-L, Movie 3). Consequently, the embryo is stuck inside the amniotic tissue and the embryo closes ventrally rather than dorsally, resulting in inside-out phenotypes.

To uncover the influence of Tc-mael on the formation of extra-embryonic membranes and patterning of anterior embryonic anlagen in more detail, we visualized the emergence of these primordia by analysing the Tc-eve, Tc-giant (Tc-gt), and Tc-zen-1 expression (Falciani et al., 1996; Patel et al., 1994) (Fig. 8). Upon Tc-mael knockdown, the Tc-zen-1 domain is smaller, indicating a reduction of the serosa primordia (Fig. 9B). However, the serosa-germ-rudiment border is still oblique.

In wild-type differentiated blastoderm stage, Tc-gt is expressed in a wedge-shaped anterior domain comprising pre-gnathal and gnathal segments (Bucher and Klingler, 2004). At that stage a secondary domain also arises at the posterior pole (Bucher and Klingler, 2004). In Tc-mael RNAi, the anterior domain is unaffected, whereas the posterior Tc-gt domain is lost (see below). In addition, the first and second primary Tc-eve domains (Tc-eve stripe 1 and stripe 2) are present in blastoderm stages, but reduced and less defined to some degree (Fig. 8B,D).

During gastrulation, extra-embryonic membranes fold over anterior and posterior embryonic margins, thereby completing the rim of a serosal window (Handel et al., 2000). Tc-zen-1 is expressed throughout serosal window formation and closure, with high expression levels at the margin of the serosa window (Fig. 8C). Upon Tc-mael knockdown, Tc-zen-1 expression is lost from the anterior rim of serosa window (Fig. 8D,G) and head primordia appear to be less condensed and are shifted towards the anterior pole of the egg (Fig. 8D). At that stage there are no obvious effects on anterior Tc-eve expression. The first and second primary Tc-eve stripes have split, while the 3rd double-segmental domain is present (Fig. 9G), reflecting that gnathal and thoracic anlagen are present. Thus, Tc-mael is required for proper serosal size regulation and head morphogenesis.

Tc-mael RNAi embryos are depleted of abdominal segments (Figs 7 and 8). The fourth primary Tc-eve domain, which corresponds to the first and second abdominal segments, does not form properly and, in addition, posterior morphology becomes highly irregular (Fig. 8G,H). Already during late differentiated blastoderm stages, the posterior Tc-gt domain is lost (Fig. 8J), indicating that posterior elongation is already affected at that stage.

To reveal whether Tc-mael is already involved in setting up a functional growth zone, we monitored the expression of Tc-wg and Tc-wntD/8 (Bolognesi et al., 2008a,b) (Fig. 9). In wild type, Wnt signalling is required for growth-zone formation and axis elongation (Beermann et al., 2011; Bolognesi et al., 2008a,b). Although in Tc-mael RNAi the terminal Tc-wg domain was basically present, the wntD/8 is notably reduced or even absent (Fig. 9C), resembling the situation in Tc-tor or Tc-tsl knockdown. Our results indicate that posterior patterning by Torso may be realized through Tc-maelstrom-dependent activation of the posterior Tc-wntD/8 domain.

Next, we monitored Tc-mael expression in both Tc-tsl- and Tc-cic-depleted embryos. As expected, we observed an expansion of Tc-mael expression in Tc-cic RNAi (Fig. 10E), whereas in Tc-tsl RNAi expression was lost (Fig. 10C). Thus, Tribolium Maelstrom is indeed a target gene of the terminal system in this beetle.

Given that Drosophila maelstrom is also required for achieving proper DV polarity (Clegg et al., 2001, 1997; Findley et al., 2003; Sato et al., 2011), we monitored DV axis formation upon Tc-mael RNAi. It has been shown before that the dorsoventral tilt of the serosa/embryo boundary depends on BMP signalling in Tribolium (Nunes da Fonseca et al., 2008; van der Zee et al., 2006). In Tc-dpp RNAi, this border becomes DV symmetric and is shifted towards the anterior. Depletion Tc-sog has opposing effects. The serosa increases at the expense of the head anlagen and the serosa-embryo boundary develops vertical to the egg axis (van der Zee et al., 2006). Thus, the loss of DV asymmetry reveals impaired dorsoventral (DV) patterning in Tribolium (van der Zee et al., 2006).

Although also in Tc-mael RNAi the serosal anlagen are reduced (Fig. 8B), the serosa-germ-rudiment border is still oblique (Fig. 8B), indicating that DV axis formation is unaffected. This is corroborated by anterior Tc-gt expression (Fig. 8J). In Tc-mael RNAi, the gt domain still shows a pronounced DV asymmetry. Thus, Tc-mael has no impact of on early anterior DV axis polarization.

Interestingly, however, we found Tc-mael to be required for posterior Tc-sog retraction (Fig. 11). In wild type, Tc-sog is initially expressed in a broad ventral domain, which subsequently clears from the posterior pole (van der Zee et al., 2006) (Fig. 11A,C). Although there is no pronounced difference in early stages, Tc-sog fails to retract from the posterior pole of Tc-mael RNAi embryos (Fig. 11D) and posterior dorsal Dpp activity (as monitored by pMAD staining) (Fig. 11F) is reduced to some degree. Consequently, posterior dorsal tissue (i.e. the dorsal amnion) lacks peak levels of Dpp activity. In wild type, Tc-iro is expressed in a domain marking the anterior border of the germ rudiment, which will give rise to the anterior amnion. In addition, Tc-iro is expressed in the dorsal amnion (Nunes da Fonseca et al., 2008). In Tc-mael RNAi embryos, the anterior of Tc-iro expression was maintained, whereas the dorsal most expression domain was absent (Fig. 11H). Thus, in Tc-mael RNAi embryos, posterior dorsal tissue is lateralized.

The depletion of Tribolium capicua results in the expansion of terminal anlagen and de-repression of terminal target genes, reflecting de-repression (i.e. gain of function) phenotypes of Torso signalling. This resembles the situation in Drosophila (Jimenez et al., 2000; Paroush et al., 1997), indicating that Torso signalling induces gene expression by also relieving Cic-mediated repression in Tribolium.

Although the functions of Tribolium capicua in terminal patterning may very well be conserved, we did not observe any obvious impact of Tc-cic on dorsoventral pattern formation. In Drosophila, the function of Cic is not restricted to Torso-RTK signalling. During Drosophila oogenesis, Cic is required to translate EGFR signalling into asymmetric pipe expression, which culminates in the establishment of a nuclear Dorsal gradient (Moussian and Roth, 2005).

In Tribolium, EGF signalling has broadly conserved roles in setting up the DV axis of the embryo (Lynch et al., 2010). The depletion of the Tribolium gurken ortholog (Tc-Tgf-alpha) results in the ventralization of embryos, resembling Tc-dpp RNAi phenotypes to some degree (van der Zee et al., 2006). In both Tc-Tgf-α and Tc-dpp RNAi, the first morphologically visible sign of DV polarity, the obliqueness of the border between serosa and germ rudiment, is lost (Lynch et al., 2010; van der Zee et al., 2006). However, as indicated by morphology and Tc-zen-1 expression, Tc-cic RNAi does not affect the obliqueness of the serosa/germ-rudiment border (Figs 2 and 3). Thus, in Tribolium there is no obvious antagonistic interaction of EGFR signalling with Cic in embryonic DV axis formation.

The expansion of the serosa in Tc-cic RNAi embryos goes along with de-repression of Tc-zen-1 (Fig. 3G,I). At first glance this situation resembles that of Drosophila, as knockdown of Dm-cic results in the lateral expansion of Dm-zen expression (Astigarraga et al., 2007; Jimenez et al., 2000; Kim et al., 2011). In Drosophila, zen is broadly activated by maternal factors (Rushlow et al., 1987). During early blastoderm stages, cic participates in the ventral and lateral repression of zen by Dorsal (Dl) (Astigarraga et al., 2007; Jimenez et al., 2000; Kim et al., 2011). At the poles, Dl-mediated repression of zen is antagonized by Torso signalling (Rusch and Levine, 1994), which likely depends on MAPK-dependent downregulation of Cic (Kim et al., 2011). However, the situation in Tribolium differs from Drosophila. The distinction between serosa and germ rudiment is primarily set up by inputs from the AP system and secondarily modulated through DV signalling. In particular, Tc-zen-1 expression and, thus, serosa size regulation depend on the combined activity of terminal signalling and Tc-orthodenticle (Kotkamp et al., 2010), whereas the obliqueness of the border between serosa and germ-rudiment is subsequently established by EFGR and BMP signalling (Lynch et al., 2010; Nunes da Fonseca et al., 2010, 2008; van der Zee et al., 2008, 2006). Given that serosal size regulation, but not the obliqueness of the border between serosa and germ rudiment is affected in Tc-cic RNAi, it appears that eggs produced after Tc-cic depletion maintain normal embryonic DV polarity.

The depletion of Tribolium maelstrom by RNAi affects axis elongation, germ-band condensation and the formation of the extra-embryonic serosa (Figs 6–8). In Drosophila, Mael coordinates oocyte polarity through interaction with components of the MTOC (Sato et al., 2011). Achieving proper cell polarity is essential for diverse processes, including cell shape changes and cell migration, asymmetric cell division, and morphogenesis (St Johnston and Ahringer, 2010). In addition, early Tribolium development depends on extensive cell-polarization events (Benton et al., 2013; Handel et al., 2000). RNAi with Tc-mael disturbs the anterior condensation of the embryonic anlagen (Figs 7 and 8). Hence, in absence of Tc-mael, cell polarity may not be established appropriately, which in turn may affect early morphogenesis. This would be equivalent to the situation in Drosophila (Clegg et al., 2001, 1997; Findley et al., 2003; Sato et al., 2011). Although we currently cannot exclude the possibility that Tc-mael RNAi phenotypes are due to impaired transposon silencing, we posit that Tc-Mael coordinates cell polarity (possibly by acting on MTOC organization) at least in serosa morphogenesis and anterior germ-band condensation in a Torso signalling-dependent manner.

Although serosa phenotypes indeed resemble aspects of Tc-tsl or Tc-tor RNAi phenotypes, we observed additional defects in head development upon Tc-mael knockdown. At first glance Tc-mael RNAi head phenotypes are not related to terminal patterning. As mentioned before, anterior Torso-signalling is involved in serosal size regulation (Schoppmeier and Schröder, 2005). The loss or reduction of this extra-embryonic membrane by Tc-tsl or Tc-zen-1 RNAi results in the expansion of the head anlagen, which is largely compensated for later in development (Schoppmeier and Schröder, 2005; van der Zee et al., 2005). Nevertheless, a small faction of larvae depleted for Tc-zen-1 or Tc-tsl do also display head defects (van der Zee et al., 2005) (F.P. and M.S., unpublished). These phenotypes are highly variable and, thus, may instead reflect defects during size compensation and head morphogenesis rather than specific functions in head patterning.

At the posterior pole, Tc-mael function is necessary for growth-zone formation. Upon Tc-mael knockdown, expression of Tc-wntD/8 is lost and posterior elongation is terminated prematurely (Figs 6 and 9). Given that Tc-wntD/8 is under the control of Torso signalling (Fig. 3) (Bolognesi et al., 2008b; Schoppmeier and Schröder, 2005), our results suggest that posterior patterning by Torso may be realized, at least in part, through Maelstrom-dependent activation or positioning of posterior Wnt domains.

On the other hand, the knock-down of Tc-wntD/8 only results in a small percentage of embryos lacking abdominal segments and additional removal of Tc-wg only enhanced the penetrance of this phenotype to some degree (Bolognesi et al., 2008a). As the effects of Tc-mael RNAi on growth-zone formation and axis elongation are considerably more severe, Tc-mael likely fulfils additional functions in posterior patterning, which may involve DV patterning.

In Tc-mael RNAi embryos, posterior Tc-sog expression fails to retract and, as a consequence, posterior dorsal tissue, i.e. the dorsal amnion, lacks peak levels of Dpp activity. Changes in Dpp activity also affect morphogenetic movements during gastrulation (Nunes da Fonseca et al., 2010; van der Zee et al., 2006). Upon Tc-dpp RNAi, for example, the germ-band is symmetrically folded inwards at the posterior pit and extends toward the anterior. In Tc-mael RNAi embryos, however, we did not observe any strong effects on posterior invagination, and posterior amniotic fold formation is normal, suggesting that posterior morphogenetic movements might be largely unaffected. Thus, even though Tc-mael also contributes – in a Torso-independent manner – to proper DV axis formation, we deem it unlikely that this function has an impact on early posterior axis polarization or growth-zone formation.

During oogenesis, Drosophila mael mutants fail to establish cytoplasmic polarity and to accumulate Gurken normally. In addition, other polarity markers (e.g. osk, BicD) fail to accumulate in the posterior ooplasm, and bcd mRNA becomes localized at both poles (Clegg et al., 2001, 1997; Findley et al., 2003; Sato et al., 2011). The Tribolium grk ortholog has no major role in AP axis formation, but rather acts in DV patterning (Lynch et al., 2010). Moreover, Tc-wntD/8 is not maternally expressed, indicating that additional potentially localized factors might be involved in wntD/8 regulation.

An important role for localized maternal determinants has long been postulated for several insect taxa, including crickets, beetles and flies. However, thus far, the existence of two morphogenetic centres (anterior and posterior) has been proven only for the long-germ wasp Nasonia vitripennis (Rosenberg et al., 2009). Tribolium uses anterior mRNA localization to some degree (Fu et al., 2012). Posterior localized maternal determinants, however, have not yet been identified (Schmitt-Engel et al., 2012). Although posterior patterning in Tribolium may depend on spatially restricted polarity cues (i.e. the terminal system) that regulate short-range acting target genes required for axis elongation, such as Wnt signalling (Beermann et al., 2011; Bolognesi et al., 2008a,b), rather than on long-range acting gradients (i.e. Nanos) (Schmitt-Engel et al., 2012), it is tempting to speculate that Tc-mael promotes growth-zone formation by localizing an as yet unknown posterior maternal determinant.

Fixation and single in situ and immunofluorescence staining were performed using standard protocols (Patel et al., 1989; Tautz and Pfeifle, 1989). Embryos were subsequently counterstained with Hoechst 33342.

Parental RNAi and dsRNA synthesis was performed as described previously (Bucher and Klingler, 2004; Bucher et al., 2002). dsRNAs were injected into female pupae at a concentration of 1 µg/µl. RNAi phenotypes were confirmed by injection of non-overlapping-dsRNA fragments. RNAi for all fragments resulted in identical phenotypes, excluding the possibility that an unrelated gene was affected (off-target effects).

First instar larvae were cleared in lactic acid/10% ethanol overnight at 60°C. Cuticles were transferred to a drop of lactic acid on a slide. The cuticular autofluorescence was detected on a Zeiss Axiophot microscope, and maximum projection images were created from z stacks using the Analysis D software (Olympus).

Sequencing was carried out on a SOLiD 4 system (Life Technologies) using 50 bp fragment reads. Mapping to the Tcas3 reference genome (http://bioinf.uni-greifswald.de/gb2/gbrowse/tribolium/) and counting were performed using LifeScope version 2.5 (Life Technologies). In order to filter out adapters, ribosomal and tRNAs, a filter reference file was created using the adapter sequences provided by Life Technologies, large and small subunit rRNA sequences from the reference, and the results of a tRNAscan-SE (v1.1) run over the reference sequence.

Samples T1 (wild type), T2 (tsl RNAi) and T3 (cic RNAi) mapped read counts were 15,517,184; 6,172,710 and 31,441,775, respectively. Counting yielded read counts for 143,603 exons in 16,543 transcripts, as provided by the Tcas3 reference. Normalization and comparison of expression levels between all three samples was performed using the differential expression analysis for sequence count data algorithm (DESeq) (Anders and Huber, 2010).

For time-lapse imaging, the EFA-nGFP-strain of Tribolium castaneum was used (Sarrazin et al., 2012). Embryos (4-6 h) were treated with 12.5% bleach twice for 30 s and then rinsed with tap water. Embryos were then mounted in a hanging drop of Sigma Halocarbon oil 700.

For time-lapse imaging, a Leica TCS SP5 II Confocal System was used. Stacks were recorded about every 12 min for an interval of 12 h at 10× magnification at 20°C. Processing of the image stacks was carried out in LAS AF Version 2.4.1 build 638 and Fiji (Schindelin et al., 2012).

We are grateful to Martin Klingler for continuous support and to Kristen Panfilio for discussions on extra-embryonic membranes. We thank the anonymous reviewers for valuable comments.