Abdominal-B and caudal inhibit the formation of specific neuroblasts in the Drosophila tail region

During the development of the central nervous system (CNS), a huge variety of different cell types has to be generated. For the establishment of functional neural circuits it is a prerequisite that characteristic numbers of each cell type become arranged in a reproducible spatial pattern. This pattern is founded during the formation and specification of neural stem cells. Subsequently, tight temporal control of their proliferation and survival as well as specification, differentiation and survival of their progeny ensures the generation of characteristic cell lineages. Drosophila melanogaster is one of the favoured models used to investigate these processes and shares many fundamental mechanisms in CNS development with vertebrate systems (for reviews see Doe et al., 1998; Thor, 1995).

In the embryonic CNS of Drosophila, neural stem cells (termed neuroblasts, NBs) can be individually identified by their characteristic time and position of delamination from the neuroectoderm (NE), and the expression of a unique combination of molecular markers (Broadus et al., 1995; Doe, 1992; Urbach and Technau, 2003). Their specification occurs already in the NE by Cartesian grid-like expression of patterning genes, providing positional information along the anteroposterior (AP) and dorsoventral (DV) axes (for reviews see Bhat, 1999; Skeath, 1999). Hence, each NB generates a unique and almost invariant cell lineage (Bossing et al., 1996; Schmid et al., 1999; Schmidt et al., 1997).

Along the AP axis the CNS is composed of segmental units (neuromeres). In the thoracic (T1-T3) and most of the abdominal segments (A1-A8) each hemineuromere originates from a repetitive set of ∼30 NBs (Broadus et al., 1995; Doe, 1992). Serially homologous NBs, exposed to similar positional cues, produce similar lineages, although some of these include subsets of cells that specifically differ between segments (reviewed by Technau et al., 2006). Neuromeres in the tail region (A9, A10) exhibit a progressively derived character (compared to the assumed developmental ground state in T2; Lewis, 1978). They are formed by only 23 NBs in A9 and by 11 NBs in A10 per hemineuromere, and thus their size is continuously decreased (Birkholz et al., 2013).

Segmental identity and patterning is under the control of the highly conserved Hox genes, each encoding a transcription factor with a DNA-binding homeodomain (e.g. reviewed by McGinnis and Krumlauf, 1992). During neurogenesis in Drosophila, Hox genes have been shown to act in various developmental stages at the level of NE, NBs and their progeny to control cell-fate specification, proliferation and/or apoptosis in a segment-specific manner (for reviews, see Reichert and Bello, 2010; Rogulja-Ortmann and Technau, 2008). Whereas Hox genes clustered in the Antennapedia complex (ANT-C) (Kaufman et al., 1980) control differentiation of the head, gnathal and anterior thoracic segments, those in the Bithorax complex (BX-C) (Lewis, 1978) shape the posterior thoracic and all of the abdominal segments (reviewed by Peifer et al., 1987). Hox gene expression domains coincide with parasegmental boundaries instead of segmental ones. A parasegment (PS) represents a fundamental genetic unit and consists of the posterior compartment of one segment (expressing the segment polarity gene engrailed (en) (DiNardo et al., 1985; Patel et al., 1989) and the anterior compartment of the next segment (En negative) (reviewed by Deutsch, 2004; Martinez-Arias and Lawrence, 1985) (e.g. PS14 consists of posterior A8 and anterior A9). The BX-C comprises the genes Ultrabithorax (Ubx), abdominal-A (abd-A) and Abdominal-B (Abd-B) (Casanova et al., 1987; Sánchez-Herrero et al., 1985; Tiong et al., 1985). Arrangement of these genes within the BX-C is collinear with their expression domains along the AP body axis (reviewed by Duncan, 1987; Lewis, 1978) and Abd-B is expressed most posteriorly (Harding et al., 1985). Abd-B contains two distinct genetic elements, which are active in different domains: the morphogenetic (m) subfunction is necessary to generate the morphological diversity of PS10-13, whereas the regulatory (r) element is needed for the identity of PS14-15 (Casanova et al., 1986).

Here we have investigated the role of the different Abd-B isoforms in shaping the most posterior neuromeres of the ventral nerve cord (VNC). We focused on a subset of four NBs and their lineages (NB2-4, NB3-3, NB6-4 and NB7-3) that express the molecular marker Eagle (Eg) (Dittrich et al., 1997; Higashijima et al., 1996). We demonstrate that the r isoform of Abd-B (Abd-B.r) acts at different levels: It confers abdominal fate to NB6-4 (producing glia exclusively), and it triggers programmed cell death (PCD) of several progeny cells within the NB3-3 lineage. Furthermore, it inhibits the formation of a specific set of NBs. In Abd-B null mutants, which show no expression of BX-C genes in PS14-15, NB3-3 and NB6-4 (producing glia plus neurons) in PS14 assume thoracic fate, and in PS15 additional NBs are formed, including NB7-3, which is never generated in PS15 of wild-type embryos. Ectopic expression of the m isoform of Abd-B (Abd-B.m) leads to a rescue of the Abd-B null mutant phenotypes, demonstrating similar potentials of both isoforms. However, Abd-B.r requires co-expression of the ParaHox gene caudal (cad) to repress the formation of NBs in PS15. The combined action of Abd-B.r and cad is also sufficient to ectopically induce posterior identity in anterior neuromeres. We conclude that Abd-B and cad are required to inhibit the formation of specific NBs and to modify particular NB lineages, in order to adjust proper size and composition of the terminal neuromeres.

The following fly strains were used: wild type (Oregon R); eg-Gal4 (Ito et al., 1995); sca-Gal4 (Klaes et al., 1994); elav-Gal4 (Lin and Goodman, 1994); UAS-cad (Moreno and Morata, 1999) (from Ulrich Schaefer, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany); UAS-cad (Hwang et al., 2002) (from Mi-Ae Yoo); en-lacZ (Hama et al., 1990) (from Alfonso Martinez-Arias, University of Cambridge, UK); Df(3R)C4 (Karch et al., 1985) (from François Karch); Df(3L)H99 (White et al., 1994), UAS-P35 (Hay et al., 1994), cad² (Macdonald and Struhl, 1986) and UAS-nGFP (all from Bloomington Stock Center); Abd-B^M1, Abd-B^M3, Abd-B^M5 (Sánchez-Herrero et al., 1985), Abd-B^D18 (Hopmann et al., 1995) and Abd-B^Uab1 (Karch et al., 1985) (all from Ernesto Sánchez-Herrero); Ubx^MX12, abd-A^M1, Abd-B^M8 triple mutant (Casanova et al., 1987), UAS-Abd-B.r (Rivas et al., 2013) and UAS-Abd-B.m (Castelli-Gair et al., 1994) (all from James Castelli-Gair Hombría). All experiments were performed at 25°C.

Embryos (staging according to Campos-Ortega and Hartenstein, 1997) were dechorionated, fixed and immunostained following previously published protocols (e.g. Patel, 1994). The following primary antibodies were used: mouse anti-Abdominal-A (1:200) (Kellerman et al., 1990) (from Ian Duncan); mouse anti-Abdominal-B (1:20) (Celniker et al., 1989), mouse anti-Invected (1:2) (Patel et al., 1989) and mouse anti-Ultrabithorax (1:20) (White and Wilcox, 1984) (all from DSHB); chicken anti-β-Galactosidase (1:1000) (Abcam); guinea pig anti-Caudal (1:400) and guinea pig anti-Runt (1:500) (Kosman et al., 1998) (from John Reinitz); rabbit anti-Caudal (1:100) (Macdonald and Struhl, 1986) (from Paul Macdonald); rabbit anti-Deadpan (1:100) (Bier et al., 1992) (from Harald Vaessin); mouse anti-Eagle (1:100) (Karcavich and Doe, 2005) (from Chris Doe); rabbit anti-Eagle (1:500) (Dittrich et al., 1997); rabbit anti-Engrailed (1:100) (Santa Cruz Biotechnology); rabbit anti-Eyeless (1:1000) (Kammermeier et al., 2001) (from Uwe Walldorf); mouse anti-GFP (1:250) (Roche); rat anti-Gooseberry-proximal (1:2) (Zhang et al., 1994) (from Robert Holmgren); rabbit anti-Miranda (1:100) (Betschinger et al., 2006) (from Juergen Knoblich); guinea pig anti-Reversed-polarity (1:10,000) (von Hilchen et al., 2013).

For in situ hybridisation we generated an RNA probe for Abd-B.m (616bp) targeting its unique N-terminal protein coding sequence (CDS). The Abd-B.r RNA probe (220 bp) is directed against two exons, which are exclusively present in all described r-specific transcripts (Fig. 3A). The probes were obtained by amplification from cDNA pAB713 (Abd-B.m) and pAB728 (Abd-B.r) (Kuziora and McGinnis, 1988) (from Ernesto Sánchez-Herrero) and cloned into pDrive (Qiagen). The following primers were used for amplification: for Abd-B.m 5′-CACTGGAGGGAGAAACACTCGC-3′ and 5′-CAACAGCAGCAGCAGCAGCAG-3′; for Abd-B.r 5′-CTGTGGGATGGGAACTGACGC-3′ and 5′-GACCAAAAGCACTACCCAAATAACTG-3′.

pDrive was linearised and probes were synthesised by a T7 RNA polymerase using FITC (Abd-B.m) or DIG (Abd-B.r) RNA labelling Mix (Roche). The hybridisation was carried out as described before (Plickert et al., 1997; Tautz and Pfeifle, 1989) and the probes were diluted 1:5000 (Abd-B.m) or 1:1000 (Abd-B.r).

As fluorescent secondary antibodies we used the DyLight (Jackson ImmunoResearch) and Alexa (Life Technologies) series. Non-fluorescent secondary antibodies were either biotinylated or alkaline phosphatase-conjugated (Jackson ImmunoResearch). All secondary antibodies were used according to the manufacturer’s protocols.

Non-fluorescent stainings were documented on a Zeiss Axioplan microscope; fluorescent confocal images were acquired on a Leica TCS SP2 or SP5 microscope. All images were processed by Adobe Photoshop and Illustrator.

A two-tailed t-test was performed for statistical significance (see Fig. 1K, Fig. 8G).

We previously provided a comprehensive map of NBs generated in the tail region and found NB numbers to be significantly reduced in segments A9 (by 28%) and A10 (by 66%), compared with thoracic and anterior abdominal segments (Birkholz et al., 2013). As Hox gene expression domains coincide with parasegmental instead of segmental boundaries, we calculated NB numbers corresponding to terminal PSs. We used the stem cell marker Deadpan (Dpn) (Bier et al., 1992) to label NBs in early stage 12 (St12e) wild-type embryos and found 31.13±0.35 (n=8) Dpn-positive cells per side in PS13, 25.88±1.13 (n=8) in PS14 and 11.88±0.64 (n=8) in PS15 (Fig. 1A,K).

As Abd-B is expressed most posteriorly of all BX-C Hox genes, we investigated its role in shaping terminal neuromeres. Abd-B^D18 is a small deficiency, removing the whole Abd-B CDS. In Abd-B^D18 mutant embryos, we found the number of Dpn-positive cells in PS15 to be increased (15±0.67, n=10), whereas the more anterior PSs show no significant differences compared with wild type (Fig. 1B,K). Thus, normal Abd-B expression is required to reduce the number of NBs in PS15. In apoptosis-deficient Df(3L)H99 mutant embryos the number of Dpn-positive cells in terminal PSs corresponds to wild type (Fig. 1C,K). This indicates that Abd-B does not reduce NB numbers by inducing PCD, but by suppressing the formation of these NBs.

To identify NBs that are affected by this inhibition, we performed antibody stainings against Eg, which in thoracic and anterior abdominal neuromeres of wild-type embryos is expressed in four NBs (NB2-4, NB3-3, NB6-4 and NB7-3) and all of their progeny cells (Dittrich et al., 1997; Higashijima et al., 1996). According to the NB map of the tail region (Birkholz et al., 2013), NB6-4 is formed in PS14 and PS15 (but rapidly downregulates Eg expression). However, NB7-3 and NB3-3 are not formed in PS15, and NB2-4 is not generated in PS14 and PS15 (see also Fig. 1D). Therefore, these NBs are candidate NBs, which could be suppressed by Abd-B. In St13 Abd-B^D18 mutants (Fig. 1E), as well as in Df(3L)H99 mutants (Fig. 1F), we found NB6-4, NB3-3 and NB2-4 to be distributed as in wild type, indicating that Abd-B does not affect formation of these NBs. Yet, their lineages displayed some changes: in Abd-B^D18 mutants, the NB6-4 lineage in PS14 (100%, n=28 hemi-PSs) and occasionally in PS13 (14%, n=28) reveals homeotic transformation into thoracic fate, as it consists of neurons in addition to glial cells (Fig. 1E) [normally only glia in abdominal NB6-4 lineages (Schmidt et al., 1997)]. The NB3-3 lineage in PS14 of St16 Abd-B^D18 mutants consists of 12 cells, compared to only six cells in wild type (not shown), corresponding to a transformation into thoracic and anterior abdominal fate (Schmidt et al., 1997). As the size of the NB3-3 lineage is also increased in PS14 of Df(3L)H99 mutants (10 cells), Abd-B seems to promote PCD in several progeny cells of NB3-3.

Strikingly, in PS15 of Abd-B^D18 mutants we found an ectopic Eg-positive cell cluster (on each side; 100%, n=28; Fig. 1E), that we never observed in wild type (Fig. 1D) or in Df(3L)H99 mutants (Fig. 1F). These cells are negative for Runt (Run) (Fig. 1G), but positive for En (Fig. 1H) and are located within the last En stripe. As they are negative for the glial marker Reversed polarity (Repo) (Fig. 1I), they do not seem to derive from NB6-4. Instead, their Eyeless (Ey) expression unambiguously identifies them as NB7-3 progeny (Fig. 1J). Furthermore, they are located in the same dorsal position as the NB7-3 daughter cells in more anterior neuromeres and their precursor delaminates at St11l-St12e (S5). The cluster consists of up to three cells in St13, five cells in St14, and four cells from St16 on, which is also typical for the NB7-3 cell lineage (Bossing et al., 1996; Karcavich and Doe, 2005; Novotny et al., 2002; Schmid et al., 1999). We therefore conclude that the ectopic Eg-positive cluster generated in Abd-B^D18 mutants corresponds to the NB7-3 lineage. Moreover, its absence in Df(3L)H99 mutants indicates that Abd-B does not act to induce PCD, but to repress formation of NB7-3 in PS15 of early wild-type embryos (see Fig. 9A). As the phenotype is also found in an Abd-B^D18 heterozygous situation, but with lower penetrance (50%) (supplementary material Fig. S1; Fig. 9A), the effects appear to be dose dependent.

In the following we use NB7-3 as a model to further elucidate the mechanism controlling restriction of NB formation in the tail region.

Interactions among Hox genes of the BX-C are characterised by ‘posterior prevalence’, which means that Hox genes that are active in posterior PSs tend to repress the expression of more anterior Hox genes (for reviews, see Duboule and Morata, 1994; Morata, 1993). Accordingly, in a St14 wild-type embryo the Abd-A protein domain reaches from PS7 to PS13a, whereas in Abd-B loss-of-function mutants, Abd-A expression is at least expanded to PS14 (Karch et al., 1990; Kellerman et al., 1990; Macias et al., 1990). To examine whether the ectopic Eg-positive cells in Abd-B mutants arise as a result of expansion of the other genes from the BX-C, we investigated their expression in the VNC. In the loss-of-function allele Abd-B^M1, and in Df(3R)C4, a chromosomal deletion of Abd-B, we found the anticipated expansion of Abd-A in the VNC, but we never observed Abd-A expression in the ectopic NB7-3 lineage of PS15, or in the NB3-3 and NB6-4 lineages of PS14 (supplementary material Fig. S2A,B). However, in Abd-B^D18 mutants we found no obvious expansion of the Abd-A domain within the VNC (compare Fig. 2A,B). Recently it has been postulated that Abd-B is not necessary for the repression of Abd-A in the VNC, but instead the iab-8 non-coding RNA (Gummalla et al., 2012), which is also affected in Abd-B^M1 and Df(3R)C4 but not in Abd-B^D18. Yet, all mutants displayed the same phenotype concerning the Eg pattern (Fig. 9A). Ubx, the most anterior gene of the BX-C, is normally expressed from PS5 to PS13 in the VNC (White and Wilcox, 1984). In Abd-B^D18 mutants we observed no posterior expansion of Ubx expression (Fig. 2C), which is in agreement with a previous report (Struhl and White, 1985). In addition, a triple mutant removing the whole BX-C revealed the same Eg pattern in PS14-15 as the Abd-B single mutants (supplementary material Fig. S3A,B; Fig. 9A).

Thus, in Abd-B mutant background expression of Abd-A or Ubx is observed neither in the ectopic NB7-3 lineage in PS15, nor in the NB6-4 and NB3-3 clusters in PS14. As a consequence of the missing BX-C input they assume thoracic fate, resembling the ground state (pPS4 + aPS5) (Lewis, 1978).

The Abd-B gene consists of at least four overlapping transcription units (Fig. 3A): Class A [nomenclature according to Zavortink and Sakonju (Zavortink and Sakonju, 1989)] is expressed in PS10-13 and is the only transcript that encodes m function. Class B and C are expressed in PS14-15, γ-transcripts only in PS15, and each of these transcript classes may contribute to r function (Boulet et al., 1991; DeLorenzi et al., 1988; Kuziora and McGinnis, 1988; Sánchez-Herrero and Crosby, 1988; Zavortink and Sakonju, 1989). To recapitulate the localisation of the genetic elements in the VNC by in situ hybridisations, we generated one probe, specific for the A-transcript (Abd-B.m) and a second one, directed against all transcripts encoding r function (Abd-B.r) (Fig. 3A). The stainings confirmed that in St12e Abd-B.m is expressed in PS13 and anterior to it, revealing a sharp posterior border (Fig. 3B), whereas Abd-B.r transcripts are exclusively expressed in PS14-15 (Fig. 3C). cDNA sequences revealed the existence of two different protein isoforms: Abd-B.m (encoded by the A-transcript) and Abd-B.r (encoded by the B, C and γ-transcripts). Abd-B.m is larger (55 kDa) and shares a common C-terminus with Abd-B.r, which harbours the homeodomain but has a unique glutamine-rich N-terminus, comprising 223 additional amino acids that are absent in the truncated r protein (30 kDa) (Fig. 3A) (Celniker et al., 1989; DeLorenzi et al., 1988; Zavortink and Sakonju, 1989). To investigate the wild-type Abd-B protein expression in distinct (Eg-positive) cells of the embryonic VNC, we used an antibody that detects both protein isoforms. As the expression domains of the Abd-B transcripts are spatially non-overlapping, it is reasonable to assume that the antibody detects the Abd-B.m or Abd-B.r protein in the respective PSs. Already at St11m, we detected abundant Abd-B.r expression in PS14, including NB3-3, and in PS15a. The expression of Abd-B.m in PS13 appeared rather weak and was restricted to the posterior, dorsolateral part of this PS (see also Delorenzi and Bienz, 1990). Neither NB2-4 nor NB3-3 in PS13 revealed expression of Abd-B.m (Fig. 3D). From delamination onwards NB6-4 and NB7-3 of PS14 showed very strong expression of Abd-B.r. However, in PS13 neither NB6-4 nor NB7-3 disclosed any Abd-B.m expression before, or directly after, delamination. The weak Abd-B.m expression in PS13 was still restricted to the posterior, dorsolateral part (Fig. 3E) and did not increase or expand in this PS before St12m (not shown) (Delorenzi and Bienz, 1990).

Next, we analysed Abd-B mutants, which specifically disrupt the function of only one isoform. Abd-B^M3 and Abd-B^M5 both solely disrupt the m function, but possess intact r function (m^-r⁺) (Casanova et al., 1986). In both mutants, we found no obvious differences in the Eg pattern (Fig. 4A; supplementary material Fig. S3C; Fig. 9A) compared to wild type (Fig. 1D). This was expected, considering that Abd-B.m is not expressed posterior to PS13 and that in Abd-B null mutants (m^-r^-) the Eg pattern is only affected in PS14 and PS15. Consequently, in Abd-B^Uab1, a mutant specific for the r isoform (m⁺r^-) (Casanova et al., 1986), we obtained almost the same phenotype as in the Abd-B loss-of-function alleles: significantly enlarged NB3-3 clusters and ectopic NB6-4 neuronal subclones in PS14, as well as ectopic NB7-3 lineages (co-expressing Eg, En and Ey) in PS15 (Fig. 4B, Fig. 9A). Thus, the loss of the Abd-B.r isoform is responsible for all described effects. It has been shown before that, in accordance with the ‘posterior prevalence’ phenomenon, Abd-B.r suppresses Abd-B.m (Casanova et al., 1986). Indeed, we observed Abd-B.m expression in PS14 and PS15 of Abd-B^Uab1 mutants. However, this expression was visible late (St11m) and dorsolaterally, just like in the unaffected PS13. Neither the neuronal NB6-4 clusters in PS14, nor the ectopic NB7-3 clones in PS15 displayed an early ectopic expression of Abd-B.m (Fig. 4C). A broader expression of Abd-B.m in PS14-15 came up in St12m (Fig. 4D), however, at a lower level compared with Abd-B.r expression in wild type (Fig. 3E) (Boulet et al., 1991; Sánchez-Herrero, 1991). Thus, Abd-B.m seems to be expressed too late in PS14 and PS15 of Abd-B^Uab1 mutants to rescue the phenotype that emerges due to the loss of Abd-B.r.

Taken together, the isoform Abd-B.r is responsible for inhibition of NB7-3 formation in PS15, for specification of abdominal NB6-4 in PS14 and for NB3-3-fate in PS14.

As Abd-B.r suppresses the formation of NB7-3 in PS15, we tested its ability to inhibit the formation of this NB in anterior PSs, when expressed ectopically. To drive the expression of UAS-Abd-B.r we used scabrous (sca)-Gal4, which from St8 onwards reveals a broad and strong expression in the NE and in the NBs that delaminate from it. Surprisingly, ectopic Abd-B.r does not remove NB7-3 (0%, n=110 thoracic and abdominal hemi-PSs) (Fig. 9B) or any other Eg-positive cell cluster in more anterior PSs (Fig. 5A). To investigate if the total number of NBs is altered in this situation, we counted Dpn-positive cells in PS4-7. We noticed only a marginal decrease in the number of NBs per hemi-PS (30.67±0.49, n=12; Fig. 5F) compared with wild type (31.25±0.46, n=8; Fig. 5E, Fig. 8G). In a few cases the ectopic expression of Abd-B.r resulted in a homeotic transformation of the thoracic NB6-4 into an abdominal fate, which is accompanied by the loss of its neuronal sublineage (13%, n=36 thoracic hemi-PSs). However, the frequency of this homeotic transformation was strongly increased when we ectopically expressed Abd-B.m (81%, n=48 thoracic hemi-PSs). Furthermore, we observed a reduction in the total number of NBs per hemi-PS (26.65±0.81, n=20; Fig. 5G, Fig. 8G), which clearly exceeds the effect of ectopic Abd-B.r. This includes efficient removal of thoracic and abdominal NB7-3 cell clusters (50%, n=176 thoracic and abdominal hemi-PSs) (Fig. 5B, Fig. 9B). Upon ectopic expression of Abd-B.m together with P35, a known suppressor of apoptosis, the NB7-3 clusters in anterior PSs could not be rescued (42%; Fig. 9B), which confirms that they do not undergo PCD (Fig. 5C). Additionally, we ectopically expressed Abd-B.m by using an embryonic lethal abnormal vision (elav) driver line, which directs expression in all neuronal tissues from St12l on, but is not expressed in the NE. This late expression of Abd-B.m was not able to remove the NB7-3 cells in thorax or abdomen (0%, n=152) (supplementary material Fig. S4; Fig. 9B). These results emphasise that a strong ectopic expression of Abd-B.m in the NE (as achieved by the sca-Gal4 driver) is required to repress the formation of NB7-3 in more anterior PSs, whereas late expression does not lead to a removal of NB7-3 cells (as in PS13 of wild type).

We also tried to rescue the Eg expression phenotype of Abd-B^D18 mutants by applying ectopic Abd-B.m in the mutant situation (using sca-Gal4). In most cases, we were able to remove the neuronal NB6-4 subclone in PS14 (73%, n=11 hemi-PSs) and the ectopic NB7-3 lineage in PS15 (82%, n=11 hemi-PSs) (Fig. 9A). The efficiency of this rescue strongly correlated with the disappearance of NB7-3 cell clusters in more anterior PSs (Fig. 5D). This suggests that Abd-B.m, if expressed early enough, can completely substitute for Abd-B.r function concerning repression of the neuronal NB6-4 sublineage and inhibition of NB formation. In wild type, Abd-B.m cannot adopt this function as its expression starts too late.

Why does ectopic Abd-B.m possess the potency to suppress formation of specific NBs, whereas ectopic Abd-B.r does not? When we compared the ability of the respective constructs to downregulate Ubx expression, we noticed that Abd-B.m was very efficient (Fig. 5I) (Castelli-Gair et al., 1994), whereas Abd-B.r was not (Fig. 5H). Perhaps ectopic Abd-B.r does not mediate a homeotic transformation, because it cannot suppress anterior Hox proteins (see Discussion).

It has been claimed that the ParaHox gene cad determines the character of the most posterior segment. Although it is not a member of the BX-C, ectopic cad expression induces analia development in anterior segments and its loss results in a homeotic transformation into the next anterior segment (Moreno and Morata, 1999). cad reveals maternal and zygotic expression, sharing a common CDS (Mlodzik et al., 1985; Mlodzik and Gehring, 1987). In St12m we observed zygotic Cad expression in the NE of PS15 (Macdonald and Struhl, 1986), where it overlaps with the expression domain of Abd-B.r, but it is also expressed posterior to this (Sánchez-Herrero and Crosby, 1988). Its anterior margin is restricted by the En stripe of PS15, and there is only a minor overlap in the expression of En and Cad (Kuhn et al., 1995) (Fig. 6A). In the VNC we found co-expression of Cad and Abd-B.r in NBs belonging to PS15, although we never observed NBs exclusively displaying Cad expression (supplementary material Fig. S5A).

Considering that Cad is co-expressed with Abd-B.r in NBs of PS15, we wondered whether Cad might also affect NB formation. In zygotic cad null mutants, cad², we observed a significant increase of NBs in PS15 (14.25±1.26, n=4; Fig. 6B) compared with wild type (11.88±0.64, n=8; Fig. 1A,K). Similar to Abd-B^D18, cad² mutants show an ectopic cluster of cells on each side of PS15, co-expressing Eg and En (Fig. 6C, Fig. 9A). By combining cad² with eg-Gal4, UAS-nGFP, we could clearly show that in late stages this cell cluster completely differentiates as a NB7-3 cell lineage, forming its typical contralateral projections (Fig. 6D) (Bossing et al., 1996; Schmid et al., 1999). However, different from Abd-B^D18, NBs in PS14 are not affected in cad² mutants (no ectopic neuronal NB6-4 sublineage or enlarged NB3-3 lineage), as expected from the lack of Cad expression in this PS. The failure to inhibit the formation of NB7-3 in PS15 of cad² mutants seems to be a recessive effect, as we noticed it only in the homozygous situation, whereas Eg expression in the heterozygotes is indistinguishable from wild type (Fig. 9A). These data show that cad is also involved in the suppression of NB formation in PS15.

As both Abd-B and Cad affect NB7-3 formation in PS15, we attempted to clarify their functional relationship.

In Abd-B^D18 (and in Abd-B^Uab1) mutants, Cad is still expressed in the ectopically formed NB7-3 in PS15 (Fig. 7A). Conversely, in cad² mutants we noticed Abd-B expression in cells of the ectopic NB7-3 lineage of PS15 (Fig. 7B), although it seems slightly delayed. Furthermore, neither ectopic Abd-B.r nor ectopic Abd-B.m was able to activate or repress Cad expression in the VNC, when driven by sca-Gal4. Moreover, ectopic cad does not change the expression domain and intensity of Abd-B (not shown). These data suggest that Abd-B and Cad do not regulate each other within the VNC, and that they act independently to suppress NB7-3 formation. In cad²/Abd-B^D18 transheterozygotes (supplementary material Fig. S5B), we observed the same phenotype as in the Abd-B^D18 heterozygous situation (supplementary material Fig. S1; Fig. 9A).

Although it is not able to reduce the expression of Abd-B, ectopic cad strongly downregulates Ubx (Fig. 8B) and Abd-A (supplementary material Fig. S6A), in this respect fulfilling the ‘posterior prevalence’ criterion for Hox genes. Ectopic cad expression (driven by sca-Gal4) was weakly able to inhibit the formation of NB7-3 in anterior PSs (17%, n=126 thoracic and abdominal hemi-PSs) (Fig. 8C, Fig. 9B). In agreement with this, the ectopic NB7-3 lineage in PS15 of Abd-B^D18 mutants was just missing in one case when we ectopically expressed cad (8%, n=13 hemi-PSs) (Fig. 8F, Fig. 9A). Additionally, ectopic cad was also able to transform thoracic NB6-4 into abdominal identity (50%, n=30 thoracic hemi-PSs) (Fig. 8C) and rescued the abdominal NB6-4 fate in PS14 of Abd-B^D18 mutants (Fig. 8F). Furthermore, like Abd-B.m, cad has the potential to inhibit the formation of further NBs (in addition to NB7-3) in more anterior PSs (27.92±0.64, n=13; supplementary material Fig. S6B; Fig. 8G), although with lower efficiency. However, its ability increases significantly when it is ectopically expressed together with Abd-B.m (24.8±0.95, n=20; supplementary material Fig. S6C; Fig. 8G). This is also reflected by repression of NB7-3 in all investigated PSs (100%, n=80 abdominal and thoracic hemi-PSs; Fig. 8D, Fig. 9B). Upon ectopic expression of Abd-B.r together with cad, which reflects their native co-expression in PS15, we observed almost the same efficiency in suppression of NB7-3 (91%, n=96 hemi-PSs; Fig. 8E, Fig. 9B) and of other NBs (25.43±1.50, n=14; supplementary material Fig. S6D; Fig. 8G), as opposed to the inefficient suppression of NB7-3 upon misexpression of Abd-B.r alone.

Taken together, although Abd-B and Cad act independently of each other, their combined expression is needed to unfold the full potential concerning inhibition of NB formation.

During the generation of the Drosophila VNC, Hox genes of the BX-C act along the AP axis during different stages of development at the level of NE, NBs and their progeny cells to control segmental patterning (reviewed by Rogulja-Ortmann and Technau, 2008). In the NE they confer segment-specific intrinsic properties to NBs before their delamination (Prokop and Technau, 1994). For example, abd-A and Abd-B specify NB6-4 to produce the abdominal variant of its lineage (only glial cells) as opposed to its thoracic variant (glia plus neurons), which represents the ground state (no input of homeotic genes) (Berger et al., 2005; this study). Accordingly, ectopic expression of Abd-B.m leads to transformation of thoracic NB6-4 into abdominal fate (loss of neuronal sublineage). Early function of Hox genes at NE and NB level also includes the determination of segment-specific subpopulations of NBs that persist at the end of embryogenesis to become postembryonic NBs and (after a period of quiescence) resume proliferation in the larva (Prokop et al., 1998; Tsuji et al., 2008). During later stages Hox genes have been shown to control the specification of particular cell types by segment-specific gene regulation and integration of temporal cues on the level of NBs and their progeny (Estacio-Gómez et al., 2013; Karlsson et al., 2010).

A further mechanism by which Hox genes modulate segmental pattern in the CNS is through regulation of PCD. At the level of differentiated progeny cells, segment-specific control of PCD occurs as a late function of Hox genes in the embryonic VNC (Miguel-Aliaga and Thor, 2004; Rogulja-Ortmann et al., 2008; Suska et al., 2011). At the level of postembryonic NBs, the end of their proliferation is scheduled by Hox genes via PCD during the third instar larval stage (Bello et al., 2003). Furthermore, a set of sex-specific NBs in the terminal region of the VNC undergoes PCD during late embryonic/early larval stages only in females (Birkholz et al., 2013). Thus, it appeared likely that PCD also contributes to the reduction of the total number of embryonic NBs in terminal neuromeres (PS14-15) compared with more anterior PSs. Indeed, localised PCD has been shown to occur at segment boundaries in the ectoderm of gnathal and terminal abdominal segments (Lohmann et al., 2002; Nassif et al., 1998), which is controlled by Hox genes (Deformed and Abd-B) via activation of pro-apoptotic genes (Hueber et al., 2007; Lohmann et al., 2002; Zhai et al., 2010). However, in terminal neuromeres of Df(3L)H99 mutant embryos (St12e) we did not observe significant differences in numbers of NBs compared to wild type, indicating that PCD does not account for the diminished number of NBs generated in the tail region. Instead, sizes of metameric units in the blastodermal tail anlagen are significantly reduced, as revealed by fate-mapping data (Hartenstein et al., 1985; Juergens, 1987; Technau and Campos-Ortega, 1985). Specification of these units is under the control of a number of genes, which affect the identity of terminal segments, including Abd-B, cad, spalt (salm - FlyBase) and forkhead (Jürgens, 1988; Jürgens and Weigel, 1988; Moreno and Morata, 1999; Sánchez-Herrero et al., 1985). One function of Abd-B and probably cad involves suppression of ventral structures in terminal segments (Kuhn et al., 1995). If this also affects the NE, the degree of its rudimentation may account for the reduced numbers of NBs that can be formed in these segments.

Our data show that Abd-B and cad are required in PS15 to suppress the formation of a specific subset of about three to four NBs per side. This could be a consequence of the contribution of these genes to restrict the size of the NE. In this case they would prevent the establishment of only those parts of the NE, which specifically give rise to these NBs. For example, as NB7-3 in more anterior PSs normally derives from the intermediate (along the DV axis) and En-positive NE, one would expect this NE domain to be suppressed by Abd-B and cad in PS15. Alternatively, this domain may be established, but Abd-B and cad may suppress the formation of NB7-3 within this region. They could do so by regulating the expression of proneural genes, which define proneural cell clusters in the NE, each singling out an individual NB (reviewed by Skeath and Carroll, 1994). Indeed, a genetic screen for Hox downstream targets revealed that ectopic expression of Abd-B results in a downregulation of lethal of scute [l(1)sc] (Hueber et al., 2007), one of the four members of the Achaete-Scute complex (Martín-Bermudo et al., 1991). In wild type, L(1)sc is expressed in the proneural cluster that gives rise to NB7-3 and for a short period in the NB itself (S. R. Myneni, G.M.T. and A.R.-O., unpublished results). However, the ectopic co-expression of cad and Abd-B.r reduces the number of NBs from ∼31 to ∼25 in anterior segments, which is significantly different from the amount observed in wild-type PS15 (∼12). This suggests that additional (so far unidentified) posterior patterning cues are required to suppress the formation of further NBs.

Abd-B.m is expressed in PS13 (and anterior to this) and Abd-B.r in PS14 and PS15, where it overlaps with the expression of Cad. Our data indicate that Abd-B.r is necessary, but not sufficient to mediate terminal fate in the VNC. Upon ectopic expression of Abd-B.r we observed no obvious transformation of an anterior PS into posterior fate, regarding total NB number, inhibition of NB7-3 formation or frequent change of NB6-4 identity. In addition, we noticed no detectable changes in Ubx expression in the VNC compared to wild type, although ectopic Abd-B.r expression was described to result in a downregulation of Ubx transcripts (Kuziora, 1993). As shown before, Abd-B.r cannot be limited to its regulatory function as a repressor of Abd-B.m, as it also possesses a morphogenetic role in activating gene expression (Ali and Bienz, 1991; Thali et al., 1988). However, a recent publication revealed that ectopic Abd-B.r has only a weak morphogenetic capacity in anterior segments, as it very inefficiently downregulates co-factors of anterior Hox genes, competing and antagonising Abd-B function (Rivas et al., 2013). Our data underline these observations and indicate that, in terms of mediating posterior identity in the VNC, Abd-B.r requires the co-expression of cad to unfold its complete morphogenetic property. This is in line with a previous report, suggesting that both Abd-B and cad are necessary to repress the formation of ventral structures in terminal PSs (Kuhn et al., 1995). Accordingly, the ectopic expression of cad or Abd-B.r alone has only minor effects on more anterior PSs (Mlodzik et al., 1990; Rivas et al., 2013; this study). Upon ectopic co-expression in the VNC they seem to contribute complementary functions: cad is able to downregulate anterior Hox genes (Ubx, abd-A), which may be a prerequisite for Abd-B.r to repress NB formation. In contrast, Abd-B.m unifies both functions. Why it is equipped with the potency to suppress NB formation, although it does not exhibit this function in its normal expression domain, remains an open question. As the Abd-B.m and the truncated Abd-B.r protein share the same C-terminal region, it may harbour the capability to suppress the formation of NBs, while the N-terminal domain of Abd-B.m may be necessary to downregulate anterior Hox genes. Thus, cad might compensate for the missing N-terminal domain of Abd-B.r, by downregulating Hox genes. At the protein level, the N-terminal domain of Abd-B.m comprises a huge amount of glutamine runs, known as M or opa repeats (Wharton et al., 1985). A similar asparagine stretch, which might be functionally equivalent, has been identified in cad (Mlodzik et al., 1985).

Drosophila cad is located on chromosome 2L and belongs to the group of dispersed ParaHox genes, resembling ancient paralogues of the ANT-C and BX-C. cad and Abd-B might have evolved by duplication from an ancestral ProtoHox gene cluster and therefore reflect evolutionary sisters (Brooke et al., 1998; reviewed by Garcia-Fernàndez, 2005). Accordingly, cad shares several features with Hox genes of the BX-C: a Hox typical hexapeptide motif N-terminal to the homeodomain and an intron at the same position as Abd-B (Gehring et al., 1994). As expected for paralogous genes, we observed overlapping domains of Cad and Abd-B.r in the VNC. Ectopic cad expression efficiently downregulates Ubx- and Abd-A expression. However, in contrast to previous observations in genital discs (Moreno and Morata, 1999), we found no evidence for a downregulation of Abd-B in the VNC, which would be an exception from the ‘posterior prevalence’ rule.

There exist three cad homologues in mammals: Cdx1, Cdx2 and Cdx4 (Ferrier et al., 2005). In mouse (Mus musculus) they are expressed collinearly and overlap in the caudal embryonic neural tube (Beck et al., 1995; Gamer and Wright, 1993; Meyer and Gruss, 1993). Mutations result in anterior homeotic transformations and reveal functional redundancy in the control of posterior identity (Chawengsaksophak et al., 2004; Chawengsaksophak et al., 1997; Subramanian et al., 1995; van den Akker et al., 2002; van Nes et al., 2006), suggesting a conserved function between Mus musculus and Drosophila.

We thank James Castelli-Gair Hombría and L. S. Shashidhara for helpful comments on the manuscript and Sudha Rani Myneni for sharing unpublished results. We are very grateful to James Castelli-Gair Hombría, Ernesto Sánchez-Herrero, François Karch, Paul Macdonald, Ulrich Schaefer, Mi-Ae Yoo, John Reinitz, Harald Vaessin, Alfonso Martinez-Arias, Ian Duncan, Chris Doe, Uwe Walldorf, Robert Holmgren, Juergen Knoblich, the Bloomington Stock Center, the Developmental Studies Hybridoma Bank (University of Iowa) and FlyBase for providing flies, antibodies, cDNAs and sequence information.