The helix-loop-helix extramacrochaetae protein is required for proper specification of many cell types in the Drosophila embryo

The function of the Drosophila extramacrochaetae (emc) gene has been characterized almost exclusively in relation to the pat-terning of the fly’s sensory organs (chaetae and other types of sensilla). Thus, the first alleles of emc were isolated in a gene-dose titration analysis (Botas et al., 1982) to find regulatory genes of the achaete-scute complex (AS-C). The AS-C comprises four proneural genes that confer to cells the ability to become neuroblasts or sensory organ mother cells (Campuzano and Modolell, 1992, review). Viable combina-tions of emc alleles generated ectopic extramacrochaetae whose number was related to the number of AS-C copies present in the flies. Consequently, emc was proposed to be a negative regulator of the AS-C (Botas et al., 1982; Moscoso del Prado and García-Bellido, 1984).

The four AS-C proneural genes and emc encode proteins bearing the helix-loop-helix (HLH) dimerizing domain (Garrell and Campuzano, 1991; Murre and Baltimore, 1992, reviews) characteristic of a large family of transcriptional reg-ulators that function as homo-or hetero-dimers. Most proteins of this family, including the proneural ones, have, adjacent to the HLH domain, a basic region necessary for DNA binding (Murre and Baltimore, 1992). Emc lacks this region and, con-sequently, it has been proposed that Emc antagonizes proneural function, and thereby sensory organ generation, by sequestering the proneural bHLH proteins in complexes incapable of effective interaction with DNA (Ellis et al., 1990; Garrell and Modolell, 1990). Experiments in vitro support this view (Van Doren et al., 1991; Martínez et al., 1993). In addition, emc insufficiency increases the accumulation in imaginal wing discs of at least one of the AS-C proneural products (that of scute (sc), Cubas et al., 1991; Cubas and Modolell, 1992), and variations in Emc levels modulate the transcriptional activity of the achaete (ac) promoter (Martínez et al., 1993; Van Doren et al., 1992). This suggests that EMC also antagonizes proneural activity by downregulating ac/sc transcription. These interactions between proneural genes and emc help tune the levels of proneural function, sharpen the spatial distribution of proneural activity in the imaginal discs epithelia, and consequently refine the positions where sensory organs arise (Cubas and Modolell, 1992; Van Doren et al., 1992; Campuzano and Modolell, 1992, review). In the mouse, the family of Id proteins, which also lack this basic region and probably are the vertebrate homologs of Emc, negatively regulate myogenesis by a mechanism similar to that of Emc in neurogenesis (Benezra et al., 1990; Jen et al., 1992; Sun et al., 1991).

Some, albeit scarce, data suggest that during embryonic and imaginal development emc plays roles other than antagonizing the proneural function. For instance, embryos harbouring the strongest lack-of-function emc alleles die and display anomalies in the cuticle (García-Alonso and García-Bellido, 1988). In the wing, clones of cells bearing the most extreme alleles are cell lethal and clones carrying weaker but still embryonic-lethal alleles have abnormal sizes and shapes and differentiate extraveins (García-Alonso and García-Bellido, 1988). These analyses suggested that emc plays roles in cell proliferation, cellular recognition and vein patterning (García-Alonso and García-Bellido, 1988). Finally, maternal Emc is needed in sex determination for the proper communication of the X:A ratio, which results in inactivation of Sxl in males (Younger-Shepherd et al., 1992).

The aim of this work has been to study the role of emc during embryonic development. This might reveal aspects of emc function whose analysis cannot be easily approached either in emc viable mutant backgrounds or in cell clones. Moreover, it might disclose interactions of emc with other proteins which are not present or whose expressions do not occur together with that of emc during imaginal development. To this end, we have analyzed the embryonic expression of emc by hybridization in situ to wild-type and mutant embryos and the embryonic phenotype of emc strong, lack-of-function alleles. We find that emc is required for a variety of developmental processes occurring in derivatives of the three embryonic layers.

emc¹ is a point mutant (Garrell and Modolell, 1990), emc^FX119 was induced by X-rays, and the deficiency emc^E12 deletes approximately ten chromosomal bands, including the emc locus. These alleles are genetically described by (García-Alonso and García-Bellido, 1988; Moscoso del Prado, 1985). emc^ip15 and emc^iX10 were obtained by P. Heitzler in a P element-mediated and an X ray-induced mutagenesis, respectively. Mutant chromosomes were kept over a TM2-lacZ balancer to allow identification of homozygous mutant embryos by X-gal staining. To minimize the effect of maternal inheritance, whenever possible mutant embryos were obtained from TM2/emc^E12 mothers.

Embryos were dechorionated in bleach, devitellinized in a mixture of heptane/methanol, washed once in methanol, several times in PBS and mounted directly in a 1:1 mixture of lactic acid and Hoyers’ medium (Van der Meer, 1977).

Hybridizations in situ to whole mounts of embryos (prestained with X-gal to reveal β-galactosidase-expressing balancers when required, Ashburner, 1989) was performed according to Tautz and Pfeifle (1989). Digoxigenin-labelled probes were prepared with an emc cDNA clone of nearly full length (Garrell and Modolell, 1990) or with a dpp cDNA clone (Padgett et al., 1987). Immunological staining of whole mounts of embryos, using the ABC system (Vector Labora-tory), were made as described in Ruiz-Gómez and Ghysen (1993).

During most of embryogenesis, emc is expressed in complex and rapidly evolving patterns. emc mRNA is already detectable in syncytial blastoderms, where it is homogeneously distributed. This mRNA is probably of maternal origin, since emc is expressed during oogenesis in the nurse and follicle cells (not shown). Fig. 1 describes the emc mRNA distribution in the subsequent developmental stages, from cellular blastoderm to the post dorsal-closure stage (stage 15, according to Campos-Ortega and Hartenstein, 1985). The mRNA conspicuously accumulates in or adjacent to regions that will undergo mor-phogenetic movements. Examples are the cephalic (Fig. 1C) and ventral furrows (Fig. 1D), the amnioproctodeal invagina-tion (Fig. 1C), the anterior midgut primordium (Fig. 1E), the dorsal folds (Fig. 1F, G), the invaginating stomodeum (Fig. 1G, I), the tracheal pits (Fig. 1J), and the intersegmental grooves (Fig. 1M). Remarkably, the mRNA is virtually absent from the ectodermal layer from which neuroblasts segregate (Fig. 1H). This suggests that, in contrast to the developing adult peripheral nervous system, emc is not paramount in controlling proneural function in the embryonic CNS. emc is nevertheless expressed at later stages in the developing ventral cord (Fig. 1N). Other regions of abundant emc expression are the meso-dermal layer (Fig. 1G,H), the visceral mesoderm (Fig. 1L,N), the rudiments of the Malpighian tubules (Fig. 1I), the tubules themselves (not shown), and the muscle attachment sites (apodemes, Fig. 1N).

We have studied several of the strongest embryonic lethal emc alleles available (emc1, emcFX119 and emcip15), either singly or in combination with a deficiency for the emc chromosomal region (emcE12). The emc mRNA expression in emcFX119 and emcip15 homozygous embryos was extremely weak, although detectable (not shown). This is consistent with the strong phenotype of these mutants and it suggests that they still have residual emc function. The DNA lesions associated with these mutations are unknown but may affect important emc regula-tory regions.

emc mRNA distribution was unaffected in emc1 embryos but its abundance was several fold higher than in the wild type (not shown). emc1 is a point mutant in which a valine residue has been replaced by glutamic acid within the HLH domain (Garrell and Modolell, 1990), a modification that impairs the function of the Emc1 protein (Martínez et al., 1993). Assuming that no other lesions are present in the emc1 gene, the increased mRNA abundance suggests a negative regulation of emc tran-scription by the Emc protein.

Phenotypes were examined in derivatives of the three embryonic layers. As a rule, the strength of the defects, although variable, could be ordered from weakest to strongest as follows: emc1 < emc1/emcE12 < emcFX119 = emcip15 = emcFX119/emcip15 < emcFX119/emcE12 = emcip15/emcE12 < emcE12/emcE12. In all cases, the phenotype of emcE12 embryos was a strongest version of the alterations observed in the hypo-morphic combinations. This suggests that the phenotype of this deficiency may correspond to the emc null condition. Unless otherwise specified, the phenotypes described below corre-spond to the emcip15/emcE12 combination.

emc mutant embryos did not hatch and their cuticle displayed multiple alterations. Thus, head involution and differentiation of the mandibular apparatus, spiracles and filzkörper could be incomplete or totally absent (Fig. 2). Denticle belts were also affected (Fig. 2B, see also below), but not so the derivatives of the maxillary and antennal segments. Other alterations are described in the legend to Fig. 2.

emc was required for the proper formation of the interseg-mental grooves. Thus, in emc mutants accumulation of Fasciclin III (Patel et al., 1987) was reduced and many grooves appeared shallower than their wild-type counterparts or were absent (Fig. 3A-C). Moreover, accumulation of Engrailed protein, which takes place at the anterior side of each groove, occurred in fewer cells than in the wild type and to lower an extent (Fig. 3D,E) or was undetectable. Furthermore, after secretion of the cuticle, most of the denticles in the first and second denticle rows, which in the abdominal segments abut the intersegmental boundaries (Campos-Ortega and Hartenstein, 1985), were missing (Fig. 3F,G). In contrast, many of the denticles of the remaining rows were present although in somewhat disorganized arrays. These findings suggest a faulty specification of the intersegmental border cells.

Consistently with the absence of emc expression in the neu-roectoderm (Fig. 1H), the early patterns of proneural clusters, neuroblasts and SMCs, as visualized by Achaete, Hunchback and Cut staining, respectively, seemed normal in emc mutants (not shown). However, just before the start of germ band retraction, the spatial distribution of neuroblasts and their descendants, the ganglion mother cells, was more disorganized than in the wild type (Fig. 4A,B). Later, when the first neural processes appeared and emc started to be strongly expressed in the ventral cord (Fig. 1L), Fasciclin II staining revealed anomalous axonal patterns in approximately 50% of emcip15/emcE12 embryos. Anomalies ranged from subtle devi-ations of axonal pathways (Fig. 4D) or localized defects in the longitudinal connectives of older embryos (Fig. 4F), to a general disorganization of the pattern (Fig. 4E). Anomalies were also found in the PNS and consisted in a few missing neurons, rare changes of neuronal identity and, in the most extreme cases, a general disorganization of the neuronal groups and axonal routing (Fig. 4I,J). Axonal misrouting could be so extreme that axons belonging to cells from different segments fasciculated together (Fig. 4G,H) In emcE12 homozygous embryos, all these anomalies were exaggerated and many segmental and intersegmental nerves were interrupted or missing (not shown). Taken together, these defects indicate that emc does not play a mayor role in the early patterning of the embryonic CNS and PNS, but is involved in neuronal differentiation.

Malpighian tubules develop from four separate outpushings of the prospective hindgut (Campos-Ortega and Hartenstein, 1985) by, initially, cell proliferation and, later, polytenization and rearrangement of the cells to elongate the tubules (Janning et al., 1986; Skaer and Martínez-Arias, 1992). In the prolifer-ative stage, an enlarged cell, the tip cell, appears at the distal part of each tubule. The tip cell is required for proliferation of the cells near to it (Skaer, 1989).

emc expression preceded and accompanied formation of the Malpighian tubules. Thus, emc mRNA accumulated in the proctodeum, before and during emergence of the tubules rudiments (Fig. 1G,I). It was present in these rudiments during the proliferation and elongation period, but not in the tip cell and the fully differentiated tubules (not shown). emc mutations drastically affected the development of the tubules. Thus, in emcip15/emcE12 embryos, each tubule rudiment had from two to eight tip-like cells (Fig. 5A-C), as identified by the tip cell marker Achaete (M. Hock and H. Skaer, personal communi-cation; see also Romani et al., 1987). These extra tip cells did not enhance proliferation, since the tubules had at most a normal number of cells. Moreover, the tubules did not elongate (Fig. 5D,E). In emcE12 deficiency embryos, tubules remained as rudiments and contained only a fraction of the final number of cells of the mature wild-type tubules (not shown). Thus, cell allocation and/or proliferation were affected.

emc mRNA accumulates in the mesoderm until the end of stage 10 (Fig. 1G,H), before the mesoderm splits into the somato-pleura and splanchnopleura, which will give rise, among other structures, to the somatic and visceral muscles, respectively (Campos-Ortega and Hartenstein, 1985). emc is no longer expressed in the somatic mesodermal cells, but it is again expressed in the visceral mesodermal cells from stage 13 to 16 (Fig. 1K,L).

emc mutations caused defects that ranged from absence of some muscles (Fig. 6A,B) to great disruptions when many muscles were missing (not shown). Muscles occasionally formed large masses spanning more that one segment and in many instances muscles detached from the epidermis (Fig. 6C) and yielded a l(1)myospheroid-like phenotype (Newman and Wright, 1981). Whereas muscle detachment is probably caused by wrong specification of the apodema (where emc mRNA accumulates, Fig. 1N), the disruption of the muscle pattern may be due to an early requirement of emc in the mesoderm. This is supported by our finding that the levels of Twist, a protein involved in mesodermal specification (Simpson, 1983; Thisse et al., 1988), were generally decreased in stage 11 emcip15/emcE12 embryos (not shown). Moreover, inaccurate specification of the cells at the segmental borders may also contribute to the extreme disorganization of the muscle patterns found in the strongest combinations (Fig. 6C) (Volk and VijayRaghavan, 1994).

During germ band shortening, the visceral mesoderm forms two lateral palisade-like bands of tightly packed cells (Campos-Ortega and Hartenstein, 1985), which accumulate Fasciclin III (Patel et al., 1987; Fig. 7A,C). These bands join the anterior and posterior midgut primordia before they fuse together (Reuter et al., 1993). emc insufficiency resulted in uneven and zigzagging mesodermal bands that contained cells that were less tightly packed, more irregularly shaped and had less Fasciclin III than the wild-type controls (Fig. 7B). These anomalies were exaggerated in embryos homozygous for the emc deficiency (Fig. 7D).

In the following stage, each visceral mesodermal band splits in the middle along its length and the two halves move apart (Fig. 7E). Then, cells from both parts spread dorsally and ventrally to cover the midgut completely (Campos-Ortega and Hartenstein, 1985; Hartenstein and Jan, 1992; Reuter and Scott, 1990). In emc embryos, these movements were abnormal. Splitting of the bands was delayed and was not neat since it occurred as an irregular formation of cells (Fig. 7F). The final distribution of the visceral mesoderm cells on the midgut was also irregular (not shown).

The visceral mesoderm seems mechanically to impose on the endoderm the three constrictions that divide the midgut (Fig. 7G) by acting as a drawstring closing a bag (Reuter and Scott, 1990). Accordingly, when visceral mesoderm is altered or missing, as in tinman and bagpipe mutants, these constric-tions do not form (Azpiazu and Frasch, 1993; Bodmer, 1993). In emc mutants they were also absent (Fig. 7H). Moreover, the midgut epithelium, of endodermal origin, was of irregular thickness in different regions (Fig. 7I,J). This anomaly may be due to emc insufficiency in the anterior and posterior midgut primordia and/or to the defective differentiation of the visceral mesoderm, which participates in the arrangement of the endo-dermal cells as an ordered epithelium (Reuter et al., 1993; Tepass and Hartenstein, 1994).

We have shown that emc is expressed from the preblastoderm up to the end of embryonic development in complex patterns that are tightly controlled in time and space and comprise derivatives of the three embryonic layers. In spite of its early expression, embryos carrying strong hypomorphic combina-tions of emc (i.e. emcip15/emcE12) or the deletion of the locus and adjacent chromosomal regions (emcE12/emcE12) do not show obvious morphological defects until shortly before germ band retraction. This may be due to the presence of emc products of maternal origin in the early embryo. Indeed, emc is expressed in ovaries (Ellis et al., 1990) and, more specifi-cally, in both the nurse and follicle cells (our observations). Moreover, a maternal contribution of emc in sex determination has been demonstrated (Younger-Shepherd et al., 1992) and at least some phenotypes of emc mutants are enhanced when the mothers are strong emc mutants (M. Ruiz-Gómez, unpublished data). Strong emc alleles including those used in this work are cell lethal in the adult epidermis, even in Minute+ clones (García-Alonso and García-Bellido, 1988; J. F. de Celis, personal communication). Consequently, we have not attempted to make germ-line clones of these alleles because lethal mutations in the epidermis are also lethal in germ-line clones (García-Bellido and Robbins, 1983). With the caveat that embryos maternally deprived of emc products might present even stronger phenotypes, the observed phenotypes in most or all of the embryonic regions where emc is expressed do define multiple and novel requirements for emc during embryonic development. They also suggests that no other gene has functions largely redundant with those of emc.

In many cases, strong emc expression precedes and accompa-nies formation of invaginations and grooves like the cephalic, ventral and transversal furrows, the stomodeal, amnioproc-todeal and anterior midgut invaginations, the Malpighian tubules primordia, the intersegmental grooves and the tracheal pits. emc expression also occurs at sites where migrations and rearrangements of cells are taking place, like the visceral mesoderm cells while these are spreading over the midgut and the cells of the Malpighian tubules while they rearrange their positions and elongate the tubules. Morphogenetic movements imply changes in proliferation, selective cell adhesion, and/or cytoskeleton-mediated changes of cell shape. The phenotypes of emc loss-of-function mutations indicate that emc is func-tionally significant for at least some of these processes, specially those occurring late, after the presumed maternal inheritance may no longer be relevant. Thus, the formation of intersegmental furrows is not properly accomplished in emc mutants. Moreover, the epidermal levels of Fasciclin III, a membrane protein probably involved in cell adhesion (Snow et al., 1989), are lower than in the wild type. Later, it becomes clear that the cells at the segmental border are defectively specified, as indicated by the reduced number of engrailed-expressing cells and the preferential absence of the ventral denticles that flank this border. Defective segmental borders are also suggested in emc embryos by the occasional failure of muscles and PNS axons to recognize them as barriers (Volk and VijayRaghavan, 1994), with the resulting crossing over of muscles (not shown) and nerves (Fig. 4H). At least some of the adult epidermal folds also require certain levels of emc function, as shown by the failure of viable hypomorphic emc clones to form the notopleural and scutelar sutures (J. F. de Celis, personal communication).

A most clear defect in cellular migration is observed in the visceral mesoderm of emc mutants. Initially, the cells in the mesodermal bands are not as tightly packed as in the wild type and have abnormal shapes, suggesting deficient cell adhesion. (Fasciclin III is expressed at reduced level in these mutants.) Later, the migration of the cells over the midgut is delayed and abnormal, and, at the end of development, they form an anomalous layer which seems incapable of driving the formation of the midgut constrictions. In the case of the Malpighian tubules, the insufficiency of emc strongly impairs cell rearrangements leading to the elongation of the tubules and, in the most extreme genetic combination, cell prolifera-tion and/or allocation is also affected. As a consequence, the tubules remain short and broad. Interestingly, the accumulation of Fasciclin II, another cell adhesion protein (Greningloh et al., 1990), decays prematurely in emcip15/emcE12 mutant tubules but not in other places like the developing central nervous system or the endodermal layer. In tracheal development, the cell divisions accompanying formation of the tracheal pits do not seem to be affected. However, the formation of the tracheal tree, which does not involve cell proliferation but cell migration and changes of cell shape (Campos-Ortega and Hartenstein, 1985), is often not fully accomplished (not shown). Similar tracheal phenotypes are found in mutants for Notch (Hartenstein et al., 1992) and for breathless (Glazer and Shilo, 1991; Klämbt et al., 1992), both genes coding for trans-membrane proteins presumably involved in cell communica-tion. Moreover, Breathless protein levels are reduced in emc mutants (our unpublished results).

emc is also expressed in regions of specialized cell-cell contact and/or cell recognition. Examples are the muscle attachment sites and the CNS during axonal pathfinding and fasciculation. At some of these sites, the lack of emc function causes alterations consistent with interference with adhesive-ness or cell recognition. An example of insufficient adhesive-ness is the l(1)myospheroid-like phenotype (Newman and Wright, 1981) found in the somatic muscles, which detach from the apodema. The defect probably resides in the epidermal part of the muscle attachment sites, since emc is expressed in them, but not in the muscles themselves. A striking example of inadequate cell recognition is the axonal misroutings observed in the PNS. Misrouting may be caused by abnormal specification of the sensory neurons or by alter-ations in the cues that the growth cones recognize during pathfinding such as the tracheae, (Giniger et al., 1993) or the myoblasts (Johansen et al., 1989). We would favor the second alternative, since emc is not obviously expressed in the devel-oping cells of the embryonic PNS. However, the occasional apparent change of neuronal fate found in emc mutants, i.e. the substitution of a chordotonal organ for an external sense organ (Fig. 4J), suggests that emc can also alter neuronal specifica-tion. The defects found in the developing axons of the CNS may also be due to abnormal axonal pathfinding. It should be stressed that, in imaginal development, emc has been shown to interact genetically with genes coding for membrane proteins like Notch, Delta and Fasciclin II (de Celis et al., 1991; L. García-Alonso, personal communication).

emc insufficiency causes a loss in the expression of most of the protein markers that we have examined. These markers are characteristic of differentiated cells and their loss probably reflects the faulty specification of the corresponding cells. This in turn would result in the observed alterations in processes as diverse as cellular adhesion and recognition, cell proliferation and cell commitment. How the Emc protein may act in the specification of the diverse cell types affected in emc mutants? In the commitment of sensory organ mother cells (Campuzano and Modolell, 1992) and possibly in sex determination (Younger-Shepherd et al., 1992), Emc antagonizes the function of a group of bHLH transcriptional regulators, the proneural proteins, most likely by sequestering them as heterodimers incapable of binding to DNA (Ellis et al., 1990; Garrell and Modolell, 1990; Martínez et al., 1993; Van Doren et al., 1991). Moreover, the Id group of mouse Emc homologs inhibits differentiation of fibroblasts by a similar mechanism (Benezra et al., 1990; Jen et al., 1992; Sun et al., 1991). Although, in addition to the HLH domain, the Emc protein bears other putative regulatory regions, such as a highly acidic domain and a glutamine-rich stretch (Garrell and Modolell, 1990), at least the second one is largely dispensable since individuals carrying only the mutant Emc Achaetous protein, which lacks this region, are viable. Consequently, Emc function may be exerted mainly through its HLH domain.

bHLH proteins control transcription of downstream genes by directly activating (or repressing) them (Jan and Jan, 1993; Olson and Klein, 1994) or, more indirectly, by stabilizing the binding to DNA of non-bHLH factors (rel-related proteins) which, in turn, directly control transcription (Jiang and Levine, 1993). In any case, Emc could act in developmental processes other than neurogenesis and sex determination by binding to bHLH proteins involved in those processes and decreasing their effective concentration, thus modulating the transcrip-tional activity of their target genes. The insufficiency of Emc would increase the effective concentration of these proteins and be equivalent to the simultaneous or sequential excess of function of the bHLH-coding genes. Thus, emc phenotypes would depend on the combinations of Emc-sensitive bHLH factors present in a given time and site. For instance, the presence of extra tip cells in the Malpighian tubules can be directly correlated with derepression of achaete function. Alternatively, the anomalies in the mesodermal derivatives may be difficult to interpret if they result from the unbalance of several bHLH factors like Twist, Nautilus, Lethal of scute and Hairy, all known to be present in the mesoderm (Hooper et al., 1989; Michelson et al., 1990; Thisse et al., 1988; F. Jiménez and A. Carmena, personal communication). The pos-tulated ability of Emc to interact with bHLH factors involved in a variety of regulatory pathways may account for the diversity of developmental processes and tissues affected by the emc insufficiency. Moreover, in mammalian cell cultures bHLH proteins can interact with b-leucine-zipper proteins (Bengal et al., 1992). This suggests that Emc may participate in networks of transcriptional regulators other than bHLH proteins and make its regulatory function even more complex.

We are most grateful to A. Martínez-Arias, M. Bate, H. Skaer, G. Morata, F. Jiménez, L. García-Alonso and M. D. Ferrés-Marcó for expert advice and comments on the manuscript; to L. García-Alonso, J. F. de Celis, F. Jiménez and H. Skaer for communicating unpub-lished results; to all the members of Juan Modolell’s laboratory and to M. Skalski for their help; to M. D. Ferrés-Marcó for the Fasciclin II stainings of the nervous system; and to C. Goodman, Y. N. Jan, A. Ferrús, S. Carroll, B.-Z. Shilo, P. Macdonald and D.P. Kiehart for antibodies, W. Gelbart for a dpp cDNA clone and P. Simpson for the emc^ip15 and emc^iX10 stocks. P. C. was a Rich Foundation postdoctoral fellow. M. R.-G. is a Ministerio de Educación y Ciencia postdoctoral fellow. This work was supported by grants from DGICYT (PB90-0085) and Comunidad Autónoma of Madrid to J. Modolell and an institutional grant from Fundación Ramón Areces.