Dynamic expression of the cell adhesion molecule fasciclin I during embryonic development in Drosophila

Fasciclin I is a 72 × l0³M_r glycoprotein that was originally identified by a monoclonal antibody (mAb) screen for molecules expressed on subsets of CNS axons in embryos of the grasshopper Schistocerca americana (Bastiani et al., 1987). Cell ablation experiments in this organism had provided evidence that CNS growth cones use specific axon bundles as guidance cues (reviewed by Goodman et al., 1984), and the screen was designed to identify surface molecules whose expression patterns correlated with a potential role in guidance. Two other molecules, fasciclin II (in grasshopper) and fasciclin III (in Drosophila melanogaster), were also identified in these types of screens (Bastiani et al., 1987; Patel et al., 1987; Harrelson and Goodman, 1988; Snow et al., 1988). The cloned sequences for fasciclin I and II were used to isolate homologs in Drosophila (Zinn et al., 1988; Grenmngloh et al., 1990, 1991).

The sequence of fasciclin II is closely related to that of the vertebrate neural adhesion molecule N-CAM (Harrelson and Goodman, 1988). All three fasciclins are capable of mediating cell adhesion in tissue culture systems (Snow et al., 1989; Elkins et al., 1990b; Grenningloh et al., 1990).

To gain insights into the functions of these molecules during neural development, mutations in the Drosophila genes encoding the three fasciclins have been isolated (Elkins et al., 1990a; Grenningloh et al., 1991; T. Elkins and C. S. Goodman, unpublished results). Apparent null mutations in the fasciclin I (JasI) and fasciclin III (fasIII) genes are homozygous viable, while a null fasciclin II (fasII) mutation causes lethality when homozygous. FasII mutations prevent formation of a specific longitudinal pathway that expresses the protein (Grenningloh et al., 1991). The other mutations do not produce alterations in the embryonic CNS that can be visualized with currently available antibody reagents. Defects in individual pathways are usually very difficult to detect, however, because few axon-specific antibodies exist. For fasciclin II, it was possible to visualize the affected longitudinal pathway in fasII mutants with the 22C10 mAb, which labels a restricted set of axons in stage 12 and 13 embryos (Grenningloh et al., 1991).

Analyses of antibody-stained embryonic CNS preparations from fasl and fasIII mutants do show that many of the axons that express these proteins follow normal pathways in the mutant embryos, suggesting that if they are involved in axon guidance their functions can be substituted for by other molecules (Elkins et al., 1990a; T. Elkins, R. Jacobs, and C.S.G., unpublished results). This apparent functional redundancy correlates with observations made on vertebrate cell adhesion molecules in tissue culture systems, where it has been found that antibodies against several different molecules are required to block axon fasciculation or neurite outgrowth (Tomaselli et al., 1986; Chang et al., 1987; Neugebauer et al., 1988).

In the case of fasciclin I, a requirement for this molecule in embryonic neural development can be uncovered by constructing a double mutant stock lacking expression of both fasciclin I and of the Drosophila homolog of the abl tyrosine kinase, which is expressed on CNS axons (Gertler et al., 1989). In fasI/abl mutant embryos, the commissural axon pathways are often completely absent. The mutant combination can also be shown to alter pathfinding by identified neurons. The fasciclin I-positive RP1 motoneurons normally extend growth cones across the midline and exit the CNS in the contralateral intersegmental nerve (ISN) pathway, but in mutant embryos their axons usually fail to cross over and instead grow out in the ipsilateral ISN (Elkins et al., 1990a). The abl mutation alone does not cause embryonic lethality, and no gross alterations in the CNS are observed in abl mutant embryos (Henkemeyer et al., 1987; Gertler et al., 1989).

The synergistic interaction between the fasI and abl mutations suggests that fasciclin I does not simply mediate adhesion, but participates in a signal transduction pathway involved in growth cone extension or guidance. This pathway may function in parallel with a pathway involving abl, so that elimination of one pathway does not cause severe defects, but removing both pathways prevents correct formation of the CNS axon array.

The sequence of the fasciclin I protein does not clearly define its function, as it is not related to other sequences in the current databases. It consists of four homologous 150 amino acid domains, and is linked to the external cell surface by a glycosyl-phosphatidyl-inositol (GPI) linkage (Zinn et al., 1988; Hortsch and Goodman, 1990). Alternative splicing of two 6–9 bp micro-exons located between the coding regions for the second and third domains produces mRNAs encoding three different isoforms of the protein in each species (McAllister et al., 1992). In the present study, we have examined in detail the pattern of expression of fasciclin I mRNA and protein during development of the Drosophila embryo. These results help to explain the origin of the fasI/abl phenotype, and may provide insights into the possible roles of this molecule in embryonic development.

Immunocytochemistry on whole-mount embryos was performed according to Patel et al (1987, 1989). Ascites preparations (1:500) or cell supernatants (1:2) of the anti-fasciclin I monoclonal antibody 6D8 (Hortsch and Goodman, 1990) were used as the primary antibody Photographs were taken using Nomarski optics on either a Zeiss Axiophot or a Nikon compound microscope.

In situ hybridization to whole mount embryos was performed essentially according to the protocol of Tautz and Pfeifle (1989). Non-radioactive probes were made by incorporating digoxigenin-dUTP (Boehringer Mannheim) either by pruning with random hexamers or by synthesizing anti-sense single-stranded DNA probe using a single primer PCR reaction. Digoxigenin was detected using a monoclonal antibody against digoxigenin coupled to alkaline phosphatase. The alkaline phosphatase reaction was carried out using NBT and X-phosphate as supplied in the Boehringer Mannheim Genius detection kit, or with a Vector Laboratories ABC kit. Photographs were taken as for antibody staining.

We examined the expression of fasciclin I RNA during development using the digoxigenin whole-mount in situ hybridization method of Tautz and Pfeifle (1989). There are two major fasciclin I transcripts, of 3.0 and 5.0 kb, which differ only in their 3’ untranslated regions. We used both a coding region probe and a probe specific to the large RNA. Identical RNA expression patterns were observed with the two probes (data not shown). Fasciclin I protein expression was visualized using horseradish peroxidase (HRP) immunohistochemistry with a mAb raised against a bacterially expressed fasciclin l/protein A fusion (Hortsch and Goodman, 1990).

Fasciclin I mRNA is maternally deposited in the egg, and the message is initially uniformly distributed. At about 1 h after egg laying (AEL) the nuclei, which have undergone 7 synchronous divisions in the central region of the egg, begin to migrate out to the periphery. The fasciclin I RNA hybridization signal becomes localized with the nuclei, so that it forms a broad annulus (data not shown). After the nuclei reach the periphery they divide three more times, and then cellularization begins. Just before cellularization, fasciclin I RNA is observed in a ring at the level of the nuclei (Fig. 1A).

Fasciclin I protein is initially uniformly distributed on the surface of the egg, and it first appears when the outlines of the forming cell membranes become evident. During the cellularization process there is a period of about 10–15 min in which fasciclin I protein is expressed at a low level in the anterior third of the egg and at a higher level in the posterior two thirds. The highest levels of protein observed at this time are on the membrane at the extreme posterior end of the egg, adjoining the pole cells (Fig. 1B). Pole cells are the germ line precursors, and segregate away from the rest of the embryo very early in development, at the time of the 9th cleavage division. The pole cells themselves do not express the protein at this time. After this stage, fasciclin I protein again becomes uniformly distributed, and at the beginning of gastrulation (2:50 h AEL, stage 6 of Campos-Ortega and Hartenstein, 1985) it is expressed at a high level on all cell membranes. During gastrulation (2:50-3:20h AEL), fasciclin I mRNA is apparently uniformly distributed (data not shown). The protein, however, is expressed at higher levels on the invaginating cells of the ventral and cephalic furrows (Fig. 1C).

Immunohistochemical analysis of the progeny of crosses in which there is no maternal fasciclin I expression (fasI^TE/fasI^TE females×+/+ males), or in which 1/2 of the progeny lack zygotic fasciclin I expression (+/fasI^TE females×/fasI^TE males) indicates that the transition between maternal and zygotic expression takes place at about the time of gastrulation (data not shown).

Immediately following gastrulation, the fast phase of germ band elongation begins. The germ band refers to the segmented region of the embryo, which gives rise to three gnathal, three thoracic, and at least nine abdominal segments. During germ band elongation the segmental primordia fold over the posterior end of the embryo and extend forward as far as 75 % egg-length (EL; measured from the posterior end). The embryo remains in the germ band extended stage for several hours (3:40–7:20 hours AEL, stages 9–11). During this time neuroblasts delaminate from the ectoderm and divide asymmetrically to produce a chain of ganglion mother cells (GMCs). Each GMC then divides once, giving rise to two neurons. This process eventually generates approximately 250 neurons per hemisegment, and these will form the segmental ganglia of the CNS.

In the 4–5 h old embryo fasciclin I RNA is localized in 14 stripes, which correspond to segmentally repeated bands of neuroepithelial cells and delaminating neuroblasts (Fig. 1D, stage 9). The protein has a similar distribution, although it appears slightly later and seems to be expressed at lower levels in the epithelial layer (Fig. 1E, stage 10). These stripes of neuroblasts and neuroepithelial cells are localized to the posterior compartment of the segment, or the anterior region of the parasegment (Martinez-Arias and Lawrence, 1985; data not shown). This was determined by double staining with anti-fasciclin I mAb and antibodies against β-galactosidase in a transgenic line containing a β-galactosidase gene under the control of the promoter of the fushi tarazu gene, which is expressed in 7 stripes at the anterior border of alternating parasegments (Hafen et al., 1984).

In later germ band extended embryos the hybridization signal moves inward into the embryo and becomes a single longitudinal stripe running the entire length of the segmented region (Fig. 1F, stage 11). Immunocytochemistry with the mAb shows that the protein has a similar distribution (data not shown). This stripe corresponds to the neuronal layer, in which many cells express fasciclin I. These cells are localized in both compartments of the segment, so a banded pattern is no longer observed. We do not know if the early neurons expressing fasciclin I are primarily derived from the fasciclin I-positive neuroblasts.

Each thoracic and abdominal segment of the embryo contains five classes of peripheral sensory neurons localized in three clusters: the ventral, lateral, and dorsal clusters (Campos-Ortega and Hartenstein, 1985; Ghysen et al., 1986). The external sensory (es) neuron and chordotonal (ch) neuron classes both have single dendrites, and there are three classes of neurons with multiple dendrites (md neurons). The precursors of these PNS neurons undergo their final mitoses at about 6:30h AEL (Bodmer et al., 1989). The dorsal cluster appears first, followed closely by the lateral and ventral clusters. These cells then extend axons into the CNS, and the mature pattern of PNS neurons appears by about 11 h AEL.

At the beginning of germ band retraction (about 7 h AEL) small, segmentally repeated clusters of fasciclin I RNA-expressing cells can be visualized under the dorsal ectoderm (data not shown). These are the dorsal cluster neurons. As germ band retraction proceeds, two other clusters of hybridizing cells appear, and all three clusters change shape and grow as the embryo develops (Fig. 1G). This embryo is at about stage 12/5 of Klambt et al. (1991). These authors have subdivided stage 12 into three substages, 12/5, 12/3, and 12/0, where the second number corresponds to the number of segments still on the dorsal side of the embryo. Fig. 2A shows in situ hybridization to neurons of the ventral and lateral clusters in a stage 14 embryo. Fig. 2B shows the lateral and dorsal clusters. All of the PNS neurons appear to express fasciclin I RNA, as do many of their support cells.

The early cells of the PNS also express fasciclin I protein. This staining is less obvious than in in situ hybridization experiments, because the surrounding ectoderm also stains with the antibody (Fig. 3C,D; stage 12/0). This ectodermal staining could be due to soluble fasciclin I which is bound to these cells (Hortsch and Goodman, 1990), as we do not see clear expression of fasciclin I RNA in the ectoderm at this stage by in situ hybridization.

Later in development fasciclin I protein is expressed on the surfaces of all cell bodies and axons in the PNS (Fig. 3B). The ectodermal expression disappears shortly after germ band retraction is completed. Fasciclin I protein is also observed on the surfaces of glial cells that enwrap the PNS axon bundles (Fig. 3A). In addition to the segmentally repeated PNS cells, there are several specialized sensory structures in the head and tail. Most or all of the cells in these organs also express fasciclin I protein (Zinn et al., 1988). Fasciclin I RNA and protein are retained by some or all of the PNS cells, albeit at reduced levels, until the embryo develops cuticle and it is no longer possible to do whole-mount staining (about 15 h AEL; Fig. 1H shows in situ hybridization to a stage 16 embryo).

CNS axonogenesis begins in stage 12/5 with the formation of the commissural tracts. The posterior commissure is established first, followed rapidly by the anterior commissure. The commissures are initially fused at the midline (see Fig. 2D), but then become separated by the midline glial cells (Klambt et al., 1991). By stage 15, a complete orthogonal axon scaffold has formed, with anterior and posterior commissures, a bilaterally symmetric pair of longitudinal axon tracts which extend the length of the embryo, and two peripheral nerve roots per hemisegment (the ISN and the segmental nerve (SN)), where motor-neuron axons exit and sensory neurons enter the CNS.

Most or all of the initial commissural axons express fasciclin I protein at approximately equal levels (Figs 2D, 4A). The ventral unpaired midline neurons (VUMs), whose growth cones and axons lie at the junction between the fused anterior and posterior commissures at stage 12/0, express fasciclin I on both their cell bodies and axons at this stage (Fig. 4A). This expression is transient, and disappears shortly after the separation of the commissures. Other early neurons expressing fasciclin I protein on their cell bodies include the aCC and RP1 cells, and a large laterally located cluster of neurons. In contrast to the transient VUM expression, these cells continue to express the protein for several hours. aCC and RP1 are motoneurons and extend axons in the ISN pathway. Fasciclin I is expressed on this pathway at high levels throughout development.

Shortly after the commissures have separated, the expression of fasciclin I on commissural axons becomes restricted to a single large bundle at the posterior edge of the posterior commissure, adjacent to the aCC cell bodies, and a smaller bundle in the middle of the anterior commissure (Fig. 4D,E). The protein disappears from the rest of the commissural axon bundles. The axons of the lateral cluster neurons continue to express fasciclin I, and most of these axons form a tight bundle which appears to join the fasciclin-I positive bundle in the posterior commissure (Fig. 4D,E). There is also a second, smaller fasciclin I-positive tract connecting these neurons to the anterior commissure. The protein is also expressed on a single bundle at the medial edge of the longitudinal tract. The changing pattern of fasciclin I protein expression in the CNS is schematically depicted in Fig. 5.

In situ hybridization shows that fasciclin I RNA is always restricted to a subset of CNS neurons (Fig. 2C), and there is no evidence that this subset is larger at early stages of CNS development. Thus, the early expression of fasciclin I protein on all commissural axons may indicate that a subset of neurons that express fasciclin I RNA (perhaps at low levels) pioneer the commissures, or that fasciclin I cleaved by an endogenous phospholipase (Hortsch and Goodman, 1990) can bind to axons of nonexpressing cells.

In addition to the early nonneural expression described above, fasciclin I RNA and protein are also observed on a variety of other cells and tissues during development. The cells at the dorsal edge of the epidermis, which grow together over the amnioserosa during the process of dorsal closure, express fasciclin I RNA and protein at high levels (Fig. 2B). This process is completed during stage 15, and shortly after dorsal closure these cells cease to express fasciclin I.

The gonads express fasciclin I RNA (Fig. 1H) and protein (data not shown), and this expression continues until cuticle has formed and the embryo is no longer accessible. We do not know precisely when this expression begins, as it is difficult to visualize the pole cells prior to their incorporation into the gonad. Pole cells do not express fasciclin I during the germ band extended stage. In the coalesced gonad of the later embryo, both pole cells and mesodermal support cells appear to express the molecule.

Fasciclin I is expressed by all three cell types (glandular epithelium, separate and common ducts) of the salivary glands. Sections of the posterior midgut and of the hindgut also express the protein. Finally, fasciclin I protein expression is observed on the anterior and posterior spiracles, which are the outlets of the tracheae. Both neural and nonneural expression of fasciclin I protein is eliminated by the fasI^TE mutation, showing that all of the mAb staining observed is due to authentic fasciclin I (data not shown).

The results described above show that the cell adhesion molecule fasciclin I is expressed in a complex pattern throughout embryonic development. The protein is uniformly expressed on nascent cell membranes at the end of the synctitial blastoderm stage. It is briefly distributed in a graded pattern in the cellular blastoderm (Fig. 1B), after which expression again becomes uniform. In the gastrulating embryo all cells express fasciclin I, but invaginating cells have higher levels of the protein on their surfaces (Fig. 1C).

During the germ band extended stage, fasciclin I expression evolves through two distinct phases. In stage 9-10 embryos, fasciclin I is localized to 14 segmentally repeated stripes of neuroepithelial cells and neuroblasts (Fig. 1D, E). During stage 11, fasciclin I RNA and protein disappear from the neuroblast layer and become concentrated in a single continuous longitudinal stripe of CNS neurons (Fig. 1F). The early cells of the PNS, which first become visible just prior to germ band retraction, also express fasciclin I RNA (Fig. 1G) and protein (Fig. 3C, D). As germ band retraction proceeds, fasciclin I protein expression on the ectoderm increases, and later disappears during stage 14 (Fig. 3; compare B with C and D). Expression on the cell bodies and axons of the PNS, in contrast, is stable; the surfaces of these cells retain fasciclin I protein until the embryo is no longer accessible to antibody.

The CNS expression pattern of fasciclin I during axonogenesis can be described as a combination of transient and stable components. During the formation of the commissures (stage 12), fasciclin I protein is found on most or all axons (Figs 2D, 4A-C). This pattern of expression lasts for less than an hour. After this time, fasciclin I is expressed at high levels only on a single bundle in each commissure, on a bundle in the longitudinal tract, and on the axons of the ISN and SN (Fig. 4D,E). The cell bodies of the VUM neurons express fasciclin I during the commissural formation phase, but become fasciclin-I negative thereafter. In contrast, the aCC, RP1, and lateral cluster neurons maintain fasciclin I expression on their cell bodies until at least stage 16 (summarized in Fig. 5).

In addition to neural expression, the germ band retracted embryo expresses fasciclin I in a variety of nonneural tissues. These include the dorsal edge of the epidermis, the salivary glands, the gonads, sections of the hindgut and posterior midgut, and the spiracles. The duration of fasciclin I expression varies between these different tissues. For instance, the dorsal epidermis expresses fasciclin I only during the process of dorsal closure. In contrast, expression in the gonads is very stable.

The expression patterns of the two other Drosophila fasciclins have many of the characteristics of the fasciclin I pattern described here. Fasciclins II and III are both dynamically expressed in neural and nonneural tissues. In the CNS, fasciclin III is transiently expressed on 5 commissural bundles and on part of the ISN pathway. Like fasciclin I, it also appears on a segmentally repeated subset of neuroepithelial cells and underlying neuronal lineages at the germ band extended stage. At this stage, it is also expressed on patches of epithelial cells near the stomodeal and proctodeal invaginations and on the visceral mesoderm. After germ band retraction, it is expressed on luminal cells of the salivary gland, and in a 14-stripe pattern on the ectoderm. These stripes coincide with the segmental grooves (Patel et al., 1987). Fasciclin II is initially expressed on a single longitudinal CNS pathway, and appears on several other pathways later in development. It is also expressed in patches of ectoderm from which the spiracles arise, as well as in a section of hindgut and in the malpighian tubules (Grenningloh et al., 1991).

These results suggest that the function of neural adhesion molecules in Drosophila embryogenesis is not restricted to nervous system development. These molecules are likely to be used for morphogenesis and adhesion in a variety of different tissues. In some cases they may form stable adhesive zones that are involved in maintaining the structure of a particular organ. In tissues where they are only transiently expressed, they may facilitate morphogenetic movements or cell signaling events.

Although grasshoppers and Drosophila develop to adulthood in very different ways, the overall structure of the early embryonic CNS is very similar in the two organisms. This resemblance extends to the axon trajectories of certain identified neurons (Thomas et al., 1984). Comparison of the expression patterns of fasciclin I in the two species might thus yield insights into its function.

In both grasshopper and Drosophila, fasciclin I is expressed at an early stage on segmentally repeated subsets of neuroepithelial cells. Later in development, all PNS cell bodies and axons express fasciclin I in both insects. In the grasshopper CNS, fasciclin I protein is localized on a single bundle in each commissure, on a bundle in the longitudinal tract, and on the axons of the ISN. This pattern is almost identical to the later expression in the Drosophila CNS, and this resemblance extends to particular identified cells such as aCC and RP1, which express the molecule in both organisms (Bastiani et al., 1987; Zinn et al., 1988; this work).

The pattern of fasciclin I expression in grasshopper and Drosophila differs in two major ways. First, fasciclin I is localized to patches of epithelium in the grasshopper limb buds. There are no counterparts to these structures in fly embryos. Secondly, the early, transient phase of CNS expression in Drosophila is not observed in grasshopper. In Drosophila, the molecule is initially expressed on all commissural axons and later becomes restricted to particular bundles. In contrast, fasciclin I is localized to specific pathways from the beginning of axonogenesis in grasshopper. This may correlate with differences in the way the commissures are established in the two insects. In Drosophila, the two commissures are initially fused and then become separated by insertion of a midline glial cell, MGM (Klambt et al., 1991). In the much larger grasshopper embryo, the anterior and posterior commissures are always separate.

The fasI/abl phenotype (absence of commissural axon tracts) is clearly not due to a failure of pathfinding by those axons that comprise the fasciclin I-positive bundles observed in later Drosophila embryos (Fig. 5), because all commissural axons are affected. Rather, it appears that there is a general defect in growth cone extension across the midline (Klambt et al., 1991). This correlates well with the early, transient expression of fasciclin I on all commissural axons reported here.

We do not know what mechanism accounts for the failure of commissural axons to cross the midline in fasl/abl embryos. The specialized midline glia and neurons are essential for commissure formation, but these seem to be in their normal positions in mutant embryos (Elkins et al., 1990a). One possibility is that commissural growth cones must recognize and interact with the VUM cells, whose growth cones and nascent axons he at the fusion point of the two commissures. In orthodenticle mutants the VUM cells die and the posterior commissure does not form (Klambt et al., 1991). Since fasciclin I has been shown to be a homophilic adhesion molecule in tissue culture (Elkins et al., 1990b), and the commissural axons and VUM cells both express the molecule at this stage, it is possible that elimination of fasciclin I-mediated growth cone adhesion to the VUMs is a component of the fasI/abl phenotype.

Because the commissures are absent in fasI/abl embryos, we cannot determine whether fasciclin I also has a role in specific pathfinding by later growth cones that might use the fasciclin I-positive bundles as guidance cues. This is analogous to the situation in slit mutant embryos, where the early nervous system collapse phenotype (due to the absence of midline slit expression) prevents the observation of a possible phenotype correlated with the later expression of slit product on CNS axons (Rothberg et al., 1990).

FasI/abl embryos are essentially normal except for their alterations in the CNS axon array (Elkins et al., 1990a). This is probably because the CNS is the major tissue in which both fasciclin I and abl are expressed. The double mutant phenotype would thus be most likely to be observed in the CNS. Fasciclin I expression outside the CNS is likely to have a functional role, however, since its pattern has been conserved over the 300 million years of evolution that separate Drosophila from grasshoppers. In order to uncover such a role, it will be necessary to find mutations in other genes that produce a synergistic phenotype when combined with fasI mutations.

A possible developmental role that is consistent with the early nonneural expression of these apparently redundant adhesion molecules could be to ensure that morphogenetic movements take place normally in all embryos. We have observed that in collections of 11–13 h old embryos from wild-type stocks, about 1% of embryos are severely abnormal. The most common phenotype observed in these embryos is a failure of germ band retraction; gastrulation/germ band extension defects are also observed. In egg collections from homozygous fasI flies, the percentage of these abnormal embryos increases to 10–12%, and, in collections from homozygous fasI/fasIII stocks, to 30–35 %. This phenotype of fasI mutants can be rescued by P-element transformants containing the fasciclin I gene (K. Zinn, T. Elkins, and C. S. Goodman, unpublished results).

This result suggests that these and other cell adhesion molecules could act as ‘correctors’ in morphogenesis. We hypothesize that although most mutant embryos lacking a particular adhesion molecule will be able to develop normally, embryos in which cells are in abnormal positions may fail to develop because they are unable to correct these defects. The fraction of embryos in this class would increase still further if multiple correcting functions are removed by mutation, because the extent of tolerable abnormality in cell position would be decreased as each molecule was eliminated (the ‘window of normality’ would shrink in size). That such a correcting mechanism actually exists is illustrated by embryos derived from mothers having different numbers of copies (1, 3 or 4) of the bicoid gene, which determines anterior cell fates. These embryos vary greatly from normal in the positions of their cephalic furrows and in the placement of pair-rule gene expression stripes, yet they all develop into normal larvae and adults (Driever and Nusslein-Vollhard, 1988).

Involvement in a correcting mechanism could provide a selection for the evolution of genes encoding new cell adhesion molecules, even if they do not initially have functions independent of those of previously existing molecules. The new adhesion proteins could then evolve new roles and patterns of expression that would allow the development of greater complexity in the nervous system.

We thank Shin-Shay Tian for performing some of the whole-mount in situ hybridization experiments, Michael Hortsch for the anti-fasciclin I mAbs and for help with initial immunocytochemistry experiments usmg these mAbs, Karen Jepson-Innes and Susan Parkhurst for help with the whole mount in situ technique, Dolores Ferres-Marco for help with immunocytochemistry, Yash Hiromi for a stock carrying the ftz-lacZ insertion on the X chromosome, Bruce Alberts for helpful discussions concerning bicoid phenotypes, and Nipam Patel for invaluable advice This work was supported by grants from the NIH to C S.G. and K.Z.