A context-dependent combination of Wnt receptors controls axis elongation and leg development in a short germ insect

The precise regulation of cell-cell communication through signalling pathways is essential for both embryogenesis and organogenesis. One significant pathway is the evolutionarily conserved Wnt signalling pathway (Janssen et al., 2010), which coordinates crucial cellular processes such as proliferation, polarity, migration, and the determination of cell fate (Habas and Dawid, 2005; Logan and Nusse, 2004; Widelitz, 2005).

The canonical or β-catenin-dependent Wnt pathway is initiated by the binding of the Wnt ligand to transmembrane receptors encoded by the Frizzled gene family. This action causes β-catenin to be stabilised in the cytoplasm and translocated into the nucleus where it activates the transcription of target genes (Bhanot et al., 1996; Cadigan and Nusse, 1997; Chen and Struhl, 1999; Hsieh et al., 1999; Rulifson et al., 2000; Wu et al., 2004; Yang-Snyder et al., 1996). Proteins of the low-density-lipoprotein (LDL) receptor-related (LRP) class act as Wnt co-receptors; in insects, they are represented by the Drosophila LDL-receptor Arrow (Wehrli et al., 2000) and in vertebrates by LRP5 and LRP6 (Pinson et al., 2000; Tamai et al., 2000).

Vertebrate and arthropod embryos rapidly elongate their body axes during early embryogenesis by adding body segments during a post-blastodermal growth phase. The central tissue controlling axial elongation (AE) is a proliferative zone at the posterior of the embryo known as the growth zone or, more generally, as the segment addition zone (Janssen et al., 2010). In arthropods, this part of the embryo is defined as the region immediately posterior to the last generated segment, the presegmental region (PSR) and the paddle-shaped posterior growth zone (Martin and Kimelman, 2009; Schröder et al., 2008). In vertebrates, the presomitic mesoderm (PSM) sheds blocks of mesodermal tissue, the somites, under the control of graded fibroblast growth factor (FGF) and Wnt signalling pathways (Aulehla et al., 2008; Dequéant and Pourquié, 2008). In arthropods, key molecules, such as caudal and torso/torsolike, as well as the Wnt ligand Wnt8 and other Wnt pathway members (armadillo/β-catenin, pangolin/TCF and arrow/LRP5/6) are required for setting up the growth zone and maintaining axis elongation (Bolognesi et al., 2008b; Bolognesi et al., 2009; Copf et al., 2004; McGregor et al., 2008; Miyawaki et al., 2004). Body axis elongation cannot be studied in Drosophila because all segments form almost simultaneously (long germ embryogenesis). Recent functional studies have demonstrated the involvement of Tribolium arrow and Tc-Wnt 8/D in axis formation (Bolognesi et al., 2008b; Bolognesi et al., 2009). Because only strong arrow^RNAi phenotypes have been analysed, the complex role of the Wnt co-receptor has only been partially resolved. How Tribolium generates cellular complexity at the Wnt receptor level is not known.

In this paper, we investigated the role of Wnt receptor genes (Tc-fz1, Tc-fz2 and Tc-fz4) and the co-receptor gene (Tc-arrow) in different processes of Tribolium embryogenesis. We first characterised Wnt-receptive sites in the Tribolium embryo by analysing the expression patterns of the Wnt receptors. We found that, as in Drosophila (Kennerdell and Carthew, 1998), the Wnt receptors Tc-Fz1 and Tc-Fz2 function redundantly. In Tribolium, they are necessary for the maintenance of a functional growth zone. In Tc-fz-1/2^RNAi embryos, axis elongation stops prematurely. In the collapsed growth zone of Tc-fz-1/2^RNAi embryos, posterior cells are still present, but the transition from pair-rule to the segmental function of Even-skipped is impaired. Tc-frizzled 1 plays a Tc-frizzled 2-independent role in leg formation, whereas Tc-frizzled 4 supports Tc-fz1 in leg development and additionally functions in gut development. We identify Tc-Arrow as the obligatory Wnt co-receptor for Wnt-dependent processes, complementing previous studies (Bolognesi et al., 2009). Finally, we discuss the dosage dependency of Wnt receptor-co-receptor interactions in body and leg axis development.

LRP, frizzled and arrow gene homologues in the Tribolium genome (http://beetlebase.org) (Kim et al., 2010) were identified using BLAST with the Drosophila protein sequences. Sequence analysis, sequence alignment and PCR primer design were performed using the programs Edit Seq, MegAlign and Primer Select from the Lasergene DNASTAR package. The phylogenetic relationship of the protein sequences was based on a Clustal-W alignment and analysed with TREE-PUZZLE (Strimmer and von Haeseler, 1996). The trees in Fig. 1 were visualised with TreeView (Page, 1996) and Dendroscope (Huson et al., 2007).

The following amino acid sequences were used for the alignment in Fig. 1 [protein (species) Accession Number (amino acid positions)].

DmFz1 (Drosophila melanogaster) CAA38458 (1>581); AgFz1 (Anopheles gambiae) XP_317942 (1>181); TcFz1 (Tribolium castaneum) NP_001164247/TC014055 (1>549); NasFz7 (Nasonia vitripennis) XP_001605802 (1>580); DrFz7a (Danio rerio) AAH68322 (1>559); DrFz7b (Danio rerio) NP_739569 (1>557); MmFz7 (Mus musculus) AAH49781 (1>750); XenFz7 (Xenopus laevis) AAF63152 (1>549).

DmFz2 (Drosophila melanogaster) AAC47273 (1>694); TcFz2 (Tribolium castaneum) EFA01325/TC003407 (1>605); DrFz5 (Danio rerio) NP_571209 (1>592); DrFz8 (Danio rerio) NP_570993 (1>579); MmFz5 (Mus musculus) NP_073558 (1>585); MmFz8 (Mus musculus) NP_032084 (1>685); AgFz2 (Anopheles gambiae) XP_311505 (1>687).

DmFz3 (Drosophila melanogaster) ABW09320 (1>646); NasFz3 (Nasonia vitripennis) XP_001600300 (1>573); AgFz3 (Anopheles gambiae) XP_310970 (1>487).

TcFz4 (Tribolium castaneum) EFA09255/TC006527 (1>527); DmFz4 (Drosophila melanogaster) NP_511068 (1>705); AgFz4 (Anopheles gambiae) XP_319612 (1>580); MmFz4 (Mus musculus) NP_032081 (1>537).

DrFz10 (Danio rerio) AAH76546 (1>580); MmFz9 (Mus musculus) AAB87503 (1>549); MmFz10 (Mus musculus) NP_780493 (1>582).

TcArr (Tribolium castaneum) XP_975483/ TC008151 (1>1580); DmArr (Drosophila melanogaster) NP_524737 (1>1678); AgArr (Anopheles gambiae) XP_320740 (1>1619); NasArr (Nasonia vitripennis) XP_001603043 (1>1634); DrLRP5 (Danio rerio) XP_696943 (1>1593); DrLRP6 (Danio rerio) NP_001128156 (1>1620); MmLRP5 (Mus musculus) NP_032539 (1>1614); MmLRP6 (Mus musculus) NP_032540 (1>1613).

TcLRP4 (Tribolium castaneum) XP_001814959/ TC007146 (1>2041); DmLRP4 (Drosophila melanogaster) NP_727914 (1>2009); AgLRP4 (Anopheles gambiae) XP_311125 (1>1740); NasLRP4 (Nasonia vitripennis) XP_001605666 (1>2084); DrLRP4 (Danio rerio) XP_697930 (1>1243); MmLRP4 (Mus musculus) EDL27564 (1>1911).

The following gene fragments were cloned and used for in situ hybridisation and dsRNA production: Tc-fz1 (accession number: NM_001170776), an 881-bp fragment corresponding to position 127-1007; Tc-fz2 (XM_963025/867-bp fragment, position 81-947); Tc-fz4 (XM_962510, 534 bp, position 822-1355), Tc-arrow (XM 970390, 694 bp, position 1712-2413). These fragments show no sequence identity of longer than 19 consecutive nucleotides within the predicted coding region of the Tribolium genome that could serve as off-targets in RNA interference. In the case of Tc-fz1, a 21-bp identity exists with a non-coding genomic region. Cross-reactivity of fz1 and fz2 was not expected because the two genes contain only short sequences of identity (12 bp). This prediction was borne out by the non-overlapping results obtained in the respective single RNAi experiments.

cDNA was synthesised from total RNA (RNeasy/Qiagen) using random hexamer primers from the Transcriptor First Strand cDNA Synthesis Kit (Roche). PCR fragments were subcloned into the pCR4 vector of the TOPO-TA Cloning Kit (Invitrogen) and sequenced by AGOWA (Berlin). Double-stranded RNA for RNA interference was produced with the MEGAscript in vitro transcription kit (Ambion) and injected into pupae or adults (parental RNAi) at several different concentrations in multiple independent experiments (see Table S1B1-6 in the supplementary material) under standard conditions (Bucher et al., 2002; Schröder et al., 2008).

The parental application of dsRNA results in the knockdown of both the maternal and zygotic transcripts; the effect is most strongly seen in the first egglays from the treated females and diminishes with successive egglays. In this way, a variety of phenotypes (‘phenotypic series’) were achieved, and weaker phenotypes were explained by a lower dose of dsRNA in the respective embryo. The absence of the Tc-fz2-RNA transcript has been confirmed by in situ hybridisation in Tc-fz1/2^RNAi embryos (see Fig. S3 in the supplementary material).

All other molecular techniques were performed according to the instructions of the suppliers or following standard protocols (Sambrook et al., 1989).

Fixation of embryos and cuticle preparations of first instar larvae were performed as previously described (Patel, 1994; Van der Meer, 1977). In situ hybridisation to RNA transcripts was performed as published (Tautz and Pfeifle, 1989). Embryos were embedded in 100% glycerol and photographed on an Axiophot Zeiss microscope. Immunolocalisation studies of the Even-skipped and Engrailed proteins were performed as previously described (Patel et al., 1994) using the respective monoclonal antibodies at a dilution of 1:100 obtained from the Developmental Studies Hybridoma Bank.

Wild-type beetles of the San Bernardino strain were reared on whole-wheat flour in a 30°C incubator; the collection of eggs and pupae was performed as previously described (Beermann et al., 2004).

Three frizzled genes and one arrow orthologue coding for Wnt receptors and a Wnt co-receptor, respectively, are present in the Tribolium genome (Fig. 1). Our phylogenetic analysis grouped the Tc-Frizzled proteins together with other insect and vertebrate Frizzled proteins into three subfamilies (Fig. 1A). Tc-Frizzled-1, the orthologue of Drosophila Frizzled, clusters with other arthropod Frizzled-1 proteins and vertebrate Frizzled-7 proteins within the Frizzled-1/7 subfamily. The insect Frizzled-2 proteins and the vertebrate Frizzled-5 and Frizzled-8 proteins constitute the Frizzled-2/5/8 subfamily. Frizzled-4 together with dipteran and vertebrate Frizzled-4 orthologues and the vertebrate Frizzled-10 proteins represent the third group. The significant amino acids necessary for binding Dishevelled (Umbhauer et al., 2000; Wu and Mlodzik, 2008) are conserved in the Frizzled C-terminal regions (Fig. 1B), reinforcing the classification of the Tribolium Frizzled proteins.

Within the Tribolium genome, the single orthologue of the Drosophila Wnt co-receptor Arrow is encoded by Tc-arrow (Tc008151). Another gene coding for a protein with the low-density-lipid (LDL) receptor related protein (LRP) motif was identified as the Tribolium LRP4 protein (Tc007146; Fig. 1C). However, only Tc-Arrow contains the multiple phosphorylation sites for Axin and Gsk3β in its cytoplasmic part, suggesting that Tc-Arrow has a similar functional role as Drosophila Arrow in Wnt signalling (see Fig. S4 in the supplementary material).

Tc-fz-2 expression started at the early blastoderm stage at which point it covered the prospective head region for 53-82% of the egg length (100%=anterior pole). Dorsally, the head-specific domain was narrower than at the ventral side (Fig. 2A,A′, bars). When the germ rudiment became visible, a thin line of Tc-fz2-expressing cells marked the dorsal border between the extra-embryonic tissue and the embryonic anlage (see Fig. S1 in the supplementary material). By the onset of gastrulation, Tc-fz2 was strongly expressed around the procephalic lobes but excluded from the posterior egg pole (Fig. 2B,B′). Once the embryo started to segment, Tc-fz2 transcripts were strongly expressed in the prospective labrum and in the antennal and mandibular segments (Fig. 2C). Posterior to the mandibular stripe and anterior to the posterior growth zone, a broad Tc-fz2 expression domain that faded out towards the posterior covered the presegmental region (PSR). This central Tc-fz2 domain did not overlap with Tc-wingless (Tc-wg) expression at the posterior end of the embryo (Fig. 2C). During AE, the emerging stripe pattern became evident (Fig. 2C-E). Tc-fz2 expression persisted in one (Fig. 2C,D), and later two, broad stripes (Fig. 2E,F) in the PSR until the embryo was fully elongated (Fig. 2F,G). The segmental Tc-fz2 stripes were out of register with the Tc-wg stripes (Fig. 2F). In the fully extended germband, Tc-fz2 was expressed in the posterior part of each body segment, mapped by the relative position of Tc-wg (Fig. 2F). In the head, Tc-fz2 expression was detected in the clypeolabral region, a region surrounding the stomodeum, and in the antennal and intercalary segments (Fig. 2G). By the time of germband shortening, the segmental pattern of Tc-fz2 expression declined in the epidermis and increased in the visceral mesoderm and the gut anlagen (Fig. 2H,I). In the hindgut, Tc-wg was co-expressed with Tc-fz2, indicating a possible interaction (Fig. 2I). No Tc-fz2 expression was observed in the larval legs, the growth zone proper or the most posterior cells (Fig. 2C-E,I).

Tc-fz4 was expressed in a dorso-proximal position in elongating legs in the coxa-trochanter region (Fig. 2J). As the legs continued to grow, this expression domain intensified, spread laterally and extended more distally without reaching the distal tip (Fig. 2K-M). When the legs had reached their full length, Tc-fz4 expression was circumferential in the proximal leg region, encompassing most of the coxal region, the future trochanter and part of the tibiotarsus. Within this expression domain, Tc-fz4 expression was modulated: transcripts were more abundant at the sites where the joints of the trochanter will form (Fig. 2N). The head appendages also expressed Tc-fz4 in dorso-proximal positions (Fig. 2K-M). In all appendages, the distal tip lacked Tc-fz4 expression throughout embryogenesis. By contrast, Tc-fz1 and Tc-arrow were ubiquitously expressed throughout embryogenesis (Fig. 2O,P).

To elucidate the functions of the different receptors, we depleted Tc-fz1 and Tc-fz2 by parental RNAi. Single Tc-fz2 RNAi knockdown resulted in viable, wild-type progeny. However, double knockdown of Tc-fz1 and Tc-fz2 resulted in axis elongation defects ranging from mild to severe (Fig. 3). In weakly affected Tc-fz1/2^RNAi individuals, larval cuticles developed only one to two abdominal segments and a remnant of the hindgut. Abdominal segments three to eight and the posterior urogomphi were missing (Fig. 3B,C). The head and the thorax with their associated appendages developed normally. More strongly affected Tc-fz1/2^RNAi individuals displayed defects that affected the thorax and appendages. Nevertheless, the pregnathal region remained only weakly affected (Fig. 3C-E). The most strongly affected cuticles, which resulted from injections of a high dsRNA concentration, were of a spherical shape (Fig. 3D,E), with only a relic of head and thorax.

To understand how the Tc-fz1/Tc-fz2^RNAi phenotype manifests during embryogenesis, we analysed the expression of marker genes prior to cuticle formation (Fig. 4). The primary pair-rule genes even-skipped and odd-skipped (Choe et al., 2006) are expressed in similar patterns in the posterior growth zone and presegmental region as well as segmental stripes in young wild-type embryos (Fig. 4A,F) (Brown et al., 1997; Choe et al., 2006; Patel et al., 1994).

The phenotypic series of the Tc-fz1/2^RNAi knockdown showed an increasing reduction of the growth zone tissue (Fig. 4B-E,G-J). Expression of Tc-Eve protein was ultimately reduced to a few cells at the posterior end, but was not completely abolished (Fig. 4E) as has been described for Tc-torso^RNAi embryos (Schoppmeier and Schröder, 2005) and Tc-arrow^RNAi embryos (Bolognesi et al., 2009). Expression of odd-skipped was reduced to one subterminal stripe (Fig. 4G,H) or was dramatically reduced in intensity but not completely eliminated (Fig. 4J). The segmental marker Engrailed (Fig. 4K-O) was expressed normally in the remaining head and thoracic segments of older embryos (Fig. 4L-O). The expression of Tc-wingless at the segmental borders of Tc-fz1/2^RNAi embryos was weak or absent, but segmental Tc-wg spots indicated the integrity of the remaining individual segments in older embryos (see Fig. S2 in the supplementary material). This finding has also been observed in Tc-arrow^RNAi embryos (Bolognesi et al., 2009).

In wild-type embryos (Fig. 4P), Tc-wnt5 was prominently expressed in the posterior growth zone, in segmental stripes during germband elongation and at the tips of appendages. In the head, the transcripts were detected in wedge-shaped domains in the head lobes and in the antennal segment (Fig. 4P). No expression of the ligand Tc-wnt5 was seen in Tc-fz1/2^RNAi embryos, except for the head-specific domain (Fig. 4Q,R).

Residual expression of the Wnt ligand Tc-wnt8 in Tc-fz1/2^RNAi embryos displaying a strong phenotype (see Fig. S2A-C in the supplementary material) confirmed the presence of the posterior-most cells of the growth zone (see Fig. S2B,C in the supplementary material). In wild-type embryos (see Fig. S2A in the supplementary material), Tc-wnt8 expression is confined to a tiny horseshoe-shaped domain at the very posterior during a short time window extending from early blastoderm to the young embryo with few segments (Bolognesi et al., 2008a).

To test the connection of the Wnt pathway to other signalling pathways possibly involved in setting up segmentation and maintaining the growth zone in Tribolium, we evaluated the expression of Tc-fgf8 in Tc-fz1/2-depleted embryos. In vertebrates, Fgf signalling is involved in somitogenesis of the embryo in a graded fashion (Delfini et al., 2005), and in Drosophila it influences the establishment of the germ layers, mesoderm development and cell migration (Klingseisen et al., 2009; McMahon et al., 2010). Although the embryonic FGF signalling sites in Tribolium have been characterised, the functional requirement for individual members of the FGF pathway remains to be shown (Beermann and Schröder, 2008). In young wild-type germ anlage Tc-fgf8 is expressed in two small stripes bordering the prospective mesoderm in the growth zone. As the segments formed, transcripts were detected in small segmental blocks on both sides of the midline, potentially the segmental mesoderm, and anteriorly between the head lobes (Fig. 4S). In Tc-fz1/2-depleted embryos, Tc-fgf8 expression was completely lost in the posterior region but persisted between the head lobes (Fig. 4T). Thus, posterior and segmental Tc-fgf8 expression depends on Tc-fz1/2-mediated Wnt signalling. This finding provides a connection between the two signalling pathways, placing Wnt signalling upstream of Fgf signalling in segmentation. The existence of posterior cells in Tc-fz1/2^RNAi embryos during AE was demonstrated by the expression of the posterior marker gene caudal (Fig. 4V-X), whereas expression of the hindgut marker gene brachyenteron (byn) was seen only in less strongly affected Tc-fz1/2^RNAi embryos (Fig. 4AA,BB).

Knockdown of the Wnt co-receptor Tc-arrow by RNA interference resulted in a spectrum of leg phenotypes ranging from mild to severe (Fig. 5). In addition, embryos with reduced function of Tc-arrow developed consecutively fewer abdominal segments (Fig. 5A-H). Embryos with the mildest phenotype developed abdominal segments A1-5 with flattened urogomphi (Fig. 5A, arrowhead) and exhibited twisted legs. Embryos with a slightly stronger phenotype (Fig. 5B) had only four abdominal segments, and the leg segments distal to the coxa were severely shortened, twisted and fused (stars in Fig. 5B). In intermediate and strong Tc-arrow^RNAi phenotypes, the number of abdominal segments and the distal part of the legs were severely reduced or even absent. In the legs, only the proximal coxa (Fig. 5C,D) remained intact (Fig. 5E). The strongest phenotypes were characterised by a complete loss of segmentation, appendages and other cuticular markers (Fig. 5F-H) (Bolognesi et al., 2009).

RNAi knockdown of Tc-Fz1 function interfered with the formation of the distal leg (Fig. 6A-E). In wild-type insects, the larval leg is stereotypically organised (Beermann et al., 2004) such that the coxa belongs to the body wall proper (Cohen et al., 1993) and is followed distally by the trochanter, the tibiotarsus and the pretarsal claw (Fig. 6A). In weakly affected Tc-fz1^RNAi larvae, the leg segments distal to the coxa appeared fused and twisted, giving them a bubbly appearance (Fig. 6D,E), whereas in stronger phenotypes, the segments distal to the coxa were fused into a club-like structure of uncertain composition (Fig. 6C). The most distal structures of the appendages, the pretarsal claw and setae of the antennae, were missing entirely (Fig. 6B,C-E). The coxa stayed intact, as shown by the presence of the coxa-specific bristles cx-1 and cx-2 (Grossmann et al., 2009). Thus, Tc-fz1-dependent Wnt signalling is crucial for appendage elongation, as was previously reported for wingless in Drosophila (Kubota et al., 2003) and Tribolium (Grossmann et al., 2009).

To gain insight into the mechanistic cause of the leg phenotype, we analysed the embryonic Tc-fz1^RNAi phenotype using marker genes for the ventral (Tc-wingless; Fig. 6F-L) and the distal leg (Tc-LIM1 and Tc-dachsous; Fig. 6M-U). In wild-type larvae, Tc-wingless was expressed in a ventral stripe in elongating legs and gnathal appendages (Fig. 6F,K). In Tc-fz1^RNAi embryos, Tc-wg expression in the distal parts of the appendages was abolished except for a ventral stripe in the coxa (Fig. 6G-J,L). Abdominal segmentation in Tc-fz1^RNAi embryos was normal (Fig. 6G).

The LIM homeobox gene Tc-Lim1 showed three distinct expression domains. The most proximal ring of expression (p in Fig. 6M,N) corresponds to the coxa, the middle ring (m in Fig. 6M,N) to the femur and the most distal ring (d in Fig. 6M,N) to the distal tibiotarsus and the pretarsal claw (Pueyo et al., 2000; Tsuji et al., 2000). Tc-Lim1 expression domains serve as markers for proximal, medial and distal positions in the embryonic leg anlage. Tc-dachsous marks the distal leg during elongation (Fig. 6Q) and is expressed in the coxa at older stages (Fig. 6R).

In Tc-fz1^RNAi embryos, both marker genes document the absence of the anlagen for the tibiotarsus and the pretarsal claw. In weaker leg phenotypes, the proximal and medial rings of Tc-Lim1 expression were still present (Fig. 6O); however, only the most proximal ring remained intact in embryos displaying a stronger leg phenotype (Fig. 6P). Here, the distal ring was absent, whereas the middle expression domain was broadened, resulting in a fused and augmented femur (Fig. 6P). In all cases, the distal ends of the remaining legs were bent (Fig. 6O,P,S-U) and malformed.

In Tc-fz1^RNAi legs, the distal expression domain of Tc-dachsous was strongly reduced (Fig. 6S) or absent (Fig. 6T,U), consistent with the Tc-fz1^RNAi cuticle leg phenotypes (Fig. 6B-E). In summary, only the coxa, i.e. the proximal part of the leg, was intact in Tc-fz1^RNAi embryos and thus confirms the function of Tc-fz1 in distal leg formation.

RNAi knockdown of both Tc-fz1 and Tc-fz4 resulted in a stronger phenotype than those seen in the Tc-fz1 single RNAi experiments. Here, in addition to weaker phenotypes that were indistinguishable from those induced by Tc-fz1-RNAi (Fig. 7A), more strongly affected legs presenting only the proximal coxa were observed (Fig. 7B). Moreover, Tc-fz1/4^RNAi larvae developed with a malformed hindgut (Fig. 7G). Interestingly, Tc-fz1/4 double knockdown led to a higher percentage of affected embryos (see Table S1 in the supplementary material) than the respective single RNAi knockdowns. In the single Tc-fz4 RNAi knockdown, no specific phenotypes above background levels were seen (see Table S1 in the supplementary material).

We have analysed the expression patterns and functions of the genes coding for the Wnt receptors (frizzled 1, 2 and 4) and the Wnt co-receptor Tc-arrow during embryogenesis of the short germ beetle Tribolium. We reveal a redundant and combinatorial code of Wnt receptors and a co-receptor that regulates segmentation, axis elongation, leg formation and gut development in Tribolium.

Whereas mammals possess ten unique Frizzled receptors, only four are present in the genomes of Drosophila, Caenorhabditis elegans and the cnidarian Nematostella (van Amerongen and Nusse, 2009). Tribolium has three frizzled orthologues, with no orthologue for Drosophila fz3. Thus, the number of Tribolium frizzled genes is not dramatically different from that of other insects or Nematostella, indicating that frizzled1 and 2 as well as the structurally similar frizzled 3 and frizzled4 genes constituted a basic Wnt receptor set in the last common ancestor of Nematostella and insects.

The Tribolium genome contains two similar genes that code for LRP orthologues. We identified Tc008151 as the single Arrow/LRP5-6 Wnt co-receptor and performed the aforementioned experiments with this gene. The second LDL receptor, TC007146, represents the Tribolium LRP4 orthologue – a putative Wnt antagonist (Fig. 1C). Prior to our analysis, the Tribolium arrow orthologue had been mistakenly assigned to the accession number Xm001814907 (Bolognesi et al., 2009), which actually represents the LRP4 orthologue (Fig. 1). As our results on Tc-arrow function overlap with those of Bolognesi et al., we presume that their published results do indeed describe the function of the Tribolium Arrow orthologue. Interfering with LRP4 function in Tribolium results in a different phenotype than that shown for Tc-arrow^RNAi (R.P. and R.S., unpublished).

In Tribolium, Tc-fz2 and Tc-fz4 show distinct expression patterns, whereas Tc-fz1 and Tc-arrow are ubiquitously expressed. Thus, Tc-fz1-mediated Wnt signalling could potentially occur throughout the entire embryo at all stages. Several tissues are marked by the expression of Tc-fz2 or Tc-fz4, indicating specific Frizzled-dependent Wnt signalling sites. fz2 expression sites are similar in Tribolium and Drosophila in late developmental stages (Bhanot et al., 1996). Exceptional for Tribolium is the strong expression in the head anlage observed at the blastoderm stage. However, no head defect was observed in Tc-fz1/2^RNAi embryos. This finding might be explained by a functional redundancy between Fz1, Fz2 and Arrow. We were able to rule out a combined involvement of Tc-fz1 or Tc-fz2 with Arrow in head development because the double RNAi experiments Tc-fz1/arrow and Tc-fz2/arrow did not result in head phenotypes (data not shown).

The Tc-fz2 expression pattern and the Tc-fz1/2 double RNAi phenotype identify the region between the posterior growth zone and the new segments as crucial for axial elongation (AE). This tissue develops in a similar position to the presomitic mesoderm (PSM) in vertebrates and is called the presegmental region (PSR) in Tribolium (Schröder et al., 2008). In both Drosophila and Tribolium, Fz1 and Fz2 function redundantly: in Drosophila during segmentation and in Tribolium within the PSR, the axis elongation phenotype was only seen when Tc-Fz1 and Tc-Fz2 were depleted simultaneously. As Wnt signalling drives segmentation from within the PSR, reception of the Wnt signal in this tissue by the Tc-Fz1 and Tc-Fz2 receptors is necessary for this process.

Our marker gene analysis showed that the posterior growth zone is not completely abolished in Tc-fz1/2^RNAi embryos, as was previously shown for Tc-arrow^RNAi embryos (Bolognesi et al., 2009). The markers Eve, odd and wnt8 (Fig. 4, see Fig. S2 in the supplementary material) are still expressed in posterior cells of the growth zone. The persistent expression of the marker gene caudal (Fig. 4V-X) proves the identity of these cells as posterior cells. Because the complete wild-type expression pattern of caudal is not retained in Tc-fz1/2^RNAi embryos, the terminal-most tissues could still contribute to axis elongation. The absence of the hindgut marker brachyenteron (Fig. 4Z,AA) in strongly affected Tc-fz1/2^RNAi embryos is in accordance with the loss of the hindgut in fz1/2^RNAi larval cuticles (Fig. 3B-F). In less strongly affected Tc-fz1/2^RNAi embryos, byn expression is still detectable (Fig. 4AA,BB). Our results show that the transition of the pair-rule genes from the primary to the segmental phase (Choe et al., 2006; Patel et al., 1994) depends on Wnt signalling. As a consequence of this loss of Wnt signalling, AE does not occur. In contrast to Tc-wnt8, Tc-wnt5 expression is completely abolished in strong Tc-fz1/2^RNAi embryos. This finding could be explained by an autoregulatory loop in the PSR involving Frizzled and Wnt5.

In vertebrates, coordinated cell movements in the PSM under the control of FGF signalling are responsible for AE (Bénazéraf et al., 2010). Because signalling pathways are connected to other signalling pathways (van Amerongen and Nusse, 2009), we hypothesised that FGF signalling might be involved in AE in Tribolium as well. Indeed, the prominent wild-type FGF expression domain is missing in Tc-fz1/2^RNAi embryos (Fig. 4T), indicating that FGF signalling depends on Wnt signalling in Tribolium and might itself be crucially required for AE. It is striking that in both vertebrates and insects, a tissue immediately posterior to the last somites or segment formed – the PSM and the PSR, respectively – fulfils analogous functions in the axial elongation process.

In Drosophila and Tribolium, wingless is required for segmentation as well as for distalisation and dorso-ventral patterning of the appendages (Cohen and Jürgens, 1989; Grossmann et al., 2009; Ober and Jockusch, 2006). For Tribolium, it was argued that the leg patterning function of wingless is only required during later embryonic stages (Grossmann et al., 2009).

We show that Tc-fz1 function is required for both proximo-distal (PD) and dorso-ventral (DV) axis formation in the leg. Because both the PD and the DV leg phenotypes are the result of parental RNAi, we could not verify a distinct time-dependency for Tc-fz1 function. Rather, our data support concentration-dependent regulation of the PD and DV axes during leg development. The Tc-fz1^RNAi leg phenotype strongly resembles the wingless^RNAi phenotypes described for Tribolium (Grossmann et al., 2009) and Drosophila (Diaz-Benjumea and Cohen, 1993; Diaz-Benjumea and Cohen, 1994). Based on bristle mapping, we propose that the best candidate ligand for Tc-Fz1 in appendage formation is Tc-Wnt1.

Remarkably, Tc-fz1/2^RNAi larvae lacking most of the abdominal segments develop normal legs (Fig. 3B). Here, the reduced Tc-fz1 function seems to be partially compensated. This result is in contrast to Tc-arrow^RNAi, in which leg formation is severely affected in larvae with a mild AE defect (Fig. 5B). Why does the knockdown of Tc-fz1 in the Tc-fz1/2 double RNAi experiment only interfere mildly with leg formation? AE is apparently very sensitive to depletion of Tc-fz1/2 function. The dose of dsRNA required for an AE phenotype hardly affects Tc-Fz1-dependent leg patterning at all. Presumably, in the double knockdown embryos, the concentration of Tc-fz1 transcripts does not sink below a critical threshold. On the other hand, AE and leg patterning require similar amounts of the co-receptor Arrow. These results support a dosage dependency of Wnt receptors and co-receptors during embryogenesis.

In larvae with weak Tc-fz1 RNAi phenotypes, lesions occur in the proximal leg, affecting the joints of the trochanter – sites of Tc-fz4 expression (Fig. 2N) – and leading to a fusion of coxa and trochanter (Fig. 6C,C′). Tc-Fz4, therefore, supports Tc-Fz1 in transducing the Wnt signal in the proximal leg (Fig. 8B). This assumption is corroborated by the finding that Tc-fz1/4 double RNAi enhances the frequency and the phenotype of this leg phenotype (see Table S1B in the supplementary material and Fig. 7G). However, Tc-Fz4 on its own is dispensable in the leg; Tc-Fz1 appears to be the crucial partner. At present, the precise function of Tc-Fz1/Fz4 in the dorso-proximal leg is speculative. The hypothesised leg differentiation centre at the coxa/trochanter boundary, postulated for Tenebrio (Huét and Lenoir-Rousseaux, 1976), is an intriguing possibility. Larvae with the strongest Tc-fz1^RNAi phenotype display loss of distal structures, including the pretarsal claw (Fig. 6C). Because Tc-fz4 is not expressed distal to the tibiotarsal region, Tc-Fz1 seems to be the single primary Wnt transducer in the distal appendage (Fig. 8B). The ligand for this signal is probably distal Wnt1, based on the leg phenotype of wg^RNAi larvae (Grossmann et al., 2009). Depending on their conformation, Frizzled receptor proteins have been proposed to signal via different downstream signalling pathways and could thus serve as multifunctional signal transducers depending on, e.g. ligand availability (Carron et al., 2003). In this way, a variety of different biological outcomes can be controlled by a small number of Wnt receptors.

Double Tc-fz1/4^RNAi embryos reveal an additional function of the two receptors in hindgut formation. Although a normal hindgut develops in the respective single RNAi experiments, the elongation of the gut tubule is impaired in Tc-fz1/4^RNAi larvae; only remnants of the hindgut-specific cuticle develop (Fig. 7G). Tc-fz1/4^RNAi larvae survive to stage L2 (Fig. 7C) but subsequently die, probably due to a non-functional hindgut and stunted mouth appendages. A similar gut phenotype has been observed in brachyenteron^RNAi experiments (Berns et al., 2008) indicating that Tc-byn and Tc-Fz4 might function within the same pathway. Indeed, byn has been shown to be a target of the Wnt signalling cascade in vertebrates (Arnold et al., 2000; Yamaguchi et al., 1999). Alternatively, elongation of the gut might be achieved via convergent extension under the combined control of Tc-frizzled 1 and Tc-frizzled 4, pointing to the involvement of the planar cell polarity pathway.

Surprisingly, wingless expression in the segments is strongly reduced in Tc-fz1/4^RNAi embryos (Fig. 7D). The continued expression of wingless during segmentation therefore seems to depend on a feedback loop involving both Frizzled1 and Frizzled4. Because the abdominal segments form normally in Tc-fz1/4^RNAi embryos (Fig. 7E,G), the loss of wingless expression does not interfere with segmentation.

The requirement of Tc-Arrow for leg formation, AE and segmentation identifies this protein as an obligatory co-receptor in all these processes. We explain the spectrum of Tc-arrow^RNAi phenotypes, ranging from weak to strong (Fig. 5), by the declining amounts of ds-Tc-arrow RNA transferred by the injected mothers to their offspring during successive egglays. This ‘phenotypic series’ might point to a dosage dependency at the level of Wnt reception. In such a scenario, leg formation and AE would be most sensitive to the loss of Tc-Arrow function, whereas segmentation is more robust and only requires low concentrations of Arrow. This hypothesis is supported by the higher percentage of weaker Tc-arrow^RNAi phenotypes in RNAi experiments with concentrations of 10 ng/μl compared with the injection series with 50 ng/μl or higher (see Table S1B6 in the supplementary material).

Interestingly, specific defects in the larva were obtained only when the ubiquitously expressed genes Tc-fz1 and Tc-arrow were knocked down, i.e., not when the genes showing distinct expression patterns (Tc-fz2 or Tc-fz4) were depleted. The acquisition of new expression domains during evolution can promote the formation of embryonic parts with novel structural and functional aspects. We speculate that such a scenario could apply to the Tc-fz2 and Tc-fz4 genes.

Previous studies analysed the function of wingless (Grossmann et al., 2009) and wnt8 (Bolognesi et al., 2008b). As no functional data for other Wnt molecules are available, the identities of the Wnt ligands that activate the Frizzled receptors during these embryological processes remain unclear. Our speculations are based on the expression patterns of the different Wnt ligands (Bolognesi et al., 2008a) (see Fig. S5 in the supplementary material) and are summarised in Fig. 8.

Further diversification of signalling outcomes could be achieved by the formation of homo- or heterodimers of Frizzled receptors via their cysteine-rich-domains. Such dimerisation events have been shown to occur in vitro for various vertebrate Frizzled proteins and other G-protein-coupled receptors (Angers et al., 2002; Carron et al., 2003; Dann et al., 2001; Milligan, 2004; Schulte and Bryja, 2007; Stiegler et al., 2009). Whether the Frizzled proteins function as independent monomers, homodimers or heterodimers in Tribolium is unknown. Fz1 and Fz2 in the PSR and Fz1 and Fz4 in the proximal leg could potentially form various receptor combinations (displayed in Fig. 8).

In conclusion, we have shown that a variety of tissue-specific outcomes are guided by a combinatorial code of three Wnt receptors and one Wnt co-receptor in Tribolium. We have identified the presegmental region in the Tribolium embryo as the necessary tissue for axis elongation at the Wnt receptor level and uncovered a network of signalling pathways within the PSR that controls this process. Our findings provide parallels to the involvement of FGF and Wnt signalling in vertebrate development (Dequéant and Pourquié, 2008). Both the FGF and Wnt signalling pathways are involved in axial patterning and serve as crucial regulators of animal embryogenesis. Future work will aim to identify the correct receptor-ligand combinations in the presegmental region and the appendages to disclose the complexity of the Wnt signalling pathway in short germ insects.

We gratefully acknowledge Rolf Reuter for continuous and generous support, T. Mader and A. Hlawa for excellent technical assistance, P. Spahn for reading drafts of the manuscript and the German Research Council (DFG) for its long-standing and continuous funding. The Even-skipped (2B8) and the Engrailed (4D9) monoclonal antibodies developed by N. Patel were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242.