‘Promoter trapping’ in Caenorhabditis elegans

How is gene expression organized during development in the nematode, Caenorhabditis elegans? One way to address this question is to examine the expression patterns for a large sample of C. elegans genes selected at random. This could reveal examples of spatially specific, cell-lineage-specific, tissue-specific or structure-specific gene expression. The largely invariant cell lineage of C. elegans and observations after laser ablation of cells suggested much of C. elegans development is directed by cell-intrinsic determinants (Chitwood and Chitwood, 1974; Sulston et al. 1983). However, these studies showed that the fates of some cells, with equivalent developmental potential, are determined by intercellular communication (Sulston and White, 1980; Sulston et al. 1983; Greenwald, 1989). Experimental manipulation of early embryos has now suggested that cell-extrinsic signals may have even more significance for C. elegans development than previously thought (Priess and Thomson, 1987). Furthermore, evidence of global positional cues for C. elegans is emerging. The phenotypes of mutations in mab-5 (Costa et al. 1988) and egl-5 (Chisholm, A. 1991) suggest these two genes could be involved in specification or interpretation of positional cues along the longitudinal axis of C. elegans. The relative importance of cell-intrinsic determinants, intercellular signals and global positional cues in C. elegans development could be revealed by a screen of gene expression patterns.

The results of a screen of gene expression patterns would be valuable in other ways. First, particular gene expression patterns could become the subject of molecular and genetic investigations of regulatory systems important for development. Second, gene expression patterns could provide cytogenetic markers for evaluation of mutations obtained in other genetic screens or of the consequences of cell ablation experiments. Finally, the molecular description of C. elegans developing from the genetic and physical maps (Edgley and Riddle, 1989; Coulson et al. 1988) could benefit from the information about gene expression patterns once matched to specific genetic loci.

Procedures for examining random gene expression patterns have already been developed for Drosophila melanogaster and the mouse. For ‘enhancer trapping’ in Drosophila, a lacZ gene with a weak promoter was transposed around the genome inside a modified P element (O’Kane and Gehring, 1987). β-galactosidase expression was directed by enhancers adjacent to the point of insertion and accurately reflected the expression patterns of nearby genes (Bellen et al. 1989). For the mouse, random integration events introduced lacZ into different chromosomal locations, as a consequence of transformation (Allen et al. 1988; Gossler et al. 1989). Two types of lacZ construct were used; one had a weak promoter and expression depended on insertion near an enhancer (‘enhancer trapping’), the other had a 3’ splice acceptor site upstream and required fusion into a transcript by splicing for expression (‘gene trapping’). Again the pattern of β-galactosidase expression depended on the site of integration. The results of these studies suggest this will be a rewarding approach for the investigation of animal development.

C. elegans would be particularly suitable to this type of analysis. The largely invariant cell lineage has been described completely (Sulston et al. 1983), allowing patterns of gene expression to be defined precisely within this framework (e.g. Way and Chalfie, 1989). Such definition facilitates comparison of expression patterns with (i) final cell fates, (ii) common patterns of cell division or (iii) other expression patterns. Likely interrelationships in the regulation of genes may be readily inferred. Furthermore, C. elegans is well-disposed to genetic analysis (Brenner, 1974), important for subsequent investigations of the generation of particular patterns.

I have developed a screen applicable to C. elegans in which gene expression patterns are sampled by transla-tionally fusing lacZ to random genes. The approach is distinct from ‘enhancer trapping’ (O’Kane and Gehring, 1987) and ‘gene trapping’ (Gossler et al 1989). Rather than inserting a reporter into an intact gene in situ, fusions have been made in vitro with 5’ gene segments. I have therefore called the approach ‘promoter trapping’. Results from an intitial screen are presented here.

C. elegans genomic DNA was prepared as previously described (Sulston and Hodgkin, 1988) and partially digested with Sau3A. 5– 10kb fragments were isolated by agarose gel electrophoresis and ligated into BamHi-digested, alkaline-phosphatase-treated pPD22.11, using conventional procedures. pPD22.ll was kindly provided by Andrew Fire (Fire et al. 1990). Ligated DNA was transformed into Escherichia coli and independent colonies were picked into 2×YT broth in 96-well multiwell plates.

For the initial screen, plasmids were assayed in pools of 96 independent constructs. Preliminary experiments suggested that with a pool of 96 constructs expression from an individual construct could be detected; β-galactosidase expression was detected in 3 out of 6 C. elegans lines transformed with a control plasmid diluted 200-fold with vector plasmid DNA and in 11 out of 12 lines when diluted 60-fold. (The control plasmid was pPD18.56, which expresses β-galactosidase in the pharynx and was provided by Andrew Fire). After pooling bacterial cutures, which had been grown overnight in multiwell plates, plasmid DNA was prepared by the alkali lysis procedure and further purified by LiCl precipitation (Spence, 1990). The DNA solution for injection was as previously described (Fire, 1986) with library plasmid DNA at 180μgml^-1 and plasmid pRF4 DNA, similarly prepared, at 20μgml^-1. pRF4, kindly provided by Craig Mello (personal communication), is a pUC19 based plasmid containing the C. elegans rol-6(su1006) gene. The rol-6(su1006) gene carries a dominant mutation which causes late larval and adult C. elegans to roll to the right as they move, a readily observable phenotype for identifying transformants. To obtain transformants with DNA integrated into the genome, a synthetic oligonucleotide (55mer) was included in the injection solution to 1mgml^-1 (Mello, C. personal communication ). The DNA was injected, as described by Craig Mello (personal communication), into the syncytial distal gonad arm of adult hermaphrodites of the wild-type C. elegans strain, N2 (Brenner, 1974). Agar plates seeded with the E. coli strain OP50 (Sulston and Hodgkin, 1988) were used to maintain N2 prior to injection but the E. coli strain TGI (Gibson, 1984) cured for the F episome was used for transformants because this strain is completely deleted for lacZ. rol-6 progeny of injected animals were transferred to fresh plates, maximally 5 per 4.5 cm plate. Animals with the rol-6 phenotype were transferred at four subsequent generations to establish transformed lines, before staining for β-galactosidase.

The histochemical staining procedure is a modification of the procedure developed by Fire (1986). Ten adult rol-6 hermaphrodites, from an established transformed line, were placed on a seeded 4.5 cm plate. 4 to 5 days later, as the bacteria were cleared, the resulting population was harvested by washing from the agar surface with water. Worms were centrifuged (30 s, 1000 g) and 3 μl aliquots were placed in each well of an 8-well glass microscope slide. A coverslip was overlaid and the worms were frozen by placing the slide on dry ice. The coverslip was flipped off with a razor blade and the slide was placed in acetone at – 20°C for 5 min. The preparation was allowed to air dry at room temperature before applying 25 μl staining solution containing X-gal. Another coverslip was applied avoiding air bubbles and sealed to the slide with nafl varnish. The slide was incubated at 37°C until sufficiently stained (up to 24 h.). Staining patterns were recorded on Kodak Ektachrome 160.

C. elegans strains were grown on ten 9 cm plates. As the bacteria were cleared, the worms were washed from the plates with M9 buffer and eggs were prepared by treatment with alkaline hypochlorite (Sulston and Hodgkin, 1988). The washed egg pellets were stored at – 20 °C until dissolution in Laemmli sample buffer (100°C, 5 min) and solubilized proteins were separated by polyacrylamide gel electrophoresis using the Laemmli discontinuous buffer system (Laemmli, 1970). Proteins were transferred to nitrocellulose using a semi-dry electroblotter with Tris-glycine-methanol buffer (Harlow and Lane, 1988). The blot was probed with a mouse monoclonal antibody to β-galactosidase (Promega) at 1/1000 dilution overnight at room temperature and a phosphatase-conjugated goat anti-mouse IgG antiserum (Sigma) at 1/1000 dilution for 1 h using BSA as carrier protein and 0.15 M NaCl, 10 mM Tris, pH 7.4 to wash. The blot was developed using BCIP/NBT as substrate for the phosphatase (Harlow and Lane, 1988).

Hizndlll DNA fragments from the inserts of isolated plasmids were purified by agarose gel electrophoresis and labeled by random priming (Feinberg and Vogelstein, 1983). The labeled DNA fragments were hybridized to the YAC polytene grid provided by Alan Coulson using standard genomic Southern hybridization conditions and exposed for autoradiography overnight.

A library of potential gene fusions was created. C. elegans genomic DNA fragments were ligated into a plasmid upstream of a lacZ reporter gene (Fig. 1) and the ligation products were used to transform E. coli. Plasmid DNA was prepared from E. coli prior to use in transformation of C. elegans. Since the lacZ gene lacks transcriptional start signals, expression of β-galactosidase in C. elegans from one of these constructs would depend on both transcriptional fusion and in-frame, translational fusion with a coding region reading out from the insert. This means that the proportion of the inserts that would drive β-galactosidase expression would be low and, therefore, constructs were assayed in pools of 96 independent plasmid constructs. The transforming DNA is thought to assemble into large extrachromosomal arrays of several hundred plasmid molecules (Stinch-comb et al. 1985), the composition of an array reflecting the composition of the injected DNA. Thus, the arrays generated in these experiments contained, on average, a few copies of each of the 96 plasmids in a pool. Established transformed lines could be passaged for repeated examination of β-galactosidase expression or frozen for future reference.

C. elegans transformant lines have been established for 45 construct pools. The results of staining these lines for β-galactosidase expression (Fig. 2) are summarized in Table 1. For many of the expression patterns, β-galactosidase is localized to the nucleus. This is not significant in these experiments because nuclear localization is encoded by the vector plasmid. The vector had been constructed in this way to aid identification of expressing cells (Fire et al. 1990).

Pool 4 gave nuclear-localized β-galactosidase staining in body wall muscle cells of adult hermaphrodites (Fig. 2B and C). The 95 body wall muscle cells of the adult are arranged as four longitudinal bands, one running in each quadrant of the body and each consisting of two rows of rhomboid-shaped mononucleate muscle cells (Sulston and Horvitz, 1977). For the transformants, the roller genotype causes the body wall to be twisted and the muscle bands form a spiral along the length of an individual.

Pool 9 caused β-galactosidase expression in specific uterine cells (Fig. 2D and E). The hermaphrodite gonad is a symmetrical, bilobed, tubular organ. Each arm consists of a syncytial ovary, an oviduct, a spermatheca and the uterus through which the two arms are continuous (Hirsh et al. 1976; Kimble and Hirsh, 1979). Fertilized eggs are retained in the uterus for a few hours of development before expulsion through the single opening, the vulva, located mid-ventrally. The β-galactosidase staining was not nuclear localized but appeared to fill the multinucleated, toroidal, endothelial cells forming part of the wall of the uterus. The staining first appeared in the last larval stage, during development of the gonad.

Pools 23 and 80 induced β-galactosidase expression in particular cells of the hypodermis. The hypodermis is a single cell layer covering the surface of the worm and secreting the cuticle (Sulston et al. 1983). Pool 23 gave expression in the very tip of both head and tail of adult hermaphrodites (Fig. 2G and H). The β-galactosidase activity appears to be nuclear localized, but not as tightly as for other patterns. Pool 80 gave nuclear-localized expression in the hypodermal seam cells of late embryos and all larval stages, but not adults (Fig. 2R). The seam cells form a continuous row along each lateral line and divide several times postembry-onically before finally fusing in the adult to produce seam syncytia.

Pool 26 caused a highly specific and reproducible, nuclear-localized staining of three cells around the rectum of late larvae and adult hermaphrodites (Fig. 2M). The stained cells are probably the three rectal epithelial cells (Sulston et al. 1983).

Pool 38 gave β-galactosidase expression in two valve structures, the intestinal– rectal valve and the spermatheca! valve (Fig. 2N and O). The intestinal– rectal valve consists of a ring of two endothelial cells, which forms part of the connection between the intestine and hypodermis, and a single toroidal sphincter muscle cell, which wraps round the valve and contracts to close it (White, 1988). The spermathecal valve consists of a single toroidal cell with four nuclei, which connects the spermatheca and uterus and through which fertilized eggs must pass, β-galactosidase expression began as each structure developed and the β-galactosidase activity probably persists thereafter. The intestinal– rectal valve staining was first observed late in embryogenesis and the spermathecal valve staining was first observed in the last larval stage. β-galactosidase activity was localized to a subcellular annular structure. In both the spermathecal valve and the anal sphincter muscle cell, a complete ring of filaments has been observed by electron microscopy, which are thought responsible for valve constriction. These fibrous structures and the rings of β-galactosidase staining have similar dimensions. The β-galactosidase fusion protein appears to be localized to the filaments.

Pool 64 gave a transformant fine which showed β-galactosidase expression in the excretory cell (Fig. 2Q). The excretory cell is the principal component of the excretory system, thought to be responsible for osmoregulation (Nelson et al. 1983). The main body of the H-shaped cell is ventral to the terminal bulb of the pharynx and arms project anteriorly and posteriorly along the lateral lines for most of the individual’s body length. The β-galactosidase activity was detected in the main body of the cell and along the posterior projections to the tail. Expression in the anterior projections may have been too weak to detect.

Pools 2, 20, 25, 40 and 42 appear to give β-galactosidase expression in neural structures. Pool 42 gave an arc of β-galactosidase activity in the region of the circumpharyngeal nerve ring (Fig. 2P). This expression pattern has been obtained reproducibly with this construct pool but the underlying structure causing this pattern is not known. The other four pools gave more tightly localized β-galactosidase expression but because the nucleus occupies much of a nerve cell body, it was not certain that the staining was nuclear-localized. Pool 25 appeared to cause staining of many cells in the neuropil of the head and tail and along the ventral cord (Fig. 2L). Pools 2 (Fig. 2A), 20 and 40 gave expression in the circumpharyngeal nerve ring only. In contrast to pools 2, 20, 25 and 42 which gave staining for all larval stages and adults, expression for pool 40 was only detected in larval stages.

Finally, pools 1, 11, 24, 64 and 68 gave restricted expression in early embryos. The timing of expression and the distribution of β-galactosidase activity suggest that the expression patterns are distinct for all five pools. None of the five pools gave expression in every cell of stained embryos. Precise determination of these expression patterns must await identification of the individual active plasmids from each pool and this has been taken furthest for pool 24 (Fig. 21,J and K and see below).

An expression pattern was likely to be due to a particular plasmid construct if that expression pattern could be generated repeatedly with a particular pool. Such observations would make unlikely the possibility that an expression pattern was the result of a unique DNA rearrangement occurring during transformation. Expressing lines arising from different daughters of the same injected individual may not necessarily be independent because an extrachromosomal array may duplicate in the gonad and become included in more than one oocyte. The degree to which an expression pattern result has been repeated varies between construct pools, e.g. only one transformed line of the two established with pool 1 gave β-galactosidase expression and this was weak. In contrast, the expression with pool 24 was very strong and obtained in ten transmitting lines produced from at least four independent transformation events. Expression patterns have been observed reproducibly, i.e. from more than one injected individual, for pools 4, 23, 24, 25, 26, 38, 40, 42 and 68.

15 of the 45 pools gave β-galactosidase expression in the pharynx and at least five distinct patterns were observed (not listed in Table 1). This high incidence of pharyngeal expression has not yet been explained but could reflect the complexity of the pharynx. A more likely explanation is that an insert containing an incomplete or constitutive promoter gives expression in the pharynx because of a cryptic pharyngeal enhancer on the vector plasmid. This could explain why no constitutive expression patterns were observed in this screen. The distinct patterns of expression within the pharynx may result from incomplete promoters in an insert affecting the expression pattern. For the moment, the pharyngeal expression patterns are not being pursued further.

β-galactosidase activity was not detected for the transformed fines obtained for 19 of the 45 pools examined. This would be expected if 1 in approximately 110 of the genomic DNA inserts can drive β-galactosidase expression. This frequency is slightly better than expected. Assuming one gene per 20 kb. (Heine and Blumenthal 1986) and an average protein size of 300 amino acids, approximately 5 per cent of the C. elegans genome encodes protein. To obtain translational fusion, the end-point of an insert fused to lacZ must be in a protein coding region and the translational reading frame and direction of translation must also be correct. Therefore, only 1 in approximately 120 constructs could be active. This figure is optimistic because although C. elegans genes tend to be compact (Emmons, 1988), a DNA fragment from the 3’ end of a large gene would not contain a complete promoter. However, restriction enzyme sites may occur less frequently in intergenic regions because of the nucleotide bias of the C. elegans genome (Sulston and Brenner 1974), and therefore, large restriction enzyme fragments produced in partial digests may be enriched for DNA fragments containing intergenic regions. DNA fragments containing mainly intergenic regions but with a small 5’ gene segment are exactly the type likely to drive β-galactosidase expression in this experiment. Therefore, the use of partial restriction enzyme digestion products, rather than truly random DNA fragments, may have contributed to the frequency of obtaining active constructs.

The existence of the plasmid DNA in transformants as an extrachromosomal array means expression patterns are incomplete. First, the extrachromosomal arrays are not transmitted to all progeny so not all animals of a transmitting line will stain. Transmission rates vary markedly between transformed lines (from approximately 5 to 95 % ) although the rates do seem to increase during the first few generations, during the establishment of transformed lines. Second, staining is not observed in all the appropriate cells of a staining individual, possibly because the array is variably lost in mitotic cell divisions or because genes on extrachromosomal arrays may become inactivated during development. A complete expression pattern may be inferred by comparing many stained individuals. Alternatively, the individual active plasmids could be identified and used to generate integrative transformants before the β-galactosidase expression patterns are precisely determined. With this in mind, the first four positive pools, pools 2, 4, 9 and 24, were selected for further study.

An efficient method for identifying the active plasmids from the pools was developed. The original bacterial clones of the library had been picked into individual wells of 8×12-well multi-well plates. Before pooling the cultures for DNA preparation, the plates had been replica-plated. The cultures were now regrown in the 8×12 arrays. Rather than pooling all the cultures of a plate as before, separate pools were made of all the cultures in individual rows and individual columns. Furthermore, pools of the corresponding rows and columns of the four plates constituting pools 2, 4, 9 and 24 were also mixed together before DNA was prepared. This procedure gave 12 new pools (labeled 1 tol2) of 32 plasmids and 8 new pools (labeled A to H) of 48 plasmids. The individual plasmid responsible for the β-galactosidase expression of each original pool should be present in two of the new pools, one prepared by pooling cultures of a single row and the other prepared by pooling cultures of a single column. The 20 new plasmid DNA pools were assayed for ability to direct β-galactosidase expression in C. elegans in the same way as the original pools.

The active plasmids of pools 4, 9 and 24 were identified in this way (Table 2) and were labeled, according to their positions in the 8×12 arrays, as pUL#4F5, pUL#9F7 and pUL#24C7, respectively. The transmitting lines established with the new pools appeared to express β-galactosidase more strongly and more completely than the lines generated in the initial screen, presumably because of the smaller pool sizes.

The expression pattern observed for the original pool 2 was not observed in this set of assays. Either the bacterial clone containing the active plasmid of pool 2 did not re-grow or the original observation was artefactual. The two expressing fines established with the original pool 2 were derived from the same injected individual and could have contained the same spuriously derived extrachromosomal array.

C. elegans was transformed with the identified plasmids, pUL#4F5, pUL#9F7 and pUL#24C7. The procedures were the same as for the initial screen except that individual plasmids were used instead of construct pools and an oligonucleotide was included in the injection solution to stimulate integrative transformation (C. Mello, personal communication). Transformants were identified as before, using the dominant marker, rol-6(su1006). β-galactosidase expression was stronger in the lines produced using purified plasmids as compared to lines produced using the plasmid pools, facilitating precise determination of expression patterns.

Five transmitting lines have been established using pUL#4F5 and all express β-galactosidase in adult body wall muscle cells, as expected. No integrative transformants have been obtained with this plasmid, which means that complete expression patterns have not yet been observed. However, probably all body wall muscle cells could express β-galactosidase from this construct. Expression was also observed in vulval muscle cells but not in pharyngeal muscle cells. The restriction of β-galactosidase activity to adult animals may accurately reflect genetic regulation or a requirement to build up detectable levels of enzyme over an extended period of time.

Attempts to introduce pUL#9F7 into the C. elegans germ-line were initially unsuccessful. A few transformed progeny showing the roller phenotype (due to the coinjection marker) were observed but were usually sterile, often rupturing at the gonad, and the few progeny occasionally produced for the next generation never showed the roller phenotype. Heritable transformation was achieved by diluting pUL#9F7 forty-fold with the vector plasmid, pPD22.11. Eight transformant lines then obtained showed the expected pattern of β-galactosidase expression, confirming correct identification of the active plasmid from pool 9.

The β-galactosidase expression pattern obtained with pUL#9F7 is quite complex. Most of the wall of the uterus is composed of just eight toroidal cells, four in each arm of the gonad. In adult hermaphrodites, β-galactosidase activity appears in two cells in each gonadal arm and may be restricted to the basolateral surface at least for the cells furthest from the vulva. β-galactosidase was also observed in the spermathecal valve cell, but here expression may be restricted to the luminal surface. Confirmation of these subcellular localizations would require immunoelectron microscopy. The subcellular distribution of β-galactosidase may explain why transformants could only be obtained with diluted pUL#9F7. Overexpression of the fusion protein could lead to disruption of the structure to which β-galactosidase is being localized. Alternatively, the insert of pUL#9F7 could contain another gene, which is toxic when present in multiple copies.

For pUL#24C7, two transmitting lines have been established and one of these shows transmission to all progeny at each generation indicating that the plasmid has integrated into the C. elegans genome. pUL#24C7 causes nuclear-localized β-galactosidase expression in a subset of cells in the 50– 200 cell embryo, the number of nuclei staining initially increasing, to maximally about 20, and then decreasing. (Precise identification of the expressing cells is underway, I.A.Hope, unpublished data). The stronger staining of these transformants revealed that β-galactosidase activity remains in two cells throughout subsequent embryonic development and for a short while after hatching. The two cells stained are Z1 and Z4, which give rise postembry-onically to the entire somatic gonad.

Expression of β-galactosidase in the screen described here should depend on translational fusion because the lacZ gene present in the vector plasmid contains no C. elegans promoter elements. Dependence on translational fusions means that the β-galactosidase of a plasmid construct and the wild-type copy of the gene to which lacZ is fused are likely to be expressed in at least a related, if not identical, pattern because all promoter elements acting on the former will also be acting on the latter. However, an active construct’s DNA insert may not contain all the relevant regulatory elements for the wild-type copy of the tagged gene. e.g. regulatory elements have been found at the 3’ end of some C. elegans genes (Ahringer and Kimble, 1991). Dependence on translational fusion also means that identification of the genes tagged with β-galactosidase will be straightforward because the junction of the fusion will be within a protein coding region. Finally, in a proportion of the constructs, β-galactosidase may be directed to a particular subcellular location by the C. elegans encoded N-terminal segment of the fusion protein and this could provide valuable information about the tagged gene.

Some evidence in support of β-galactosidase expression being dependent on translational fusion in these experiments has already been described. First, the frequency with which active constructs were obtained (1 in 110) is close to that predicted, assuming a requirement for translational fusion. Second, for some constructs β-galactosidase was apparently directed to particular subcellular structures by the N-terminai segment of the translational fusion protein, e.g. the valve expression of pool 38, the uterus wall expression of pool 9 and possibly the excretory cell expression of pool 64.

Direct evidence for translational fusion has been obtained for the identified plasmid pUL#24C7. A western blot of embryonic proteins from the integrative transformant line containing pUL#24C7 was probed with a monoclonal antibody to β-galactosidase (Fig. 3). The major, transformant-specific protein had an estimated mass of 140×10³M_r, clearly larger than the wildtype β-galactosidase of 116×10³M_r. The size difference is likely to be due to a C. elegans encoded N-terminal segment of approximately 200 amino acids. The yield of the fusion protein was estimated as 0.1% of protein extracted from C. elegans embryos.

A fusion protein has not yet been clearly detected on a western blot for pUL#4F5 or pUL#9F7. This may be because the non-integrative transformants so far obtained with these plasmids do not express sufficient fusion protein.

A physical map of the C. elegans genome has been assembled (Coulson et al. 1986). The map consists primarily of a set of cosmid, genomic DNA clones which have been ordered by comparing the sizes of HindIII– Sau3A restriction fragments, a technique called ‘fingerprinting’. pUL#4F5, pUL#9F7 and pUL#24C7 all have inserts of approximately 6.5 kb containing 2 HindIII sites. The four HindIIl to Sau3A restriction fragments they possess are insufficient to define a unique genomic location using the fingerprinting technique alone.

However, the genomic locations of the inserts could be determined using ‘YAC polytene grids’ kindly provided by John Sulston and Alan Coulson. Large C. elegans genomic DNA fragments have been cloned in yeast artificial chromosome (YAC) vectors and positioned on the physical genome map (Coulson el al. 1988). DNAs from 960 such clones have been placed in order with respect to the genome on a nylon membrane, creating a YAC polytene grid. The DNA on such a membrane is thought to represent at least 90 % of the C. elegans genome with an average two-fold redundancy. Radiolabeled DNA fragments from the inserts of the active plasmids were hybridized to the YAC polytene grids and YACs containing the same DNA fragments as the inserts were identified (Fig. 4). The genomic locations of the inserts were thus defined to approximately 100 kb. These results in combination with the fingerprint comparisons (performed by Ratna Shownkeen and interpreted by Alan Coulson) provided the precise genomic locations.

Details of the locations for the three plasmid inserts on the physical genome map are presented in Table 3. All three fusions appear to have defined new genetic loci. This is not surprising because only 830 of the 5000 genes estimated to constitute the C. elegans genome (Emmons, 1988) have so far been identified by mutation (Edgley and Riddle, 1989).

I have developed a screen of gene expression patterns for C. elegans. Earlier attempts to develop a screen similar to the Drosophila enhancer trap but using Tc1, the best characterized transposable element of C. elegans, were unsuccessful (I. A. Hope, unpublished data). Hence, an alternative approach was used. Expression patterns directed by 5’ segments from a random sample of C. elegans genes were examined using lacZ as a reporter. The promoter-reporter fusions were created in vitro and assayed after transformation into C. elegans. A preliminary screen, designed to establish the procedures, and the expression patterns obtained therein, are described here. Although the procedures are labour intensive, a larger screen, e.g. to examine hundreds of expression patterns, is feasible. The results of a larger screen would address the question initially posed concerning the general organization of gene expression necessary for the coordination of C. elegans development.

A wide range of expression patterns has already been obtained. Examples of β-galactosidase expression were observed from early embryogenesis, throughout development, to the mature adult. Also, β-galactosidase expression occurred in a variety of cell types including muscle cells (e.g. pool 4), nerve cells (e.g. pool 25), cells of the gonad (e.g. pool 9), hypodermal cells (e.g. pool 80) and undifferentiated cells (e.g. pool 24). The range of expression patterns obtained suggest many C. elegans genes could be examined with this approach.

For C. elegans, the genomic location of tagged genes can be rapidly defined, as described here for three selected pools. For each expression pattern, an active plasmid construct was identified. Then, the origin of the genomic DNA insert of each plasmid was determined using the C. elegans physical genome map. In this way, expression patterns may be matched with genes previously characterized by other genetic and molecular approaches. However, the majority of C. elegans genes have so far received little if any attention and most expression patterns would define new genetic loci. A large scale screen of gene expression patterns will add valuable information to the growing body of knowledge, based on the genetic and physical genome maps (Edgley and Riddle, 1989; Coulson et al. 1988), which is intended to lead to a thorough molecular description of C. elegans.

For many of the expression patterns described here, β-galactosidase activity appears in a structure as that structure differentiates, suggesting the tagged gene probably encodes a terminal differentiation product. Some of the expression patterns of this class arouse particular interest.

For transformant lines generated with pool 38, β-galactosidase appears to be localized to a filamentous ring which may be responsible for constriction in the intestinal– rectal and spermathecal valves. The tagged gene probably encodes a component of this ring and this observation estabfishes a molecular link between these morphologically similar structures. Are there other components or does assembly involve a single highly specialized protein? What is the structure of this ring that confers the properties of a valve to these cells?

pUL#9F7 of pool 9 induces β-galactosidase expression in the uterus. Most of the uterine wall is composed of a small set of apparently very similar cells (J. White, personal communication), yet β-galactosidase is expressed in only some of these cells. What is the basis of this distinction and is it physiologically significant? The histochemical staining pattern suggests the β-galactosidase is localized to different cell surfaces at different positions along the gonad. If real, how is this differential subcellular localization achieved and is it functionally important?

Subcellular localization makes a β-galactosidase pattern particularly striking. It will be important to determine that the subcellular distributions of the unfused C. elegans’ proteins and the β-galactosidase fusion proteins under discussion here are the same. These subcellular distributions of β-galactosidase have been observed because translational fusion is required for reporter expression. The vector plasmid used here, pPD22.11, encodes a nuclear-localization signal to ease identification of expressing cells (Fire et al. 1990). Future screens may be performed with the nuclear localization signal deleted so that subcellular localization signals of tagged genes may be more readily revealed.

Expression patterns of terminal differentiation products are the end result of developmental control pathways. Expression in early embryos, prior to differentiation, would result from regulation at intermediate steps along these pathways. An example of an expression pattern of this type has recently been described with CeMyoD which is encoded by a C. elegans gene related to the vertebrate MyoD gene family (Krause et al. 1990). CeMyoD is expressed from early embryogenesis in cells that produce body wall muscle cells.

In the screen described here, five construct pools (pools 1, 11, 24, 64 and 68) gave β-galactosidase expression in early embryos. Expression pattern characterization has been taken furthest for pool 24. The active plasmid, pUL#24C7, was identified and a transformant line carrying copies of the plasmid integrated into the genome was obtained. The identities of the cells expressing β-galactosidase during early embryogenesis have not yet been determined but during late embryogenesis and for a short while after hatching β-galactosidase expression is restricted to two cells, called Z1 and Z4.

Z1 and Z4 give rise to the entire somatic gonad (Kimble and Hirsh, 1979). These are highly specialized cells that migrate posteriorly, after their birth, to become closely associated with Z2 and Z3, the germline progenitor cells. A basal lamina surrounds these four cells, which constitute the gonad primordium. Z1 and Z4 are vital for the survival and postembryonic division of Z2 and Z3. Postembryonically, Z1 and Z4 divide to produce the cells constituting the uterus, spermathecae and oviducts in the hermaphrodite and the cells constituting the vas deferens and seminal vesicle in the male. Furthermore, Z1 and Z4 produce the anchor cell, which induces the formation of the vulva from the hermaphrodite hypodermal precursor cells, and the distal tip cells, which control the proliferation and differentiation of germ cells. Investigation of the expression pattern directed by pUL#24C7 may reveal how the identities of Z1 and Z4 are determined.

Distinctions between screens of gene expression patterns, as described here, and conventional genetic screens should be addressed. In a conventional genetic screen, mutagenesis disrupts gene activities and the resulting phenotypes form the basis of gene selection. In a screen of gene expression patterns, the time and location of gene product synthesis form the basis of selection; the function of a gene product is not readily inferred purely from when and where it is produced. This distinction between the two types of screen should also affect the direction of subsequent analysis of selected genes. For a conventional genetic screen, the basis of gene selection would be function and subsequent investigation would be primarily to determine what activity the gene product has and what it acts on. Usually this involves cloning the gene first. Starting from a gene expression pattern, the emphasis should be primarily in the opposite direction, to determine how and by what the expression pattern is generated. Genes of interest are cloned as an inherent part of the approach described here. Indeed, a plasmid containing a β-galactosidase tagged gene would be an immediate subject for molecular manipulation to define cis-acting elements. Furthermore, a C. elegans fine transformed with a β-galactosidase tagged gene would be an ideal subject for genetic analysis to identify trans-acting factors. With these considerations in mind, I intend to investigate further the early embryonic expression patterns, characterizing them to cellular resolution and investigating the mechanisms by which they are generated.

I thank Andrew Fire and Craig Mello for plasmid DNAs and technical advice, Alan Coulson and Ratna Shownkeen for the YAC polytene grids and fingerprint analysis and Rebecca Wrigley for technical assistance during part of this work. I am appreciative of advice and comments on the work and the manuscript provided by many of my LMB colleagues. I would particularly like to thank Tom Barnes, Siegfried Hekimi, Jonathan Hodgkin, Roger Hoskins, David Livingstone, Andrew Spence, John Sulston and John White for their suggestions.