The novel homeodomain gene buttonless specifies differentiation and axonal guidance functions of Drosophila dorsal median cells

The homeodomain DNA-binding motif, initially discovered in Drosophila, is now understood to play a major role in the development of most multicellular eukaryotes. The first home-odomain genes identified in Drosophila were associated with mutant phenotypes ranging from homeotic transformations of segment identity to repetitive segmentation defects to gross defects in embryonic polarity. Extensive genetic and molecular studies indicate that these homeodomain genes contribute to the establishment and maintenance of a detailed positional reference system; homeodomain genes operating at this level appear to influence the development of a variety of cell types in multiple germ layers.

In contrast, several more recently identified homeodomain genes appear to function in a more restricted range of cell types. For example, the Drosophila homeodomain genes H2.0, msh-2, tinman, and bagpipe are expressed primarily in mesoderm (Barad et al., 1988; Bodmer et al., 1990; Azpiazu and Frasch, 1993), expression of the bsh homeodomain gene is restricted to a subset of neurons in the Drosophila brain (Jones et al., 1993) and the rough gene functions in photoreceptor cell differentiation in the developing retina (Saint et al., 1988; Tomlinson et al., 1988). In other organisms, the home-odomain genes mox-1 and mox-2 are expressed only in vertebrate mesoderm and its derivatives (Candia et al., 1992), Csx expression is restricted to the developing vertebrate heart (Komuro and Izumo, 1993), and mec-3 functions in a particular set of sensory neurons of the nematode (Way and Chalfie, 1988). Homeodomain genes in this second class thus appear to be more strongly associated with the commitment to or imple-mentation of specific differentiation pathways as opposed to specification of positional information.

We report here the characterization of buttonless (btn), a novel Drosophila homeodomain gene which falls into the second group. Its expression is exquisitely specific, being restricted to a single pair of cells per segment in the trunk of the embryo. These cells, the dorsal median (DM) cells, are closely associated with the central nervous system and appear biochemically and morphologically similar to the Drosophila cultured cell line from which the btn gene was isolated. DM cell bodies are tightly paired at the dorsal midline of the ventral cord while each cell’s single large process projects laterally to the muscle attachment sites in the body wall. In embryos lacking btn gene function the DM cells fail to form and, as a consequence, axons that normally project laterally along the DM cell processes to form the transverse nerve instead appear to wander along the surface of the ventral cord. These observations indicate that the btn gene function is critical for the specification of DM cell fate and for transverse nerve pathfinding.

Total RNA was prepared according to standard procedures (Sambrook et al., 1989) from the shibire cell line (Simcox, 1981) and from an isolate of this parent line stably transfected with a construct carrying the open reading from for the Ultrabithorax open reading frame under control of the Drosophila metallothionein promoter. Poly(A)⁺ RNA was isolated using oligo(dT)-cellulose (Boehringer Mannheim) and a cDNA library (2×10⁶ independent clones) was prepared from the induced stably transfected line using the Lambda Zap cDNA cloning kit (Stratagene). The probe used to screen this library was an end-labelled 1024-fold degenerate oligonucleotide corresponding to a widely conserved region of the third helix of the homeodomain (Burglin et al., 1989). Hybridization to filters proceeded in 6× SSC, 5× Denhardt’s solution, 0.05% sodium pyrophosphate, 1% SDS and 100 µg/ml salmon sperm DNA at 42°C; washes were in 6× SSC, 0.05% sodium pyrophosphate and 0.2% SDS at room temperature, then at 50°C. A btn cDNA clone isolated from this library was used as a probe to screen a genomic library (Maniatis et al., 1978) and a 4-8 hour embryonic cDNA library (Brown and Kafatos, 1988).

Sequence determination utilized the dideoxy termination method with Sequenase (US Biochemical). For the largest embryonic btn cDNA (650 bp) flanking vector primers and synthetic primers were used. For genomic sequence, nested exonuclease III deletions spanning a 2.4 kb HindIII/EcoRI genomic fragment were generated using the Erase-a-Base kit (Promega). The deduced amino acid sequence of btn protein was used to search the NBRF and SWISS-PROT data bases using the FASTA algorithm.

Hybridization to polytene chromosome squashes was as described (Langer-Safer et al., 1982) using the alkaline phosphatase-based DNA detection system (GIBCO/BRL). A probe was generated from the 6.5 kb EcoRI btn genomic fragment (from 0 to 6.5 in Fig. 1) using the random hexamer priming kit (Boehringer Mannheim) in the presence of biotinylated dUTP (ENZO Biochemicals). The btn sequences hybridized near polytene division 94B, which is the location of P-element insert in P[rib(94B)] described by Karpen et al. (1988).

Southern hybridization with the 6.5 kb EcoRI btn genomic fragment as a probe established that the P element was inserted near btn coding sequences and its precise location was determined by isolation of sequences flanking the P[rib(94B)] insert using inverse PCR (Ochman et al., 1990). Briefly, two primers, GTATACTTCG-GTAAGCTTCGGCTAT and CGAAATGCGTCGTTTAGAGC-AGCAG, which hybridize to the 5′ and 3′ portions of a region near the 5′ end of the P element, were used as primers in a PCR reaction with 1.5 µg of DNA from the homozygous P[rib7(94B)] line. Prior to PCR, this DNA was digested with Sau3A, ligated overnight in a 300 µl reaction at 4°C, and digested with AseI.

The P-element insert in P[rib(94B)] was mobilized by exposure to transposase from the P[Δ2-3 ry⁺] Sb chromosome. Excision events were identified by loss of the ry⁺ marker present in P[rib(94B)]. One of the chromosomes recovered was homozygous lethal and carried a deletion of 3.3 kb that removed the entire btn transcription unit (see Fig. 1). The extent of this deletion was inferred by Southern hybridization to DNA digested with various restriction endonucleases from heterozygous btn individuals using the btn cDNA and a 6.5 kb EcoRI fragment (from 0 to 6.5 kb in Fig. 1) as probes.

Analysis of btn RNA expression

For northern analysis, 5 µg of poly(A)⁺ RNA derived from different developmental stage embryos or tissue culture cells was electrophoresed on 1.5% formaldehyde gels and transferred to nitrocellulose membranes (S&S or Costa). Filters were hybridized with radiolabeled btn cDNA and washed according to standard procedures (Sambrook et al., 1989).

In situ hybridization to whole embryos was performed as described by Tautz and Pfeifle (1989). Probes were generated by random hexamer priming using digoxigenin-dUTP and detected with the Genius Kit (Boehringer Mannheim). Stained embryos were prepared for sectioning by embedding and freezing in OCT solution (Tissue Tec), and 10 µm sections were prepared with the Microm 500 OM-Cryostat.

Embryos and larvae were fixed and immunostained as described (Patel, 1994). Embryos were viewed on a Zeiss Axioplan microscope and photographed on Ektachrome 64 slide film or captured electronically using a Sony DXC-760MD camera and a Nuvista Plus capture board. Slides were digitized using a Kodak RFS 3500 scanner. Figures were arranged using Adobe Photoshop and Aldus Pagemaker and printed using a Kodak XL7700 dye-sublimation printer.

A 1.85 kb EcoRI-Pst1 genomic fragment, which contains 5′ flanking sequences and part of of the btn ORF (Fig. 2), was first subcloned into the EcoRI-PstI site of Bluescript (Stratagene). The EcoRI-BamHI fragment of the resulting construct was then subcloned into the EcoRI-BamHI site of the pCasperβgal vector (Thummel et al., 1988) to create an in-frame fusion to a lacZ reporter gene. This construct, pbtn1.85/lacZ, was coinjected with p25.7wc (Karess and Rubin, 1984) into W¹¹¹⁸ embryos at a DNA concentration of 500 µg/ml and 100 µg/ml, respectively. Germ-line transformation was performed essentially as described (Rubin and Spradling, 1982).

For cell fate analysis in the btn mutant background, embryos were collected from a stable line P[btn1.85-lacZ]/P[btn1.85-lacZ]; btn^r3/TM3, P[ftz-lacZ]. btn mutant embryos from this strain can be identified by the absence of the ftz expression pattern.

Mutant btn embryos were collected from btn/P144 parents. The P144 enhancer-detector line carries a lacZ gene with strong expression in the central and peripheral nervous systems of late stage embryos (unpublished observations), and homozygous btn embryos therefore can be unambiguosly identified by the absence of lacZ antibody staining.

The shibire cultured cell line (Simcox, 1981) was maintained in Schneider’s medium (Gibco/BRL) supplemented with 10% fetal calf serum (Gibco/BRL) and an antibiotic/antimycotic mixture (Gibco/BRL).

The buttonless (btn) gene was first isolated in cDNA form using as hybridization probe a degenerate 23-base oligonu-cleotide corresponding to a widely conserved region from within the third helix of the homeodomain (Burglin et al., 1989). The library screened with this probe derived from a cultured Drosophila cell line called shibire (originally established from shibire mutant embryos; Simcox, 1981) in which expression of the Ultrabithorax homeotic protein had been experimentally induced (see Materials and Methods). From an initial screen of 150,000 clones, five of eight positives contained genuine homeodomain coding sequences, of which one corresponded to Ubx, one to another homeotic gene, Abdominal-B, and one to the previously described mesoderm-specific homeodomain gene, H2.0 (Barad et al., 1988). The remaining two homeodomain clones represented the gene but-tonless (btn), named for its expression pattern and mutant phenotype (see below). Further cDNA and genomic clones were isolated using the initial btn cDNA clone from the shibire cell line as a probe; only three embryonic cDNA clones were recovered from screening of a million colonies from an embryonic cDNA library, consistent with a low overall abundance of btn message within the embryo (see below).

A restriction map of the btn genomic region is shown in Fig.1, with the location and structure of the btn transcription unit as determined by comparison of genomic and btn cDNA sequences (Figs 1 and 2). The btn transcript, ∼600-700 bases in length (Fig. 3), is difficult to detect and appears to derive from a single exon. The homeodomain coding sequences lie within an open reading frame (ORF) capable of encoding a 158 residue protein; this is the largest ORF within a fully sequenced 607 bp cDNA, and is preceded by two in frame chain termination codons just upstream of the initiating methionine codon. A stretch of 12 adenines at the 3′ end of the cDNA is preceded by several imperfect matches to the polyadenylation signal AAUAAA. We can not distinguish between the possibilities that this stretch of adenines reflects a genuine polyadenylation event or a site of internal priming by oligo(dT) during cDNA synthesis; neither possibility affects the extent of the ORF, which ends 86 bases upstream.

Outside the homeodomain, the predicted btn protein shares no significant homologies with other proteins in the databases (see Materials and Methods). The btn homeodomain itself is most related to a divergent family of vertebrate homeobox genes (Fig. 3B), sharing amino acid identities of 77-78% with the Mox-1 and Mox-2 genes in the mouse (Candia et al., 1992), the Gax gene in the rat (Gorski et al., 1993) and the Xelmox2 gene in Xenopus (Candia et al., 1992). All of these homeodomains, including that of btn, contain a lysine residue at position 3 (K3). The majority of other homeodomains carry an arginine residue (R3) at this position, which is known to make a specific contact with the second of four bases within the core of the homeodomain recognition site (TAAT). The Abd-B homeodomain of Drosophila, which also carries a lysine residue at position 3, has been shown to prefer a T at the second position within the core (TTAT), but this difference in specificity also involves other residues (Ekker et al., 1994). The most closely related Drosophila homeodomain (63% identity) is that encoded by the homeotic gene proboscipedia (Cribbs et al., 1992); the Antennapedia homedomain is 53% identical. To our knowledge, the btn mRNA encodes the smallest reported homeodomain protein.

Nothern blot hybridization indicates that the btn transcript is expressed throughout the embryonic period beginning sometime after 4 hours of development, but is only weakly detectable by blotting during the larval, pupal and adult periods (Fig. 3). Expression is first detectable by in situ hybridization in a group of two to four midline cells in each of the three thoracic and in the first seven abdominal segments at stage 11 (the extended germ band at 5G to 6 hours of development; Fig. 4A). As determined by simultaneous detection of the fushi tarazu and btn mRNAs (not shown), these cells are located between parasegment boundaries, approximately at the position of future segment boundaries. These cells are rounded in shape and are located along the dorsal midline of the meso-dermal bridge that links the two halves of the mesoderm (Fig. 4B), thus indicating their mesodermal origin (Bate, 1993). A mesodermal origin for DM cells was also suggested by Beer et al. (1987) based on the fates of marked mesodermal cells transplanted into unmarked hosts. During germ-band retraction, the bodies of the btn-expresssing cells remain at the midline but begin to extend lateral processes until, by stage 14, the cells appear well differentiated and their processes extend beyond the ventral cord to the sites of muscle attachment (Fig. 4C-G). The number of cells expressing btn at this stage is reduced from four to two per segment by an as yet unresolved mechanism that could involve cell fusion or cell death. We shall refer to the mature, differentiated, btn-expressing cells as dorsal median (DM) cells. Note that in contrast to anterior DM cells, which display a characteristic oval shape with lateral processes, the cell bodies of posterior DM cells form a triangle with processes that extend anteriorly and laterally from the vertices (Fig. 4E,H).

The distinctive morphological characteristics of the DM cells have been noted previously (Caudy et al., 1986). These cells have also been shown to express neuroglian (Bieber et al., 1989) and extracellular matrix proteins such as laminin, glutactin and collagen IV (Olson et al., 1990), molecules that appear to function in cell-cell interactions and adhesion.

Localization of the btn gene to polytene chromosome region 94B by in situ hybridization (data not shown) revealed an absence of nearby mutations or deficiencies, but did permit identification of a promising strain carrying the single P-element insertion P[rib7(94B)] (Karpen et al., 1988). The site of the [rib7](94B) insertion was mapped by Southern hybridization and sequence analysis to a location 1717 bp upstream from the btn translation initiation codon (Figs 1, 2A; see Materials and Methods). This insertion, which carried a ry⁺ eye color marker but otherwise was associated with no phenotype, was exposed to transposase from the P[Δ2-3] to generate btn mutations for phenotypic analysis. Of approximately 65 ry^‐ excision events, two were associated with homozygous lethality; one of these, designated btn, carried a deletion from 0 to +3.2 kb (see Fig. 1). This region includes the entire btn transcription unit, and this deletion therefore constitutes a null mutation.

To study the role of btn in DM cell development, we immunostained embryos from btn heterozygous parents and dissected to expose the CNS. The monoclonal antibody BP102 (Patel, 1994) was used to highlight the commissures and longitudinal connectives of the CNS and antibodies directed against β-galactosidase were used to identify homozygous btn embryos (by absence of β-galactosidase expression from a P-element reporter on the balancer chromosome; see Materials and Methods). In wild-type and heterozygous embryos, DM cells are located between commissures along the segment boundary, and are clearly visible under Nomarski optics (Fig. 5A). In all 20 of the btn homozygotes identified by absence of β-galactosidase staining, however, DM cells were completely absent (Fig. 5B). To rule out the possibility of artifacts generated during dissection of btn mutants, embryos were viewed without dissection after immunostaining with antiglutactin antibody to visualize the DM cells. Homozygotes were once again identified by use of a β-galactosidase-marked balancer and the DM cells were absent in all homozygous mutants (Fig. 5C,D). Fig. 5E,F illustrate the staining by glutactin of a structure located just ventral to the DM cells. This staining was unaffected by the btn mutation and thus serves as a positive control for glutactin staining. The name buttonless for this mutation derives from the apparent loss of DM cells, which resemble a row of buttons when present.

To understand how btn is involved in DM cell determination or differentiation, we examined the effects of the btn mutation upon expression of DM cell markers. Although several lacZ reporter strains express β-galactosidase in the DM cells, the expression patterns are complex or not well characterized with respect to the early stages of DM cell formation (Hartenstein and Jan, 1992; Nelson and Laughon, 1993; Wharton and Crews, 1993). We therefore constructed and characterized a reporter using promoter elements from the btn gene itself. This construct, P[btn1.85-lacZ], contained 1.85 kb of btn genomic sequence, including upstream sequences and the first 32 codons of the btn ORF as an in-frame fusion to lacZ. Three independent transgenic lines harboring this construct generated β-galactosidase expression in an embryonic pattern nearly indistinguishable from that of the btn RNA (Figs 4F-H, 6A). These strains were also useful for characterization of larval btn expression, which is very similar to that of the embryo, including cells with triangular morphology in the posterior segment (Fig. 4H). In embryos and larvae, the β-galactosidase expression is cytoplasmic, suggesting that the presumed nuclear localization signal for the btn protein is situated carboxy-terminal to the first 32 residues.

Expression of this reporter in btn mutant embryos (identified by absence of the marked balancer chromosome) is like that in wild type at stage 11, when normal btn expression first appears in the DM cell precursors (see Fig. 4A). Shortly there-after, however, the level of β-galactosidase expression declines in the mutants relative to wild type. In addition, DM precursor cells in btn mutants fail to maintain the characteristic position and shape of the DM cells (Fig. 6A,B), although the ultimate fate of these precursor cells in btn mutant embryos is uncertain since β-galactosidase expression from the reporter construct is lost by stage 14. These results indicate that btn is required for DM cell differentiation. In contrast, initial commitment to the DM cell fate occurs in the absence of btn function, as indicated by normal initiation of btn reporter expression in btn mutant embryos.

Having developed the ability to ablate the DM cells in Drosophila genetically, we looked to cellular studies of several other larger insects for clues regarding potential function of the DM cells. The muscle pioneer cells of the grasshopper, like Drosophila DM cells, are also mesodermal in origin, have large cell bodies and long processes, and occur in pairs immediately dorsal to each segment of the developing CNS (Ho et al., 1983). In contrast to Drosophila DM cells, the muscle pioneers of the grasshopper appear to serve as a site of recruitment for other mesodermal cells that eventually form the trans-verse muscle, a muscle not seen in Drosophila. In Manduca, a pair of non-neuronal cells that look like muscle pioneer cells also occur dorsal to the CNS at the posterior edge of a cluster of non-neuronal cells called ‘the strap’ (Carr and Taghert, 1988). What is common to these cells in the grasshopper and Manduca is that they appear early during development and are thought to prefigure the growth of the transverse nerve, which courses laterally toward the periphery in each segment.

The transverse nerve of Manduca carries the axons of two motor neurons and 16 neuroendocrine neurons (Carr and Taghert, 1988); a significant proportion of these axons enter the transverse nerve after first coursing posteriorly along the midline within the median nerve, then bifurcating laterally when they encounter the transverse nerve. In Drosophila the median nerve also appears to grow posteriorly along the midline and then bifurcates to contribute to the transverse nerve, which also has been shown to carry axons of identified neurosecretory cells (Zitnan et al., 1993). The median nerve and the transverse nerve can both be visualized by antibody staining for Fasciclin II (Grenningloh et al., 1991; Van Vactor et al., 1993), and we have used this antibody in combination with anti-β-galactosidase antibodies in embryos carrying the btn-lacZ reporter to simultaneously mark these nerves and the DM cells (Fig. 7A). This double stain confirms that the median nerve bifurcates at the point where it encounters the DM cell bodies and that the path followed by the transverse nerve precisely follows the path of the DM cell processes. Anti-Fasciclin II staining of an embryo not carrying the btn-lacZ reporter (Fig. 7C) more clearly illustrates the bifurcation of the median nerve at precisely the point where it meets with and contributes to the transverse nerve; the morphology of these nerves in relation to the DM cells is schematically illustrated in Fig. 8A. The close correspondence between the paths of the DM cell processes and the transverse nerve, particularly in the posterior segments where the DM cell processes form a triangular shape, strongly supports the notion that DM cells may aid in transverse nerve pathfinding.

If DM cell processes serve as a guiding substratum for trans-verse nerve outgrowth, loss of DM cell processes should disrupt formation of the transverse nerve. Indeed, we find that transverse nerves are not present in btn mutant embryos at stage 14, with only the median nerve visible at its outgrowth stage (Fig. 7B). By stage 16, abnormally long and thick axon bundles expressing Fasciclin II can be seen extending along the midline at the dorsal surface of the ventral cord (Fig. 7D). These bundles, although of varying lengths and thicknesses, were observed reproducibly in all 15 mutant embryos characterized at this later stage of development. We presume that these bundles result from outgrowth and fasciculation of axons that contribute to the median nerve in multiple segments (Fig. 8B). These results thus suggest that DM cells are necessary for bifurcation of the median nerve and lateral outgrowth of the transverse nerve.

The btn gene was isolated from a cell line of unusual mor-phology that was established from shibire mutant embryos (Kosada and Ikeda, 1983; Simcox et al., 1985). These cells display a monopolar or bipolar cell morphology with extended processes that resemble those of DM cells (Fig. 9A). Since the initial source of btn sequences was a cDNA library made from cells artificially induced to express Ubx protein, we confirmed by northern blotting that btn mRNA is normally expressed in the shibire cell line and does not require transcriptional activation by Ubx protein (Fig. 9B). In contrast, the btn transcript is not expressed in the Schneider line 2 cells, which display a characteristic round morphology (Schneider, 1972). Given the exquisite specificity of btn expression in the embryo and the obvious morphological similarities between DM cells and the shibire cells, we suggest that the shibire cell line may be representative of DM cells in embryos.

The absence of DM cells in btn mutant embryos indicates an essential role for btn function at some stage during the devel-opment of these cells. This role appears not to be the initial commitment to DM cell fate since expression of the btn-lacZ reporter gene is initiated normally in btn mutants. Expression of the btn-lacZ reporter gene represents the earliest sign of commitment to DM cell fate, and we thus conclude that activ-ities other than those encoded by btn are required for initial commitment to the DM cell fate. The twist (twi) gene appears to be a good candidate for encoding such an activity since it encodes a member of the helix-loop-helix family of transcrip-tion factors (Thisse et al., 1988) and is expressed in cells of the mesodermal bridge (Bate, 1993). Consistent with this idea, we find a total of 6 consensus binding sites (CTNNAG; Ip et al., 1992) for the twi protein within a 300 bp region of genomic sequence immediately upstream of the btn cDNA, and a total of 10 sites within the larger 1.7 kb segment that suffices to confer DM cell-specific expression. Transcription factors encoded by other genes such as dorsal and tinman (Azpiazu and Frasch, 1993; Bodmer et al., 1990; Ip and Levine, 1992; Steward, 1987), which are involved in early specification of mesodermal fates, may also be involved in btn activation, but their expression in cells of the mesodermal bridge has not been documented clearly.

The expression within the mesoderm of regulators, such as twist, dorsal and tinman, is far too general to account for the highly specific and restricted pattern of btn gene expression. A candidate gene for a role in restriction of btn expression is suggested by the work of Hartenstein et al. (1992), who observed that DM cells differentiate in approximately twice the wild-type number in Notch mutant embryos. The phenocritical period for hypertrophy of the DM cells in the temperature-sensitive Notch mutant occurs between 4 and 6 hours after fer tilization (Hartenstein et al., 1992), which coincides with and possibly precedes the time when btn is first expressed. These observations suggest that lateral inhibition mechanisms associated with the function of Notch and other neurogenic genes may restrict the number of cells initially committed to the DM cell fate, possibly by restricting expression of btn.

Although initiation of btn-lacZ reporter expression does not require wild-type btn function, later maintenance of btn expression does. Because we were unable to trace the fate of the DM cell precursors in btn mutant embryos, we can not rule out the possibility that loss of btn reporter expression in btn mutants may be an indirect result of a series of other events, such as failure to differentiate and ultimately death and phago-cytosis of the DM cell precursors. The alternative possibility, that maintenance of btn gene expression requires a shift to autologous regulation by the btn product, is consistent with the extremely A/T rich character of genomic sequences immedi-ately upstream of the btn cDNA and with the presence of many motifs reminescent of homeodomain binding sites (for review see Levine and Hoey, 1988; McGinnis and Krumlauf, 1992).

Wild-type btn function is also required for DM cell differentiation, and we therefore presume a second regulatory function of the btn protein product to be the activation, whether direct or indirect, of genes required for the differentiation and function of DM cells. Such genes might encode specialized cytoskeletal proteins and cell surface proteins involved in the outgrowth and maintenance of the unusually large DM cell processes. In addition, other target genes for btn might encode extracellular matrix proteins and cell surface proteins that play a later role in axon guidance (see below). Expression of btn thus appears to serve the dual functions of autogenous maintenance and implementation of the DM cell differentiation pathway. The execution of both functions by a single regulatory protein represents an economical mechanism for simultaneously maintaining and implementing a differentiated state.

The evidence indicative of a role for DM cells in axon guidance is threefold. First, the extending processes of DM cells are readily visible at stage 12 and are complete by stage 14, the earliest stage at which the transverse nerve can be detected (see Fig. 6A, data not shown). DM cell differentiation thus precedes the bifurcation of the median nerve and lateral outgrowth of the transverse nerve. Second, simultaneous labeling of DM cells with β-galactosidase and of the transverse nerve with anti-Fasciclin II antibodies demonstrates that axons of the transverse nerve grow in close association with processes of the DM cells (Fig. 8A). Finally, genetic ablation of the DM cells results in a failure of the median nerve to bifurcate and in loss of the transverse nerve. The DM cell thus represents a single identified cell capable of serving as a template for guidance of axonal growth.

The mechanisms by which bifurcation of the median nerve and lateral outgrowth of the transverse nerve are guided may involve recognition of cell-surface cues from DM cell processes. The expression by DM cells of many extracellular matrix proteins including laminin (Montell and Goodman, 1989), glutactin, entactin and collagen IV (Olson et al., 1990) is consistent with this possibility. Drosophila laminin in particular can promote cell adhesion and differentiation of primary Drosophila embryo cells (Volk et al., 1990). Similarly, vertebrate laminin has been shown to interact with cell surface receptors such as integrin receptors and with other extracellular matrix proteins to promote adhesion, neurite outgrowth, differentiation and other functions (for review see Adams and Watt, 1993).

In addition to extracellular matrix proteins, DM cells also express neuroglian, a member of the immunoglobulin super-family with homology to the vertebrate neural adhesion molecule L1 (Moos et al., 1988; Rathjen and Schachner, 1984). L1 can interact with N-CAM (neural cell adhesion molecule), another member of the immunoglobulin superfamily (Barthels et al., 1987; Cunningham et al., 1987; Hoffman et al., 1982) with similar structural organization to Drosophila Fasciclin II and 25-30% amino acid identity in the extracellular domains (Grenningloh et al., 1990). The homologies between these interacting vertebrate molecules and their Drosophila counter-parts suggest the possibility that Fasciclin II expressed on axons of the median and transverse nerves may interact with neuroglian expressed on the surfaces of DM cell processes to promote adhesion and guide axon outgrowth. Loss of this inter-action in the btn mutant might also account for the abnormal path and thickness of the median nerve, since in the absence of neuroglian the homophilic adhesion properties of Fasciclin II (Grenningloh et al., 1990) might lead to abnormal fasciculation of axons expressing it.

Isolation of the btn gene has provided a unique view of the commitment and differentiation of an unusual and highly specialized group of cells originating from within the mesoderm. Mutations in the btn gene have further provided us with the means to cleanly ablate DM cells, and this in turn has revealed the role of DM cells as simple cellular cues in axonal pathfind ing. Other demonstrations of axonal guidance by single identified cells generally have relied upon laser ablation in organisms with large, easily identified cells and relatively simple nervous systems. In the grasshopper, for example, laser ablation of a specific muscle pioneer cell in limbs (Ball et al., 1985) or of the segmental boundary cell and the intersegmental nerve root glia (Bastiani and Goodman, 1986) cause the axons of specific identified neurons to be misrouted. In the leech ablation of the axonal runway cell results in loss of the sex nerve (Jellies and Kristan, 1988). Alongside these others, the role of DM cells in axonal pathfinding provides one of the simplest examples of axonal guidance by a single identified cell yet described. The identification of a cultured cell line representative of DM cells should provide novel opportunities for study of the cellular and molecular properties that underlie the unique morphology and function of DM cells. In addition, tests of specific gene function in DM cells should be facilitated by genetic tools such as the btn mutation and the DM cell-specific promoter element.

We thank Gary Karpen for the P[rib7(94B)] line, A. Simcox for the shibire cultured cell line, C. Goodman for anti-Fasciclin II antibody and J. Fessler for anti-glutactin antibody. We also thank M. Bate, M. Caudy, P. Taghert, M. Schubiger and N. Tublitz for helpful discussions. N. H. P. was supported by grants from NIH and a Mcknight Scholars Award.