dorsotonals/homothorax, the Drosophila homologue of meis1, interacts with extradenticle in patterning of the embryonic PNS

The embryonic peripheral nervous system (PNS) of Drosophila comprises approximately 600 neurons and 1200 associated cells, organized in a highly stereotyped segment-specific pattern (Campos-Ortega and Hartenstein, 1985; Ghysen et al., 1986). Five distinct patterns of sensory organs are present in different thoracic and abdominal segments of the embryo, seen in segments T1, T2-T3, A1-A7, A8 and A9. Each of the abdominal hemisegments (A1-A7) contains 44 PNS neurons organized in four clusters along the dorsoventral axis: 12 neurons in the dorsal cluster, 12 neurons in the lateral cluster and 20 neurons in the two ventral clusters (Fig. 2A). Three main types of neurons are present in the PNS: external sensory (ES) neurons that innervate mechanoreceptors and chemoreceptors in the cuticle, chordotonal (Ch) neurons that innervate stretch receptors located subepidermally, and multiple dendritic neurons.

Mutations in many genes that affect pattern formation in the embryonic PNS have been identified (for a review see Jan and Jan, 1993 and references therein; Salzberg et al., 1994, 1997; Kania et al., 1995). Phenotypic analysis of such mutations and molecular characterization of some of the corresponding genes shed light on specific aspects of pattern formation in the PNS, mainly the regulation of cell number and cell fate within the PNS lineages (for reviews see Posakony, 1994; Campos-Ortega, 1995; Hassan and Vaessin, 1996; Doe, 1996); however, many questions remained open. For example, what are the mechanisms that govern the localization of neurons to specific positions within a segment? What are the mechanisms that generate the diversity among PNS patterns of different segments?

To address these questions we focused on a specific group of sensory neurons, the five lateral Ch (LCh5) neurons of abdominal segments A1-A7. These neurons and the three dorsal chordotonal neurons of thoracic segment T2-T3 (DCh3) are the only sensory neurons originating in the posterior compartment of the corresponding segments (Hartenstein, 1987). In addition to the difference in their number (three versus five) and position (dorsal versus lateral), the DCh3 and LCh5 neurons differ from each other in their orientation. Whereas the DCh3 neurons are oriented such that their dendrites point ventrally, the LCh5 neurons are oriented such that their dendrites point dorsally (Fig. 2B). The precursors of the LCh5 and DCh3 neurons are born in a similar dorso-lateral position. Later in development the abdominal precursors, or their descendants, are thought to move ventrally and rotate to assume a lateral position and proper orientation (Salzberg et al., 1994). The three Ch neurons of the eighth abdominal segment (LCh3 in A8, Fig. 2A) also originate in a dorso-lateral position similar to that of the LCh5 and DCh3 neurons, but exhibit a dorso-lateral position in the mature PNS and a DCh3/LCh5 intermediate orientation.

The mechanisms controlling the migration of Ch neurons from their dorsal place of origin to a lateral position are not known. However, several assumptions can be made based on phenotypic analysis of various mutations that disrupt this process. First, one may assume that the ‘homeotic identity’ of the segment in which the neurons originate is crucial. The DCh3 neurons that originate in a similar position to LCh5, but in thoracic segments, remain associated with the dorsal PNS cluster and do not migrate. In the absence of abd-A activity, the LCh5 neurons are transformed into DCh3 neurons and remain associated with the dorsal PNS cluster (Heuer and Kaufman, 1992; Salzberg et al., 1994). In cut mutant embryos, in which ES organs are transformed into Ch organs, the ectopic Ch neurons remain in their dorsal position and do not move ventrally (Bodmer et al., 1987; Blochlinger et al., 1988). In embryos that overexpress the cut gene under the regulation of the heat shock promoter (Blochlinger et al., 1991) the LCh5 neurons are transformed into ES neurons and often fail to migrate ventrally, even though they are born in a proper posterior position. Altogether, these observations suggest that the identity of the neuronal cells and the place in which they are born within the segment (anterior versus posterior compartment) play an important role in determining their ability to acquire a lateral position.

Here we describe the cloning and phenotypic analysis of the homothorax (hth) gene, which is allelic to dorsotonals (dtl). Loss-of-function alleles of hth lead to a ‘dorsal chordotonals’ phenotype that is somewhat similar to the abd-A phenotype, but do not affect abd-A expression. This observation suggests that hth may function as a modulator of abd-A activity. The hth gene encodes a homeobox-containing protein of the meis1 family that is required for the nuclear localization of Extradenticle (EXD), a known modulator of homeotic protein activity. The EXD protein itself is required for maintaining normal levels of the HTH protein. Phenotypic analyses revealed that loss-of-function mutations in exd lead to an hth- like phenotype in the PNS. Based on our observations we propose that hth, exd and abd-A interact genetically in patterning of the embryonic PNS.

For excision mutagenesis in the dtl/hth locus, males of the genotype yw; P{lacZ, w+}1422-4/TM6 were crossed to yw; Ki P{ry+, Δ2-3}99B females. Individual P{lacZ, w+}1422-4/Ki P{ry+,Δ2-3}99B male progeny were mated to yw; Ki H/TM6B,D³ virgins. Male progeny with white eyes were tested for complementation with P{lacZ, w+}1422-4/TM6B females. Approximately 150 excisions were isolated. Genomic DNA of the excision strains was analyzed by PCR to determine the molecular nature of the excision event. The primers used were the 31 bp of the P-element inverted repeat 5′-CGACGGGACCACCTTATGTTATTTCATCATG-3′ and two genomic primers located 180 bp upstream and 270 bp downstream of the P{lacZ, w+}1422-4 insertion: 5′-TTGTTGCATTCTGAACAG-AGCGGGGCAG-3′ and 5′-GCGCGCGATTTCCAAAGTGTTT-CTTGTG-3′, respectively. Lethal excision strains that did not contain P-element sequences (based on the PCR data) were further characterized by Southern analysis to identify deletions caused by imprecise excision events. Seven strains harboring such deletions were identified (Table 1).

The mutant homeotic alleles iab-2^D24 and Df(3R)Ubx¹⁰⁹ are described in Lindsley and Zimm (1992) and the Flybase (1997). abd-A^K622 is a hypomorphic allele described in Salzberg et al. (1994). Embryos lacking both maternal and zygotic exd expression were generated using yw; exd^XP11f FRT 18D/OVO^D2FRT 18D; hs-FLP38/+ mothers as described in Rauskolb et al. (1993). exd− embryos were recognized by the abnormal shape of the clypeolabrum. The UAS-hth strain carries a UAS-hth transgene on the second chromosome and is described in Pai et al. (1998).

To determine the lethal phase of the various hth alleles, flies of the balanced mutant strains were crossed out to wild-type flies. Eggs were collected from heterozygous hth/+ parents, and 100-200 eggs of each strain were monitored.

Plasmid rescue from fly genomic DNA was performed as described in Wilson et al. (1989). Northern and Southern analysis, library screening and other molecular procedures were performed using standard techniques (Sambrook et al., 1989). Two genomic libraries were screened: a λDASH II ry⁵⁰⁶/ry⁵⁰⁶ library (kindly provided by Hugo Bellen) and a λCh4A library (Maniatis et al., 1978). For the isolation of embryonic cDNA clones, a 9-to 12-hour embryonic library (Zinn et al., 1988) was screened. Multiple cDNA clones were isolated, four of which were either fully or partially sequenced.

For the preparation of digoxigenin-labeled hth probe, a 0.88 kb XbaI-SalI fragment from hth cDNA was used as a template. The probe was prepared as described in the DNA labeling and detection kits (nonradioactive) from Boehringer Mannheim. In situ hybridization to whole-mount embryos was performed as described by Tautz and Pfeifle (1989).

Rabbit-anti-HTH serum (AS1924) was raised against a bacterially produced HTH protein. The ORF of hth was amplified by PCR and cloned into the BamHI site of the pET15B vector (Novagen) downstream to the 6×His tag (Pai et al., 1998). The His-HTH fusion protein was produced in E. coli strain BL21.DE3 following the pET System Protocol (Novagen). The His-HTH protein was purified on a Ni²⁺-NTA-Agarose column (Qiagen) following the manufacturer’s protocol. The eluted protein was renatured by dialysis into 50 mM Tris-HCl, pH 7.4, 500 mM NaCl, 0.05% Tween-20 and 20% glycerol. The purified protein (100-150 μg) was mixed with Freund’s adjuvant and injected five times, once a month, to New Zealand rabbits. The preimmune and immune sera were tested on embryos and western blots.

The AS1924 serum was used in a 1:1,000 dilution for whole-mount embryo staining, and 1:5,000 for western blot analysis. Other antibodies used for the analysis of embryonic phenotypes are listed below. mAb 22C10 (Goodman et al., 1984) was used in a 1:20 dilution. Monoclonal anti-Cut antibody (Blochlinger et al., 1990) was used in a 1:20 dilution. mAb BP102 (Seeger et al., 1993) was obtained from the Developmental Studies Hybridoma Bank at the University of Iowa and was used in a 1:50 dilution. mAb 21A6 was obtained from Dr. S. Benzer and used in a 1:10 dilution. mAb B11M (anti-EXD) (Aspland and White, 1997) was used in a 1:5 dilution. mAb Dmabd-A.1 (anti-Abd-A) (Kellerman et al., 1990) was used in a 1:1,000 dilution. Anti-Atonal serum (Jarman et al., 1994) was used in a 1:5,000 dilution. Secondary antibodies were biotinylated anti-mouse IgG, or anti-rabbit IgG from Vector Laboratories. Signal detection was carried out with Vecta-Stain Elite ABC-HRP kit (Vector Laboratories). Prior to use, all polyclonal sera were preabsorbed to 0-to 20-hour-old embryos for 2 hours at room temperature, or overnight at 4°C. Embryos were fixed and processed for whole-mount antibody staining using standard techniques (Patel, 1994). Stained embryos were cleared with 70% glycerol, mounted and examined using Nomarski optics on a Zeiss Axioskop. To prepare protein extracts for Western analysis embryos were homogenized in 50 mM Tris-HCl, pH 7.6, 150 mM NaCl and 1% NP-40. Protein were separated by 10% SDS-PAGE and transferred to nitrocellulose membrane. Signal detection was performed with HRP-conjugated goat anti-rabbit IgG (Sigma) using enhanced chemiluminescense (Amersham Corp.).

Two embryonic lethal EMS-induced alleles of dtl (dtl^H321and dtl^J186) were isolated in a mutagenesis screen aimed at identifying genes required for normal pattern formation in the embryonic PNS (Salzberg et al., 1994). Two insertional alleles, dtl^P1422-4 and dtl^P1323-7, were identified subsequently in a similar screen (Salzberg et al., 1997). Both P-element insertions were mapped to cytological position 86C1-4 (Salzberg et al., 1997). The P-elements inserted in dtl^P1422-4 and dtl^P1323-7 are of the P{lacZ, w⁺} type carrying the white gene as a dominant eye color marker. The pigmentation in the eyes of dtl^P1422-4 and dtl^P1323-7 flies is restricted to the posterior region of the eye, suggesting the presence of a genomic silencer in the vicinity of the insertions. This unique eye pigmentation pattern was identified previously in a screen for silencers in the Drosophila genome (posterior insertions P1 and P2; Sun et al., 1995). Complementation tests between lethal posterior insertions and lethal dtl alleles revealed that they belong to the same complementation group. In addition, complementation tests between lethal dtl and hth alleles (Jürgens et al., 1984; hth^5E, obtained from R. Mann) revealed that these mutations are also allelic. Therefore, we refer to all the mutations in this complementation group as hth alleles. Additional mutations in the hth locus were generated by imprecise excisions of the P{lacZ, w⁺}1422-4 insertion. The various hth alleles form an allelic series ranging from mild to very severe (Table 1).

Mutations in hth have pleiotropic effects on embryonic development. Cuticle preparations of loss-of-function alleles of hth reveal defects in segmentation and head involution (Fig. 1A-C). The head skeleton was reduced or absent. Thoracic segments appeared deranged and the thoracic denticle belts were eliminated. Abdominal denticle belts appeared more dispersed and less differentiated than normal and exhibited a weak engrailed-like phenotype. In addition, the width of the denticle belts in anterior abdominal segments appeared reduced and similar to that of the seventh or eighth abdominal segment. The central nervous system (CNS) of hth mutant embryos was examined using mAb BP102 (Seeger et al., 1993). Two distinct phenotypes were evident in the ventral nerve cord (VNC) of mutant embryos: substantial widening of the VNC in thoracic segments and abnormal scaffold of CNS axons throughout the VNC (Fig. 1D,E). As shown in Fig. 1E, the longitudinal pathways were absent or reduced in all thoracic and abdominal segments. The anterior commissure was present, but the posterior commissure was often reduced (arrow in Fig. 1E). The spacing between the anterior and posterior commissures was also reduced, most notably in thoracic segments (white arrowheads in Fig. 1E). Another phenotype observed in the CNS of hth mutant embryos is the outgrowth of multiple nerve roots on each side of the CNS as compared to two nerve roots in wild-type embryos (black arrowheads in Fig. 1E).

In wild-type embryos the LCh5 neurons are located invariably in the lateral PNS cluster of abdominal segments A1-A7 (Fig. 2A,B). In contrast, these neurons were situated in a more dorsal position in 25% or 36% of the abdominal segment in the PNS of embryos homozygous for hth^H321 (n=91) or hth^J186 (n=56) respectively. This phenotype was also observed in more than 30% of abdominal segments in homozygous hth^P1422 or hth^P1323 embryos (Fig. 2C). The affected Ch neurons remained associated with the dorsal PNS cluster, or occasionally, were positioned between the dorsal and lateral PNS clusters. The orientation of the affected neurons was also abnormal. Whenever the affected LCh5 neurons remained associated with the dorsal PNS cluster their dendrites pointed ventrally or posteriorly instead of dorsally (Fig. 2C). The ‘dorsal chordotonals’ phenotype could be detected in all the abdominal segments in varying frequencies. In weak alleles, such as hth^P1422 or hth^4-4, it was observed more frequently in the posterior abdominal segments (A5-A7). Stronger alleles affected all the abdominal segments in similar frequencies. Weak hth alleles did not affect any PNS neurons other than the LCh5 neurons.

Strong hypomorphic mutations in hth, such as hth^K1-8 or hth^64-1, affected not only the position and orientation of the LCh5 neurons, but also caused a reduction in their number. Only three dorsal Ch neurons were observed in nearly 100% of abdominal segments of homozygous hth^K1-8or hth^64-1 embryos (Fig. 2D). Most of the affected neurons remained associated with the dorsal PNS cluster and their dendrites pointed ventrally. In spite of their abnormal location and orientation, the affected Ch neurons appeared fully differentiated, as judged by their overall morphology and the presence of normal-looking scolopales at the tip of their dendrites (Fig. 2E). Based on anti-Atonal staining we concluded that the precursors of the LCh5 neurons are born in a normal dorso-lateral position in hth mutant embryos (data not shown). Due to the lack of appropriate markers we could not detect the ‘dorsal chordotonals’ phenotype in mutant embryos earlier than stage 12. In addition to the ‘dorsal chordotonals’ phenotype, we observed a reduction in the total number of PNS neurons and defects in the axonal pathways in the PNS of strong hypomorphic hth mutants (Fig. 2D). One of the most severe alleles (hth^64-1) caused on average a loss of 3-4 dorsal PNS neurons, 3-4 lateral PNS neurons and 10-12 ventral PNS neurons. The remaining neurons are not organized in a highly stereotyped pattern and it was difficult to asses their identity. However, based on anti-Cut staining we concluded that in the dorsal cluster one dorsal ES neuron and 2-3 Cut-negative MD neurons are lost (data not shown). The ventral Ch neurons were only rarely lost in hth^64-1 mutant embryos (arrowhead in Fig. 2D).

The number, position and orientation of the LCh5 neurons affected by hth mutations resembled the number position and orientation of the thoracic DCh3 neurons, or to lesser extent, the position and orientation of the A8-LCh3 neurons (see Discussion). A similar phenotype was observed in embryos homozygous for mutations in the homeotic selector gene abd-A (Heuer and Kaufman, 1992; Salzberg et al., 1994). In the absence of abd-A activity, the LCh5 neurons were transformed into DCh3 neurons (Heuer and Kaufman, 1992), and as such they remained associated with the dorsal PNS cluster and their dendrites pointed ventrally (Fig. 2F). Only rarely were the transformed neurons observed in a lateral or dorso-lateral position in embryos homozygous for a null allele of abd-A (iab-2^D24, 4-5%, n=105). In embryos homozygous for both the iab-2^D24and the hth^P1422 mutations, the LCh5 neurons exhibited a DCh3-like morphology and number. However, quite surprisingly these neurons were positioned lateraly in 17-19% of the abdominal segments (n=173 or 149 in two independent double mutant strains). This observation suggests that, at least in some aspects, mutations in hth and abd-A have opposite effects on Ch neuron development.

Since the PNS phenotype associated with loss of hth function suggested a homeotic transformation of LCh5 neurons towards the identity of DCh3 or A8-LCh3 neurons, which do not depend on abd-A for their development, we examined the expression pattern of the Abd-A protein in hth mutant embryos. Abd-A is normally expressed in the ectoderm of abdominal segments from PS7 to the anterior region of PS13 (Karch et al., 1990). In addition, Abd-A is expressed in the LCh5 neurons of segments A1-A7 (Fig. 3A) and in the VNC in segments A2-A7 (Karch et al., 1990). The spatial distribution of the Abd-A protein is not altered in the ectoderm or CNS of embryos homozygous for the hth^K1-8 allele as compared to wild-type embryos, although a slight reduction in the level of the protein was observed (Fig. 3). Since the LCh5 neurons are positioned aberrantly in hth^K1-8 mutants, we double-stained these embryos with an antibody against Abd-A and with mAb 21A6, which labels the scolopales of Ch neurons. This double-labeling experiment allowed us to identify unambiguously the misplaced LCh5 neurons and establish whether or not they express Abd-A. We found that the abdominal Ch neurons express the abd-A gene in the absence of hth activity (data not shown). We therefore concluded that hth may be required for the activity of Abd-A rather than its expression.

Mutations in the exd gene were shown to cause homeotic transformations in the embryonic cuticle without altering homeotic gene expression (Peifer and Wieschaus, 1990). The EXD protein affects the morphological consequences of homeotic gene activity by modulating the DNA binding specificity of homeotic proteins (Rauskolb et al., 1993; van Dijk and Murre, 1994; Chan et al., 1994). The involvement of exd in homeotic determination and pattern formation in the PNS was not addressed however, and the PNS pattern of embryos lacking normal exd activity has not been described previously.

To determine whether exd is required for normal pattern formation in the embryonic PNS we examined embryos lacking either zygotic, or maternal, or both zygotic and maternal expression of exd using mAb 22C10 as a marker. Embryos lacking zygotic expression of exd were obtained from attached-X mothers (C(1)DX, y f /Y) and exd^XP11/Dp(1:Y)Y#1 males (Rauskolb et al., 1993). All of the embryos from this cross contained high levels of maternally derived EXD, but one quarter of them lacked zygotic exd expression. Embryos lacking a maternal contribution of exd were derived from exd^XP11/exd^XP11 germline clones (Rauskolb et al., 1993). Half of these embryos received a wild-type copy of exd from their fathers. The other half did not receive a copy of exd from their fathers and thus lacked both maternal and zygotic exd expression.

Embryos lacking only zygotic expression of exd appeared normal when stained with mAb 22C10 and did not exhibit any obvious defects in their PNS. In contrast, embryos with normal zygotic expression, but no maternal contribution of exd, exhibited a ‘dorsal chordotonal’ phenotype in approximately 5% of the examined abdominal segments (n=77; Fig. 2G). Neurons other than LCh5 were not affected. This phenotype appears identical to the PNS phenotype of weak hypomorphic alleles of hth, although the expressivity was lower. Embryos lacking both maternal and zygotic expression of exd exhibited a PNS phenotype very similar to that caused by strong loss-of-function hth alleles (Fig. 2H). In these embryos, the LCh5 neurons were positioned in the dorsal PNS cluster in more than 65% of the abdominal segments and their number was reduced. A reduction in the number of other PNS neurons and axonal pathway defects were also observed (Fig. 2H).

To clone hth, genomic fragments flanking the P{lacZ, w+}1422-4 insertion site were plasmid-rescued and used as probes to screen genomic and embryonic cDNA libraries. Genomic phage clones that span more than 25 kb were isolated and a partial restriction map of the hth locus was constructed (data not shown). Southern analysis and comparisons of cDNA and genomic sequences revealed that the hth transcription unit spans more than 25 kb of genomic DNA and contains at least four exons. Multiple embryonic cDNA clones were isolated and sequenced. The embryonic cDNA clones contained a single large open reading frame (ORF) of 1461 nucleotides, identical to the ORF reported by Pai et al. (1998). No alternatively spliced variants were identified. Rescue experiments demonstrated that the identified ORF corresponds to the hth gene and can rescue the lethality associated with the hth^P1422 mutation (Pai et al., 1998).

The predicted translation product is a protein of 487 amino acids of M_r 52.8×10³. The predicted HTH protein displays extensive sequence similarity with a family of PBX-related homeobox-containing proteins that includes the murine proto-oncogene Meis-1 (Moskow et al., 1995), the murine Meis1-related proteins (Nakamura et al., 1996a; Steelman et al., 1997), xMeis1 from Xenopus laevis and the human MEIS1 protein (Smith et al., 1997) (Fig. 4A). The highest similarity between the HTH and Meis-1 proteins is found in the homeodomain, with up to 96% identity over a 69-amino-acid stretch (Fig. 4B). The similarity between these proteins extends beyond the homeodomain and spans almost their entire length. Most notably, a stretch of 120 amino acids in the amino-terminal region of the proteins exhibits 84% of sequence identity (Fig. 4B). This domain, which does not share sequence similarity with any other protein in the available data bases, is referred to as MH-box (Meis-homology box, or Meis-HTH box).

Developmental northern blot analysis with hth cDNA probes revealed two major transcripts that are 4.0-4.5 kb and 6.0-6.5 kb long (Fig. 4C). The gene is transcribed only zygotically. Transcripts of hth are relatively abundant in mRNA samples of 3-to 15-hour-old embryos, first instar larvae and pupae. These transcripts are less abundant in mRNA samples of second instar larvae and are present in very low levels in 0-to 3-hour-old embryos, third instar larvae and adults (Fig. 4C).

A polyclonal antiserum was raised against a bacterially expressed HTH protein. The anti-HTH serum recognized a single protein with an approximate molecular mass of 50 kDa in extracts prepared from Drosophila embryos (Fig. 4D).

To determine the spatial pattern of hth expression we performed in situ hybridization to whole embryos using hth cDNA as a probe. hth expression begins at cellular blastoderm. It is expressed in the region that gives rise to the segmented portion of the embryo and is excluded from the anterior and posterior tips (Fig. 5A). During gastrulation, hth expression is detected throughout the ectoderm, but is still excluded from the procephalon (Fig. 5B). Starting at stage 11, high levels of hth mRNA are detected in the thoracic region whereas hth expression in the abdominal segments start to decline (Fig. 5C,D). Two of the head segments, the mandible and maxilla, express moderate levels of hth, whereas expression is absent from the labium. At the same time hth expression becomes evident in the developing visceral mesoderm and a very high level of expression appears in the clypeolabrum. At stages 13-14, the levels of hth transcripts remain high in the head and thorax and decline further in the abdomen. hth expression appears also in the rudiments of the Malpighian tubules and the developing central nervous system (Fig. 5E). During stages 15-17 the ectodermal expression of hth declines whereas expression in the CNS becomes more prominent (Fig. 5F). The expression of hth in the CNS is graded; it is high in the anterior portion of the VNC and weak in its posterior portion.

To test whether any of the enhancer detectors inserted in the hth locus is a good reporter of hth expression we stained embryos of all the insertional strains with an anti-β-galactosidase mAb. We found that the P{lacZ, w+}1422-4 insertion is an excellent reporter of hth expression (data not shown).

To determine the distribution of the HTH protein, we stained whole embryos with anti-HTH serum. The distribution of the HTH protein was very similar to the distribution of hth mRNA (compare Figs 5 and 6), although the protein was not detected prior to the beginning of gastrulation (stage 6). During gastrulation (stages 6-8), the HTH protein was localized to the cytoplasm (Fig. 6B). At stage 9 (germband extension) HTH started to accumulate in nuclei in a spatially regulated fashion (Fig. 6C-G). Nuclear localization was detected in the ectoderm, and in specific cells within the thoracic portion of the VNC (see also Fig. 8E) and within the visceral mesoderm (data not shown). To determine whether strong loss-of-function alleles of hth abolish hth transcription or protein production, we examined the distribution of hth mRNA and protein in mutant embryos. As shown in Fig. 6H, hth^K1-8/hth^K1-8 embryos lack detectable levels of the HTH protein. These embryos lacked also detectable levels of hth transcripts (data not shown).

Starting at germband extension, and throughout the rest of embryonic development, hth expression is modulated in a segment-specific fashion. Most notable is the repression of hth expression in the ectoderm and VNC of abdominal segments during late stages of embryonic development (Fig. 7A). The genes of the homeotic complexes are the major regulators of segmental identity (reviewed by Lewis, 1978). We therefore examined hth expression in embryos deficient for the abd-A gene (iab^D24), or the abd-A and Ubx genes together (Df(3R)Ubx¹⁰⁹). In the absence of abd-A activity hth expression was derepressed in the abdominal ectoderm in cells along the segment boundaries (Fig. 7B-D). In the absence of both abd-A and Ubx, the derepression was more prominent; hth was expressed in ectodermal cells throughout the segment (Fig. 7E). In addition, a uniform level of the HTH protein was observed in all the thoracic and abdominal neuromers of Df(3R)Ubx¹⁰⁹ homozygous embryos (Fig. 8M).

The spatial distribution of the HTH protein in wild-type embryos, from stage 10 until the end of embryonic development, is very similar to the expression pattern of the EXD protein (Rauskolb et al., 1993; Flegel et al., 1993; Aspland and White, 1997). Unlike exd, hth is not expressed maternally. The subcellular localization of HTH also resembles the subcellular localization of EXD (Mann and Abu-Shaar, 1996; Aspland and White, 1997).

The similar PNS phenotypes associated with mutations in the hth and exd genes, and the colocalization of the HTH and EXD proteins, prompted us to test whether one of these proteins regulates the other protein’s spatial distribution. To determine whether EXD is required for proper distribution of HTH we stained exd-deficient embryos with anti-HTH serum. We found that embryos lacking both maternal and zygotic expression of EXD exhibited very low levels of HTH-specific immunoreactivity as compared to their siblings (with zygotic expression of exd) or wild-type embryos (Fig. 8A,B and E,G). These results demonstrate that EXD is required for maintaining normal levels of the HTH protein in the embryo. We do not know at present whether EXD is required for hth transcription, protein production or stabilization of the HTH protein.

To determine whether HTH is required for proper distribution of EXD we stained hth mutant embryos with anti-EXD antibody. The overall level of the EXD protein was only slightly reduced in homozygous hth^K1-8/hth^K1-8 embryos as compared to hth^K1-8/+ embryos from the same preparation (Fig. 8C,D). However, a dramatic change was observed in the subcellular localization of EXD. In homozygous hth^K1-8/ hth^K1-8 embryos the EXD protein was localized mainly to the cytoplasm and did not accumulate in nuclei (Fig. 8C,D and F,I,J). This data indicates that HTH is required, either directly or indirectly, for the nuclear localization of EXD.

To establish whether hth expression is sufficient for driving nuclear localization of EXD, we examined the distribution of the EXD protein in embryos expressing hth ectopically. To drive ectopic hth expression we crossed flies carrying a UAS-hth transgene (Pai et al., 1998) to flies of the 69B-GAL4 strain, which drives GAL4 expression in multiple embryonic tissues (Brand and Perrimon, 1993). We found that in most embryonic tissues, ectopic expression of hth resulted in ectopic nuclear localization of EXD (Fig. 8K,L). In embryos lacking abd-A and Ubx activity, both HTH and EXD were expressed in all the abdominal neuromers and exhibited very similar patterns of nuclear localization (Fig. 8M,N). These data indicate that HTH is sufficient to drive nuclear localization of EXD. Further experiments are required to establish whether the ectopic HTH protein induces both expression and nuclear localization of EXD, or leads only to a transport of cytoplasmic EXD to the nucleus.

Loss-of-function mutations in the hth gene lead to a dorsal instead of lateral localization of the abdominal LCh5 neurons. In addition, the number of the affected neurons is reduced from five to three and their orientation is altered. Similar, but not identical effects on the number, position and orientation of these neurons are caused by mutations in the abd-A gene that transform the LCh5 neurons into DCh3 neurons. One possibility is therefore that mutations in hth cause some kind of homeotic transformation of the LCh5 neurons toward the identity of DCh3 neurons. Another possibility is that mutations in hth transform the LCh5 neurons into A8-LCh3 neurons, which exhibit an intermediate DCh3/LCh5 position and orientation (Fig. 2A). Whether the ‘dorsal chordotonals’ phenotype of hth reflects a transformation of the LCh5 neurons into a DCh3 or A8-LCh3 identity remains an open question. The position of the affected neurons within the segment suggests a transformation towards a DCh3 identity. However, the observation that posterior segments are more susceptible to this phenotype, and the cuticular phenotype caused by hth mutations, supports a posterior transformation toward an A8-LCh3 identity. In addition, the observation that abd-A/hth double mutants exhibit an intermediate phenotype may support a posterior transformation by hth mutations as opposed to anterior transformation by abd-A mutations. In any case, both the thoracic DCh3 neurons and the A8-LCh3 neurons differ from the LCh5 neurons in that they do not depend on abd-A for their development. The thoracic DCh3 neurons require the Ubx gene for their development (Heuer and Kaufman, 1992). It is not known which homeotic genes are required for the development of the A8-LCh3 neurons (abd-B mutations have no effect on the pattern of the PNS (Heuer and Kaufman, 1992).

The cuticular phenotype of hth mutant embryos is similar to the cuticular phenotype of embryos that lack maternal and zygotic expression of exd (Peifer and Wieschaus, 1990). This cuticular phenotype is not consistent with the chordotonals phenotype of hth (or exd) in terms of which segments are transformed phenotypically to which. However, although sensory organs and epidermal structures are derived from the same ectodermal tissue, the PNS phenotypes of homeotic mutations do not always conform with simple homeotic transformations. For example, whereas mutations in the Antennapedia gene transform the cuticle of thoracic segments T2 and T3 to a T1 identity (Wakimoto and Kaufman, 1981), no parallel phenotype is observed in the PNS (Heuer and Kaufman, 1992). Moreover, null mutations in the abd-A gene transform the cuticle of segments A2-A7 toward the identity of A1 (Sanchez-Herrero et al., 1985), but the LCh5 neurons are transformed into DCh3 neurons of T3 identity (Karch et al., 1990; Heuer and Kaufman, 1992; Salzberg et al., 1994). It is therefore possible that exd and hth lead to posterior homeotic transformations in the cuticle of abdominal segments (A1 to A3, A2 to A5 etc, see Peifer and Wieschaus, 1990), but lead to a different type of transformation of the LCh5 neurons (i.e. A1-A7 to A8).

The precursors of the LCh5 neurons originate from five individual ectodermal cells in a two-phase process that requires the activity of the rhomboid gene and the EGF-receptor signaling pathway (Okabe and Okano, 1997). The first three precursors are determined by the activity of atonal and the neurogenic genes. In the second phase, the three precursors express the rhomboid gene, which induces the recruitment of two additional precursors through the activity of the EGF-receptor pathway. Thus, an alternative possibility we need to consider is that hth affects the EGF-dependent Ch neural induction. The loss of ES and MD neurons in the anterior compartment of abdominal segments, which is caused by strong hypomorphic alleles of hth, cannot be correlated with a simple homeotic transformation. Further analyses are required to establish whether hth is required for the formation of these neurons, for their survival, or for the determination of their segmental identity.

Zygotic expression of exd is not required for normal pattern formation in the PNS, indicating that exd is required for this process prior to stage 10 when zygotic expression can be detected for the first time (Rauskolb et al., 1993). The observation that maternal expression of exd is necessary for normal PNS development further supports this notion, because maternally derived exd mRNA does not persist in the embryo beyond stage 8 (Rauskolb et al., 1993). Whereas zygotic expression of exd is not absolutely required for normal PNS development, it can partially compensate for the loss of maternally expressed exd. These data suggest that the developmental time window during which exd is required for proper pattern formation in the PNS is between stages 8-10. During the same time window, both EXD and HTH start to accumulate in nuclei. Starting at stage 11, the level of EXD and HTH expression in the abdominal segments start to decrease. Thus, EXD and HTH are required for LCh5 development prior to, and/or concurrently with, the formation of the first LCh5 precursors. We did not detect hth expression in the LCh5 precursors themselves after their delamination; however, the ectodermal cells from which the precursors are derived do express it.

Although the process of Ch organ migration and rotation is not understood, we can divide the system conceptually into two components: the neuronal cells and their environment (or the receiving and signaling components of the pathway, respectively). We can think of two scenarios that are not mutual exclusive. One is that hth affects the homeotic identity of the LCh5 neurons themselves. The other possibility is that hth affects the environment in which these neurons form and migrate. In midgut development abd-A and Ubx, which are expressed in neighboring parasegments of the visceral mesoderm, were shown to regulate dpp and wingless expression, which affect the underlying endoderm (Manak et al., 1995). It is possible that the influence of HTH and EXD on Abd-A activity in the ectoderm affects signaling molecules such as Wingless and DPP, which in turn affect the localization of the Ch neurons. For example the dpp gene was recently shown to control tracheal cell migration along the dorso-ventral axis of the embryo (Vincent et al., 1997). Support for this idea comes from phenotypic analysis of hth mutations in adult flies. Loss of hth activity in eye imaginal discs results in ectopic eye formation in the ventral head tissue, whereas ectopic expression of hth suppressed normal eye development (Pai et al., 1998). These phenotypes are consistent with a role for HTH in activating Wingless and/or repressing DPP signaling (Ma and Moses, 1995; Treisman and Rubin, 1995; Chanut and Heberlein, 1997; Pignoni and Zipurski, 1997). In addition, wingless expression was reduced in the ectoderm of stage 11 hth⁻ embryos (data not shown).

The relationship between hth and exd is probably not that simple, as both of them are required for each other’s normal distribution in the embryo. HTH is required for nuclear localization of EXD, but we do not know whether EXD is required for nuclear localization of HTH. When HTH was ectopically expressed in the embryo it generally assumed a nuclear localization. However, in such experiments a concomitant ectopic nuclear localization of EXD was observed. The subcellular distribution of ectopically expressed HTH in an exd⁻ background has not been established yet and the level of endogenous HTH is not maintained in such a genetic background. hth transcription starts at cellular blastoderm, long before EXD becomes nuclear. Therefore it is not likely that exd is required for hth transcription. Moreover, HTH production is not completely abolished in the absence of exd. One possible scenario is that a direct interaction between EXD and HTH stabilizes the latter.

Is the nuclear localization of EXD the only role of HTH? Further experiments are required to answer this question. However, the data gathered so far suggests that if not the only role, this is probably the major role of hth. The loss-of-function phenotypes of hth and exd are extremely similar both in the embryo and in adults (Pai et al., 1998). We did not observe any phenotype in hth mutants that was not detected in exd mutants. In addition, the dpp enhancers known to be directly regulated by Ubx and Abd-A (Capovilla et al., 1994; M. Capovilla and J. Botas, unpublished) responded similarly in vivo to the loss of either exd or hth function (Chan et al., 1994; M. Capovilla, personal communication). As a homeodomain-containing protein, HTH is expected to bind DNA and affect transcription of effector genes. The mammalian homologue of hth, meis1, was shown to be coactivated with the homeotic genes Hoxa7 and Hoxa9 in murine myeloid leukemia formation (Nakamura et al., 1996b). However it is not known whether meis1 cooperates with homeotic genes under normal physiological conditions. Whether HTH functions (like EXD) as a transcription regulator on its own and whether it interacts directly with homeotic genes, or binds to homeotic genes in a complex with EXD, are some of the most interesting questions remaining to be elucidated.

We are truly grateful to Hugo Bellen in whose laboratory this project was initiated, and we thank Sharyl Martini for technical assistance in the initial stages of this work. We thank S. Benzer, M. Capovilla, S. Cohen, D. Duncan, I. Duncan, A. Jarman, R. Mann and R. A. H. White for antibodies and fly strains. We thank M. Capovilla and J. Botas for sharing unpublished data, and M. Capovilla for a helpful discussion. This work was supported by grants to A. S. from the ‘Israel Cancer Association’ and ‘The Israel Science Foundation’ founded by the Israel Academy of Sciences and Humanities – Charles H. Revson Foundation, and by grants to Y. H. S. from the National Science Council (NSC-86-2313-B-001-003, NSC-87-2311-B-001-106), Taiwan, Republic of China. C.-Y. P. is supported by a National Science Council Postdoctoral Fellowship.

The regulation of EXD nuclear localization by HTH has also been reported.