Distal-less is a downstream gene of Deformed required for ventral maxillary identity

In Drosophila melanogaster, the assignment of stable and different anterior-posterior axial identities in the gnathal, thoracic and abdominal regions, eventually to be realized in morphological variation in different segments, is under the control of the homeotic selector genes (Lewis, 1978; Kaufman, 1983; Lawrence, 1984). Most of the known Drosophila homeotic selector genes reside within the Antennapedia and Bithorax gene complexes. The Antennapedia complex (ANT-C) selector genes regulate the identities of the posterior head and thoracic segments (Kaufman et al., 1990), while the Bithorax complex (BX-C) selector genes regulate the identities of posterior thoracic and abdominal segments (Lewis, 1978; Sanchez-Herrero et al., 1985; Duncan, 1987). The ANT-C and BX-C homeotic selectors, collectively referred to as the HOM-C or HOM genes, have been cloned and extensively characterized (reviewed by Kaufman et al., 1990; Akam, 1989; McGinnis and Krumlauf, 1992; Cribbs et al., 1992). Each of the eight HOM genes plays a pivotal role in determining the identity of a few hundred embryonic cells and their descendants. We wish to address the question of how the relatively crude spatial identities specified by HOM proteins are refined to smaller spatial domains.

The limits of HOM gene activation within specific parasegmental boundaries in early embryos are initially specified by the maternal, gap, pair-rule and segment polarity genes. These four gene classes also produce the overall polarity and basic metameric pattern of the embryo (Nüsslein-Volhard and Wieschaus, 1980; Martinez-Arias and Lawrence, 1985; Ingham and Martinez-Arias, 1986; White and Lehman, 1986; Akam, 1987; Riley et al., 1987; Irish et al., 1989; Harding and Levine, 1988; Jack et al., 1988, Jack and McGinnis, 1990). After their activation in unique domains, HOM proteins presumably dictate cellular identities by regulating the spatial and temporal expression patterns of many downstream genes. They do so at least in part by acting as DNA binding transcriptional regulators, each binding DNA via a sixty amino acid homeodomain (Desplan et al., 1985, 1988; Jaynes and O’Farrell, 1988; Thali et al., 1988; Beachy et al., 1988; Hoey and Levine, 1988; Hayashi and Scott, 1990).

Some of the downstream genes directly targeted by HOM proteins are likely to have effector or ‘realizator’ functions (Garcia-Bellido, 1977), and would include genes that regulate cell movement, shape and mitotic orientation, and genes involved in cell-cell communication. It seems likely that another class of direct downstream targets would include genes whose products are involved in further spatial subdivisions of the large metameric primordia specified by HOM proteins. Some of these putative sub-segmental identity proteins might be required to act in combination with HOM proteins, and some might function independently. Each HOM protein presumably controls the developmental fates of cells within its expression domain by acting in combination with many gene products, some of which are likely to be other transcription factors.

Since few HOM downstream genes have as yet been identified, it is not known how many such genes each HOM protein directly activates or represses or to what extent individual downstream genes are targeted by more than one HOM protein. The best characterized regulatory targets of the HOM proteins are other genes of the HOM complex. For example, each HOM gene expressed in the thorax and abdomen is transcriptionally repressed by more posterior HOM-C proteins (Struhl, 1983; Hafen et al., 1984; Struhl and White, 1985). Also, the HOM genes Deformed and labial positively regulate their own transcription in epidermal cells, and Ultrabithorax (Ubx) autoregulates in visceral mesoderm cells (Bienz and Tremml, 1988; Kuziora and McGinnis, 1988; Chouinard and Kaufman, 1991).

Excepting cross-regulation and autoregulation, surprisingly few genes are known to be HOM-regulated. These include decapentaplegic (dpp), encoding a TGF-β homolog, regulated by Ubx and abdominal-A (abd-A) within the visceral mesoderm, and wingless (wg), a homolog of the mammalian proto-oncogene int-1, also regulated by abd-A within the visceral mesoderm (Reuter et al., 1990; Immergluck et al., 1990). In addition, two transcripts currently known as 35 and 48 are regulated in a Ubx-dependent fashion in the central nervous system (CNS) and ventral-lateral epidermis (Gould et al. 1990). Finally, the spalt-major gene is down-regulated by Antp in imaginal disk cells (Wagner-Bernholz et al., 1992).

The HOM gene Deformed (Dfd) is required for the identities of the maxillary and mandibular segments of the embryonic head, and is particularly important in the cells that give rise to the maxillary cirri (noninnervated triangular papillae located above the mouth), the ventral organs and the mouth hooks. Dfd null mutants are missing these structures, and ectopic expression of Dfd protein in embryos induces the development of ectopic cirri, ventral organs and mouth hooks in the labial and thoracic segments. During head involution Dfd also plays a role in orchestrating the intricate cell movements involved in the spatial positioning of head structures (Kaufman, 1983; Merrill et al., 1987; Regulski et al., 1987).

In this report we show that Dfd activates transcription from the homeobox gene Distal-less (Dll) within the maxillary segment. Dll is required for the development of the distal regions of all segmented adult appendages; legs, antennae, maxillary palps, labium, labrum and proboscis (Cohen, 1990; Cohen and Jürgens, 1989; Cohen et al., 1989; Sato, 1984; Sunkel and Whittle, 1987). A number of larval structures also require Dll for their formation (Sunkel and Whittle, 1987, Cohen and Jürgens, 1989). These include the labral, the antennal, the maxillary, and the labial sense organs of the head, the Keilin’s sense organs of the thorax and some minor elements of the head skeleton. As shown here, Dll function is also required for the development of a subset of the maxillary cirri, and the regulatory effect of Dfd on Dll is exerted within cirri progenitor cells. An enhancer from a 3′ region of the Dll locus controls the expression pattern of Dll in cirri primordia, and contains a regulatory element that can be specifically activated in response to Dfd protein expression. The regulation of Dll by Dfd may be an example of a HOM protein activating a downstream target gene whose product will combinatorially interact with its HOM regulator at later stages to define a sub-segmental identity.

For single antigen detection in whole-mount embryos, embryos were collected for 1 hour and aged at 25°C to the desired stage of development. Embryos were then harvested, dechorionated, fixed with paraformaldehyde, and stained with rabbit anti-Dfd or mouse anti-β-galactosidase antibodies as previously described (Jack et al., 1988; Bergson and McGinnis, 1990). To detect both Dfd and β-galactosidase proteins in hsp70-Dfd embryos, 30 minute collections were aged 2.5-3 hours at 25°C to bring the embryos to the cellular blastoderm stage of development, heat shocked for 1 hour at 37°C, and aged for 7 hours at 25°C. The primary antibodies, a polyclonal rabbit anti-β-galactosidase (1:500) (Promega) and a polyclonal mouse anti-Dfd (1:500), were added simultaneously; detection was with separate secondary antibodies, biotinylated goat anti-mouse IgG (1:500; Cappel) and alkaline-phosphatase-conjugated goat anti-rabbit IgG (1:500; Jackson Immunoresearch). Adsorption times, buffers, and wash regimens were performed as in Jack et al. (1988). Dfd antigen was detected using the Vector Labs ‘ABC’ Horseradish Peroxidase detection kit without the addition of CoCl₂ to the developing solution. β-galactosidase antigen was detected using an alkaline phosphatase detection buffer (Vector Labs). After dehydration, singly-stained embryos were cleared with two rinses of methyl salicylate and mounted in a 1:1 mixture of methyl salicylate and Permount (Fisher) and doubly stained embryos were cleared with two rinses of xylene and mounted in Permount.

The localization of transcripts by whole-mount in situ hybridization followed the method of Tautz and Pfeifle (1989). The probe used for Dll was a 1.4 kb EcoR1 cDNA fragment containing 1.2 kb of coding sequence (Cohen et al., 1989).

Larval cuticular preparations were done as previously described (Van der Meer, 1977). Dll mutant larvae with non-involuted heads were generated by crossing desired mutations (e.g. Dll^SA1and Dll⁵) into a backgound containing the hsp70-Dfd/UbxHD construct (Kuziora and McGinnis, 1989) and subjecting embryos to a mild heat shock (10-20 minutes at 37°C) at 6-8 hours after egg lay, and then allowing the embryos to age for 24-36 hours at 25°C. This treatment of hsp70-Dfd/UbxHD embryos causes almost complete lack of head involution, but induces little or no homeotic transformation of head segments. Cirri were counted on a total of 22 Dll^SA1; hsp70-Dfd/UbxHD larvae. The maxillary lobes of these animals had an average of 5 cirri missing from the dorsal row (range 4-6), and an average of 7 cirri missing from the ventral row (range 6-8). Similar results were obtained with Dll⁵; hsp70-Dfd/UbxHD larvae.

A 5.8 kb HindIII fragment containing the Dll ventral-lateral maxillary enhancer (ETD6) was subcloned into HZ50PL, a P-element injection vector containing an hsp70 basal promoter fused to the lacZ gene (Hiromi and Gehring, 1987). This construct was coinjected with the helper plasmid p 25.7wc into embryos of the ry⁵⁰⁶ strain (Rubin and Spradling, 1982; Pirotta et al., 1988). Three independent transgenic strains, ETD6-57, 69, and 71 were tested for expression patterns in wild-type and mutant backgrounds.

Dll expression evolves from a single anterior-dorsal patch at the cellular blastoderm stage to the complex pattern shown in Fig. 1B by late stage 12 of development (Cohen, 1990). Of particular importance to this study is the domain of Dll expression within epidermal cells of the ventrallateral maxillary segment, which is initiated during early germ band retraction. The ventral-lateral maxillary Dll-expressing cells are entirely contained within the larger Dfd expression domain in the maxillary segment of stage 12 embryos (Fig. 1A). Fate mapping studies have indicated that the ventral-lateral maxillary region will give rise to the cirri, mouth hooks, and the ventral (sense) organs of the larval head (Turner and Mahowald, 1979; Jürgens et al., 1986).

To test whether Dfd function is involved in the establishment of the normal expression pattern of Dll in the maxillary segment, we performed in situ hybridizations to detect Dll transcripts both in Dfd mutant embryos and in hsp70-Dfd embryos. The Dll expression pattern in a Dfd mutant embryo is shown in Fig. 1C. Dll transcripts are no longer detectable within the ventral-lateral maxillary cells and are ectopically expressed in several mandibular segment cells. This regulatory relationship is non-reciprocal, as Dfd expression is not detectably altered in Dll mutant embryos (data not shown).

hsp70-Dfd embryos are homozygous for a P element insertion containing a Dfd cDNA fused to a heat inducible promoter. Heat-shocking these embryos for 1 hour at 37°C during the cellular blastoderm stage of development induces ectopic expression of Dfd protein that persists (via autoregulation) in the antennal segment and ventral regions of the labial, thoracic, and abdominal segments (Kuziora and McGinnis, 1988). Dll expression in hsp70-Dfd embryos is shown in Fig. 1D. The Dll pattern is strikingly altered in head and thoracic segments. Dll is no longer activated within the antennal segment (which correlates with a deletion of the antennal sense organ in hsp70-Dfd larvae; unpublished results), and the labial pattern closely resembles that of the maxillary segment, including an ectopic ventral-lateral domain of Dll transcription. In thoracic segments, the wild-type Dll expression domains in the posterior-lateral epidermis (which will give rise to Keilin’s organs; Cohen, 1990) are repressed with a variable penetrance. In addition, Dll transcripts are ectopically expressed in novel thoracic domains that resemble the ventral-lateral maxillary Dll pattern in their more ventral positioning on the D/V axis.

From these experiments, we concluded that Dfd function is required for the ventral-lateral maxillary domain of Dll transcription, and Dfd expression is sufficient to induce ectopic ‘ventral-lateral domains’ of Dll expression in the labial and thoracic segments.

The maxillary cuticular structures that are deleted in Dfd mutant larvae are the cirri, ventral organs, and mouth hooks (Merrill et al., 1987; Regulski et al., 1987). The maxillary sense organ, which develops from a dorsal-anterior region of the maxillary segment that persistently expresses Dll but not Dfd transcripts and protein (Fig. 1A,B), is less strongly affected in Dfd mutants. The fact that the Dll ventral maxillary expression domain is dependent on Dfd suggested the possibility that Dll expression in this region might mediate part of the Dfd morphogenetic function. Previous studies have not provided a detailed picture of ventral maxillary development in Dll mutants due to the difficulty of scoring structures deriving from this region on involuted or partially involuted heads. To determine if any ventral maxillary derived structures are dependent on Dll expression, we analyzed the cuticular phenotype of Dll mutant larvae in which head involution was prevented (see Materials and Methods).

When displayed on the surface of a non-involuted maxillary lobe, the structures that develop from the ventral maxillary segment are much easier to visualize. These structures; cirri, mouth hook, and ventral organ, are shown in Fig. 2 on a wild-type (A) and non-involuted larval head (B). The mouth hook develops from the most ventral region of the lobe while the ventral organ and two rows of cirri develop from the adjacent ventral-lateral region (Fig. 1A). The anterior row of cirri in the top of the frame of Fig. 2B (which will become the more dorsal row on the pseudocephalon if head involution takes place) contains 9-10 cirri, and the posterior row contains approx. 14 cirri. The ventral organ, consisting of a small dome shaped structure flanked by a smaller papilla, is located just under the two anterior row papillae farthest from the mouth hook.

Dll mutant larvae have normal mouth hooks and ventral organs, but are missing an average of five cirri from the anterior row, and seven cirri from the posterior row (Fig. 2C). In addition, the ventral organ often includes extra papillae of unknown origin. As the ventral organ and associated cirri are still present in Dll mutants, the missing cirri apparently derive from the most ventral region adjacent to the mouth hook.

Two chromosomal rearrangment breakpoints that interrupt the Dll locus, diagrammed in Fig. 3, provide a hint as to the location of sequences that control the ventral-lateral maxillary expression domain of Dll. Both breakpoints are located 3′ of the Dll coding regions, and transcripts are still produced from the Dll loci residing on the rearranged chromosomes. The Dll^J chromosome has a breakpoint that maps within a 1.7 kb EcoRI fragment that spans sequences from approx. 2 to 4 kb downstream of the Dll transcription unit, while the Dll^B breakpoint maps within a region that includes the 3′ exons of the Dll gene (Cohen et al., 1989). Stage 12 embryos that are homozygous for the Dll^J chromosome exhibit a normal or near normal pattern of Dll transcripts in the ventral-lateral maxillary domain (Fig. 4D). On the other hand, embryos that are homozygous for the Dll^B chromosome do not accumulate Dll transcripts within ventral-lateral maxillary cells, although the remainder of the Dll expression pattern appears to be normal (Fig. 4C). This suggests that a Dll regulatory element required for the ventral-lateral maxillary domain of Dll transcription maps in the 3′ sequences between the Dll^B and Dll^J breakpoints.

One possible function for the interval between the Dll^B and Dll^J breakpoints is that of a ventral maxillary-specific transcriptional enhancer. To test this, a 5.8 kb HindIII fragment, designated ETD6 (Fig. 3), was cloned upstream of the basal promoter in the lacZ reporter vector HZ50PL (Hiromi and Gehring, 1987). Transgenic embryos carrying this construct (ETD6 strains) express β-galactosidase in ventral-lateral maxillary cells in a pattern that exhibits a spatial and temporal overlap with the normal Dll ventral maxillary expression domain (Figs 4A, 5).

Dll transcription within the ventral-lateral maxillary segment is initiated during early germ band retraction and includes 20-25 cells by the end of germ band retraction (Fig. 5A). Interestingly, this number is approximately the same as the number of cirri (approx. 23-24) that typically develop from ventral-lateral maxillary epidermis. In comparably staged embryos, the ETD6 enhancer directs β-galactosidase expression in approximately 16 cells that overlap the most ventral cells in the Dll ventral-lateral maxillary domain (Fig. 5B). During head involution (stage 14), the maxillary segment undergoes a slight rotation, and ventral maxillary cells come to occupy anterior positions relative to the formerly dorsal (now posterior) cells. It is during this stage that the first morphological signs of cirri appear, as some of the ventral-lateral maxillary cells gradually become organized into two orderly rows. The ventral row of cirri cells, derived from a relatively posterior position in the ventral-lateral maxillary lobe, and the dorsal row of cirri cells (anteriorly derived) can be first visualized between stages 14 and 15 (Turner and Mahowald, 1979). During this period, Dll transcripts are expressed in these apparent cirri precursors (Fig. 5C; Turner and Mahowald, 1979). The ETD6 enhancer expression pattern is limited to the same group of cells (Fig. 5D).

During stages 16 and 17 the cirri-producing cells become clearly visible, now localized on the anterior-ventral aspect of the pseudocephalon. Expression of β-galactosidase from the ETD6 enhancer, at stage 16, is detected in approximately 16 cirri-producing cells, 8-10 cells in the dorsal row and approx. 6 cells in the ventral row. In both rows, the ETD6-expressing cells include those that are closest to the developing mouth hooks and thus correspond to positions that do not produce cirri in Dll mutant larvae. In stage 17 larvae, ETD6 expression persists in the proximal cirri (nearest the mouth hooks) of the dorsal row, but appears to be gradually extinguished in the ventral row (Fig. 5E). Dll expression, as measured by whole-mount in situ hybridization, appears to be extinguished in cirri at some time during stage 16-17. The Dll transcript patterns in cirri during these very late stages resembles the ETD6 expression pattern, but is difficult to accurately define due to the low resolution of in situ hybridization, coupled with the spatial clustering of other Dll-expressing structures (antennal sense organ, maxillary sense organ) in the pseudocephalon of late stage embryos (data not shown).

The spatial overlap of the ETD6 enhancer expression with the ventral maxillary domain of the Dll pattern suggested that this 3′ regulatory element supplies most of the ventrallateral maxillary pattern of Dll. To test whether the ETD6 enhancer is regulated in a Dfd-dependent manner, the ETD6 expression pattern was analyzed in Dfd mutant and hsp70-Dfd embryos. As mentioned earlier, the ETD6 expression pattern is localized to Dfd-protein-expressing cells in the ventral-lateral maxillary epidermis, as is shown in the doubly stained embryo in Fig. 6B. In Dfd mutant embryos, the ETD6 enhancer is not active at any stage of embryonic development (data not shown).

In hsp70-Dfd embryos, heat shocked for 1 hour at the cellular blastoderm stage and aged to retracting germ band stage, the ETD6 enhancer (blue) is ectopically activated in the antennal, labial and thoracic segments within cells ectopically expressing Dfd (brown) (Fig. 6C). The pattern of ectopic activation of ETD6 is very similar to the pattern of ectopic activation of ventral-lateral patches of Dll transcription in hsp70-Dfd embryos after heat shock (Fig. 1D).Note that even in the labial and thoracic segments only a subset of the cells that ectopically express the Dfd protein also express the 3′ enhancer. In addition, although Dfd protein is also ectopically expressed in abdominal segments, the ETD6 enhancer is very rarely activated in the abdomen, and if so, only in one or two cells.

The foregoing results suggest that Dll may be an important downstream mediator of the Dfd morphogenetic function in the maxillary segment. To test whether Dll is equally important when Dfd is expressed ectopically, we placed the hsp70-Dfd construct in Dll mutant genetic backgrounds. As previously mentioned, ectopic expression from the heat inducible Dfd cDNA induces maxillary cirri in labial and thoracic segments. Dll mutations did not completely revert this homeotic transformation, as some ectopic cirri are still induced in hsp70-Dfd larval cuticle (data not shown).

A detailed analysis of the cirri produced in hs70-Dfd larvae may provide an explanation for this lack of requirement for Dll in ectopic cirri development. We note that among the four to six cirri typically induced by hsp70-Dfd in the labial segment and T1, a ventral organ is also often present. In the strongest transformations two rows of ectopic cirri are present, each with four to six cirri, with one row containing a ventral organ (data not shown). This indicates that some of the ectopic cirri formed in hs70-Dfd larvae correspond to the cirri most distant from the mouth hooks. As shown in Fig. 2C, these are the cirri in the maxillary segment that are not dependent on Dll function. These results indicate that although both Dfd and Dll are required for the development of the ventral-most cirri, Dfd can act either independently or in combination with other factors to induce the formation of ectopic cirri.

These experiments indicate that the homeobox containing gene Dll is one of the downstream transcription units regulated by the HOM gene Dfd. This regulation takes place in embryonic ventral-lateral maxillary epidermal cells, which represent a small subset of the cells expressing Dfd. Dll function is required in these cells for the development of the specialized maxillary structures known as cirri, thus mediating an important part of the segmental identity function of Dfd. The regulatory influence of Dfd on Dll is exerted principally through an enhancer element that resides in a 3 ′ region of the Dll locus. At present, the limits of this Dfd-dependent ventral-lateral maxillary enhancer are defined by a HindIII site at the left end of the 5.8 kb ETD6 test construct, and the Dll^J breakpoint that maps 1-3 kilobases away (Fig. 7).

Obviously, Dfd expression alone is not sufficient for the activation of the Dfd-dependent Dll 3′ enhancer since the enhancer is initially activated in retracting germ band embryos, hours after the establishment of Dfd protein expression in blastoderm stage embryos. In addition, the 3′ enhancer is activated in only a fraction of the epidermal cells that express Dfd protein (Jack et al., 1988; Mahaffey et al., 1989). Thus other unknown factors must limit the expression of the Dll 3′ enhancer to ventral-lateral epidermal cells and either prevent its expression prior to germ band retraction, or promote it thereafter. These factors are likely to be present in other segments of the animal as Dll transcription can be induced in ectopic ventral-lateral patches in hsp70-Dfd embryos. Whether the Dll 3′ enhancer is directly activated by the Dfd protein is as yet unknown, but this regulatory element is an excellent candidate for a direct downstream target of the Dfd protein due to its lack of activation in Dfd mutant embryos and its ectopic activation in hsp70-Dfd embryos. Immunoprecipitation assays using Dfd protein overexpressed and purified from E. coli (Regulski et al., 1991; Dessain et al., 1992) indicates that the ETD6 enhancer has significant in vitro binding affinity for Dfd protein (unpublished results). However, much more information about the location and function of important sequences within the ETD6 enhancer will be required before it is known whether the 3′ enhancer is directly regulated by Dfd protein.

Dll function is necessary only for the development of the cirri that develop adjacent to the mouth hook. Although we can say that these cirri most likely derive from the more ventral regions of the Dfd dependent Dll expression domain in the maxillary segment, there are no obvious morphological (or compartmental) subdivisions that separate the affected from unaffected cirri. The only difference we have detected between the affected and unaffected cirri is in their differential expression of the ETD6 enhancer construct. After stage 14 of development the Dll 3′ enhancer is progressively inactivated in a distal to proximal direction (proximal cirri cells being those closest to the animal’s ventral midline), first in the ventral row of cirri and then in the dorsal row. Thus, relative to the unaffected cirri of each row, the affected cirri-producing cells exhibit the most persistent expression of the Dll 3′ enhancer. Perhaps Dll expression is only required during the late stages of cirri development, so that those cells that express both Dll and Dfd at late stages (Dfd is expressed in all of the cirri cells well into stage 17, Malicki et al., 1992) show a mutant phentype when lacking Dll function.

As Dll is persistently expressed in combination with Dfd in the ventral-lateral maxillary epidermis, and both are required for the morphological development of this region, the two homeodomain transcription factors produced from these loci are likely to represent an important part of a combinatorial morphogenetic code that assigns a specific subsegmental identity within the maxillary segment. Other factors must certainly be required for the complete combinatorial code that is sufficient to generate ventral-lateral maxillary identity, and this code is likely composed of both Dfd-dependent and Dfd-independent factors. Our results, along with those of Wagner-Bernholz et al. (1992), and Vachon et al. (1992), suggest that there is at least one intermediate level of transcription factor genes involved in spatial and/or cell-type identity in the genetic pathway between homeotic genes and the ‘effector’ gene products which must eventually be activated to achieve the final differentiated state. The recent work of Vachon et al. (1992) has shown that Dll transcription is repressed by the Drosophila HOM proteins Ubx and abd-A in the abdominal region of the embryo. Interestingly, the DNA element that mediates the Ubx/abd-A regulatory effect maps a considerable distance upstream of the first Dll exon, whereas the ETD6 element that mediates the Dfd regulatory effect maps in 3′ sequences (Fig. 7). Thus, some of the intermediate regulators like Dll may be common targets of several homeotic genes through different regulatory elements.

The mouse genome contains genes that are structural homologs of Dfd and Dll (Price et al., 1991; Porteus et al., 1991; Robinson et al., 1991; McGinnis and Krumlauf, 1992). Dollé et al. (1992) and Bulfone et al. (1992) report that transcriptional expression from two of the mouse Dll homologs (Dlx-1 and Dlx-2) is initiated in midstage embryos in spatially restricted domains within the mesenchyme of the four branchial arches. In the fourth branchial arch, Dlx expression overlaps with the expression patterns of mouse Dfd-like genes (Hunt and Krumlauf, 1991), which were initiated in this region at earlier stages of embryogenesis. Thus it is possible that during the process of facial development, a regulatory relationship between mammalian Dll-like and Dfd-like genes may exist that resembles the Dfd regulation of Dll in the Drosophila head.

We thank Nadine McGinnis for much experimental assistance with the early stages of this work, and many members of the laboratory for their commments, help and criticism. These experiments were supported by a grant to W. M. from the NIH (HD28315) and a grant to S. M. C. from HHMI.