Neural enhancer-like elements as specific cell markers in Drosophila

The structural and biochemical complexity of the nervous system is somehow coded for by the genome. How this is done, and how much genetic information is required to generate this complexity, is not known. Estimates of the number of unique mRNA sequences suggest that many more genes are expressed in the nervous system than in any other tissue (Chikaraishi, 1979; Chaudhari & Hahn, 1983; Milner & Sutcliffe, 1983). However, we know nothing of the distribution of these unique transcripts, most of which are present at a low abundance, and therefore we have no idea about their function. Monoclonal antibodies specific for small subsets of neurones have been described in many species (Barnstable, 1980; Zipser & McKay, 1981; McKay & Hockfield, 1982; Miller & Benzer, 1983), suggesting a high level of molecular specification. However, it is at present not possible to estimate the total antigenic diversity present in a system. Furthermore, this diversity might reflect the expression of many genes, or may result from variable processing or post-transcriptional modifications of only a few genes.

A more direct approach to the genetics of the nervous system would be to search for genes that are specifically expressed in subsets of neurones. It should then be possible to estimate the number of such genes, to analyse their distribution and how they are controlled, and to assess their role in neural development and function. This approach classically involves a screen for mutations producing some sort of defect in the system to be studied. In the case of the nervous system, defects such as the absence of a small subset of neurones would probably lead to early lethality but would otherwise be very difficult to detect. This probably explains the paucity of mutants known to alter the structure or connectivity of the nervous system, relative to the wealth of mutations that affect the development of the epidermis.

A method has been recently developed that allows the detection of transcriptional regulators present in the genome of Drosophila (O’Kane & Gehring, 1987) or of the mouse (Allen et al. 1988). This method is based on the fact that many regulatory sites in eucaryotes act at a distance of up to several thousand base pairs of the promoter, and in either orientation. If a weakly transcribed reporter gene is inserted at random locations in the genome, it may then reveal the properties of whichever regulatory site happens to be near the site of insertion. In this paper, we describe four of the Drosophila transgenic lines isolated previously by O’Kane & Gehring (1987) where the reporter gene, lacZ of Escherichia coli, is expressed in subsets of peripheral neurones. We show that this method can reveal exquisitely specific patterns of expression, and generate potentially useful cell markers. The markers described here have enabled us to trace the early development of the peripheral nervous system.

We found that the sensitivity of the galactosidase assay based on the cleavage of the chromogenic substrate X-gal is very similar to that obtained by immunostaining using mouse antigalactosidase (Promega) revealed by the biotin-avidin-peroxidase system (Vector labs). Furthermore, the X-gal staining achieves a better contrast since there is no background even after prolonged incubation with the substrate. On the other hand, the X-gal-stained embryos are mounted in a glycerolbased medium while immunostained embryos can be dehydrated and cleared, resulting in better optical properties. For the present work, we relied mostly on X-gal staining, since this method is much simpler and cheaper than immunostaining, and all the figures are of X-gal-stained preparations. We wish to stress three aspects of our protocol for X-gal staining: first, the embryos need not be devitellinized, second, the disodium phosphate solution should be prepared just before use, and third, the usual X-gal solvent, dimethyl-formamide, should not be used.

All chemicals are Merck except X-gal (Boehringer).

Phosphate–citrate buffer, pH 7·6: 9 vol. of a fresh solution of Na₂HPO₄ 0·2 M (0·89g Na₂HPO₄. 2H₂O in 25ml water), 1 vol. of citric acid 0·1 M, 10 vol. of water. The pH of the buffer is not critical from 7 to 8.

Fixative: 1 vol. of formalin (= 37 % formaldehyde solution), 10 vol. of phosphate–citrate buffer. Using freshly prepared paraformaldehyde instead of formalin makes no difference.

Incubation buffer: phosphate–citrate buffer containing 5 mm each of potassium ferri- and ferro-cyanides and 0·02 % Triton. This is obtained by adding 33 mg of K₃Fe(CN₆), 42 mg of K₄Fe(CN₆) and 0·2 ml of 2% triton X-100 to 20 ml of freshly prepared phosphate–citrate buffer.

Mounting medium: let 7 g of gelatin swell for a few min in 42 ml of water, dissolve in boiling water bath. Add 63 g of glycerol and a crystal of phenol as bactericide. Keep at 4 °C. To use, warm and keep at 45 °C.

Dechorionate the embryos in bleach, rinse and transfer to a 1:1 mixture of fixative and heptane in a tube large enough for the embryos to form a single layer at the interphase. Shake vigorously, let rest 5 min, and repeat the shaking–resting two more times (15 min fixation). Discard the bulk of the heptane (upper phase) and replace the fixative twice (lower phase) with phosphate–citrate buffer. Transfer the embryos with a Pasteur pipette in a dissection dish, blow gently to evaporate the remaining heptane. As soon as no heptane is left, replace the phosphate-citrate by incubation buffer, sink the embryos and transfer into a plastic well (96- or 24-well plates). Add a saturating amount of X-gal powder (less than 1 mg will be more than adequate for 0·2 ml of incubation medium) and incubate in a moist chamber at 30 °C. For embryos homozygous for the different transformant chromosomes described here, overnight incubation will give an intense staining. Heterozygous embryos usually require two days. Incubation can be prolonged for several days if needed (the enzyme appears to remain active up to at least five days at 30 °C in the incubation buffer). Check periodically that undissolved powder remains in the medium. If many embryos are stained in a small volume of incubation buffer, the buffer should be renewed periodically as the ferri/ferrocyanides are used up in the process, eventually resulting in a discoloration of the normally yellow buffer and the diffusion of the blue reaction product. Stained embryos can be kept for several month in incubation buffer at 4 °C.

Imaginai discs are dissected from mature larvae and transferred immediately to the fixative. After 15 min fixation, the discs are rinsed in phosphate–citrate buffer, transferred to 0·2 ml incubation buffer in a plastic well and enough X-gal powder is added to saturate the buffer. The presence of triton is essential even when staining imaginai discs, presumably to increase the solubility of X-gal. Incubation is carried over at 30 °C for at least two days.

Replace the incubation buffer with a 1:2 mixture of fresh incubation medium and of glycerol. The embryos will first shrink and recover their normal shape after a few min. Put a small drop of warm mounting medium on a slide, position the embryo, let the drop set, check the positioning of the embryo and apply a small (5×5 mm) coverslip. These preparation can be kept at room temperature for at least several months. Pictures were taken with a red filter and with or without Nomarski optics on a Zeiss microscope equipped with a regular × 16 or a neofluar immersion × 25 objective.

The four transformant lines described in this paper were isolated by O’Kane & Gehring (1987). The position of the P[lac,ry⁺]A inserts was determined by in situ hybridization to salivary gland chromosomes with the probe Carnegie20 (Rubin & Spradling, 1983). Hybridization of nick-translated biotin-labelled probe was performed as described by LagerSafer et al. (1982). 5-biotin(19)-2’ -deoxyuridine-5’ triphosphate was obtained from Calbiochem. Detek-I-hrp (Enzo Biochemicals), a complex of streptavidin and biotinylated horseradish peroxidase, was used for the detection of the hybridized probe. The positions of the inserts are, for the transformant line A2: 43E1-5, for A18: 85A1-4, for A31: 4B1-4, for A37: 80A.

The combination AS-C⁻; A37 was obtained by crossing Df(I)260.1/FM4 females with homozygous A37 males (see Lindsley & Zimm, 1987, for a description of the deficiencies and Lindsley & Grell, 1968, for a description of the other chromosomes used in this work). In late embryos, the absence of the cuticular derivatives of external sense organs allowed us to identify the AS-C⁻ embryos with certainty. In early embryos such as those shown in Fig. 7, we had no independent markers to identify the embryos that were Df(I)260.1/Y and therefore lacked AS-C entirely. However, the phenotypes illustrated in Fig. 7D-F and present in about one fourth of the embryos are so strikingly different from normal that we feel confident they correspond to AS-C⁻ embryos.

Two combinations of da⁻-, A37 were analysed, using either the mutation da^IIB31 (Caudy et al. 1988) or the deficiency Df(2L)J27 which removes the da gene (Sandler, 1977). In the first case, we used a da^IIB31/CyO; A37/A37 line established by C. Dambly-Chaudiere. One fourth of the embryos showed a slightly twisted phenotype at the germ band elongation stage, either because of the da mutation or more likely because of some other mutation on the da^IIB31 chromosome. Whatever the reason, this allowed us to identify the homozygous da embryos at the critical early stages of A37 expression. In the case of the deficiency, we crossed Df(2L)J27/ + ; A37/+ males and females. Homozygous Df(2L)J27 embryos do not show the twisted phenotype; however, the different stages of mutant phenotypes observed in homozygous da^IIB31 embryos were also observed in about a fourth of the stained embryos, confirming that the lack of precursor staining in da^IIB31 embryos is the result of the inactivation of the da gene. Fig. 9 shows the results obtained with the deficiency, since the slight twisting of the da^IIB31 embryos makes focusing difficult.

Two A37stg recombinants were used, one kindly given to us by Y. N. Jan and L. Y. Jan and another one obtained by selecting ca recombinants from females heterozygous for a multiply marked rucuca chromosome carrying the allele stg^3M53, and for A37. An A 18stg recombinant was obtained by a similar cross.

We have used the descriptions of Campos-Ortega & Hartenstein (1985) and of Wieschaus & Niisslein-Volhard (1986) to define the developmental stages of the embryos. The transition from one stage to the next is progressive, though, and therefore giving the stage at which a given pattern is observed is mostly indicative. Thus we occasionally had to resort to expressions such as ‘slightly later’ or ‘at about the same time’ to define more precisely the time relationship between two events.

We analysed four transformant lines carrying the transposon P[lac,ry⁺]A that show lacZ expression in at least some peripheral neurones. These lines were designated P [lac,ry⁺]A2, or A2 in short, P[lac,ry⁺]A18 or A18, P[lac,ry⁺]A31 or A31 and P[lac,ry⁺]A37 or A37. The line A18 was briefly described previously as ‘insertion 18’ (O’Kane & Gehring, 1987). The location of the inserts in these lines are given in Materials and methods. In the following, we will describe in some detail the pattern of expression of lacZ in these lines. As a background for this description, Fig. 1 represents the entire complement of peripheral neurones in two representative body segments of the late embryo, the metathoracic (T3) and first abdominal (A1) segments (Ghysen et al. 1986). The pattern in T2 is identical to that of T3, and A2–A7 are identical to Al. Five types of peripheral neurones can be distinguished: ch neurones that innervate chordotonal organs (blackened in the Figure), es neurones that innervate external sense organs (shaded in Fig. 1), da neurones that develop dendritic arborizations, bd neurones that are bipolar and tr neurones that innervate tracheal branches (Bodmer & Jan, 1987). The neurones are distributed in four loose clusters within each segment, from ventral to dorsal: v, v’, 1 and d. Only those neurones and sense organs that will be referred to in the following sections are named.

The overall pattern of expression of lacZ in A37 embryos after dorsal closure (stage 15, see Materials and methods) is shown in Fig. 2A (dorsolateral view) and 2B (lateral view of another embryo). Most of the labelled cells belong to the sensory system. Fig. 2A illustrates the presence of labelled cells at locations that correspond to the external sense organs. Clear examples are the dorsalmost abdominal papillae p8 and p9 and the four spiracular sense organs (sp) in Fig. 2A, and the abdominal papillae p4 and p5 in Fig. 2B. In addition clusters of labelled cells are also found at the positions expected for the dorsal (d) and lateral (1) clusters of neurones (Fig. 2A and B), as well as at the level of the ventral clusters v′ and v (partly in focus in Fig. 2B).

The lateral region of one abdominal segment (box in Fig. 2B) is shown at a higher magnification in a different embryo of the same age, in Fig. 3. Panel 3A focuses at the level of the cuticle to show the sensory hairs hl and h2, and the position of papilla p7, not visible in this picture. The second panel focuses on the epidermis to show intensely labelled cells under each sense organ (arrows). The third panel focuses under the epidermis and shows the neurones lesA, lesB and lesC (see also Fig. 1) innervating, respectively, hl, p7 and h2 (Dambly-Chaudière & Ghysen, 1986). Also labelled are the neighbouring da neurones, the row of five lateral chordotonal neurones (lch5) and their sheath cells (lsh5), the isolated chordotonal neurone v′chl, and part of the dorsal cluster.

As far as we could see, no known peripheral neurone remains unlabelled (at least in all the cases where individual neurones can be identified). It is more difficult to decide whether all non-neuronal cells that contribute to the formation of sense organs are also labelled, in part because our knowledge of these cells is still limited. In the case of the external sense organs, the dendrite of the neurone is surrounded by three support cells. The innermost of these (thecogen cell) ensheathes the tip of the dendrite, the middle one (trichogen cell) forms the sensory process, hair or papilla, and the outer one (tormogen) forms the socket (refs in Blochlinger et al. 1988 and Hartenstein, 1988). Only one of these three support cells, possibly the tormogen, expresses lacZ intensely enough to be impossible to miss in the transformant A37 (Fig. 2A, cells marked p8 and p9). However, in the case of the dorsalmost sense organ, the papilla p9 in the abdominal segments, the absence of other sense organs around makes it possible to detect two additional labelled cells (Fig. 4A, arrowheads) that presumably correspond to the other two support cells (see also Blochlinger et al. 1988). At this stage of development, the corresponding neurone, desD, is included in the dorsal cluster of neurones (d in Fig. 4B).

Some non-neuronal labelled cells remain unidentified, possibly because of the uncertainties on the number and localization of the support cells. For example, the small cluster labelled with a question mark in Fig. 3C probably corresponds to the papilla p6, which is located at this site and is innervated by two neurones that belong to the ventral cluster v′. However, we do not know how many cells contribute to this particular sense organ, which is different from all other papillae in that it is doubly innervated. Likewise the pair of epidermal cells labelled by a double question mark in Fig. 3B seems to be associated with the pentascolopidial chordotonal organ (lch5); in later embryos they are elongated and lie parallel to the lsh5 cells. This is surprising in view of the fact that the five elements seem identical, so that one would expect to find five associated epidermal cells; however, it has recently been shown that a monoclonal antibody specifically recognizes four of the five lch5 neurones, demonstrating that the five elements are definitely not identical (Bodmer et al. 1987).

The uncertainties on the number of cells that form a larval sense organ make it difficult to decide whether lacZ is expressed in other, non-sensory cells as well. We addressed this question by analysing the expression of lacZ in A37 embryos deleted for the achaete–scute gene complex, AS-C. Embryos that are deleted for AS-C lack all external sense organs and most of their peripheral neurones (Dambly-Chaudière & Ghysen, 1987). They retain only the chordotonal neurones and a few other peripheral neurones such as the dorsal bipolar neurone, dbd. The effect of the deletion is particularly clear on the dorsal cluster of the abdominal segments. This cluster contains 11 neurones: 5es, 5da and 1bd. In AS-C⁻ embryos, this cluster is reduced to one cell, the dbd neurone, or occasionally two.

The dorsal region of an A37 embryo (boxed in Fig. 2A) is shown at a higher magnification in Fig. 4A,B. In addition to the dorsal cluster of neurones and the support cells described previously, a number of epidermal cells are very weakly labelled in a region that extends approximately from the dorsal cluster to p8, in the posterior half of each segment (outlined in panel A). This low level of galactosidase activity can also be detected in Fig. 2.

The same region of an AS-C⁻ embryo is shown Fig. 4C,D. The very weakly labelled epidermal cells are still present (Fig. 4C). However, the p8 and p9 support cells, as well as all but one cell of the dorsal cluster, have disappeared. The remaining labelled cell in Fig. 4D is the dbd neurone, which is never affected by the AS-C⁻ deletion. In the lateral and ventral regions also, the deletion of AS-C removes most of the labelled cells but does not affect the labelling associated with the chordotonal organs (not shown).

The first signs of lacZ expression are observed during stage 7 (end of gastrulation and beginning of germ band extension, see Materials and methods) in the head region (Fig. 5A). At this stage, the expression is restricted to a cluster of 10–15 contiguous epidermal cells on either side, in or close to the prospective hypopharyngeal region (Jurgens et al. 1986). During stage 8 (germ band elongation), a second stripe of contiguous epidermal cells becomes positive, dorsal and anterior to the cephalic furrow. This stripe soon broadens laterally while the labelling disappears dorsally, resulting in the formation of a second head cluster that probably corresponds to the origin of the antennal organ (Fig. 5B). At stage 9, small clusters of labelled contiguous epidermal cells can be detected in the body segments (Fig. 5C). During stage 10, a new cluster appears at the most anterior tip of the head, probably corresponding to the clypeolabrum, and 13 clusters of large subepidermal cells form in the two gnathal segments G2 (maxillary) and G3 (labial), in the three thoracic and in the first eight abdominal segments (Fig. 5D,E). Segmental differences can already be detected between the gnathal and following segments, and at the posterior end of the body, where A8, A9 and A10 differ from all previous segments (Fig. 5E). During stage 11, all clusters increase in size, and soon comprise tens of cells, located both within and under the epidermis (Fig. 5F,G). The A8, A9 and A10 clusters remain noticeably smaller than all others. Later development will include the merging of the antennal and maxillary clusters, presumably to form the antenno–maxillary sensory complex, the retraction of the germ band and the arrangement of broad masses of labelled cells in the precise patterns illustrated Figs 5H and 2A,B.

We have followed in more detail the emergence of the clusters in the body segments (Fig. 6). The first signs of lacZ expression appear almost simultaneously in all segments, near the end of stage 9, as one or a few adjacent epidermal cells become weakly positive (Fig. 6A). The region of lac expression spreads progressively, the cells at the border of the cluster being less labelled than those in the centre (arrows, Fig. 6B). Early during stage 10, one large intensely labelled cell appears under each cluster (arrowheads, Fig. 6B), just posterior to the parasegmental groves that form at the same time. This cell will be called P. Soon thereafter, another subepidermal cell, A, is detected in each segment, anterior to the previous one (Fig. 6C, arrows). This leads to the regular A/P pattern illustrated in Fig. 6D for the three thoracic segments. The epidermis above the A cell shows no detectable lac expression. Up to this stage, the labial segment, G3, and the segment A8 lag behind the other body segments, except for a very dorsal cell specific to A8 that will eventually give rise to the spiracular sense organ (open arrowhead in panels B, C and E). In the next stage, the single A cell somehow becomes a pair of anterior cells (Fig. 6D, A pair in segments A3 and A5). Eventually an A pair is found in all segments (Fig. 6E). At about the same time the early pattern of P cells also evolves into pairs of posterior cells (Fig. 6E, arrows, and 6F). There is no strict sequence in the transition of A and P cells to pairs of cells, and the posterior pair may occasionally arise before the anterior one, although the other sequence is more usual. Elongated figures suggestive of a dividing cell are often seen (Fig. 6D, arrowhead with question mark). However, an increase in the number of labelled cells is also seen when A37 is combined to the mutation string (stg, Jürgens et al. 1984) which prevents cell divisions after the blastoderm stage (Y. N. Jan, personal communication; Tearle & Nüsslein-Volhard, 1987). The A31stg pattern is shown at three different times in Fig. 7A,B and C. The early pattern of A37 expression (Fig. 7A) is relatively normal, except for the large size of the cells which reflects the lack of postblastoderm mitoses. Later on, pairs of cells appear in some segments (Fig. 7B, arrows). The pattern soon becomes very abnormal (Fig. 7C) and does not change further, possibly because all available cells are already specified. This result suggests that the formation of an anterior pair does not result from the division of the initial A cell, but from the activation of lacZ in a second cell near the initial A cell.

By the time all segments show an anterior and a posterior pair of lac-expressing cells, still during stage 10, two additional labelled cells are detected in each segment, one dorsal (D) and one ventral (V). The pattern at this stage is shown in Fig. 6F, where the anterior and posterior pairs and the dorsal and ventral cells have been marked in segments T3 and A6. The D and V cells are soon converted into pairs of cells. At this stage, each segment ideally contains four pairs of subepidermal cells, in addition to the epidermal cluster. However, slight and apparently random differences between the segments, and between the different regions of one segment, make this stage already fairly complex, and we have not examined the later stages in great detail.

The early development of lacZ expression in normal and AS-C⁻ embryos is compared in Fig. 8, A–C (normal) and D–F (mutant). In the body segments, the absence of AS-C has no effect on the appearance of the early epidermal clusters or of the large subepidermal P cell that forms under each cluster. However, the A cell never forms (or forms but never expresses lacZ). This is particularly obvious in the mutant embryo shown in Fig. 8D, where a pair of P cells is present in each segment: by this time, all segments also contain one or two A cells in normal embryos (Fig. 8A). The dorsalmost cells of A8, precursors to the spiracular sense organs, are also absent in the mutant (sp in Fig. 8A,C). At later stages, it would seem that the formation of the dorsal cells is impaired in the mutant, but not of the ventral pair (compare Fig. 8E, mutant embryo, with 8B, normal: D, dorsal cells; V, ventral cells). Just before the beginning of germ band retraction, the phenotype of the mutant embryos can be correlated with the final phenotype (Fig. 8F): in the thoracic segments, a small group of dorsal cells (arrowheads) and one ventral cell (arrows) can be detected in the mutant, corresponding, respectively, to the dorsal ch3 and ventral ch1 chordotonal organs. All of the lateral cells are absent. In the abdominal segments, the dorsal cluster has not formed and the lateral cluster is essentially restricted to the lateral chordotonal organ, lch5. A substantial reduction is also obvious in the gnathal segments, G2 and G3, and probably in the head as well, but these have not been analysed.

It has been shown recently that.the deletion or inactivation of the gene daughterless (da) results in the absence of all known peripheral neurones and sense organs in homozygous embryos (Caudy et al. 1988). Thus, in addition to its maternal effect on sex determination (Bell, 1954; Cline, 1976) da is also zygotically required for the formation of the peripheral nervous system. This requirement appears very specific, in that all other cuticular derivatives of the epidermis form normally. We have examined the effect of da^IIB31, an embryonic lethal allele that effectively prevents the formation of sense organs, or of a deficiency that removes the da gene, Df(2L)J27, on the A37 pattern of labelling. Fig. 9 shows four stages in the expression of lacZ in A37 embryos (left column) and in da⁻; A37 embryos (right column). All embryos have been overstained in order to reveal the clusters of weakly labelled epidermal cells in the body segments. These appear normally in da⁻ embryos (Fig. 9E), and remain positive until after the retraction of the germ band (Fig. 9H), at which time the posterior edge of each cluster is seen to coincide with the intersegmental grove. Besides these cells, however, no lacZ expression can be detected in the body segments, from the time of appearance of the A and P cells (compare Fig. 9A and E) to the final stage where the pattern is complete (Fig. 9D and H).

The sense organs of the adult differentiate during metamorphosis. In the wing disc, the first differentiating neurones appear shortly after pupariation (Murray et al. 1984, Jan et al. 1985). The morphological differentiation of the outer support cells (trichogen and tormogen) becomes detectable more than 10 h later (Lees & Waddington, 1943). However, the determination of at least some sense organs occurs much before pupariation (Garcia-Bellido & Merriam, 1971). We have examined the wing discs of mature larvae and observed positive cells (Fig. 10A,B). The fate map of the wing disc has been so accurately described (Bryant, 1975) that we can easily associate some of the labelling to known sense organs. For example, the double row of cells that stops abruptly at the middle of the disc (arrow, Fig. 10A) delineates exactly the anterior margin of the wing, where two or three rows of sensory bristles will form. The group of cells shown by the arrowhead in Fig. 10B corresponds to the large cluster of campaniform sensilla on the dorsal radius, while the smaller group labelled with an asterisk in Fig. 10A corresponds to the smaller cluster on the ventral radius (these two regions of the disk will become apposed during metamorphosis as the wing pouch evaginates). The pairs of cells marked by the open arrowheads in Fig. 10A and B presumably correspond to the ACV/E1 and L3.2/E2 pairs of neurones that will later differentiate along the third vein of the future wing (Murray et al. 1984). A few cells are also labelled at reproducible locations within the presumptive notum, probably the precursors of the precisely located macrochaetes. Indeed, scute mutations that specifically remove some of the macrochaetes on the notum also remove some of these positive cells (F. Huang, personal communication). Many positive cells can also be detected in the mature leg disc (Fig. 10D, arrows), even though before pupariation this disc contains only six or seven larval neurones (Jan et al. 1985). In addition to the peripheral neurones, a small subset of central neurones also express lacZ in late larvae (Fig. 10C,D, arrowheads).

The pattern of expression of lacZ in embryos of this line is shown Fig. 2, C (dorsolateral view) and D (lateral view of another stage-15 embryo). All lateral and dorsal chordotonal organs of the body segments are positive: the dorsal triscolopidial chordotonal organ (dch3) in the thoracic segments, the lateral pentascolopidial organ (lch5) in the abdominal segments, the lateral triscolopidial organ (lch3) specific of Tl, and the chordotonal organs of segments A8, A9 and A10 (out of focus in the Figs). Each chordotonal organ comprises one neurone and three support cells (Hartenstein, 1988), much as the neurones that innervate external sense organs (reviewed in Bodmer et al. 1987). In the transformant A18, lacZ is expressed in the neurone and at least two support cells. The highest level of expression is always in the support cell that is closest to the neurone, the sheath or scolopale cell. This is clearly seen in the abdominal lch5, where the row of five sheath cells (sh) is more intensely labelled than the closely apposed row of five neurones (ch, Fig. 2C,D). Another support cell, the cap cell, is also consistently labelled (cc, Fig. 2C). The ventral chordotonal organs, one in each thoracic and three in each abdominal hemisegment, are more weakly labelled and only one or two cells can usually be detected, instead of the three cells that are consistently labelled in the dorsal/lateral organs. In addition to the chordotonal organs, a few more cells are moderately positive. At least some of them are support cells of the external sense organs, which are supposed to be homologous to the sheath cells of the chordotonal organs (compare Fig. 2C,D and A,B). Finally, pericardial cells around the dorsal vessel (heart), which extends anteroposteriorly under the dorsal midline, are also labelled (marked AS in Fig. 2C).

The first signs of lacZ expression, in the body segments, occur at about the same time as in A37, early during stage 10. This early expression occurs with a marked bisegmental periodicity in stripes of epidermal cells in the dorsal half of each segment (Fig. 11A). These appear similar to the weakly labelled epidermal clusters in the transformant A37 (see Fig. 8E). In addition, the cells of the amnioserosa also express lacZ (Fig. 11C, AS). Later on, intensely labelled cells appear at the edge of each stripe (arrowheads in Fig. 11B), resulting in a regular pattern of very weak stripes with two labelled cells at the edge (Fig. 11C). This pair of cells is probably the same as the posterior pair in the A37 transformant, as judged from their position relative to the tracheal pits that form at this stage (Fig. 11D). There is not much change in this pattern until the germ band retracts (Fig. 11E,F). When retraction is nearly complete several changes take place (Fig. 11G). The number of labelled cells increases and segmental differences become obvious: T1 differs from all following segments, A8 and A9 differ from all preceding segments and the dorsalization in T2–3 relative to Al–7 is just beginning. These changes proceed very fast, so that when the germ band is fully retracted the final pattern is essentially obtained (Fig. 11H). The next step is dorsal closure, during which the amnioserosa cells are internalized and become incorporated in the heart as pericardial cells (Campos-Ortega & Hartenstein, 1985). This step is easily followed in this transformant by virtue of the lacZ expression in the cells of the amnioserosa (compare Figs 11H and 2C).

In this line, lacZ is expressed in a very elaborate pattern, in the cephalic region of late stage 11 (end of the extended germ band stage, Fig. 12A). In the body segments, lacZ is weakly expressed by one or two peripheral cells in each segment (arrows, Fig. 12A) and by one pair of cells in each segmental hemiganglion (Fig. 12B). The number of positive cells in the central nervous system increases progressively during the retraction of the germ band, up to about twelve after the germ band has retracted and dorsal closure has occurred (Fig. 12D,E). This number further increases up to several dozens in the late embryo, with reproducible differences in the level of expression in different neurones. The majority of central neurones remain unlabelled even after the CNS has condensed, late in embryogenesis.

In addition to a subset of central neurones, a few peripheral neurones are labelled in the dorsally closed embryo (Fig. 12C). Three cells can be detected in most segments: one lateral cell, intensely labelled in Al to A7, less so in the thoracic segments and duplicated in A8; one dorsolateral cell, moderately labelled in all segments from T2 to A8, and one dorsal cell that is more weakly positive and cannot be seen in Fig. 12C. The region outlined in Fig. 12C is shown at a higher magnification in Fig. 13B. The intensely labelled lateral cell is almost certainly the lateral bipolar neurone, as judged from its position relative to the pentascolopidial organ (dotted outline) and depth under the epidermis. The dorsolateral cell is in fact a closely apposed pair of cells, one of which is the dorsal bipolar neurone, dbd (Fig. 13A,B, open arrowhead) that establishes the dorsolateral nerve trunk (arrowheads). The nature of the second labelled cell is not known. Finally, the dorsal neurone belongs to the dorsal cluster (Fig. 13A). The fact that there is one and only one positive neurone in each dorsal cluster, together with its relatively constant, posterior location within the cluster strongly suggest that it is the same dorsal neurone that is positive in all segments.

In addition to these peripheral neurones of the body segments, a number of other cells also express lacZ in this transformant. In the head, unidentified cells in the antennal lobe and the anterior parts of the maxillary and labial lobes are labelled strongly. Weaker labelling is present in the clypeolabrum, the mandibular lobe and the optic plaque. The Malpighian tubules and a short length of the hindgut are also labelled (out of focus in Figs 12C and 13A). Unidentified cells at the base of the posterior spiracles are labelled strongly (dorsally at the posterior end of the embryo, Fig. 12C,D).

A number of totally unrelated cell types express lacZ in this line. The strongest labelling is found in several cells in or around the antennomaxillary complex, the hypophysis and the epiphysis (Jürgens et al. 1986). Cells in the antennal lobe are labelled from stage 13, followed by the cells in the labial lobe and clypeolabrum. In the peripheral nervous system of the body segments, lacZ is expressed in two cells per hemisegment, from late stage 11 onwards. One cell is the most ventral neurone of the dorsal cluster (Fig. 14D), the other is one neurone that belongs to the ventral v′ cluster (not shown). In the central nervous system, fourteen clusters along the ventral midline are labelled from stage 11 onwards (Fig. 14A,B). These might be midline neurones or glial cells (Crews et al. 1988). In the epidermis, stripes of cells under the abdominal setal belts A1–A7 are marked (Fig. 14C). There are fewer cells in Al than in the following segments, consistent with the reduced number of denticles in Al. The absence of labelling under the A8 denticles is more surprising, as the A8 denticles do not appear very different from those in the preceding abdominal segments. In later embryos, some parts of the gut are also labelled: some of the midgut, some of the hindgut and the outer layer of the proventriculus (Fig. 14C).

The use of randomly inserted reporter sequences to detect in situ enhancer-like elements in the genome of Drosophila has been described previously (O’Kane & Gehring, 1987). In this report, we describe four such transformants where all or part of the peripheral nervous system expresses the inserted gene, the bacterial lacZ. In all cases the P-lacZ fusion protein coded for by P[lac,ry⁺]A appears to be located predominantly in the nucleus of most cells where it is expressed. As other lacZ fusion proteins are known to be localized cytoplasmically (e.g. Doe et al. 1988) the nuclear localization that we observe may result from the 125 amino acids of the P transposase (Rio et al. 1986) that are present in the expected fusion protein.

One of the transformants, A37, has been analysed in some detail since it might allow the identification of precursor cells to the peripheral nervous system at about 5 h of development, much before the time when sensory neurones and support cells begin to differentiate (approximately 10 h of development, CamposOrtega & Hartenstein, 1985; Ghysen et al. 1986; Hartenstein, 1988).

In late A37 embryos, lacZ is expressed in most and probably all peripheral neurones, as well as in the support cells that will differentiate the non-neuronal parts of the sense organs. As expected, this expression is completely suppressed in homozygous da⁻ embryos, where all known peripheral sense organs and peripheral neurones are absent. Likewise in embryos deleted for the AS-C genes, which lack a subset of the peripheral nervous system, the pattern of lacZ expression is accordingly reduced. Thus there is no doubt that in the late embryo, the expression of lacZ is typical of the cells that form the sense organs.

Before we examine the pattern of lac expression in early A37 embryos, we must mention two exceptions to the rule that only sense-related cells express lacZ in the mid- and late stages of A37 embryogenesis. The first exception are the very weakly labelled epidermal clusters detected in Fig. 4A,C and Fig. 8E–H, and also present in the transformant A18 (e.g. Fig. 11A). Very similar patterns have been observed in transformants for other constructs also using lacZ (W. Bender, personal communication), as well as in transformants for several Ubx–lacZ fusion constructs. In the latter case this early pattern was thought to be part of a ‘basal pattern’ dependent on Ubx regulatory sequences (Bienz et al. 1988). In a screen of several hundred inserts using a different P-lacZ transposon, P[1ArB], some 30 % of insertions gave a similar pattern (C. Wilson, C. O’Kane, H. Bellen, U. Grossniklaus, R. K. Pearson & W. J. Gehring, unpublished observations), suggesting that the control of this pattern might reside within lacZ or other elements of the transposon itself and be revealed whenever the construct inserts in a ‘permissive’ genomic site. The second clear exception to the association of lacZ expression with sense organs is a row of five or six cells on each side of the anal slit. These cells do not correspond to any identified sense organ or any neurone, and like the weak epidermal clusters they remain unaffected in homozygous da⁻embryos.

Whether the early positive cells in A37 embryos are precursors to the sense organs is more difficult to establish. This is because the pattern of lacZ expression at intermediate stages is very complex, making it impossible to identify with any certainty individual ZacZ-expressing cells. Since the histochemical reaction requires previous fixation of the embryos, we could not follow the fate of positive cells either. Thus we have had to rely on indirect evidence to assess the relation between the very early and the late patterns of expression.

A simple approach to the relation between the early cells that express lacZ and the late sensory cells is to ask if the former are also affected by mutations that are known to remove all or part of the peripheral nervous system in late embryos. We observed that the absence of da gene activity prevents the appearance of the early Zee-expressing A and P cells, as well as all the subsequent development of the A37 pattern. This is unlikely to reflect a generalized defect in the ectoderm of early da⁻ embryos, since the inactivation of da has no detectable effects on non-innervated epidermal derivatives such as the dorsal hairs or the ventral denticles. On the other hand, the role of da in the formation of sense organs is not understood, nor is the relation between the zygotic requirement for peripheral neurogenesis and the maternal requirement for sex determination. In order to guard against the possibility that the effect of da is due to some peculiarity of this gene, we have also examined the effect of removing the achaete–scute gene complex, AS-C.

Genetic and molecular analyses have lead to the conclusion that AS-C genes are responsible for the decision of ectodermal cells to develop into adult sense organs (Garcia-Bellido & Santamaria, 1978; Campuzano et al. 1986). The AS-C contains several genes, some of which are expressed in a segmentally repeated complex pattern in the lateral ectoderm of gastrulating embryos (Cabrera et al. 1987; Romani et al. 1987), consistent with the idea that these genes are also involved in the determination of the larval sense organs which are formed during embryogenesis. The deletion of AS-C removes only part of the peripheral nervous system (all es and most da neurones), but leaves the chordotonal organs intact. We observed that, in A37 embryos deleted of AS-C, the P cells appear normally in all segments, but neither the A cell nor the A pair can be detected. Since the AS-C genes appear very specifically involved in the development of the nervous system, the early effect we observe makes it likely that the A cell is related to sense organs that require AS-C genes (e.g. the external sense organs). On the other hand, the fact that the P cell is absent in da⁻ embryos but not in AS-C⁻ embryos suggests that this cell is related to the sense organs that do not require AS-C, in particular the chordotonal organs. In this context it is interesting to mention that the analysis of embryos with defective segmentation has revealed that most and possibly all external sense organs are derived from the anterior region of each segment, while at least the dorsal and lateral chordotonal organs are posterior derivatives (Hartenstein, 1987).

A second argument supporting the idea that the early ZacZ-expressing cells in A37 are precursors to the sense organs is provided by the early pattern of expression in transformant A18 embryos, where lacZ is preferentially expressed in chordotonal organs at later stages. The initial expression of lacZ in A18 embryos occurs during stage 10 in a pair of cells in each segment. This pair of cells has the same location and appearance as the posterior pair that is detected in A37 embryos of the same age. Furthermore, this early pattern of A18 expression disappears in da⁻ embryos but not in AS-C⁻ embryos (not shown), strongly suggesting that the early pair detected in A18 is identical to the posterior pair in A37.

Thus all our data are consistent with the idea that, during stage 10, a pair of cells is singled out in the posterior region of each body segment, as precursors to some of the posterior (presumably chordotonal) sense organs. These cells express lacZ both in A18 and A37 and require da but not the AS-C. At nearly the same time a second pair of cells is singled out in the anterior region, as precursors to some of the anterior sense organs. These cells express lacZ in A37 but not in A18, and they require both da and the AS-C genes. Slightly later additional cells located dorsally and ventrally also express lacZ; the contribution of these precursors to the final pattern has not been assessed.

If this interpretation of the patterns we observed is correct, it implies that the central and the peripheral nervous systems are formed nearly simultaneously, since we observe the first precursors to sense organs early during stage 10, while the central neuroblasts segregate from the epidermis from stage 9 to stage 11 (Hartenstein & Campos-Ortega, 1984). The expression of lac in A37 would then follow very closely (by less than 1 h) the beginning of AS-C expression in the lateral ectoderm of the body segments (Cabrera et al. 1987; Romani et al. 1987), and precede by several hours the first signs of peripheral neurogenesis that could be detected so far (Campos-Ortega & Hartenstein, 1985; Ghysen et al. 1986; Bier et al. 1988).

We do not know the lineage relationship between these early precursors and the final complement of neurones and support cells. According to our observations on string embryos, the early anterior and posterior pairs do not seem to arise from a division of the A and P precursors. If the development of larval sense organs in the fly is similar to the situation described in other insects (Bate, 1976), we would expect each precursor to form one sense organ; however, we cannot rule out that at least some of the precursors give rise to more than one sense organ.

In view of the conclusion that A37 marks the precursors to the larval sense organs early during embryogenesis, it is interesting that in imaginai discs positive cells are found at the locations where adult sense organs will later differentiate. Here again some of the Zac-expressing cells are removed by scute mutations that suppress adult sense organs, supporting the view that A37 marks cells that are committed to form a sense organ, be it larval or adult.

The results presented here confirm that the P-ZacZ fusion is a very useful tool for generating cell markers in Drosophila. Such markers are easier to generate than monoclonal antibodies, and it is easier to isolate the genes near the site of P-lacZ insertion than to isolate the gene coding for the antigen recognized by a monoclonal antibody. While the absence of a plasmid replication origin in P[lac,ry⁺]A makes the construction of a genomic library necessary for cloning, several P-lacZ fusion constructs containing a plasmid origin within the transposon have now been constructed, thus making cloning much faster and simpler (C. Wilson, R. K. Pearson, H. Bellen, C. O’Kane, U. Grossniklaus & W. J. Gehring, unpublished experiments). Furthermore, the availability of a stable genomic source of P transposase (Robertson et al. 1988) makes it possible now to generate new insertions on a large scale simply by performing a few genetic crosses. Given that the markers described here were recovered from a relatively small genetic screen (49 insertions), it seems likely that the large screens that are now feasible will generate a large set of highly specific cell markers. These should prove invaluable for the analysis of the developmental mechanisms that generate and specify the reproducible patterns and great variety of cell types in the peripheral and central nervous system.

The analysis of the da; A37 embryos was originally done as part of the study of the daughterless mutation in collaboration with Yuh-Nung Jan, Lily Yeh Jan and Christine Dambly-Chaudière. We are grateful to them for sharing observations and for giving us the stg mutant and a A37stg recombinant chromosome. We also thank Françoise Huang for telling us about the phosphate effect described in the Materials and methods, and for communicating her unpublished results on A37 imaginai discs in scute mutants. We thank Walter Gehring for his support, hospitality and advice for one of us (C.O’K.). A.G. is chercheur qualifié of the Fonds National de la Recherche Scientifique (Belgium). C.O’K. was supported by an EMBO long-term fellowship and by a fellowship from the Roche Research Foundation.