The hedgehog gene product, secreted from engrailed-expressing neuroectoderm, is required for the formation of post-S1 neuroblasts in rows 2, 5 and 6. The hedgehog protein functions not only as a paracrine but also as an autocrine factor and its transient action on the neuroectoderm 1–2 hours (at 18°C) prior to neuroblast delamination is necessary and sufficient to form normal neuroblasts. In contrast to epidermal development, hedgehog expression required for neuroblast formation is regulated by neither engrailed nor wingless. hedgehog and wingless bestow composite positional cues on the neuroectodermal regions for S2-S4 neuroblasts at virtually the same time and, consequently, post-S1 neuroblasts in different rows can acquire different positional values along the anterior-posterior axis. The average number of proneural cells for each of three eagle-positive S4-S5 neuroblasts was found to be 5–9, the same for S1 NBs. As with wingless (Chu-LaGraff et al., Neuron 15, 1041-1051, 1995), huckebein expression in putative proneural regions for certain post-S1 neuro-blasts is under the control of hedgehog. hedgehog and wingless are involved in separate, parallel pathways and loss of either is compensated for by the other in NB 7–3 formation. NBs 6–4 and 7–3, arising from the engrailed domain, were also found to be specified by the differential expression of two homeobox genes, gooseberry-distal and engrailed.
In Drosophila, the central nervous system (CNS) develops through neural precursor or neuroblast (NB) formation, with delamination occurring at five waves, S1-S5 (Doe, 1992). At S5, the final stage of NB formation, each hemisegment contains a subepidermal layer consisting of 30 NBs arranged in a stereotyped spatial pattern. Each NB generates several ganglion mother cells (GMCs), which divide once to produce postmitotic neurons and/or glia.
Examination of cell ablation in grasshopper and in vitro culture in Drosophila (Doe and Goodman, 1985; Huff et al., 1989; Lüer and Technau, 1992) suggested that the unique properties of neurons and glia are due to intrinsic identity rather than circumstantial factors. The identity of neurons is deter-mined by parental GMCs whose fates are controlled by parental NBs. NB identity appears dependent on the location of the neuroectodermal region from which a given NB is derived.
Initially, neuroectoderm is an equipotential two-dimensional sheet on which discrete proneural clusters are formed (reviewed in Doe and Goodman, 1993). Proneural genes including the achaete-scute complex (AS-C) are expressed in stereotyped positions and bestow general potential for following neural fate on the cells expressing them (Skeath et al., 1992, 1994). Within each cluster, only one cell is selected as an NB by lateral specification, while the remaining follow epidermal fate (for review, Campos-ortega, 1993). Embryos lacking neurogenic gene expression show neural hyperplasia whereas proneural gene mutations cause neural hypoplasia (Lehmann et al., 1983; Jimenez and Campos- ortega, 1990).
Pair-rule and segment polarity genes may quite likely provide the neuroectoderm with positional cues along the anterior/posterior axis. Pair-rule genes may regulate the expression of proneural genes for S1 NB development, since achaete (ac) and scute expression is controlled positively and negatively by fushitarazu (ftz) and odd-skipped (Skeath et al., 1992). Patel et al. (1989a) observed severe CNS defects associated with segment polarity gene mutations. For example, patched (ptc) mutants exhibit occasional lack or duplication of a class of NBs. gooseberry-distal (gsb-d) is required for S1, row-5 NB specification (Zhang et al., 1994; Skeath et al., 1995). Chu-LaGraff and Doe (1993) showed wingless (wg) to be essential for the formation and/or specification of NBs adjacent to wg-expressing domains. wg function required for NB development appears distinct from that for epidermal development.
Here, we show that hedgehog (hh), a segment polarity gene coding for a secretory protein (Hh; Mohler and Vani, 1992; Lee et al., 1992; Tabata et al., 1992; Tashiro et al., 1993), is essential for formation of most post-S1 NBs derived from engrailed (en)-expressing neuroectodermal regions and their immediate neighbors. A transient action of paracrine and/or autocrine Hh on the neuroectoderm prior to NB delamination is necessary and sufficient to form normal NBs. In NB formation, Wg and Hh appear involved in separate, parallel pathways and, in particular, in NB 7-3 formation, endogenous Wg and Hh may mutually compensate for the loss of each other. The combined activity of gsb-d and en is also shown necessary for the specifications of NBs 6-4 and 7-3.
MATERIAL AND METHODS
Unless noted otherwise, hhIIO (a strong allele; Mohler, 1988) was used as the hh mutant, while a temperature-sensitive hh allele, hh9K (Mohler, 1988), was used for temperature-shift experiments. hh13C (a presumed null allele; Mohler, 1988) was also used for some experiments. Other mutants examined were wgIL (Chu-LaGraff and Doe, 1993), Df(2R)gsbIIX (a deletion of gsb-d and gooseberry-proximal (gsb-p)), enE (a deletion of en and invected; Tabata et al., 1995) and Df(1)N8 (a deletion of Notch (N)). As molecular markers, enhancer trap lines 5953 (huckebein-lacZ (hkb-lacZ); Doe, 1992), H162 (seven-up-lacZ (svp-lacZ); Mlodzik et al., 1990) and 17en40 (wg-lacZ; Kassis et at., 1992) were used, while K42 is a line having a kinesin-lacZ gene whose expression is regulated by the eagle (eg) enhancer (Higashijima et al., 1996). hhIIO was introduced into the third chromosome of K42 or 5953 by recombination. Embryos homozygous for hhIIO were identified using a balancer, TM3 ftz-lacZ. Embryos homozygous for wgILhhIIO were identified based on the fact that wg hh embryos have no en expression in either the ectoderm or midline cells at S5 (Bejsovec and Wieschaus, 1993).
Immunohistochemisty was carried out as described by Higashijima et al. (1996). Unless otherwise noted, staged embryos raised at 27°C were collected. Developmental stages and stages of NB formation were according to Campos- ortega and Hartenstein (1985), and Doe (1992). NB identity was based on the expression of molecular markers, cell position and morphology (Broadus et al., 1995). Primary antibodies used were: mouse anti-En (Patel et al., 1989b), mAb16F12 (anti-Gsb-d; Zhang et al., 1994), mouse anti-Ac (Skeath and Carroll, 1992), rat anti-RK2 (Repo; Campbell et al., 1994), mAbBP102 (Klämbt et al., 1991), rabbit polyclonal anti-β-gal (Escherichia coli β-galactosidase; Cappel), and mouse mono-clonal anti-β-gal (Promega). To detect Eagle, rabbit polyclonal anti-Eagle antiserum, LU1, was used (Matsuzaki et al., unpublished data). Secondary antibodies used are: AP-conjugated anti-rabbit (Cappel) and anti-mouse (Promega), and biotin-conjugated anti-rabbit, anti-rat and anti-mouse (Vector) antibodies. ABC-HRP kit (Vector) and DAB/NiCl2 were used for signal amplification. Double labeling was carried out by a combination of AP and ABC-HRP reactions, or ABC-HRP reactions with DAB/NiCl2 and those without NiCl2. The substrate for AP was NBT/BCIP and that for HRP was DAB.
Embryos produced by crossing of hhIIO K42/TM3 ftz-lacZ and hh9K/TM3 ftz-lacZ flies were used for temperature shift-up and shift-down experiments along with transient hh activation and inactivation experiments. One hour after egg laying (AEL) at 18°C or 30 minutes AEL at 29°C, eggs were collected onto a wet mesh, which was then put in a Petri dish. Several Petri dishes were incubated simultaneously by floating in a water bath with an appropriate temperature. By transferring dishes to another water bath with different temperature, both incubation time and temperature could be easily regulated. At 14 –16 hours AEL at 18°C or 7 –8 hours AEL at 29°C, embryos were fixed and stained with anti-β-gal antibodies. Embryos lacking lacZ expression regulated by the ftz-promoter were collected and expression of eg-kinesin-lacZ was examined. Cuticlar patterns at 18°C and 29°C, respectively, were examined at 49 and 24.5 hours AEL.
Requirement of hh for post-S1 NB formation in rows 2, 5 and 6
The ventral ganglia consist of three types of segments, two of which, thoracic and abdominal, have been shown to have similar NB patterns (Doe, 1992; Higashijima et al., 1996). Attention in this study was directed to abdominal segments. At S5 (late stage 11), the final stage of NB development, the NB layer of each hemisegment consists of 6 rows (rows 2 –7), each containing 3-6 NBs (Fig. 1F). For simplicity en-negative NB 1-1 and en-positive NB 1-2 were assumed to belong to rows 2 and 7, respectively. NBs in rows 6 and 7 are derivatives of neuroectodermal cells expressing hh and en (Doe, 1992).
Examination was first made of final NB patterns at S5, using svp-lacZ as a marker. Virtually all NBs (about 30) could be identified in the wild-type hemisegment by svp-lacZ-staining (Fig. 1A; Doe, 1992). Fig. 1B shows the hh mutant hemiseg-ment to consist of about 20 NBs and two large gaps (non-NB regions) in rows 2, 5 and 6. Superimposition of wild-type and mutant patterns suggest that about 2/3 of NBs arising normally in rows 2, 5 and 6 are absent from the hh mutant hemisegment without extensive dislocation of the remaining NBs. Since 1/3 of NBs in rows 2, 5 and 6 belong to S1 NBs (see Table 1) and all S1 NBs appear to arise normally in the hh mutant (see below), virtually all and only post-S1 NBs in rows 2, 5 and 6 would be probably abolished in the hh mutant.
Individual NBs abolished in the hh mutant may be identified using various molecular markers of NBs (Doe and Goodman, 1993). Four NBs (eg-NBs), 6-4, 7-3, 2-4 and 3-3, can be specifically marked by eg-kinesin-lacZ (Fig. 2C; Higashijima et al., 1996). eg-NBs are lined up along the anterior/posterior axis and this facilitates clarification of the hh effect on NB formation or specification in different rows. In contrast to the wild type (Fig. 2A,C), no eg-NB corresponding in position to NB 2-4 was recognized in the hh hemisegment at S4 and S5 (early and late stage 11; Fig. 2B,D). Similarly, no eg expression was found at 90% of NB 6-4 positions in the hh hemisegments (Fig. 2C,D). These findings appear consistent with the notion that the hh mutant cannot form NBs 2-4 and 6-4, post-S1 NBs in rows 2 and 6 (see Table 1).
hkb-lacZ is a marker specifically expressed at S4 in three row-2 NBs (2-1, 2-2, 2-4), two row-4 NBs (4-2 and 4-4), and a row-5 NB, 5-4, (Figs 1E, 2E; Doe, 1992; Chu-LaGraff et al., 1995). At S5, three additional hkb-lacZ-positive NBs, 4-3, 5-5 and 7-3, newly delaminate (Figs 1F, 2G). In the hh mutant, no hkb-lacZ expression was detected at positions corresponding to those of row-2 and row-5 post-S1 NBs positive to hkb-lacZ in the wild type (Fig. 2F,H), while no loss of hkb-lacZ-positive NBs in rows 4 and 7 was detected.
In an hh background, the expression of gsb-d, wg-lacZ and en in the NB layer was found to have changed. gsb-d expression at S3 suggested that post-S1 NBs 6-1, 6-2 and 6-4 are absent from about 60, 50, 90% of hhIIO hemisegments, respectively (Fig. 2M,N). Similarly, a comparison of Fig. 2P with Fig. 2O suggests the failure of the formation of three row-5 NBs 5-1, 5-4 and 5-5 arising at S4 or S5. The absence of these row-5 NBs was also demonstrated by wg-lacZ expression at S5 (Fig. 2K,L). In contrast, en expression at S4 shows S1-S2 NBs in row 7 (7-1, 7-2, 7-4 and 1-2) to form normally in the hh mutant (Fig. 2Q).
hh is also required for the development of GP, an S3 glial precursor in row 2 (see Fig. 1D), which divides symmetrically to produce longitudinal glia (LG) and may not be included at S5 in the NB layer (Fig. 1F; Jacobs et al., 1989). reversed polarity (repo) is normally observed in all glial cells (Fig. 2V; Campbell et al., 1994). Thus, altered repo expression at stage 16 may be an indication of the absence of LG in the hh mutant (Fig. 2W). The frequent disruption of longitudinal connectives in the hh mutant may be due in part to the loss of LG (Fig. 2X,Y).
It should be noted here that, as summarized in Table 1, all post-S1 NBs in rows 2, 5 and 6 require hh for formation. In contrast, all S1 NBs formed independently of hh, and exhibited normal gene expression patterns at S1: putative NBs 5-2, 5-3, 5-6 and 7-1 expressed gsb-d, putative NBs 7-1 and 7-4 expressed en, and ac was expressed in putative NBs 7-1, 7-4 and 3-5, and MP2 (data not shown; see Fig. 1C).
Transient Hh action on the neuroectoderm necessary and sufficient to form hh-dependent NBs
To determine the critical period sensitive to hh activity (CPSH) in NB development, temperature shift-up and shift-down experiments were carried out using hh9K, which produces a temperature-sensitive Hh (Mohler, 1988; Porter et al., 1995). Marking with eg-kinesin-lacZ makes it possible to follow the development of two hh-dependent NBs, 6-4 and 2-4, which are S3 and S4 NBs, respectively. At permissive temperature (18°C), hh9K/hhIIO embryos exhibited eg-NB patterns similar to those of the wild type (Fig. 3C). At a non-permissive tem-perature (29°C), NB 2-4 could not be detected in 96% of hh9K/hhIIO hemisegments (n=72), while mg glia, a putative progeny of NB 6-4 (Higashijima et al., 1996), was present in 37%, suggesting the leakiness of hh9K/hhIIO in NB 6-4 formation (Fig. 3F).
Incubation temperature was changed at various times and the numbers of mg glia or NB 2-4 were scored at stages 12 or 13 (Fig. 3A,B). Shift-up experiments (see filled circles) suggested that the Hh activity after 8.5 hours AEL at 18°C (equivalent to 4 hours AEL at 29°C) is dispensable for the formation of NBs 6-4 (Fig. 3A) and 2-4 (Fig. 3B). Shift-down experiments (see open circles) showed that Hh produced during 0-3 hours AEL at 29°C (0-6.5 hours at 18°C) is not essential for the formation of NBs 6-4 and 2-4. It may thus follow that CPSHs for the formation of NBs 6-4 and 2-4 are virtually identical to each other and range from 6.5 to 8.5 hours AEL at 18°C or 3 to 4 hours AEL at 29°C. This was further confirmed by transient Hh inactivation and activation experiments (lower margins of Figs 3A,B,D,E). NB 2-4 and mg glia were found in 80% hemisegments of embryos producing active Hh only during CPSH, while the absence of active Hh from CPSH brought about a considerable reduction of the fraction of hemisegments with NB 2-4 or mg. Thus, it is concluded that Hh produced during 6.5-8.5 hours AEL at 18°C (or 3-4 hours at 29°C) is necessary and virtually sufficient to form NBs 6-4 and 2-4 normally. Since S3 and S4 NBs, respectively, begin to delaminate at 9.5 and 10.5 hours AEL at 18°C (see upper margins of Fig. 3A,B), it would appear that Hh activity is required 1-2 hours (at 18°C) prior to the delamination of NBs 6-4 and 2-4, and hence target cells for Hh are not NBs but neuroectodermal precursors.
hh9K/hhIIO embryos in which Hh was inactivated only during CPSH for NB formation exhibited a cuticular pattern similar to that of the wild type (compare Fig. 3H with Fig. 3G). The cuticular pattern of embryos having active Hh only during the CPSH was indistinguishable from that of hh mutant embryos (Fig. 3I,J). Thus, Hh secreted during the CPSH for NBs 2-4 and 6-4 formation is not required for normal cuticular formation, previously shown to require the hh activity from 2.5 to 7 hours AEL at 25°C (equivalent to 5-14 hours AEL at 18°C; Mohler, 1988). Deduced CPSH for NBs 6-4 and 2-4 is presumed to overlap the earliest but least required part of the period in which hh activity must be available for epidermal development.
Size estimation of proneural regions of eg-NBs and alteration of hkb-lacZ expression in putative proneural regions in the hh mutant
To further clarify Hh functions in NB formation, the size and locations of proneural regions for post-S1 NBs should be deter-mined. However, for most post-S1 NBs, no proneural genes have been identified to date. Thus, we first estimated the numbers of putative proneural cells for three eg-NBs.
All cells in a given proneural region are considered equipotential and hence should become NBs with identical gene expression in the absence of N (Struhl et al., 1993). Since, in three of four eg-NBs (NBs 2-4, 3-3 and 7-3), eg RNA expression is initiated during NB delamination (Higashijima et al., 1996), nearly all N-mutant NBs derived from proneural regions for these three eg-NBs should express eg RNA and, thus, it should be possible to identify them by eg-kinesin-lacZ expression. In NB 6-4, eg RNA is expressed only at the last stage of NB development (Higashijima et al., 1996). As shown in Fig. 2T, approximately 20 eg-positive NBs, making up four aggregates, were found at S5 in the hemisegment of N embryos produced by N/+ parents. In hh N double mutants, two eg-NB clusters, putative derivatives of hh-dependent NBs 6-4 and 2-4 proneural regions, disappeared (Fig. 2U). This made it possible to determine the average number of proneural cells for each of the three eg-NBs (2-4, 7-3 and 3-3) as 5-9, the same for S1 NBs (5-7; Skeath and Carroll, 1992).
In the wild-type background, hkb-lacZ is expressed not only in a particular set of NBs but also in their presumptive proneural regions in the neuroectoderm (Doe, 1992). Consis-tent with this notion, 6-8 ectodermal cells were hkb-lacZ-positive in the presumptive proneural region for NB 2-4 (Fig. 2I). Study was thus made to determine whether hh has any effect on hkb-lacZ expression in the neuroectoderm. At S4, three row-2 NBs, two row-4 NBs, one row-5 NB and their putative proneural regions were hkb-lacZ-positive (Fig. 2E, I). At least in the presumptive proneural region for NB 2-4, hkb RNA expression is initiated at stage 9 just prior to NB delamination (Doe, 1992; Chu-LaGraff et al., 1995), thus suggesting that, in the NB 2-4 proneural region, CPSH occurs consider-ably prior to the hkb RNA expression period. Fig. 2F, J shows that, in the hh mutant, hkb-lacZ expression is completely abolished in neuroectodermal cells and NBs in rows 2 and 5 throughout development, indicating Hh to be requisite for inducing hkb expression in all presumptive proneural cells for hh-dependent NBs.
Regulation of the formation and/or specifications of hh-dependent eg-NBs by wg, en and gsb-d
Chu-LaGraff and Doe (1993) showed that, in wg mutant embryos, post-S1 NBs in rows 4 and 6 cannot form normally. We confirmed this in the eg-NB system (Fig. 2R). In a wg background, no eg-positive NB was detected at any NB 6-4 position while no change in eg expression in NBs 7-3, 2-4 and 3-3 was observed. Since NBs 3-3 and 7-3 were present also in the hh mutant (Fig. 2B, D), hh and wg may not be essential at the same time for the formation of NBs 3-3 and 7-3. Both hh and wg are required for NB 6-4 formation, while only hh is essential for NB 2-4 formation.
Fig. 2S shows the hh wg mutant hemisegment to always contain NB 3-3 but to lack NB 7-3 virtually completely. This suggests that neither wg nor hh is required for NB 3-3 formation while both are involved in NB 7-3 formation in a redundant manner. In NB 7-3 formation, wg and hh appear to function in separate, parallel pathways so that at least one is always essential for NB 7-3 formation. The alternative require-ment of hh and wg for NB 7-3 formation suggests that, in contrast to epidermal development (Ingham and Hidalgo, 1993), hh and wg are not regulated interdependently. The absence of wg results in the almost complete loss of all three row-6 NBs (Chu-LaGraff and Doe, 1993). About half of NBs 6-1 and 6-2 and 5-10% of NB 6-4 cells are still present in hh embryos (see Fig. 2 legend). The partial absence of row-6 NBs from hh mutants suggests that both Hh and Wg signals are required for row-6 NB formation; the loss of Hh activity is, however, partially compensated for by the activity of Wg or other unknown factors. The requirement of hh for NBs 7-3 and 6-4 formation in a wg background may indicate that Hh serves as an autocrine factor in the formation of NBs 7-3 and 6-4, since their neuroectodermal precursors are hh-positive (Fig. 5A).
A homeobox gene gsb-d has been shown essential in S1 NB specification. Loss of gsb-d, expressed in row 5 neuroectoder-mal cells at S1, causes row 5 NBs to be transformed into row 3 NBs in S1 development (Skeath et al., 1995). Since gsb-d is also expressed at stages other than S1 in all post-S1 NBs and neuroectodermal cells in rows 5 and 6 and another homeobox gene, en, is constitutively expressed in all NBs and neuroecto-dermal cells in rows 6 and 7 (see Fig. 1F), examination was made of the effects of gsb-d and en on the development of hh-dependent, post-S1 eg-NBs. The distribution of Eg was examined using anti-Eg antiserum. As shown in Fig. 4B, in the gsb-d mutant, Eg expression in row 6 was apparently altered. No Eg expression was detected at the NB 6-4 position but, instead, 72% of gsb-d mutant hemisegments (n=107) contained a new row-6 NB, similar in properties to NB 7-3 (Fig. 4B). This NB delaminated at S5 with the authentic NB 7-3, expressed both Eg and hkb-lacZ and divided quasi-symmetrically as also noted for NB 7-3 (Fig. 4C-E). Note that NB 6-4 delaminates at S3 and does not express hkb-lacZ. The putative progeny of NB 7-3 (EW and GW neurons; Higashijima et al., 1996) was also duplicated (data not shown). The absence of gsb-d would thus appear to cause the transformation of row 6 neuroectoderm into row 7 neuroectoderm so that NB 7-3 cell duplication can occur.
In the en mutant, Eg expression at NBs 6-4 and 7-3 positions disappeared in 100 and 81% of the hemisegments, respectively (n=89; Fig. 4F). No apparent effect on Eg expression in NBs 2-4 and 3-3 was detected, suggesting that the production of Hh required for NB 2-4 formation is unrelated to en. Eg-positive, row-7 NBs in 19% of en mutant hemisegments are somewhat larger than authentic NB 7-3, and, unlike NB 7-3, divide asymmetrically to generate progeny with no apparent morphological relationship to EW or GW neurons (data not shown). Abnormality in en mutant NB 7-3 lineage has been reported recently by Lundell et al. (1996).
Taken together, these results indicate four segment polarity genes, hh, wg, gsb-d and en to all function in concert to determine the formation and specifications of three hh-dependent eg-NBs (6-4, 7-3 and 2-4). The development of NB 3-3, hh-independent, however, is totally unrelated to any of these segment polarity genes (Fig. 5A).
This study shows hh to be essential for NB development in CNS. Secreted Hh acts on neuroectodermal cells adjacent to and within the hh/en-expressing domain to regulate the formation of eleven post-S1 NBs including a glial precursor. hh is not the sole element required for post-S1 NB formation. The present and other studies (Chu-LaGraff and Doe, 1993) indicate the functions of hh, wg, gsb-d and en to be essential in combination for the formation and specifications of post-S1 NBs. Unlike epidermal development for which interdependent expression of hh, wg and en is essential (Ingham and Hidalgo, 1993), hh expression required for NB formation is controlled by neither wg nor en, and thus the hh activity required for NB formation is likely only that regulated at the earliest stage by pair-rule genes (Lee et al., 1992).
Regulation of post-S1 NB formation by composite positional cues bestowed by Hh and Wg
Chu-LaGraff and Doe (1993) showed Wg to function in a manner similar to Hh in the formation of post-S1 NBs in rows 4 and 6. Interestingly, the critical period sensitive to wg activity for NB 4-2 (an S2 NB) is included in CPSH for eg-positive S3 and S4 NBs (see upper margins of Fig. 3A,B). Thus, Hh and Wg may endow neuroectodermal regions, from which S2-S4 NBs in rows 2 and 4-6 are singled out, with positional cues essential for NB formation at almost the same time (Fig. 5).
We showed all post-S1 NBs in rows 2, 5 and 6 to be hh-dependent. NB 7-3 in row 7 requires hh for formation only in the absence of wg activity. Chu-LaGraff and Doe (1993) showed wg to be essential for the formation of all post-S1 NBs in rows 4 and 6. Thus, as far as the formation of post-S1 NBs other than two row-7 NBs (1-2 and 7-2) is concerned, five different row-dependent combinations of positional cues are given by Hh and Wg (Fig. 5A). Row-4 and row-5 NB formation requires Wg and Hh, respectively. Both Wg and Hh are essential for row-6 NB formation. In NB 7-3 formation, Hh and Wg are functionally redundant to each other. Row-2 NB formation requires only Hh. Neither Hh nor Wg are required for row-3 NB formation. At present, we do not know the effects of hh on NBs 1-2 and 7-2 formation in the wg mutant. Thus, positional cues controlling NB formation along the anterior/posterior axis are quite likely to be provided mainly by hh and wg. S1 NB formation is totally independent of hh and wg, consistent with the notion that AS-C expression in S1 NBs is regulated by pair-rule genes but not by segment polarity genes (Doe and Goodman, 1993)
Possible function of Hh upstream of proneural genes and neurogenic genes
AS-C genes are essential for proneural fate determination in many nervous systems in Drosophila. However, in CNS, AS-C serves as proneural genes only for 25% of NBs, most being S1 NBs (Skeath et al., 1992). Although the present results suggest that each post-S1 NB is derived from its own proneural region, similar in size to those for S1 NBs (see Fig. 2I, T), no proneural genes for most post-S1 NBs have been reported. Thus, at present, whether hh functions upstream, in parallel or downstream of putative proneural genes for post-S1 NBs remains unclear. However, in the case of S1 NB formation, AS-C expression occurs in a 1-1.5 hour period (at 18°C) just before delamination (see Fig. 5B; Skeath and Carroll, 1992) and this suggests that, in S3 and S4 NB formation, CPSH occurs prior to proneural gene expression and, hence, Hh functions upstream of proneural genes (see Fig. 5B).
Hartenstein et al. (1994) noted the expression of hsp-Notch (intra), a hsp-promoter-driven gain-of-function form of N, during 4-5 hours AEL at 25°C to prevent the segregation of most post-S2 NBs. Considering the lag time required for effective translation, the expression period of Notch (intra) may correspond to 9-11 hours AEL at 18°C, during which S3 and S4 NBs delaminate. As schematically shown in Fig. 5B, a period of N-dependency may follow CPSH.
At least in NB 2-4, the expression of hkb and eg, respec-tively, occurs shortly before and concomitant with NB delam-ination (see Fig. 5B). These genes are implicated in axon pathfindings in certain progeny neurons (Chu-LaGraff et al., 1995; Higashijima et al., 1996). Our results indicate hkb and eg to be situated downstream of the Hh pathway.
Control of post-S1-NB formation and specifications by concerted action of four segment polarity genes, hh, wg, gsb-d and en
gsb-d and en may not be involved in proneural fate acquisition, since, unlike hh and wg mutants, no appreciable gap regions in the NB layer were found in gsb-d or en mutants (unpublished data). Analysis of eg-NBs rather suggests that the absence of gsb-d in row 6 causes row-6 to be transformed into row-7 (see Fig. 4B), and this appears consistent with the finding that paired serotonergic neurons, putative derivatives of NB 7-3, are doubled in gsb-d mutant embryos (Patel et al., 1989; Lundell et al., 1996). At S1 stage, row-5 NBs in the gsb-d mutant are transformed into row-3 NBs, whereas ubiquitous gsb-d expression generates the opposite transformation (Skeath et al., 1995). en may be essential for the acquisition of row-6 and row-7 eg-NB identity, since the absence of en resulted in 100 and 81% loss of eg-NBs in rows 6 and 7, respectively (see Fig. 4F). 19% of NBs at NB 7-3 positions, still capable of expressing eg in the absence of en, were shown not to be of the NB 7-3 type (unpublished data). These NBs may possibly be relatives of eg-positive NB 2-4, in consideration of their locations and asymmetry in cell division. Thus, the identity of four hh-dependent neuroectodermal rows (rows 5-7 and 2) may be controlled through the concerted action of two homeobox genes en and gsb-d .
Due to the loss of NBs, no NB-fate alteration occurred in the hh mutant. But, hh may also be involved in NB specification, since (1) hh is essential for putative proneural cells to express hkb and acquire ability for eg expression on delami-nation and (2) wg has been shown to be involved in specification of NB 4-2 (Chu-LaGraff and Doe, 1993).
Unlinked transcriptional regulation of hh, wg, en and gsb-d
In contrast to epidermal development, hh activity required for NB formation is unrelated to either en or wg, since an hh-dependent NB 2-4 is normally produced in en and wg mutants (Figs 4F, 2R). However, Hh and Wg pathways required for NB formation may have in common with the late Hh/Wg system certain components required for epidermal development, since our preliminary experiments indicated that, as in epidermal development (Forbes et al., 1993; Siegfried et al., 1994), the hh mutant phenotype (loss of NB 2-4) is suppressed by an additional patched mutation and the phenotypes of porcupine and dishevelled are identical to that of the wg mutant (loss of NB 6-4).
We thank T. Kojima and S. Higashijima for discussion and technical advises, and R. A. Hargis for critical reading the manuscript. We also thank C. Q. Doe, Y. Hiromi, J. Kassis, J. Mohler, T. Tabata, C. S. Goodman, R. Holmgren, H. Okano, A. Tomlinson and S. B. Carroll for providing us fly strains and/or antibodies. This work was supported in part by grants from Ministry of Science, Culture and Education of Japan to K. S.