Molecular characterization of the lethal of scute genetic function

Subdivision 1B of the X chromosome of Drosophila melanogaster contains at least six genetic functions involved in the development of the central (CNS) and peripheral (PNS) nervous systems. Three of these functions affect major aspects of CNS development (Jiménez and Campos-Ortega, 1979, 1987; White, 1980). One of them is lethal of scute (l’sc), which has been localized within the achaete-scute complex (AS-C) (García-Bellido, 1979).

Developmental studies and mosaic analysis had suggested that the l’sc function is required for CNS development (García-Bellido and Santamaría,1978). More precisely, a deletion of the l’sc region of the AS-C causes slight defects in neuroblast (NB) segregation and cell death in the developing embryonic CNS, which lead to the loss of neurons, gross anatomical malformation of the ventral cord and embryonic lethality (Jiménez and Campos-Ortega, 1987; Martín-Bermudo et al., 1991). Other genes of the AS-C, namely achaete (ac), scute (sc) and asense (ase), were shown to be mainly involved in PNS development, both in the embryo and in the adult (García-Bellido, 1979; Dambly-Chaudière and Ghysen, 1987; Marí-Beffa et al., 1991; Domínguez and Campuzano, 1993), although their deletion also affects CNS development, when l’sc is simultaneously absent (Jiménez and Campos-Ortega, 1987).

Genetically, the l’sc function is defined by the breakpoints of In(1)sc⁴ and In(1)sc⁹ (see Fig. 5 and Muller, 1955; García-Bellido, 1979), for it is uncovered by the synthetic chromosomal deletion obtained after recombination of both inversions. However, an extensive search for EMS-induced lethal mutations at the distal region of the X chromosome failed to demonstrate a lethal function within the AS-C (J. Lim and K. White, personal communication). This negative result may indicate that l’sc is a relatively small target for mutagenesis or that l’sc is, in itself, a complex locus.

The molecular cloning of the AS-C has shown that the breakpoints of In(1)sc⁴ and In(1)sc⁹ are separated by 20 kb. Within this stretch of DNA, one transcription unit has been detected (Campuzano et al., 1985); it gives rise to the T3 RNA which, similarly to other genes of the AS-C, encodes a basic-helix-loop-helix protein (bHLH) (Villares and Cabrera, 1987; Alonso and Cabrera, 1988; González et al., 1989; and the present work). This RNA has been assumed to correspond to the l’sc function (Campuzano et al., 1985). Thus, the T3 RNA accumulates in all known neurogenic regions of the embryo (Cabrera et al., 1987; Romani et al., 1987), and the corresponding protein is found in proneural clusters of the neuroectoderm and in the segregating NBs (Cabrera, 1990; Martín-Bermudo et al., 1991). However, conclusive evidence for T3 RNA involvement in the l’sc function is still lacking. It is also unknown how much of the T3 gene flanking regions, or of more distant regions of the AS-C, are necessary for this function.

To resolve the above issues, we have first improved on the molecular characterization of the T3 transcription unit, by defining its transcriptional origin and termination and by sequencing both a near to full-length cDNA and a genomic fragment with nearly one kilobase of upstream sequences. Secondly, we have shown, by P-mediated transformation experiments, that short fragments of DNA containing the T3 transcribed sequences rescue, although in only a fraction of the mutant embryos, the lethality and the CNS mutant phenotype associated with l’sc deletions. However, these transgenes, in a l’sc-background, are expressed in a simpler pattern and at lower levels than the wild-type T3 gene, suggesting that additional sequences are necessary for wild-type expression. Indeed, the analysis of T3 expression in a series of AS-C mutations has revealed the existence of multiple cis-regulatory regions, distributed along the entire complex, required for the proper activation of T3 in specific proneural clusters of the embryonic neuroectoderm. These results unambiguosly define the T3 transcription unit as the main, or exclusive, component of the l’sc function, and demonstrate that this gene, like the ac and sc genes, has highly complex cis-regulatory regions.

Subclones of the 3.2 kb EcoRI fragment (coordinates +18.1 to +21.3 of the AS-C physical map Campuzano et al., 1985) that contains the T3 structural gene and of a near to full-length cDNA (Campuzano et al., 1985) were obtained in M13 (Messing et al., 1977) and pBluescript KS(+) (Stratagene) vectors. Progressive deletions of these clones were prepared by exonuclease III treatment (Henikoff, 1984). Either single- or double-stranded templates were sequenced by the dideoxy chain termination method (Sanger et al., 1977). Part of the sequence was obtained using the modification of Tabor and Richardson (1987) to use chemically modified T7 DNA polymerase. Synthetic oligonu-cleotides of defined sequence were used to extend sequences and fill in gaps. Consensus sequences were assembled and analyzed with the help of the Staden (1980) and University of Wisconsin GCG software packages (Devereux et al., 1984). Sequences were determined in both strands.

An RNAse protection experiment was performed, as described in Melton et al. (1984) and Campuzano et al. (1986), with a singlestranded RNA probe, homologous to the DNA contained between the HindIII and PstI sites located at positions 307 and 1189 of the T3 sequence (Fig. 1). This indicated that the T3 RNA and the probe were complementary for approximately 221 nucleotides towards the left of the PstI site and suggested that the origin of transcription was in the vicinity of T973. This was confirmed with primer extension experiments, which showed that transcription starts on A972. These experiments were performed according to Bensi et al. (1985). A synthetic oligonucleotide complementary to the sequence extending from T999 to C1019 was labeled with T4 polynucleotide kinase and [γ-³²P]ATP (3000 Ci/mmole) and hybridized at 60°C for 30 minutes in 25 mM Tris, pH 7.7, 75 mM KCl, 3 mM MgCl2 and 5 mM dithiothreitol with 50 mg of poly(A)⁺ RNA from 0-12 hour old embryos. The hybridized primer was extended with 10 units of AMV reverse transcriptase and deoxynucleotide triphosphates by incubation at 42°C for 1 hour. The length of the synthesized DNA was 48 nucleotides.

p3.2T3contains a genomic DNA fragment of 3.2 kb EcoRI-EcoRI, (coordinates +21.3 and +18.1 from the AS-C molecular map, Campuzano et al., 1985). It was inserted in the EcoRI cloning site of the transformation vector pPS×D1AAdh⁺.

p5.9T3 contains a genomic fragment of 5.9 kb XbaI-EcoRI (coordinates +23.3 to +18.1). It comprises 2.7 additional kilobases in the 5’ region of the former construct. The genomic fragment was cloned into the XbaI site of the transformation vector pC20.1ry⁺.

P-element-mediated transformation into Adh^fn6cn; ry⁵⁰⁶ and ry⁵⁰⁶ embryos was performed according to Rubin and Spradling (1982), using 0.5 mg/ml of either p3.2T3or p5.9T3 plasmid DNA and 0.15 mg/ml of p π25.7wc DNA, to provide for transposase (Karess and Rubin, 1984).

All AS-C mutant strains used in this work are from the collection of A. García-Bellido and are described in Lindsley and Zimm (1992). The Df(1)sc^4Lsc^9R, labelled with y and w, was synthesized anew, by recombination of viable In(1)sc⁴ and In(1)sc⁹ chromosomes, to avoid modifiers that might have accumulated in older stocks of the deficiency. In spite of this precaution, the viability of Df(1)sc^4Lsc^9R/y2Y61l males was only 7%, as compared to that of +/y2Y61l males, even though the duplication y2Y61l covers the entire AS-C and completely rescues the lethality of Df(1)sc¹⁹(ac-sc-l’sc-) flies (data not shown). The impaired viability of Df(1)sc^4Lsc^9R/y2Y61l flies should, therefore, be attributed to deleterious factors mapping outside of the deficiency, most probably to the proximal breakpoints of the two inversions used to synthesize the deficiency. For this reason, the estimated percentage of adult viability of Df(1)sc^4Lsc^9R flies rescued with T3 transgenes (Table 1) has been corrected by a factor that takes into consideration the reduced viability of the Df(1)sc^4Lsc^9R/y2Y61l control flies.

Dechorionated embryos were fixed for 20 minutes in heptane saturated with 4% formaldehyde in PBS, and vitelline membranes were removed by methanol treatment. All subsequent incubations were performed in PBS, 0.1% Tween 20, 0.1% bovine serum albumin. For the analysis of CNS phenotypes in AS-C deficiencies rescued with T3 transgenes, embryos were first treated with a rabbit anti-horseradish peroxidase (HRP) antibody (Cappel) (Jan and Jan, 1982), then with a HRP-labelled secondary antibody, and finally stained with DAB and H2O2. For the analysis of l’sc expression in AS-C mutants, embryos were incubated with a rat anti-l’sc antibody (Martín-Bermudo et al., 1991), followed by a biotin-labelled secondary antibody and the Vectastain Elite ABC kit (Vectorlabs). Before the staining reaction, the embryos were incubated with 1% glutaraldehyde in PBS for 5 minutes and washed for 20 minutes, treatment that substantially reduces the background. DAB staining was performed in the presence of Ni and Co ions.

Plasmid DNA, fly RNA and single-stranded RNA probe preparations were performed as described (Maniatis et al., 1982; Melton et al., 1984; Campuzano et al., 1985).

The T3 gene gives rise to a 1.1 kb RNA (Campuzano et al., 1985; and Fig. 5). The sequence of a genomic fragment of T3 DNA that comprises an open reading frame has previously been reported (Alonso and Cabrera, 1988). We have further characterized the T3 transcription unit by independently sequencing a larger segment of genomic DNA and a near to full-length cDNA, and by determining its origin of transcription. The results are shown in Fig. 1. RNAse protection and primer extension experiments showed that transcription starts on A972. Thirty nucleotides upstream from this point, a rather unconventional TATA box (TATTTAAA) is found. Seven CANNTG motifs, putative binding sites for bHLH transcriptional regulators (Blackwell and Weintraub, 1990), are located in the upstream region, which suggests that T3 transcription may be regulated by its own protein and/or by other bHLH proteins (see below). The sequences comprising and surrounding the E-boxes at positions 255 and 809 are reminiscent of snail protein-binding sites (Ip et al., 1992). snail represses l’sc transcrip-tion in the presumptive embryonic mesoderm (Kosman et al., 1991).

Genetic and molecular data suggest that the T3 transcription unit, the only one detected within the 20 kb of DNA uncovered by the synthetic chromosomal deletion Df(1)sc^4Lsc^9R, is responsible for the l’sc function (García-Bellido, 1979; Campuzano et al., 1985; Jiménez and Campos-Ortega, 1987). This is the smallest available deficiency of the l’sc gene for which no point mutations have yet been isolated. To obtain more conclusive evidence for the identity between the T3 and the l’sc genes, we prepared, by P-element-mediated transformation, Drosophila lines carrying a 3.2 kb DNA fragment that contains the T3 transcribed region, 0.9 kb of the upstream region and 1.2 kb of the downstream region. Four of such transformants (lines 3.2T3.1-4) were tested for their ability to rescue the embryonic lethality associated with the Df(1)sc^4Lsc^9R. One copy of any of the transgenes rescued the embryonic lethality of a fraction of the l’sc embryos, up to 26% with line 3.2T3.1 (not shown). Two to four copies improved rescuing to 48% (Table 1, cross 1). However, only a few of the l’sc-larvae reached adulthood, being the adult viability 8% when compared to the viability of Df(1)sc^4Lsc^9R/y2Y61l control flies (see Materials and Methods). The rescue of the embryonic lethality of the larger deletion Df(1)sc¹⁹(ac-, sc-, l’sc) with two to four copies of the transgene (Table 1, cross 2) was 47%. In this case, all larvae died before the third instar, probably because of the absence of the ac and sc genes, which should aggravate the effects of the inefficient rescue of l’sc.

To attempt improving the rescue achieved with the 3.2T3 transformants, we prepared fifteen new transformant lines (5.9T3.1-15) with a fragment that provided 2.7 kb more DNA in the T3 upstream region, that is, a total of 3.6 kb. With one copy of this transgene, line 5.9T3.1 was the most efficient in yielding Df(1)sc^4Lsc^9R viable adults (not shown). One to two copies of this 5.9T3.1 gene allowed hatching of 48% of l’sc-embryos (Table 1, cross 3). Some of them reached adulthood, the adult viability being nearly 20% when compared to that of Df(1)sc^4Lsc^9R/y²Y61l control flies. This indicates that, although the 5.9T3.1 insertion rescues l’sc-lethality more efficiently than any of the 3.2T3 insertions, its rescuing capacity is still far from complete.

We also investigated the capacity of the 3.2T3 transgene to rescue the neural phenotype of l’sc deletions. Since a fraction of the rescued Df(1)sc^4Lsc^9R embryos reached adulthood, we assumed that the gross CNS morphology of at least that fraction of embryos should be fairly normal. Therefore, they were not further analyzed. Df(1)sc¹⁹ embryos show a partial hypoplasy of the CNS (compare Fig. 2A and B), which is largely due to the lack of l’sc (Jiménez and Campos-Ortega, 1987). Two to four doses of the 3.2T3 gene substantially improved this mutant phenotype (Fig. 2C). Similarly, the severe neural hypoplasy of Df(1)260.1 embryos (Fig. 2D), which lack the entire AS-C, was also considerably rescued (Fig. 2E).

We also analyzed the expression of the 3.2T3 and 5.9T3 transgenes in Df(1)sc^4Lsc^9R embryos. In the wild type, T3 starts to be expressed in the blastoderm (stage 5), in a pattern that changes during gastrulation (Cabrera, 1990; Martín-Bermudo et al., 1991). By the end of stage 8, a well-defined pattern of groups of ectodermal cells (proneural clusters) and the S1 NBs that delaminate from them, express T3. Later, in stage 9, T3 expression occurs in new proneural clusters that give rise to the S2 NBs. In contrast, during these stages, the transgenes were apparently not expressed (not shown). T3 protein accumulation from the transgenes was first detectable in stage 9, both in NBs and, to a lesser extent, in the overlaying neuroectoderm (Fig. 3A,B,D). Moreover, this late expression was much weaker than that of the wildtype gene and it was difficult to appreciate its characteristically defined spatial pattern (Fig. 3C,E).

In summary, we believe that the rescue of the lethality and of the neural phenotype (although incomplete, most likely because of the imperfect expression of the transduced genes), together with previous genetic and molecular evidence, unequivocally confirm that the T3 gene is responsible for the l’sc function. Accordingly, we shall refer from now on to the AS-C T3 gene as the l’sc gene.

The low level and simplified patterns of expression of the l’sc transgenes suggest that they do not contain all the regulatory elements necessary for wild-type expression. These elements should lie somewhere else within the 90 kb that spans the AS-C. In fact, expression of the sc gene in imaginal discs appears to be driven by discrete cis-regulatory elements, scattered through a length of 50 kb (Ruiz-Gómez and Modolell, 1987). Similar c/s-regulatory sites have been proposed to direct embryonic ac expression both in the CNS (Skeath et al., 1992) and in the PNS (Ruiz-Gómez and Ghysen, 1993). To test for the presence of regulatory elements for l’sc, we analyzed its expression in ASC mutants. Most of the mutations studied are small deletions or chromosomal rearrangements (inversions or translocations) that either remove portions of the AS-C or physically separate them from the l’sc coding region, thus uncoupling presumptive cis-regulatory elements from the basal promoter (Fig. 5A). For the sake of simplicity, we have only examined L’SC protein accumulation in the NBs that delaminate during the first two waves of segregation and in their corresponding proneural clusters. All these NBs, except one, accumulate L’SC to different extents in the wild type (Martín-Bermudo et al., 1991; and Fig. 4A). Fig. 4B-D shows L’SC accumulation in a few of the cases analyzed. In most mutants, l’sc expression is abolished in a particular subset of proneural clusters and corresponding NBs. Some light staining in these NBs and overlying ectoderm is first detected by late stage 9, but always at much lower levels than in the wild type. This late expression is similar to that found in Df(1)sc^4Lsc^9R embryos carrying l’sc transgenes (Fig. 3A-B). Table 2 summarizes the analysis of 22 AS-C mutations. Two main features stand out: first, mutations that map proximal or distal to the l’sc structural gene suppress its expression in different subsets of NBs; and second, NBs affected by any given breakpoint are also affected by all other breakpoints that map between it and the structural gene. These results strongly suggest the presence, at both sides of the transcribed sequence, of discrete cis-regulatory regions that activate l’sc in specific proneural clusters and, as a consequence, in their corresponding NBs.

The data allow definition of at least five regions containing cis-regulatory elements that drive l’sc expression in twelve proneural clusters (Fig. 5). Three l’sc-expressing NBs, NBs 3-5, 7-1 and 7-4, also express ac and sc (Martín-Bermudo et al., 1991; Skeath et al., 1992; Ruiz-Gómez and Ghysen, 1993). Interestingly, control regions A and B, required for wild-type expression of l’sc in these NBs, also seem to contain enhancer elements for ac (and probably sc) expression in the same NBs (Skeath et al., 1992, for region A; and Ruiz-Gómez, unpublished observations, for region B). The possibility thus arises that, in these NBs, either l’sc activates ac and/or sc, or vice versa. In fact, in imaginal discs, ac and sc can self-activate and cross-activate each other (Martínez and Modolell, 1991; Skeath and Carroll, 1991). Df(1)sc^4Lsc^9R, which removes l’sc, still allows ac/sc expression in these NBs, demonstrating that their initial activation is l’sc-independent, although l’sc appears to enhance the level of ac expression in NBs 7-1 and 7-4 (Skeath et al., 1992; and our own observations). Moreover, NBs 7-1 and 7-4 express l’sc, but not ac, in In(1)sc^SI mutants (Table 2 and data not shown), indicating that l’sc expression in these two NBs is at least ac independent. The same can be suggested for NB 3-5, for, in In(1)sc^V2, it expresses l’sc normally (Table 2), but not ac (Skeath et al., 1992). Therefore, regions A and B may contain independent enhancers for ac/sc and l’sc, or, alternatively, the same enhancers act simultaneously on the promoters of the three genes.

We have shown that two transduced AS-C T3 genes, containing the transcribed region and relatively short flanking sequences, rescue the lethality of a l’sc deficiency, albeit with low penetrance. Likewise, we have shown that the smaller transgene rescues to a great extent the neural phenotype associated with larger deletions of the AS-C. These phenotypes are largely due to the lack of the l’sc function (Jiménez and Campos-Ortega, 1987). Therefore, our results unequivocally demonstrate, confirming previous indirect evidence (see Introduction), that the AS-C T3 gene and the genetically defined l’sc function are the same entity.

Remarkably, l’sc mutants can reach adulthood, even though the amounts of L’SC protein provided by the transgenes are substantially lower than normal and its distribution does not conform to the wild-type pattern. This indicates that the deployment of specific quantities of L’SC is not essential for its function. Moreover, in many AS-C mutations l’sc expression is practically abolished in specific NBs and flies are still viable. These data are consistent with the view that l’sc function is partially dispensable and that it can be largely replaced by that of other proneural genes (Martín-Bermudo et al., 1991). This in turn implies that the function of the different proneural genes is essentially the same, namely, to confer upon groups of ectodermal cells a capacity to become neural precursors. Only the extent of expression of each gene, as measured both by the number of proneural clusters that express it and by the level of its expression in each cluster, would establish its relative importance for the proneural function. On the other hand, if the l ‘sc function is to a great extent replaceable, this argues against its participation, together with other AS-C genes, in the specification of particular NB fates, according to a proposed model based in a combinatorial mechanism of action (Cabrera et al., 1987; Cabrera, 1992).

The wild-type expression of l’sc undergoes dynamic changes during embryonic development, suggesting that its control is complex (Cabrera, 1990; Martín-Bermudo et al., 1991). Indeed, the analysis of l’sc expression in AS-C mutations has revealed at least five cis-controlling regions scattered throughout the gene complex. These regions contain the presumptive enhancers that activate l’sc in twelve proneural clusters that give rise to S1 and S2 NBs. Presumably, other enhancers exist to drive l’sc expression in the proneural clusters that give rise to later segregating NBs and to sensory mother cells (SMCs) of the peripheral nervous system, where l’sc is also expressed (Romani et al., 1987; and our observations). What is the need for so many cis-regulatory elements acting on the l’sc promotor? The reason may be the positional and temporal accuracy with which neural precursors delaminate from the neuroectoderm. Each precursor arises from a small, proneural group of cells defined by the activation, in response to positional cues, of one or more proneural genes. These positional cues should be different, at least in part, for every particular cluster. As l’sc expression helps define most of the proneural clusters within the neuroectoderm, its promotor should respond to a large variety of cues and, hence, the need for a substantial number of enhancer elements.

It seems paradoxical that, despite such complex regulation, most NBs can still segregate in the absence of l’sc (Martín-Bermudo et al., 1991). A most likely explanation is the expression of other proneural genes in the same proneural clusters. For example, ac and sc may partially substitute for the absence of l’sc in the proneural clusters of NBs 3-5, 7-1 and 7-4, where these three genes are normally coexpressed. However, there is no correspondence between the frequency with which a particular NB is absent in a l’sc deficiency and the extent of AS-C gene expression in that NB. Again as an example, NB 7-1 is the one most commonly absent in Df(1)sc4^Lsc9^R embryos (Martín-Bermudo et al., 1991), even though it coexpresses ac/sc, whereas NB 5-2, which expresses l’sc, but neither ac nor sc, segregates normally in the same mutant embryos. This indicates that the requirement for l’sc is not necessarily higher in those NBs where ac and sc are not expressed. It further suggests,as previously noted (Jiménez and Campos-Ortega, 1990), that the AS-C genes do not account for all the proneural activity in the CNS. If l’sc can be replaced in many sites by other proneural genes, why is it still expressed in most proneural clusters? We suggest that the phenotypic criteria so far used to characterize AS-C mutations, namely, neuroblast segregation and viability, probably do not reveal the full extent of the l’sc function. It has been shown for the imaginal SMCs that the levels of proneural products and, in some cases, the specific combination of these proteins can be critical for the correct implementation of the neural differentiation programme (Domínguez and Campuzano, 1993). Proneural insufficiency may be compatible with SMC emergence, but it can cause malformations of the resulting sensory organ. Similarly, l’sc insufficiency may subtly damage the NBs progeny and still be compatible with fly viability.

On the other hand, there may be sites where l’sc expression is truly redundant. Expression in these sites could be evolutionarily maintained by constraints imposed by the specificity and organization of the enhancers within the ASC. If an enhancer directs expression in more than one site or if it interacts with other enhancers, so that the final expression is not merely the sum of the individual expressions, it may be impossible to delete expression in sites where l’sc is entirely redundant without removing it from places where it is necessary. Thus, these enhancers and the patterns of expression they direct would have been conserved in evolution despite the true redundant character of the expression at some sites.

It is likely that the same positional cues activate different genes in the same proneural cluster. In agreement with this, the cis-acting elements that control ac and l’sc expression in NBs 3-5, 7-1 and 7-4 are found in the same AS-C regions and most probably are the same for both genes. It is interesting that other l’sc-expressing NBs (for example, NBs 52 and 5-6), whose corresponding enhancers are located proximal to the l’sc transcription unit, do not express sc (Skeath et al., 1992), although this gene is activated, in the wing disc, by enhancers located in the same region. And, conversely, l’sc is not activated in the proneural clusters of the wing disc by those sc-specific enhancers (S. Romani, unpublished observations). This indicates that the sc and l’sc promoters have different responsiveness to the set of enhancers that are intermingled within the proximal region of the AS-C, or that the same enhancers alternatively activate one or the other promoter depending on the tissue. Strong candidates to activate the cis-regulatory elements that drive expression of AS-C genes are the axis-patterning genes that operate early in embryonic development. Thus, the pattern of l’sc expression in the neuroectoderm undergoes specific modifications in mutants for the pair-rule genes (Martín-Bermudo et al., 1991) and that of ac is set up by these and the dorsoventral coordinate genes (Skeath et al., 1992).

Another level of regulation of l’sc expression is suggested by the presence, upstream from the origin of transcription,of a cluster of E-boxes, putative binding sites for bHLH transcription factors. In the wild type, the initial l’sc expression, most likely mediated by the enhancers referred to above, is followed by a progressive accumulation of the protein in the proneural clusters and in the delaminating NBs (Martín-Bermudo et al., 1991). Such accumulation could be explained by a positive feedback regulation of the L’SC protein on its own promoter or by cross-activation by other bHLH proteins. Indeed, E-boxes participate in the ac selfstimulation and its activation by sc that occur in the imaginal discs (Martínez and Modolell, 1991; Van Doren et al., 1992; Martínez et al., 1993). The late and weak L’SC accumulation in NBs and in neuroectodermal cells observed in l’sc-embryos carrying a transgene, or in NBs of AS-C mutants, might also be explained by similar positive regulation involving the E-boxes that would amplify a basal l’sc expression.

We thank our laboratory colleagues for constructive criticism of the manuscript and to A. González-Reyes for help in computer drawing of Fig. 5. MDMB, FG and MD were predoctoral fellows from the Ministerio de Educación y Ciencia, SR was a predoctoral fellow of Instituto de Cooperación Iberoamericana and MRG was a postdoctoral fellow of the Consejo Superior de Investigaciones Científicas. This work was supported by grants from Dirección General de Investigación Científica y Técnica (PB90-0082 to FJ, and PR84-0212 and PB90-0085 to JM), Comunidad Autónoma of Madrid to JM, and an institutional grant from Fundación Ramón Areces to the Centro de Biolog’a Molecular.