A genetic cascade involving klumpfuss, nab and castor specifies the abdominal leucokinergic neurons in the Drosophila CNS

During development of the central nervous systems (CNS) of vertebrates and invertebrates, limited numbers of progenitor cells give rise to vast numbers of differentiated neurons. Each neuron acquires a specific identity, manifested in particular by the position of its soma, its axonal projection and neurotransmitter type. Many of these characteristic traits are acquired during development, but how multiple transcription regulators cooperate to specify a given cell fate is poorly understood (reviewed by Jessell, 2000; Shirasaki and Pfaff, 2002). A central goal of neuroscience is to explain the development of the nervous system in terms of gene functions and genetic mechanisms. Because of the relative simplicity of its CNS, Drosophila is an excellent model system for the study of CNS development.

The Drosophila CNS can be subdivided into the brain and ventral nerve cord (VNC). The VNC arises from the neuroectoderm, a sheet of cells located ventrolaterally on both sides of the embryo (for a review, see Skeath and Thor, 2003). Two sets of patterning genes expressed along the anterior-posterior and dorsal-ventral axes divide the neuroectoderm into a Cartesian grid of neural equivalence groups. From these cell groups, individual cells delaminate to become the progenitor cells of the CNS, the neuroblasts (NBs). The delamination process occurs in five sequential waves and results in the formation of an invariant pattern of 30 NBs per hemisegment (Campos-Ortega and Hartenstein, 1985), with each NB acquiring a unique fate based on its position in the grid (Bossing et al., 1996; Schmid et al., 1999; Schmidt et al., 1997). After their formation, NBs undergo a stereotyped and invariant pattern of asymmetric cell divisions that maintains the NB and produces a secondary cell, the ganglion mother cell (GMC). Typically, each GMC divides once more to produce two postmitotic cells that differentiate as neurons or glia. The number of GMCs generated by each NB and the fate of the postmitotic cells are NB specific; thus, each NB creates a unique and reproducible lineage (Udolph et al., 1993).

Ablation experiments in the grasshopper have demonstrated that the fates of each GMC and its daughter cells are determined by their birth order from the parental NB (for a review, see Pearson and Doe, 2004). Recently, part of the underlying genetic mechanism controlling such temporal transitions in progenitor competence has been revealed (Brody and Odenwald, 2000; Cleary and Doe, 2006; Grosskortenhaus et al., 2005; Grosskortenhaus et al., 2006; Isshiki et al., 2001; Kambadur et al., 1998; Novotny et al., 2002; Pearson and Doe, 2003; Tran and Doe, 2008; Tsuji et al., 2008). These studies have identified the temporal identity genes, which are sequentially expressed in the NB and act to specify temporal windows in which the different GMCs are produced. Although NB and GMC determinants, such as spatial and temporal regulators, are important for establishing neuronal identity, it is unlikely that they directly control the expression of the unique characteristics of postmitotic neurons. The functions of many other genes that are specifically expressed in subsets of NBs, GMCs and neurons are unknown, and there are very few cases in which the progenitor NB of a neuron, and the genetic mechanisms underlying its specification, are known. In addition, the lack of availability of terminal molecular markers with restricted expression makes it difficult to unequivocally identify specific subsets of neurons.

The Drosophila embryonic/larval VNC contains ~10,000 cells (Schmid et al., 1999). Of these, ~150 are peptidergic, as defined by the expression of one of ~30 identified neuropeptides (Park et al., 2008). Because of their highly restricted expression, neuropeptides have emerged as particularly useful markers for addressing the molecular genetic mechanisms controlling neuronal subtype specification in both vertebrates and invertebrates (Brohl et al., 2008; Thor, 2008). In particular, studies of one subgroup of peptidergic neurons, the thoracic Apterous neurons, have resulted in a detailed understanding of how spatial and temporal cues are translated into combinatorial regulatory codes of cell fate determinants (Allan et al., 2005; Allan et al., 2003; Baumgardt et al., 2009; Baumgardt et al., 2007; Miguel-Aliaga et al., 2004). To understand whether or not such genetic cascades are commonly used during neuronal subtype specification, we have initiated studies of another peptidergic cell type: the Drosophila abdominal leucokinergic neurons.

The abdominal leucokinergic neurons (ABLKs) belong to the set of neurons that synthesize the neuropeptide Leucokinin (Lk). Lk is widespread among invertebrates and is known to stimulate fluid secretion by Malpighian tubules, a type of excretory and osmoregulatory system found in arthropods (Hewes and Taghert, 2001; Nassel, 2002; Terhzaz et al., 1999). More recently, it has been shown that the Lk pathway is also involved in meal size regulation (Al-Anzi et al., 2010). The larval leucokinergic system is formed by four different sets of neurons, which are easily identified in the CNS of first instar larvae by immunostaining with an anti-Lk antibody: anterior LK (ALK), lateral-horn LK (LHLK), subesophageal LK (SELK) or abdominal LK (ABLK) (Fig. 1A-B) (Cantera and Nassel, 1992; Herrero et al., 2003). We have focused on the ABLKs, which form a group of 14 cells, one per hemineuromere, in the abdominal ganglia (A1-7). These neurons are detected by immunostaining with specific anti-Lk antibodies during late embryonic development, from stage 18 and onward during larval and pupal development, as well as in adult flies. In this report, we describe our initial approach aimed at identifying the genetic mechanisms involved in specification of the ABLKs. First, we identify the NB progenitor of ABLKs. Second, we find that the ABLKs are generated within a castor (cas)/grainyhead (grh) temporal window, and show that cas is crucial for the specification of ABLKs. Third, we find that the genes nab and klumpfuss (klu) act both upstream and downstream of cas, suggesting that cas has two temporal windows and that they both play important roles in the ABLK specification process. We also find that Nab functions to prevent the repression of the ABLK fate by the Sqz transcription factor. Finally, we find that the ABLK sibling cell probably dies by apoptosis and that the Notch pathway is required to prevent the ABLK from similarly undergoing apoptosis. These findings help set the stage for in-depth comparative analyses of cell fate specification within the Drosophila CNS.

We used the following alleles to analyze wild-type and mutant phenotypes: Df(2L)ED773 (pdm⁻), Df(3L)H99, cas^D1, cas^D4, Chip^e5.5, Chip^9.6, crol⁰⁴⁴¹⁸, da¹, dimm^rev4, dimm^rev8, dve^01D01W-L186, el^3.3.1 noc^D64, esg^35Ce-1, grh³⁷⁷⁶, grn⁷², hb^P1 hb^FB (this genetic combination removes Hb CNS expression, but, as the larvae do not survive to stage 18, we use UAS-hbRNAi in combination with UAS-dicer), htl^AB42, jumu^11.683, klu^212IR51C, kr¹ kr^CD (to remove Kr CNS expression), lz⁸¹⁵, mam^GA345, mld⁹², mld⁴⁷, nab^SH143, nab^R52, numb¹, pnr^v1, pnr^vx6, rn²⁰, spdo^G104, Df(3R)Dl-KX23 (referred to as sqz^Df), sqz^ie, sqz^Gal4, stc⁰⁵⁴⁴¹, tup¹, vg^83b27r, vn¹⁰⁵⁶⁷, wor¹, zfh1⁰⁰⁸⁶⁵.

lacZ lines: eve-lacZ, gsb⁰¹¹⁵⁵-lacZ, hkb⁵⁹⁵³-lacZ, klu⁰⁹⁰³⁶-lacZ, lbe^K-lacZ [this transgenic line contains a 2 kb genomic fragment of the regulatory region of lbe driving lacZ, and reproduces the pattern of expression of lbe (De Graeve et al., 2004)], mirr-lacZ, unpg^r37-lacZ, wg-lacZ.

Gal4 lines: elav-Gal4, en-Gal4, wg^MD758-Gal4 (gift from M. Calleja,, Centro de Biología Molecular, Madrid, Spain), ins-Gal4, wor-Gal4.

UAS lines: UAS-dicer on chromosomes II and III, UAS-cas, UAS-grh^15M, UAS-hb^F4A, UAS-klu, UAS-Nintra, UAS-nab, UAS-p35, UAS-sgg^CA, UAS-hbRNAi (Vienna Drosophila RNAi Center #44892), UAS-osaRNAi, and UAS-zfh2RNAi (http://flybase.org/).

We used the following antibodies at the dilutions indicated: rabbit anti-Lk (1:50; gift from D. Nässel, Stockholm University, Stockholm, Sweden); mouse anti-β-galactosidase (1:2000; Promega); rat anti-Gsb-d (1:3; gift from R. Holmgren, Northwestern University, Evanston, IL, USA); guinea pig anti-Cas (1:500) and guinea pig anti-Hb (1:200) (gifts from T. Isshiki, National Institute of Genetics, Mishima, Japan); rabbit anti-Dpn (1:500; Bier et al., 1992); mouse anti-En (1:50; Developmental Studies Hybridoma Bank #4D9); rabbit anti-Nab (1:1000) and rabbit anti-Pdm1 (1:1000) (Terriente et al., 2007); and rabbit anti-Runt (1:500; gift from A. Brand, University of Cambridge, Cambridge, UK).

Embryos were staged according to Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein, 1985). Lk expression is first detected in ventral ganglia at embryonic stage 18, which corresponds to late stage 17 (from 18 hours after egg laying until hatching); at this stage the mouth hooks are prominent and the trachea are air filled. For antibody staining, the CNS of stage 18 embryos or early first instar larvae were dissected in PBT (PBS with 0.3% Tween20), fixed for 20 minutes in 4% formaldehyde (Polysciences #04018) and processed with antibodies in BBT-250 (PBT with 0.1% BSA and 250 mM NaCl). Slides were mounted in Vectashield. Embryos from earlier stages were stained as whole-mount preparations using the same protocol.

Embryos of genotype y w UAS-FLP¹²²; Act5C >y⁺>lacZ/wg-Gal4; tub-Gal80^ts/+ were collected for 2 hours at 25°C and shifted to 17°C. At 7 hours of embryonic development they were moved to 30°C for 2 hours, then shifted back to 17°C, and, as third instar larvae, dissected and stained with anti-Lk and anti-β-galactosidase. Cells expressing wg-Gal4 between 7 and 9 hours of embryonic development activate flp-mediated recombination of the Act5C>y⁺>lacZ cassette and they and all their progeny permanently express β-galactosidase (Act5C>lacZ) (Struhl and Basler, 1993).

We first wanted to identify the progenitor NB that gives rise to the ABLK neurons. So far, none of the lineages that generate the leucokinergic neurons has been identified, and their locations in the CNS suggest that they arise from different progenitors. Work over the last two decades has yielded a detailed lineage map of most of the 30 NBs generated in each hemisegment, providing a set of genetic markers that permit unambiguous identification of the different NBs in the embryonic ventral ganglia from stage eight to 11 (Bossing et al., 1996; Doe, 1992; Prokop and Technau, 1991; Schmid et al., 1999; Schmidt et al., 1997).

Based on morphological data, it has been suggested that ABLKs arise from NB2-4 (Schmid et al., 1999). Nevertheless, we consider this conclusion mistaken because we found that ABLKs did not express genetic markers such as mirror (mirr) and Even-skipped (Eve) found in NB2-4 (Fig. 2A-B). Instead, we found that ABLKs expressed gooseberry (gsb) (a neuroblast row five and six marker) but did not express engrailed (en) (a row six and seven marker), suggesting a row five origin. wingless (wg) is expressed in row five NBs, so to confirm that the ABLK progenitor NB belongs to row five, we traced the lineage of the cells expressing wg at late stage 11 (see Materials and methods). To this end we induced early permanent expression of β-galactosidase in all row five cells, including NB5-5, and thereby labeled all of their progeny. These experiments confirmed that the ABLK was born from NB row five (Fig. 2C-E,N).

Given the limited migration of neurons within the VNC and the lateral position of ABLKs within row five, this pointed to NB5-4, 5-5 or 5-6 as the likely progenitor NB. To distinguish between these possibilities, we used markers specific for different row five NBs. We found that ABLKs did not express ladybird early (lbe), a specific marker for NB5-6 and its progeny (Fig. 2F) (De Graeve et al., 2004), but did express huckebein (hkb), a marker for NB5-4 and 5-5 (Fig. 2G) (Bossing et al., 1996; Chu-LaGraff et al., 1995), as well as unplugged (unpg), which was expressed in NB5-5 but not in NB5-4 (Fig. 2H) (Doe, 1992).

Since Lk expression is first detected in ABLKs during late embryonic development (from stage 18 onward), the present coexpression analysis was carried out in the ventral ganglia of first instar larvae (L1). Although it has been reported that expression of gsb is lineage specific (Buenzow and Holmgren, 1995), it is not known if expression of the other genetic markers used to identify NBs at stage 11 changes late in embryogenesis. In the absence of markers that allow us to detect ABLKs at early times, we used unpg expression to trace the development of the ABLK progenitor NB. We followed the expression of unpg-lacZ in NB5-5 from its initial activation at stage 11 to stage 18, when we were first able to detect Lk expression in the ABLKs (Fig. 2I-M). Using this marker, NB5-5 was first detected at stage 11, and the expanding cluster of cells expressing unpg-lacZ could be followed into L1, when one of the cells coexpresses Lk. From these results we conclude that the ABLKs are generated by NB5-5 (Fig. 2O).

NB5-5 is one of the NBs that delaminates later in embryogenesis, at late stage 11 (Doe, 1992). A previous analysis of NB lineages by in vivo DiI (a lipophilic fluorescent tracer) labeling reported a single abdominal NB5-5 clone, of eight to 11 cells (Schmid et al., 1999). This clone included local interneurons and neurosecretory cells extending along a branch of the segmental nerve (SNa). This previous study described the axonal projections of these neurons at stages 15-17, but as Lk expression commences in embryonic stage 18, it is not surprising that none of the axonal projections described coincided with the axonal projection of the ABLKs observed in larval stages. The ABLKs further belong to a lateral cluster of four dimm-expressing cells referred to as the peptidergic lateral cluster (PLC) (Miguel-Aliaga et al., 2004; Park et al., 2008). The Lk cells extend along the SNa, project across muscle eight, then follow the intersegmental nerve (ISN) out and terminate on the alary muscles (Cantera and Nassel, 1992; Landgraf and Thor, 2005).

The fates of the different GMCs and neurons generated by each NB are in part controlled by the sequential expression of temporal identity genes (Brody and Odenwald, 2000; Cleary and Doe, 2006; Grosskortenhaus et al., 2005; Grosskortenhaus et al., 2006; Isshiki et al., 2001; Kambadur et al., 1998; Novotny et al., 2002; Pearson and Doe, 2003; Tran and Doe, 2008; Tsuji et al., 2008). The transcription factors Hunchback (Hb), Krüppel (Kr), Nubbin (Nub) and Pdm2 (henceforth Pdm), Castor (Cas) and Grainyhead (Grh) are transiently expressed in a temporal sequence in NBs, and their expression is correlated with the production of distinct cell types. GMCs are born intermittently as the NBs progress through the sequence of temporal genes, and the GMCs and their neuronal progeny initially maintain the gene expression profile of the NB when the GMC was born (Isshiki et al., 2001).

Although all NBs are believed to progress through the same sequence of temporal factors, it is unclear whether late-born NBs, such as NB5-5, actually express early temporal genes. To determine the sequence of expression of temporal factors in NB5-5, we used Deadpan (Dpn), an NB-specific marker, and gsb-lacZ, which labels NBs in rows five and six. In line with NB5-5 being late-born, we did not observe expression of the early temporal genes Hb or Kr in the newborn NB5-5 at late stage 11. It initially expressed Pdm, but this expression was rapidly lost in early stage 12, when Cas expression started (Fig. 3A-E); at this stage we did not observe Grh expression. Finally, we observed coexpression of Cas and Grh at stage 13 (Fig. 3F). Thus, as might be anticipated from its late birth, NB5-5 expresses a truncated sequence of temporal genes, initiating at Pdm, i.e. downstream of Hb and Kr.

To identify the temporal window in which the ABLKs are born, we followed the expression of Lk in the CNS of first instar larvae in which expression of Hb (wor-Gal4>UAS-hbRNAi UAS-dicer), Kr (kr¹ kr^CD), Pdm [Df(2L)ED773], Cas (cas^D1) or Grh (grh³⁷⁶) was compromised. In agreement with our observation that NB5-5 does not express early temporal genes, we observed little or no change in the number of ABLKs in larvae in which the expression of hb was knocked down, or in Kr mutant larvae (Fig. 4A-C; Table 1). Moreover, in spite of the fact that NB5-5 expresses Pdm, we found no effect upon Lk in pdm mutants (Fig. 4D). By contrast, we observed a complete absence of ABLKs in cas mutants and a pronounced reduction in grh mutants (Fig. 4E,F; Table 1). These results suggest that ABLKs are not generated in the Pdm window but rather in the Cas/Grh temporal window (Fig. 4L). In agreement with this, we found that Cas is expressed in the ABLKs (Fig. 4K), although we did not observe expression of Grh in the ABLKs in first instar larvae (data not shown).

To test whether the cas and grh temporal genes might be sufficient to trigger ectopic Lk expression, we analyzed the effect of misexpressing cas and grh with a pan-neural driver elav-Gal4. We found that cas misexpression significantly increased the number of ABLKs (Fig. 4G; Table 1). These extra ABLKs appeared in proximity to the normal ABLK cells. By contrast, grh misexpression did not affect the number of ABLKs (Fig. 4H). The basic helix-loop-helix (bHLH) protein Dimmed (Dimm) controls neuroendocrine cell differentiation (Allan et al., 2005; Hamanaka et al., 2010; Hewes et al., 2003), is expressed in the ABLKs (Park et al., 2008), and in dimm mutants the Lk expression in the ABLKs is downregulated (Table 1) (Hewes et al., 2003). Similar to previous studies (Baumgardt et al., 2007), misexpression of dimm alone had no effect upon the number of ABLKs (Table 1). However, recent findings indicate that dimm can act potently with grh to trigger ectopic neuropeptide expression (Baumgardt et al., 2009), and, similarly, we find here that co-misexpression of grh and dimm significantly increased the number of ABLKs (elav-Gal4>UAS-grh UAS-dimm; Fig. 4I; Table 1). By contrast, co-misexpression of cas and dimm did not increase the phenotype observed, misexpressing only cas (not shown). We also co-misexpressed cas and grh (elav-Gal4>UAS-cas UAS-grh) and observed that the number of ABLKs does not increase more than when misexpressing only cas (Fig. 4J; Table 1); considering that Cas activates grh, this result is not surprising (Baumgardt et al., 2009). However, when we misexpressed cas in grh mutants (elav-Gal4>UAS-cas grh^IM), strikingly, we observed extra ABLKs (Table 1; data not shown) (see Discussion).

The GMC is programmed to divide only once and generate two invariant cells that will differentiate as neurons or glia, or undergo apoptosis (reviewed by Karcavich, 2005). To determine the fate of the ABLK sibling cell, we labeled Lk expression in ganglia in which programmed cell death (PCD) was repressed by the overexpression of the baculovirus caspase inhibitor p35 (elav-Gal4>UAS-p35) (Hay et al., 1994). We observed that the ABLK neuron was duplicated in many hemisegments (Fig. 5A; Table 1). A stronger result was obtained in individuals homozygous for the deficiency Df(3L)H99; deletion of this region completely abolishes PCD (White et al., 1994) (Fig. 5B; Table 1). These results indicate that the ABLK sibling cell dies by apoptosis but forms an ABLK if apoptosis is inhibited.

It has been shown in several lineages that Notch signaling between the two GMC daughter cells is required for them to assume different cell fates (Lundell et al., 2003; Schuldt and Brand, 1999; Skeath and Doe, 1998; Spana and Doe, 1996). It has also been reported that Notch signaling drives PCD in postmitotic cells in a lineage-specific manner (Karcavich and Doe, 2005; Lundell et al., 2003; Novotny et al., 2002). To study the effect of removing Notch on ABLK fate we looked at the expression of Lk in sanpodo (spdo) mutants. Spdo is required for Notch signaling during asymmetric cell division but permits Notch signaling during early neurogenesis (Babaoglan et al., 2009; Dye et al., 1998; O'Connor-Giles and Skeath, 2003; Skeath and Doe, 1998). In spdo^G104 mutant ganglia, the number of ABLKs was strongly reduced (Fig. 5C; Table 1). The same result was obtained in mastermind (mam) mutant embryos (Fig. 5D). Mam is a transcriptional co-factor that interacts with the Notch intracellular domain (Petcherski and Kimble, 2000a; Petcherski and Kimble, 2000b). We next overexpressed a constitutively active form of Notch (elav-Gal4>UAS-Nintra) (Rebay et al., 1993; Struhl et al., 1993) and observed an increased number of ABLKs (Fig. 5E). The same result was obtained in numb mutant embryos (Fig. 5F; Table 1).

With respect to these results, we consider two alternative scenarios: one is that Notch signaling is required to activate ABLK neuronal fate which then represses apoptosis; the other, that Notch signaling does not play an instructive role in ABLK specification, but directly represses PCD in one of the two cells that have the potential to acquire the ABLK fate. Our results support the second hypothesis as p35-overexpressing embryos contained extra ABLK. To test this notion, we analyzed the expression of Lk in embryos in which Notch signaling was compromised and PCD was simultaneously repressed (mam^GA345 elav-Gal4>UAS-p35), and observed extra ABLKs (Fig. 5G; Table 1). This result suggests that (1) once PCD is suppressed, Notch signaling has no function and (2) that the specification of the ABLK fate does not require Notch signaling. Thus, we conclude that both of the postmitotic sibling cells have the potential to be ABLKs, but are fated to die unless Notch signaling is activated and rescues one of them (Fig. 5I).

We have observed that cas misexpression duplicates the number of ABLKs (Table 1; χ=14.6), indicating that an additional cell per hemisegment takes on the ABLK fate. The same result was observed in Df(3L)H99 embryos (χ=13.2), indicating in this case that the lethality of the ABLK sibling cell is rescued. When we misexpressed cas in Df(3L)H99 embryos, we expected to obtain four times the number of ABLKs, and, although in some hemineuromers we found this result, as an average their number only tripled (Fig. 5H; χ=19). We consider that these three cells probably correspond to the normal ABLK, its sibling that is rescued by the lack of cell death, and an extra ABLK originated by the cas misexpression that has no sibling cell.

There is increasing evidence that neuronal fate specification is a multistep process involving combinatorial gene expression codes that specify neuronal properties (Allan et al., 2005; Baumgardt et al., 2007; Certel and Thor, 2004; Garces and Thor, 2006). In order to identify genes involved in specification of the ABLK fate, we analyzed the expression of Lk in embryos mutant for genes known to be required for CNS development. We identified a number of mutants in which the pattern of ABLKs was not altered: crooked legs (crol), defective proventriculus (dve), lozenge (lz), osa, rotund (rn), seven up (svp), tailup (tup) and vestigial (vg) (data not shown). There were other mutants in which the number of ABLKs was reduced: chip, daughterless (da), elbow/no oceli (el/noc), escargot (esg), grain (grn), heartless (htl), jumeau (jumu), molting defective (mld), pannier (pnr), shuttle craft (stc), vein (vn), worniu (wor), zinc finger homeodomain 1 (zfh1) and zinc finger homeodomain 2 (zfh2) (Fig. 6A-H; Table 1); it is possible that the maternal effect was the cause of the weak phenotype observed in these mutations. In some mutants ABLKs were completely absent: nab and klumpfuss (klu) (Fig. 7A-B; Table 1). We focused our analysis on these last two genes.

nab encodes a nuclear co-factor without a DNA-binding domain (Svaren et al., 1996), whereas klu encodes a zinc-finger protein (Klein and Campos-Ortega, 1997). Both nab and klu are expressed in ABLKs at stage 18 (Fig. 7C-D); klu is expressed in the newborn NB5-5 in late stage 11 (Yang et al., 1997), but we did not observe expression of Nab in NB5-5 at this stage (data not shown). Overexpression of nab with a pan-neural driver (elav-Gal4>UAS-nab) increased the number of ABLKs (Fig. 7E; Table 1) and these extra ABLKs expressed Cas (Fig. 7F). In cas mutants, there is a complete loss of nab RNA and protein expression, with the exception of Nab midline expression (Baumgardt et al., 2009; Clements et al., 2003). We expressed nab in a cas mutant (elav-G4>UAS-nab cas^D1/cas^D3), but found no evidence of rescue (Fig. 7G). We next overexpressed Cas in a nab mutant (elav-G4>UAS-cas nab^R52) and found ectopic ABLKs (Fig. 7H). To determine whether or not these ectopic ABLKs belong to the NB5-5 lineage, i.e. if Cas overexpression could rescue nab, we analyzed the expression of gsb, lbe and Runt. We observed that, similar to the wild-type ABLKs, all the extra ABLKs expressed gsb and Runt, but did not express lbe (Fig. 7O-P). These results indicate that they are generated by a row five or six NB that is not NB5-6. As all these neurons appear clustered, we conclude that they are most probably generated by NB5-5 and, therefore, that Cas acts either in parallel to, or downstream of, Nab in ABLK specification (see Discussion).

Next, we wondered whether Nab and Klu could crossregulate one another at the level of transcription, but observed that expression of neither gene was altered in embryos mutant for the other (data not shown).

We analyzed the phenotype resulting from klu misexpression (elav-Gal4>UAS-klu) and, surprisingly, found that all the ABLKs were missing (Fig. 7I). Since Klu is involved in PCD during retinal development (Rusconi et al., 2004; Wildonger et al., 2005), we attempted to rescue the phenotype of klu mutants with elav-Gal4>UAS-p35 or Df(3L)H99 and found that the pattern of ABLKs was partially restored (Fig. 7J), indicating that klu overexpression has a deleterious effect.

To determine whether the expression of klu depends on cas, we analyzed the expression of klu in the cas mutant (klu-lacZ cas^D1/cas^D3); klu expression appeared normal in midline cells but failed to spread to NBs (Fig. 7K-L). We next assessed whether cas misexpression rescues the klu phenotype of a lack of ABLKs (elav-Gal4>UAS-casklu^212IR51C) and indeed we observed up to 24 ABLKs per hemisegment (Fig. 7M; χ=21), which suggests that Cas acts either in parallel to, or downstream of, Klu. We also observed that nab overexpression in klu mutant embryos (elavG4>UAS-nab klu^212IR51C) rescued expression of Lk (Fig. 7N; Table 1).

nab plays a role in specification of the FMRFamide-expressing Tv neuron of the dorsal apterous cluster and physically interacts with the zinc-finger transcription factor Squeeze (Sqz) (Baumgardt et al., 2009; Terriente et al., 2007). Since sqz is expressed in the ABLKs (Herrero et al., 2007), we tested whether the ABLK pattern was affected in sqz mutants and in nab sqz double mutants. No phenotype was observed in sqz embryos but, surprisingly, nab sqz showed an almost wild-type pattern of ABLKs (Fig. 7Q) (see below).

To better understand the mechanisms involved in the generation of cellular diversity in the CNS, we performed an analysis of the specification of ABLK neuronal fate. The small number of ABLKs present in the ventral ganglia of first instar larvae is easily identifiable. Although the lack of molecular markers does not permit identification of the complete repertoire of fates generated by the ABLK progenitor NB, we have been able to draw significant conclusions concerning the specification of ABLK neuronal fate.

Recent findings on NB5-6 demonstrate that Cas and Grh act as crucial temporal genes to specify several cell fates at the end of this lineage (Baumgardt et al., 2009). Our findings with NB5-5 reveal similar roles for Cas and Grh, and indicate that the ABLKs are specified in a Cas/Grh temporal window. We have observed that cas mutants generate no ABLKs, that cas misexpression leads to clusters of two to four ABLKs per hemisegment, and that Cas is expressed in all the ABLKs. Thus, our data confirm that cas plays a role as a temporal identity gene, which remains compatible with its proposed role as a switching temporal factor (Tran and Doe, 2008).

The proposed role of grh as a temporal identity gene remains open to question. It has also been reported that it is required to regulate mitotic activity and apoptosis of post-embryonic NBs (Almeida and Bray, 2005; Cenci and Gould, 2005; Maurange et al., 2008). However, recent evidence has emerged indicating that Grh also temporally regulates FMRFamide neuropeptide cell fate and can act in a combinatorial manner with dimm and apterous to trigger ectopic FMRFamide expression (Baumgardt et al., 2009). Similarly, we find here that Grh is required for correct specification of the ABLKs. Together, these results suggest that Grh also plays an instructive role in ABLK specification. Thus, many neuropeptidergic neurons are generated late in several lineages, and depend upon the late temporal genes cas and grh for their specification.

We have shown that NB5-5 does not express the temporal genes hb and Kr, and genetic analysis confirms that these two genes are not required for specification of ABLK fate. We have observed that NB5-5 initially expresses Pdm at the time of delamination in late stage 11. Pdm is downregulated at early stage 12, when Cas is activated, and there is a brief period in which both proteins can be detected. The lack of molecular markers does not permit us to determine whether the Pdm/Cas coexpression stage generates a GMC.

It is of interest to note that the phenotype observed misexpressing cas with NB-specific drivers was very mild (inscuteable-Gal4: χ=8.1, n=18; worniu-Gal4: χ=7.3, n=28) compared with that obtained using a pan-neuronal driver (elav-Gal4: χ=14.6). NB5-5 expresses cas soon after delamination and generates six to nine neurons (Schmid et al., 1999; Schmidt et al., 1997). This suggests that NB5-5 probably has a broad Cas temporal window. Thus, the phenotype obtained upon misexpressing cas with elav-Gal4 indicates that Cas might have a later requirement in postmitotic cells that generates subtemporal windows. Consistent with this interpretation, cas misexpression rescues the grh phenotype of loss of ABLKs, which also suggests that Grh, in addition to being required as a temporal factor, would be indirectly required to activate cas expression in postmitotic cells.

The Notch pathway is involved in many cell fate decisions in neural development (reviewed by Cau and Blader, 2009). Here, we have shown that the ABLK and its sibling are equivalent cells committed to die, and that activation of the Notch pathway in the ABLK prevents its death. A similar situation has been described for specification of the anterior and posterior Corner Cells (CaCC/pCC) neurons in the grasshopper NB1-1 lineage, in which the siblings start as equivalent cells and interaction between them leads to different fates (Kuwada and Goodman, 1985). By contrast, activation of Notch in the NB7-3 lineage drives PCD (Karcavich and Doe, 2005). Here, activation of the Notch pathway, or misexpression of p35 in the sibling cell, is sufficient to generate two ABLK neurons. A systematic analysis of the lineage of apoptotic cells in embryos in which apoptosis is prevented has shown that the lineage of abdominal NB5-5 contains twice the normal number of cells, but that they have wild-type-like axonal projections (Rogulja-Ortmann et al., 2007). These findings are in agreement with ours. We conclude that in this lineage, Notch does not play an instructive role in specifying ABLK neuronal fate, but influences a fate decision by regulating the competence to respond to a program of cell death.

We have identified a set of mutants that produce an altered number of ABLKs. In most cases the effect is very mild. Among the mutants with the strongest phenotypes we highlight jumu, nab and klu. The jumu phenotype was expected because it has been shown that Jumu is required in the NB4-2 lineage for normal segregation of Numb in the asymmetric cell divisions (Cheah et al., 2000). Consistent with this interpretation, the fact that the phenotype of jumu in the NB5-5 lineage is similar to that seen in spdo explains its phenotype and indicates that in jumu embryos Notch is off in both siblings.

nab and klu embryos display a strong reduction in the number of ABLKs, suggesting that both genes have direct roles in ABLK specification. Interestingly, Cas activates the expression of both genes via repression of Pdm. The lack of availability of markers for identifying ABLKs in earlier stages did not permit us to establish whether Nab and Klu are required in the NB or in postmitotic cells.

We did observe that misexpression of cas in nab embryos showed ectopic ABLKs, suggesting that Cas acts either parallel to, or downstream of, Nab. The lack of molecular markers specific to the NB5-5 lineage does not allow us to determine whether all ectopic ABLKs are generated by the NB5-5 or by other lineages. Nevertheless, several results suggest that, most probably, all of them are produced by NB5-5. First, it has been observed that neurons that belong to one lineage form a coherent cluster (Dumstrei et al., 2003), and we find that ectopic ABLKs appear clustered (see Fig. 4G and insets in Fig. 7H,M). Second, all of them express gsb, which labels rows five and six NB, and do not express lbe, an NB5-6-specific marker. Third, ABLKs are the unique cells expressing Lk in the ventral ganglion. However, we have not observed downregulation of cas in nab mutants, and the same has been reported in the better characterized lineages of NB3-3 and NB5-6 (Baumgardt et al., 2009; Tsuji et al., 2008); together, these results indicate that the molecular relationship between Cas and Nab requires a more complex interpretation than a linear genetic cascade.

klu encodes a zinc-finger protein but does not appear to interact directly with Nab (Clements et al., 2003), and we did not find evidence that nab and klu regulate each other. Surprisingly, we have found that nab misexpression rescues the phenotype of a lack of ABLKs observed in klu. By contrast, we have found that cas misexpression produces more ABLKs in grh, nab or klu (21, 25 and 21, respectively) than in wild-type background (14.6). As proposed above, these results suggest that Cas plays a role in postmitotic cells that is crucial for ABLK specification.

We have observed that the sqz phenotype is epistatic over the nab phenotype. Thus, although nab embryos have no ABLKs, sqz and nab sqz show a normal pattern of ABLKs. It has been shown by pull-down assay that Nab physically interacts with Sqz (Terriente et al., 2007), and in vertebrates the Nab homologs act as transcriptional co-factors. Since both genes, sqz and nab, are expressed in the ABLKs, we propose that the function of Sqz in NB5-5 lineage is to repress the ABLK fate. In normal development, as both genes are expressed in the ABLKs, Nab binds to Sqz and blocks its repressor activity; in nab embryos Sqz represses the ABLK fate, but in nab sqz the pattern is wild-type because there is no repression by Sqz. This intimate interplay between Sqz and Nab is also found in the NB 5-6 linage, in which sqz is first required to activate cell fate determinants, and then acts with nab to suppress the same determinants (Baumgardt et al., 2009).

The findings reported here extend our understanding of the mechanisms of ABLK specification. However, more precise analysis of the genes and the mechanisms involved in specification of the different cell fates in the NB5-5 lineage will require additional molecular markers. This would permit us to identify the different neurons generated from this NB and the genes required to specify their various fates.

We are grateful to A. Brand, M. Calleja, J. Castelli-Gair, W. Chia, S. Cohen, C. Doe, R. Hewes, Y. Hiromi, R. Holmgren, T. Isshiki, D. Nässel, W. Odenwall, P. Taghert and G, Technau for flies and reagents and to P. Herrero for comments on the manuscript. This work was supported by grants from the Ministerio de Educación y Ciencia (BFU2005-00116 and BFU2005-00116), the Ministerio de Ciencia e Innovación (BFU2008-04683-C02-01 and CSD2007-00008) and an institutional grant from the Fundación Ramón Areces to the CBM-SO to F.J.D.-B., and by the Swedish Research Council, the Swedish Strategic Research Foundation, the Knut and Alice Wallenberg foundation, the Swedish “Hjärnfonden”, “Cancerfonden” and the Swedish Royal Academy of Sciences, to S.T.