The neuroblast timer gene nubbin exhibits functional redundancy with gap genes to regulate segment identity in Tribolium

The gap gene network of Drosophila is arguably one of the best characterised gene regulatory networks in developmental biology. Gap genes mediate two central processes in Drosophila segmentation – the formation of segment boundaries and the assignment of segment identities – through direct regulation of pair-rule and Hox genes, respectively (reviewed by Jaeger, 2011). Homologs of many Drosophila gap genes also regulate segment patterning in other insect species (Bucher and Klingler, 2004; Cerny et al., 2005; Liu and Kaufman, 2004b; Liu and Patel, 2010; Marques-Souza et al., 2008; Mito et al., 2005, 2006). Recent attention has therefore turned to understanding how gap genes interact and function outside of Drosophila, in order to better understand the origins and evolution of this important gene network.

In Drosophila, the gap genes are thought of as markers for spatial domains, regulated initially by gradients of maternal factors, and then by cross-regulation within the gap gene network itself (Jaeger, 2011). However, recent work, particularly in the red flour beetle Tribolium castaneum, leads to a rather different way of viewing these same genes. In Tribolium and other sequentially segmenting insects, segments are added progressively, from anterior to posterior, from a segment addition zone (SAZ) at the posterior of the extending germ band (Clark et al., 2019). Gap genes are sequentially activated in the SAZ, so that cells persisting in this region experience a temporal sequence of gap gene expression (Boos et al., 2018; Zhu et al., 2017) (Fig. 1A). As each cell exits the SAZ, its gap gene expression is stabilised (Zhu et al., 2017), creating a spatial pattern of gap gene expression along the anterior-posterior axis of the trunk. The gap genes may therefore provide a timer for the maturation of cells with different axial identities from the segment addition zone (Bucher and Klingler, 2004; Cerny et al., 2005; Clark et al., 2019).

This model of the gap gene network has many similarities to the neuroblast timer network that regulates embryonic neural patterning in insects (Clark et al., 2019; Doe, 2017). The insect nervous system is produced by neural stem cells known as neuroblasts, each of which gives rise to a range of different cell types in a stereotyped order. In embryonic neuroblasts of Drosophila, this order is directed by the sequential expression of the neuroblast timer genes hunchback (hb), Krüppel (Kr), nubbin (nub), castor (cas) and grainyhead (grh) (reviewed by Brody and Odenwald, 2005) (Fig. 1B). Homologues of hb, nub and cas are expressed in the same relative order in some vertebrate neural stem cells, where they regulate the fate of neurons derived from their progeny (Alsiö et al., 2013; Elliott et al., 2008; Javed et al., 2020; Mattar and Cayouette, 2015; Mattar et al., 2018). This suggests that the roles of these genes in neural development are deeply conserved.

Parallels between the neuroblast timer series and the gap gene network have long been noted (Isshiki et al., 2001; Peel et al., 2005), giving rise to the hypothesis that elements of the neuroblast timer network may have been co-opted from neuroblasts for use in insect axial patterning (Peel et al., 2005). The first two genes in the neuroblast timer series, hb and Kr, are also canonical gap genes in Drosophila and Tribolium (Boos et al., 2018; Cerny et al., 2005; Marques-Souza et al., 2008). However, the next two genes in the neuroblast timer series, nub and cas, are not canonical gap genes in Drosophila (Jaeger, 2011). The canonical gap genes acting posteriorly to Kr in Drosophila – Dm-knirps (kni) and Dm-giant (gt) – and in Tribolium – Tc-gt and Tc-mille-pattes – are not components of the neuroblast timer series (Fig. 1D).

Although nub and cas are not canonical gap genes, they do show some intriguing similarities to gap genes. Dm-nub (and its closely linked paralogue Dm-pdm2) and Dm-cas are expressed in the Drosophila blastoderm during segment patterning, in spatial domains that follow in sequence behind Dm-hb and Dm-Kr (Cockerill et al., 1993; Isshiki et al., 2001) (Fig. 1C). Ectopic expression of Dm-nub or Dm-pdm2 results in gap-like segment deletions (Cockerill et al., 1993); however, neither gene appears to regulate the canonical gap genes (Cockerill et al., 1993), and deletion of both genes generates only incompletely penetrant and variable segment fusions (Cockerill et al., 1993; Ma et al., 1998). Dm-cas is not known to have any role in segmentation (e.g. Mellerick et al., 1992).

Data from sequentially segmenting insects has identified further parallels. A homologue of nub is necessary for the correct specification of abdominal segment identity in the bug Oncopeltus (Hrycaj et al., 2008), although not in the cricket Acheta (Turchyn et al., 2011). In Tribolium, Tc-nub and Tc-cas are also expressed in the segment addition zone (Biffar and Stollewerk, 2014), but parental RNA interference has failed to identify any role for Tc-nub in segmentation (E. Raymond and A. Peel, personal communication); functional analyses have not been carried out for Tc-cas.

In this article, we examine whether Tc-nub and Tc-cas form part of a temporal sequence of gene expression during segmentation in Tribolium, and ask whether either regulates segment addition or the assignment of segment identities. Our functional analyses demonstrate a clear role for Tc-nub in the assignment of abdominal segment identity. This role is partially redundant with that of other abdominal gap genes, explaining why it has not been identified previously. Our findings strengthen the hypothesis that elements of the gap gene network may have been recruited for a timing role in axial patterning from a pre-existing role in neural development.

We first examined whether the genes of the neuroblast timer series are expressed in temporal order in the SAZ of Tribolium during segment addition. We used hybridisation chain reaction (HCR) RNA in situ hybridisation (ISH) (Choi et al., 2018) to examine the expression patterns of Tc-hb, Tc-Kr, Tc-nub and Tc-cas in Tribolium embryos spanning the stages of segment addition [8-22 h after egg lay (AEL) at 30°C]. We found that these four genes are expressed sequentially in the SAZ in largely the same order as they are expressed in neuroblasts, and that this sequential expression results in their being expressed in spatial order along the anterior-to-posterior (AP) axis of the embryonic trunk (Fig. 2). Tc-hb mRNA is initially distributed broadly across the blastoderm (Wolff et al., 1995) (Fig. 2A), becoming lost from the posterior tip of the embryo as Tc-Kr expression emerges (Boos et al., 2018) (Fig. 2B). Tc-nub becomes expressed at the posterior tip of the embryo shortly afterwards, correlating with loss of Tc-Kr expression in the same region (Fig. 2C,D). Tc-cas becomes expressed in the SAZ midway through germband extension, in a domain overlapping the posterior of the Tc-nub domain (Fig. 2E,F). Finally, a second domain of Tc-hb becomes expressed in the posterior SAZ and remains expressed in the SAZ until the end of segment addition (Fig. 2E-H). This re-expression of hb after cas is not observed during neurogenesis in either Drosophila or Tribolium (Biffar and Stollewerk, 2014; Doe, 2017), so is a distinctive feature of the SAZ. Each of these genes is also expressed in the neurectoderm and/or neuroblasts in differentiating segments, as well as in the tissue at the extreme posterior of the embryo – the presumptive hindgut epithelium (Benton, 2018) – after segmentation is complete (Fig. S1).

To characterise the expression dynamics of Tc-nub and Tc-cas in more detail, we next examined the expression of both genes against expression of the segment polarity gene Tc-wingless (Tc-wg) (Nagy and Carroll, 1994) in embryos spanning the course of segment addition. Tc-wg stripes form sequentially in the trunk over the course of segment addition and can therefore be used as a proxy for developmental stage. Each Tc-wg stripe marks the posterior boundary of a parasegment (PS), and has been assigned a number that reflects its relationship to that parasegment (e.g. wg6 sits at the posterior of PS6; the first trunk Tc-wg stripe is designated wg0, as it sits at the posterior of PS0).

Tc-nub is expressed at the late blastoderm stage in two patches overlying the ocular Tc-wg stripes, and is first expressed at the posterior pole shortly afterwards (Fig. 3A). By the time wg2 has formed, the embryo has condensed to form a germband, and the posterior domain of Tc-nub expression has expanded to encompass the posterior one-third of the SAZ (Fig. 3A). The anterior border of this broad gap-like domain abuts wg3 in the ectoderm, but is shifted posteriorly in the mesoderm, abutting wg5 (Fig. S2). Ectodermal expression is weaker anteriorly, and stronger in the posterior SAZ. After the formation of wg6, Tc-nub expression begins to fade in the posterior SAZ, and the posterior boundary of Tc-nub eventually overlaps with the posterior boundary of wg12 (wg12p; Fig. 3A). Tc-nub is therefore expressed in the SAZ during the patterning of PS4-PS12 (posterior compartment of T1 to anterior compartment of A7, inclusive). This overlaps extensively with the expression domains of the gap genes Tc-mlpt and Tc-gt, and the gap-like gene Tc-kni (Fig. 3C).

Tc-cas expression is not detectable in the embryo until after the germband has formed. As wg4 and wg5 are forming, Tc-cas becomes expressed weakly first in the primordium of the labrum and then in the SAZ (Fig. 3B). Expression in the SAZ is excluded from the mesoderm (Fig. S2). The anterior border of this domain abuts the posterior boundary of wg6 (Fig. 3B). By the time wg8 has formed, expression of Tc-cas in the SAZ becomes modulated in a pair-rule pattern; the strongest domains of expression appear to overlap the primordia for PS9 and PS11 immediately after the formation of wg10 (Fig. 3B). Expression of Tc-cas subsequently fades in the posterior SAZ and its posterior boundary eventually overlaps with the posterior boundary of wg12 (wg12p; Fig. 3B). This means that Tc-cas is expressed in the SAZ during the patterning of PS7-12 (posterior compartment of A3 to anterior compartment of A7, inclusive), overlapping Tc-nub and Tc-mlpt (Fig. 3C). In maturing segments outside of the SAZ, Tc-cas expression fades and is lost (Fig. 3B). There is an additional domain of Tc-cas that forms after the emergence of wg10, overlapping the posterior terminal domain of Tc-wg (Fig. 3B).

Tc-nub later becomes broadly expressed in the ectoderm outside of the SAZ, with slightly stronger expression in the developing neurectoderm (Biffar and Stollewerk, 2014). Both Tc-nub and Tc-cas are expressed in neuroblasts (Biffar and Stollewerk, 2014) and the limb buds (Fig. 3A,B).

We next aimed to determine whether Tc-nub or Tc-cas have a role in axial patterning in Tribolium. To do this, we knocked down the expression of each gene by parental and embryonic RNA interference (pRNAi and eRNAi, respectively).

We found that pRNAi and eRNAi against Tc-nub (2 μg/μl dsRNA) resulted in a subtle abdominal segment transformation in a small percentage of the embryos that survived to the point of cuticle formation. Specifically, 2.9% (pRNAi) and 12.1% (eRNAi) of cuticles displayed a ‘nub’ (an ectopic, ventrolateral protrusion of cuticle, lacking joints or claws) on either side of the first abdominal segment, A1 (Fig. 4B, Tables S1 and S2). This phenotype was never observed in GFP pRNAi or eRNAi controls (Tables S1 and S2; Fisher's exact tests: P=0.02798 and P=5.853×10⁻⁶, respectively). Similar nubs form following pRNAi against Tc-abdominal-A (Tc-abd-A, also known as Tc-Abdominal; Stuart et al., 1993), and have been interpreted as homeotic transformations of the posterior compartment of an abdominal segment to the posterior compartment of the third thoracic segment, T3 (Lewis et al., 2000). This would make each nub developmentally akin to the posterior compartment of a thoracic leg. We examined the expression of Tc-abd-A after Tc-nub pRNAi, and found that most of the embryos examined (7/9) showed a downregulation of Tc-abd-A expression in the anterior of PS7, which gives rise to A1p (Fig. S3).

Neither pRNAi nor eRNAi against Tc-cas had any consistent effects on cuticular morphology or on segment patterning in embryos (n=116 and 89, respectively; Fig. 4C and Tables S1 and S2).

Intriguingly, we found that pRNAi, but not eRNAi, against either Tc-nub or Tc-cas significantly reduced the proportion of eggs that developed to the stage of cuticle formation compared with GFP injection controls (Fig. S4, Tables S1 and S2). Only ∼40-45% of eggs developed cuticle after 1 μg/μl Tc-nub or Tc-cas pRNAi, compared with 83% of eggs in GFP controls (Fisher's exact test: P<2.2×10⁻⁶). Tc-nub and Tc-cas are both expressed in ovarioles of adult female Tribolium (Fig. S5) and may therefore have roles in oogenesis or early embryogenesis. Dm-cas is known to be required for the proper formation of follicular cells in Drosophila (Chang et al., 2013) but Dm-nub does not seem to be expressed in ovaries (Celniker et al., 2009).

We also found that those embryos that do develop to the stage of cuticle formation after pRNAi against Tc-nub or Tc-cas are significantly less likely to hatch than GFP controls, despite their relatively normal external morphology (Fig. S4, Table S1). Specifically, 8-12% of embryos that develop cuticle after 1 μg/μl Tc-nub or Tc-cas pRNAi go on to hatch, compared with 94% in GFP controls (Fisher's exact test: P<2.2×10⁻⁶). The failure of otherwise ‘normal’ larvae to hatch could be a result of defects in the nervous system. Both Tc-nub and Tc-cas are expressed in neuroblasts in Tribolium (Biffar and Stollewerk, 2014), and Dm-cas mutants with otherwise normal cuticles also fail to hatch, presumably because of disruption to the nervous system (Mellerick et al., 1992). Together, our data show that Tc-nub and Tc-cas are likely involved in oogenesis and neurogenesis, and that Tc-nub affects specification of segment identity.

The spatially restricted and weakly penetrant homeotic phenotype observed after Tc-nub RNAi contrasts with the expression of this gene across the majority of the abdomen. We hypothesised that the function of Tc-nub might be obscured in RNAi experiments by redundancy with co-expressed genes. Two promising candidate genes for redundant function are Tc-giant (Tc-gt) and Tc-knirps (Tc-kni), both of which are transiently co-expressed with Tc-nub in the SAZ (Fig. 3C, Fig. S6). Tc-gt is considered a gap gene in Tribolium, as its knockdown affects thoracic segment identity and abdominal segment formation (Bucher and Klingler, 2004). In contrast, Tc-kni is not considered to be a gap gene, as its knockdown results in the deletion of only one segment boundary in the head, with no effects on segment identity (Cerny et al., 2008; Peel et al., 2013).

To determine whether Tc-nub acts redundantly with Tc-gt and/or Tc-kni to regulate abdominal segment patterning, we performed single-, double- and triple-knockdowns of these genes. We used eRNAi to avoid any possible negative effects of parental Tc-nub knockdown on oogenesis. Single knockdowns of Tc-kni and Tc-gt produced phenotypes largely consistent with previous reports (Bucher and Klingler, 2004; Cerny et al., 2008; Peel et al., 2013) (Fig. 4D-E′). The notable exception was that 11% of cuticles formed after Tc-gt eRNAi also displayed disrupted leg formation on segment T3 and ectopic legs, similarly disrupted, on segment A1 (Fig. 4E′, Table S2). This difference may be due to eRNAi causing stronger knockdown phenotypes than pRNAi, as has been observed previously (Benton et al., 2019).

While knockdown of Tc-nub or Tc-gt alone resulted in only a low frequency of homeotic transformations restricted to A1, and knockdown of Tc-kni had no effect on abdominal segment identity, we found that combinatorial knockdown of two or more of these genes generated a higher frequency of abdominal transformations than would be expected additively, often of greater severity than those observed in single knockdowns (Fig. 4; Table S2).

Knocking down all three genes together produced the most severe phenotypes. Ninety-four percent of cuticles developing from embryos injected with all three dsRNAs formed jointed clawed legs on at least one abdominal segment (Fig. 4I,J, Table S2). These cuticles had an average of four extra pairs of partial or complete legs (not including the maxillary and labial legs induced by Tc-gt knockdown), and a maximum of seven extra pairs (Table S2), indicating homeotic transformation of up to seven abdominal segments.

In addition to homeotic transformations, Tc-gt knockdowns result in truncations of the posterior abdomen with a very high penetrance (Fig. S7, Table S3; Bucher and Klingler, 2004). Knocking down Tc-kni and/or Tc-nub in addition to Tc-gt did not increase the penetrance or severity of these embryonic truncations (Fig. S7). Moreover, the frequency of truncations observed after eRNAi against Tc-nub + Tc-kni did not differ significantly from GFP controls (Fig. S7). These data suggest that the truncations observed after knockdown of Tc-nub + Tc-gt or knockdown of Tc-nub + Tc-gt + Tc-kni result primarily from loss of Tc-gt, and that neither Tc-nub nor Tc-kni plays any substantial role in segment addition. One caveat to this conclusion is that a higher proportion of triple knockdown embryos died before forming cuticle, when compared with double knockdowns (Table S2), and we observed that many triple knockdown embryos displayed severely disrupted patterning of Tc-wg stripes (e.g. see Fig. 6). Therefore, it may be that functional reduction/removal of all three genes has severe effects on the process of segment addition, or on other aspects of embryonic growth, that are masked by embryonic death.

Tc-gt, Tc-nub and Tc-kni are also co-expressed during head patterning. However, knocking down two or all three of these genes in parallel did not increase the penetrance or severity of head phenotypes – rather, knockdown effects were additive (Fig. S8), as might be expected if all three genes act independently.

Development of partial or complete legs on abdominal segments has also been observed in double knockdowns of two abdominal Hox genes, Tc-abd-A and Tc-Ultrabithorax (Tc-Ubx, also known as Tc-Ultrathorax; Bennett et al., 1999; Lewis et al., 2000). To determine whether these Hox genes are misexpressed after eRNAi against Tc-nub, Tc-kni and Tc-gt, we performed HCR in situ hybridisation in embryos midway through segment addition (16-17 h AEL). This time point is shortly after the period during which Tc-nub, Tc-gt and Tc-kni are co-expressed, and should, therefore, reveal the immediate effects of knockdown on Hox gene expression.

In wild-type embryos, expression of Tc-Ubx and Tc-abdA is detectable in the SAZ after the formation of the wg2 and wg4 stripes, respectively (Bennett et al., 1999; Shippy et al., 1998). Accordingly, we observed strong expression of both genes in the SAZ of control embryos (injected with GFP dsRNA) immediately after the formation of wg6 (Fig. 5A). In contrast, similarly staged embryos injected with Tc-nub, Tc-kni and Tc-gt dsRNA did not express Tc-Ubx or Tc-abdA (Fig. 5B). This loss of Hox gene expression is consistent with the dramatic abdominal phenotypes observed after triple eRNAi. The antennal and mandibular Tc-Wg stripes (wg0 and wg1) were deleted or highly disorganised in triple knockdowns, consistent with the head cuticle phenotypes observed in Fig. S7.

We hypothesised that the repression of abdominal Hox genes observed after triple eRNAi might result from misregulation and expansion of other gap genes. The anterior borders of Dm-Ubx and Dm-abd-A expression in Drosophila are set primarily via direct repression by Dm-hb and Dm-Kr, respectively (Casares and Sánchez-Herrero, 1995; White and Lehmann, 1986). Therefore, we used HCR in situ hybridisation to examine the expression of both Tc-hb and Tc-Kr in embryos fixed at 16-17 h AEL following eRNAi against Tc-nub, Tc-gt and Tc-kni.

We observed alterations in the pattern of Tc-Kr, but not Tc-hb, expression in embryos after simultaneous knockdown of Tc-nub, Tc-gt and Tc-kni (Fig. 6; Fig. S9). In wild-type embryos, Tc-Kr is expressed throughout the SAZ at the blastoderm stage, but becomes cleared from the posterior half of the SAZ during early germband formation (Cerny et al., 2008). This means that the SAZ is largely cleared of Tc-Kr expression by the time that the second trunk Tc-wg stripe (wg1) is formed (Tidswell, 2020). In contrast, triple-knockdown embryos with as many as four Tc-wg stripes showed little or no clearing of Tc-Kr expression in the SAZ (Fig. 6B,C). This means that after triple eRNAi, Tc-Kr, but not Tc-hb, is ectopically expressed in the SAZ.

Together, these data suggest that Tc-nub, Tc-gt and Tc-kni redundantly repress Tc-Kr expression, and that, in their absence, Tc-Kr expression expands into the abdominal primordia. We propose that this expansion leads to the repression of abdominal Hox genes, and subsequently to abdominal segment transformations.

In addition to double and triple knockdowns of Tc-nub with Tc-gt and/or Tc-kni, we also performed simultaneous knockdown of Tc-nub and Tc-cas to determine whether they might play a redundant role in the posterior abdomen. Double Tc-nub + Tc-cas knockdowns do not display any posterior abdominal phenotypes, but 10/19 (52%) of cuticles examined exhibited defects in leg morphology. Specifically, the pretarsi, or claws, of the thoracic legs were almost entirely abolished (Fig. S10A-C). Tc-nub is expressed in the leg joints (Fig. S10D), as has been observed in other insect species (Li and Popadić, 2004; Turchyn et al., 2011). We observed that Tc-cas is also expressed in the developing legs, at both the proximal and distal ends (Fig. S10D). This is, to our knowledge, the first evidence suggesting that cas functions in arthropod limb development.

In this study, we have shown that the genes hb, Kr, nub and cas are expressed sequentially in the SAZ of Tribolium, as they are in Drosophila neuroblasts. We have also shown that Tc-Nub plays a role in axial patterning, acting redundantly with the abdominal gap proteins Tc-Gt and Tc-Kni to repress Tc-Kr expression, and thereby to establish normal abdominal Hox gene expression. Our findings provide support for the theory that the neuroblast timer network was co-opted for axial patterning.

Our combinatorial knockdown experiments indicate that Tc-Nub, Tc-Gt and Tc-Kni all contribute to the repression of Tc-Kr in the abdomen. Dm-Gt and Dm-Kni are known to repress Dm-Kr expression in Drosophila (Jaeger, 2011), and Tc-Gt has long been suspected to regulate Tc-Kr expression in Tribolium (Bucher and Klingler, 2004; Cerny et al., 2005). However, this is, to our knowledge, the first evidence that Kni regulates Kr expression in a non-drosophilid insect (Jaeger, 2011) and that Nub can repress Kr in the context of arthropod segment patterning.

Tc-nub, Tc-kni and Tc-gt seem to display ‘distributed redundancy’, i.e. they have different but overlapping roles, so that if one gene is lost, the others can at least partially compensate for it (Wagner, 2005). There are obvious reasons why the gap gene network might benefit from being robust to mutation. These genes regulate some of the earliest and most crucial elements of the insect body plan (segment boundaries and segment identities), and complete disruption of gap gene function is lethal (Jürgens et al., 1984; Nüsslein-Volhard et al., 1984; Wieschaus et al., 1984). The overlapping functions of Tc-nub, Tc-gt and Tc-kni may also be important for fine-tuning the expression dynamics of Tc-Kr, allowing for more precise regulation of the overlapping Hox gene domains in the posterior thorax and anterior abdomen.

Nub may also regulate Kr expression during axial patterning in other insect species, with varying degrees of redundancy with Gt and/or Kni. In Oncopeltus, pRNAi against Oc-nub results in prominent abdominal segment transformations arising from downregulation of Oc-abd-A expression (Hrycaj et al., 2008). We have shown that similar phenotypes arise in Tribolium from ectopic expression of Kr. Knockdown of Oc-gt or Oc-kni has no obvious effect on Oc-Kr expression (Ben-David and Chipman, 2010), suggesting that Nub may play a more central role than Gt and Kni in regulating Kr expression in Oncopeltus. In contrast, deletion of both Drosophila nubbin paralogues has little effect on the gap domain of Dm-Kr expression (Cockerill et al., 1993), despite the fact that Dm-Nub is able to repress Dm-Kr in neuroblasts (Grosskortenhaus, 2006; Tran and Doe, 2008). In this species, Gt and Kni may therefore have a more prominent role in Kr regulation than Nub. Intriguingly, we have observed subtle misexpression of Dm-abd-A expression in Drosophila embryos lacking both nubbin paralogues (Tidswell, 2020), in contrast to previous reports (Hrycaj et al., 2008). It therefore seems likely that Dm-Nub is able to repress Dm-Kr expression in the context of the gap gene network, but that this interaction is weak and/or masked by redundancy with Dm-Gt and Dm-Kni.

Subtle alterations in network interactions, even while the overall output of the network is conserved (known as developmental systems drift), are a common feature of the gap gene network (Crombach et al., 2016; Wunderlich et al., 2015). Investigating the functional overlap between Nub, Gt and Kni in additional insect species, with different modes of segmentation, and at strategic points in the insect phylogeny, will help to determine when and how the function of these genes has drifted over evolutionary time. This represents a promising framework for studying gene regulatory network evolution.

It is striking that the phenotypes observed after simultaneous knockdown of Tc-nub, Tc-kni and Tc-gt are very reminiscent of those observed after knockdown of the gap gene Tc-mille-pattes (Tc-mlpt). Both treatments lead to the expansion of Tc-Kr expression into the SAZ, and the formation of ectopic legs on presumptive abdominal segments (Savard et al., 2006). It may be that Tc-Mlpt is required for the expression not just of Tc-gt (Savard et al., 2006), but also of Tc-kni and Tc-nub, in the SAZ. Knockdown of Tc-mlpt expression would then effectively phenocopy a triple knockdown of Tc-nub, Tc-gt and Tc-kni in the SAZ. This hypothesis places Tc-mlpt in a crucial position in the gap gene network and warrants further investigation.

It is intriguing to note that the expression domains of the first three neuroblast timer genes, Tc-hb, Tc-Kr and Tc-nub, align approximately with the three trunk tagma in Tribolium (gnathum, thorax and abdomen, respectively), save that they are shifted anteriorly to align with parasegment boundaries, and Tc-nub covers most but not all of the abdominal parasegments (Fig. 7A).

Functional data also support the importance of this gene-tagma pattern (Fig. 7A′). Tc-Hb represses thoracic and abdominal Hox genes (Marques-Souza et al., 2008), allowing gnathal Hox genes to be expressed. Tc-Kr represses gnathal (Cerny et al., 2005) and abdominal Hox genes, allowing the thoracic Hox genes to be expressed. Finally, Tc-Nub, in tandem with Tc-Gt and Tc-Kni, represses Tc-Kr expression, which, in the absence of Tc-Hb, allows for abdominal Hox genes to become expressed. This minimal network could therefore provide enough information to lay down the basic functional divisions of the insect axis (although not, of course, the fine details of individual segment identity).

This observation is particularly intriguing because gap gene regulation of Hox genes is thought to pre-date gap gene regulation of segment boundary positions (Clark et al., 2019). However, we lack a detailed understanding of how the gap gene network as a whole contributes to Hox gene regulation in sequentially segmenting insects such as Tribolium. A comprehensive molecular dissection of Hox gene regulation in Tribolium is required to test this hypothesis.

The idea that the neuroblast timer network might be used for axial patterning in insects was first suggested when Dm-hb, Dm-Kr, Dm-nub and Dm-cas were found to be expressed, in that order, along the AP axis of the Drosophila embryo (Isshiki et al., 2001). We have shown that the genes of the neuroblast timer network are also expressed during axial patterning in the sequentially segmenting insect Tribolium, and that Tc-nub has a clear function during this process.

The roles of hb, nub and cas in the neuroblast timer network long predate their roles in axial patterning. Homologues of all three genes (Ikaros, PouF2 and Casz1, respectively) are expressed sequentially in neural and/or retinal stem cells in mammals and promote the formation of a temporal sequence of different daughter cell types (Alsiö et al., 2013; Elliott et al., 2008; Javed et al., 2020; Mattar and Cayouette, 2015; Mattar et al., 2015). In contrast, there is no evidence that any of these genes play a role in segment formation or axial patterning outside of the arthropods. Even within the non-insect arthropods, there are species that express hb and/or Kr in their neuroblasts but not in the SAZ (Chipman and Stollewerk, 2006; Kontarakis et al., 2006). From these observations, we can infer that at least hb, nub and cas were most likely recruited to a role in axial patterning from an ancestral role in neural patterning.

Beyond the broad similarities presented in this paper, we have also identified some key differences in the order of expression and function of neuroblast timer genes in neuroblasts and in the SAZ. First, hb is expressed in the SAZ after cas, something that is not observed in neuroblasts. This posterior domain of hb is conserved in a range of insect lineages (Jaeger, 2011; Liu and Kaufman, 2004a; Marques-Souza et al., 2008; Mito et al., 2005), and has been hypothesised to influence the duration of segmentation (Nakao, 2016). It seems likely, then, that it is a significant component of the timer network in the SAZ. Furthermore, although we found a clear role for nub in regulating axial identities, we have found no such role for cas. Unlike the other neuroblast timer genes, expression of cas in the SAZ of Tribolium is modulated in a complex pair-rule pattern, arguing against its regulating axial identity across a broad, continuous region of the SAZ. The function of the ‘gap-like’ domain of cas expression in Drosophila also remains mysterious, as Dm-cas mutants appear normal outside their neural defects (Mellerick et al., 1992). It may be that cas, like nub, acts redundantly with other genes to exert its influence on axial identity; that it has lost the ability to regulate axial identity in Tribolium and Drosophila; or that it never had such a role. Analysis of cas expression and function in the SAZs of other insect species may help to distinguish between these possibilities.

It is worth asking whether the regulatory interactions that drive sequential expression of Tc-hb, Tc-Kr, Tc-nub and Tc-cas in the SAZ are the same as in Drosophila neuroblasts. Sequential expression of the neuroblast timer genes in Drosophila neuroblasts depends largely on cross-regulatory interactions between their gene products, including feed-forward activation, feedback repression and ‘next-plus-one’ repression (Averbukh et al., 2018; Doe, 2017; Isshiki et al., 2001; Nakajima et al., 2010; Rossi et al., 2017) (Fig. 7B). Our understanding of Tribolium gap gene interactions is fragmentary, but this network may share some of its regulatory interactions with neuroblasts (Fig. 7B′).

However, there are also obvious differences between the two networks. The transition between Tc-hb and Tc-Kr expression in the SAZ appears to be mediated entirely by interactions within the network (Marques-Souza et al., 2008), while in neuroblasts, the transition between Dm-hb and Dm-Kr expression is driven by the nuclear receptor Dm-Seven-Up in a cytokinesis-dependent manner (Benito-Sipos et al., 2011; Grosskortenhaus et al., 2005; Kanai et al., 2005; Mettler et al., 2006). Furthermore, the timing and extent of Tc-Kr expression in the SAZ is influenced by gap genes that are not expressed in neuroblasts, such as Tc-gt and Tc-kni (Bucher and Klingler, 2004; Cerny et al., 2008; this article).

By demonstrating sequential expression of the neuroblast timer genes in the SAZ of Tribolium, and revealing that Tc-nub is able to repress the expression of Tc-Kr to influence Hox gene expression, our findings provide strong support for the hypothesis that the neuroblast timer network has been co-opted for axial patterning during the evolution of insects. These findings will provide a basis for future studies examining the evolution and structure of the gap gene network in insects.

Tribolium castaneum strain San Bernadino beetles (provided by A. Peel, University of Leeds, UK) were reared on organic wholemeal flour (Doves Farm Foods, Hungerford, UK) supplemented with fast action dried yeast (Sainsbury's, London, UK) and the antifungal agent Fumagilin-B (Medivet) at 30°C, as described in the Beetle Book v1.2 (Bucher, 2009). Egg lays were performed on strong white organic bread flour (Doves Farm Foods, Hungerford, UK). Incubators were maintained between 40-60% relative humidity where possible, and no day/night cycle was used (beetles were kept in the dark).

Tribolium were allowed to lay on white flour for 24 h and their eggs were then collected using a sieve with a 200 μM mesh size (Retsch test sieve 200 mm×50 mm). Collected eggs were transferred into small mesh baskets (with a mesh aperture of 250 μm) and were rinsed several times in double-distilled H₂O to remove all traces of flour. Their chorions were then removed by washing twice in bleach diluted with double-distilled H₂O to a final concentration of 2.5% (v/v) hypochlorite, for 30-45 s. After further rinsing in double-distilled H₂O, dechorionated embryos were fixed as described by Marques-Souza et al. (2008), except a 0.68 mm ID (internal diameter) needle was used to enhance devitellinisation rather than a 0.9 mm ID needle. Fixed and devitellinised embryos were stored in 100% methanol at −20°C.

Ovaries were removed from adult female beetles in PBS using forceps. Dissected ovaries were transferred directly into 4% formaldehyde in PBT (PBS+0.01% Tween) on ice. An equal volume of heptane was added, and the tubes then rocked on a nutator for 20 min to allow for fixation. The ovaries were then rinsed several times in PBT and then washed into 100% methanol for storage at −20°C.

Plasmids containing clones for GFP, Tc-nub, Tc-cas, Tc-gt, Tc-kni and Tc-odd were provided by A. Peel (University of Leeds, UK) and R. Sharma (University of Cambridge, UK) (clone sequences provided in Table S4). All dsRNA fragments used were computationally predicted to have a low potential for off-target gene silencing using the default search parameters of Deqor version 3.0 (i.e. the quality score of all potentially cross-silencing siRNAs was >5) (Henschel et al., 2004). dsRNA was synthesised from PCR products using T7 polymerase, and was purified using phenol chloroform precipitation. Purified dsRNA was resuspended in RNase-free water and injected into Tribolium adults or eggs at a concentration of 1-4 μg/μl. Unless specified otherwise, single knockdowns were carried out using 2 μg/μl of dsRNA, while double and triple knockdowns used the component dsRNAs mixed to a final concentration of 1 μg/μl each (the viscosity of the injection fluid became difficult to work with above 4 μg/μl).

All injections for RNAi were carried out using a Pico-injector system (Medical Systems). Parental RNAi was carried out by injecting dsRNA into the dorsal surface of the abdomen under the elytra of adult female beetles as described by Posnien et al. (2009). Males were introduced to the injected females the day after injection, and eggs were collected starting from 1 week after injection. Eggs were collected and fixed regularly (every 1-2 days) as described above for 3-4 weeks after injection.

Embryonic microinjection for eRNAi was carried out using a method adapted from Benton (2018). One- to 2-h-old eggs were transferred into small mesh baskets (with a mesh aperture of 250 μm) and rinsed several times in double-distilled H₂O. Chorions were removed by washing twice in bleach, diluted with double-distilled H₂O to a final concentration of ∼0.6% (v/v) hypochlorite, for 30-45 s. Eggs were rinsed again and then healthy looking eggs were lined up on coverslips and allowed to dry. Eggs were covered with a 1:1 mix of Halocarbon oil 700 and Halocarbon oil 27 (Sigma Aldrich) and dsRNA was injected into the anterior pole (to reduce the risk of damage to the posterior segment addition zone). The coverslip was turned over on to a Lumox culture dish as described by Benton (2018), except that glass ‘feet’ ∼0.6 mm high (made from strips of #1.5 coverslip) were attached to the coverslip at either end of the injected rows of eggs, to prevent them from being pressed against the membrane. Injected eggs were then stored in plastic chambers with wet paper towel (to maintain humidity) and reared at 30°C.

For fixation, injected embryos were aged for the appropriate length of time then injected with PBT+10% formaldehyde (v/v) and left to fix at room temperature for 1 h. They were then transferred using an eyelash hair to Eppendorf tubes and fixed for an additional hour in a 1:1 mix of heptane and PBT+4% formaldehyde (v/v). The aqueous layer was removed and 100% ice-cold methanol added. Germbands were manually dissected away from the remainder of the yolk, chorion and vitelline membrane in PBS, and then stored in 100% methanol at −20°C until required.

Version 3.0 HCR probes (20 pairs per gene) and fluorescently labelled hairpins were produced by Molecular Instruments. Probe template sequences were taken from NCBI (Tc-hb, NM_001044628.1; Tc-Kr, NM_001039438.2; Tc-nub, XM_015979462.1; Tc-cas, XM_015980923.1; Tc-wg, NM_001114350.1; Tc-Ubx, XM_008203013.2; Tc-abd-A, NM_001039429.1). All required buffers were made according to the instructions provided by Molecular Instruments, with the one exception that the percentage of dextran sulphate in the probe hybridisation and amplification buffers was reduced from 10% (v/v) to 5% (v/v) to reduce viscosity and improve retention of embryos during washes.

Fixed embryos or ovaries were prepared for hybridisation chain reaction (HCR) in situ hybridisation (ISH) by removing methanol and replacing it with 1 ml of PBT containing 4% formaldehyde. Tubes were rocked on the nutator for 30 min to allow for additional fixing and rehydration to occur. The HCR ISH was then carried out as per the Molecular Instruments HCR v3.0 protocol for whole-mount fruit fly embryos, with the exception that hybridisation steps were carried out in 100 rather than 200 μl of hybridisation buffer, and the volume of probe added was adjusted to give the same final concentration (4 nM). Additionally, 1 ng/μl DAPI was added to the first 30 min wash on the final day so that nuclear staining could be carried out in parallel. After washing, embryos or ovaries were transferred first into 25% (v/v) glycerol and then into 50% (v/v) glycerol before being stored at 4°C to stiffen and clear for mounting.

Blastoderm stage embryos were mounted in glass-bottomed Petri dishes (Cellvis), and dissected germbands and whole ovarioles on glass slides, in ProLong Gold Antifade Mountant (Thermofisher Scientific) as per the manufacturer's instructions. Most mounted embryos and ovarioles were imaged using an Olympus FV3000 confocal microscope and associated FLUOVIEW software at the Department of Zoology Imaging Facility (University of Cambridge). 12-bit z-stacks of entire embryos and ovarioles were taken using a UPLSAPO 20× objective lens (no immersion, NA=0.75) with a z step-size of 3-5 μM and a pixel dwell time of 2 μs. Z-stacks spanned the entire depth (from ventral to dorsal surface) of flat-mounted embryos, and approximately half of the depth of blastoderm-stage embryos. A minority of embryos (several pictured in Fig. 1 and all pictured in Fig. S1) were imaged prior to the installation of the Olympus FV3000 microscope, using a Leica SP5 inverted confocal microscope at the Department of Zoology Imaging Facility (University of Cambridge). 16 bit z-stacks of embryos pictured in Fig. 1 were taken using a 11506191 20× objective lens (no immersion, NA=0.7), with a z step-size of 1-3 μm. 16 bit z-stacks of the posterior gut regions pictured in Fig. S1 were taken using a 11506192 63× objective lens (oil immersion, NA=1.4), with a z step-size of 0.3-0.5 μM. All images were taken with a scan format of 1024×1024 pixels. A 405 laser was used to visualise DAPI, and 488, 561, 594 and 633 (Leica) or 640 (Olympus) lasers were used to visualise fluorescently tagged HCR ISH hairpins.

Embryos and larvae were processed for cuticle preparation either upon hatching, or after 7-10 days if they failed to hatch in this time. Embryos and larvae of uninjected embryos were first rinsed in 2.5% (v/v) bleach and then in double-distilled H₂O to remove any remaining chorion and debris. Injected embryos and larvae were dissected out of their chorions manually, and washed in methanol and then heptane (1 h each) to remove the halocarbon oil. Embryos or larvae were then transferred to a glass slide, covered with a 1:1 mix of Hoyer's medium (Dahmann, 2008):lactic acid and a coverslip, and heated at 60°C overnight. Cuticles were imaged with an excitation wavelength of 633 nm on an Olympus FV3000 confocal microscope in the Department of Zoology (University of Cambridge). Overview images of entire cuticles were taken using a UPLSAPO 10× objective lens (no immersion, NA=0.4) with a z step-size of 4-5 μm, while close-up images of cuticles (i.e. the inset in Fig. 4B, and the cuticle images in Figs S7 and S9) were taken using with a UPLSAPO 20× objective lens (no immersion, NA=0.75) with a z step-size of 1-2 μm. All images were taken as 12 bit, with a scan format of 1024×1024 pixels and a pixel dwell time of 2 μs.

Images and z-stacks were stitched using the Olympus FV3000 FLUOVIEW software. Additional image processing was carried out in Fiji (Schindelin et al., 2012). To correct for subtle misalignment of the dichroic mirrors on the confocal microscope used for imaging, channels of images taken using the Olympus FV3000 confocal microscope were realigned using the ‘Olympus FV3000 dichroic mirror offsets’ Fiji plug-in by Matthew Wayland (https://github.com/WaylandM/dichroic-mirror-offsets). Unless otherwise stated, all images of embryos are maximum projections of confocal z-stacks. Fiji was also used to adjust image brightness and contrast, and to rotate, crop and re-slice images where necessary, in accordance with guidelines presented by Schmied and Jambor (2021). Fixed intensity values were applied for qualitative comparisons of signal intensity between images. We used the ChrisLUTs LUT package for ImageJ (Christophe Leterrier and Scott Harden; https://github.com/cleterrier/ChrisLUTs) for presenting selected confocal images. Figures were assembled in the open source vector graphics editor Inkscape (https://inkscape.org/).

Statistical analysis was carried out using R (v4.1.0) (R Core Team, 2021a) and RStudio (v1.4.1106) (R Core Team, 2021b). Bayesian logistic regressions were carried out using the bayesglm function in the arm (v1.11-2) package (Gelman and Su, 2020).

We thank A. Peel and E. Clark for discussions that helped to inspire this project, and for advice and input throughout the work. We also thank A. Peel and R. Sharma for providing beetles, plasmids and dsRNA, and for help with rearing and knockdowns. We thank S. Taylor for assistance with ovary dissection, and the Imaging Facility at the Department of Zoology (University of Cambridge) for confocal imaging support.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199719