Pseudohyphal growth in both haploid and diploid strains of Saccharomyces cerevisiae reflects concerted changes in different cellular processes: budding pattern, cell elongation and cell adhesion. These changes are triggered by environmental signals and are controlled by several pathways which act in parallel. Nitrogen deprivation, and possibly other stresses, activate a MAP kinase cascade which has the transcription factor Ste12 as its final target. A cAMP-dependent pathway, in which the protein kinase Tpk2 plays a specific role, is also required for the morphogenetic switch. Both pathways contribute to modulate the expression of the MUC1/FLO11 gene which encodes a cell-surface flocculin required for pseudohyphal and invasive growth. The MAP kinase cascade could also control the activity of the cyclin/Cdc28 complexes which affect both the budding pattern of yeast and cell elongation. A further protein which stimulates filamentous growth in S. cerevisiae is Phd1; although its mode of action is unknown, it may be regulated by a cAMP-dependent protein kinase, as occurs with the homologous protein Efg1 from Candida albicans, which is required for the formation of true hyphae. Morphogenesis in different yeast genera share common elements, but there are also important differences. Although a complete picture cannot yet be drawn, partial models may be proposed for the interaction of the regulatory pathways, both in the case of S. cerevisiae and in that of C. albicans.
The well known baker's or brewer's yeast, Saccharomyces cerevisiae, grows usually as single budding cells; Pasteur referred to them as ‘globules’. Although scattered observations on the ability of S. cerevisiae to grow in a filamentous form were made long ago , the fact that most laboratory strains grow exclusively in the unicellular form  meant that dimorphism in this yeast did not receive significant attention until the last decade. In 1992 Gimeno et al.  reported that, in a particular diploid strain of S. cerevisiae, nitrogen starvation caused formation of pseudohyphae and led to filamentous growth. During pseudohyphal growth (PH growth) the cells become elongated, budding occurs synchronously in unipolar fashion and the buds do not separate, producing the chains of cells which are called pseudohyphae (Fig. 1). Some S. cerevisiae haploid strains may also show a change in morphology, for instance, after extended growth on YPD plates . This has been called invasive growth because the filaments penetrate within the agar, below the colony. Although I will use in this review the now well established terms of PH growth for diploids and invasive growth for haploids, this does not mean that diploids cannot invade or that haploids are unable to form pseudohyphae. Both diploids and haploids form pseudohyphae, although in diploids these filaments extend on the surface of the agar away from the colony, while for haploids on YPD the filaments do not extend significantly. In addition, diploids invade agar even more strongly than haploids: their filaments resist vigorous washing of cells from agar surfaces, whereas haploids do not support such a harsh treatment but are not removed by a gentle washing [5–7].
The ability to form pseudohyphae upon nutrient deprivation has been lost in most laboratory strains which were selected for their tendency to form cells which separate easily . In the wild this ability could provide a selective advantage, as it would facilitate foraging for scarce nutrients. In addition, the capacity for developing pseudohyphae in response to stress  may allow escape from a harmful environment.
Although dimorphism is less prominent in S. cerevisiae than in other yeasts, the wide range of tools available for the study of S. cerevisiae has made this yeast the organism of choice to study the factors involved in the switch from unicellular to PH growth. Information on these factors is rapidly increasing and many important features of the process have just been uncovered. Therefore recent reviews on the subject [2,9,10] are already becoming outdated. In this review I have tried to present, in a systematic way, what is already established about this process and what are the obscure points which remain to be investigated.
2Two signaling pathways which act coordinately
The morphogenetic switch in S. cerevisiae requires the cooperation of two different signaling pathways, a MAP kinase cascade and a cAMP-dependent pathway, whose elements are described in detail below. A central element in the switch appears to be the GTP-binding protein Ras2. When this protein is constitutively activated by a RAS2Val19 mutation, which causes decreased GTPase activity, it can trigger enhanced PH growth in diploids [3,11] and invasive growth in haploids . It should be noted, however, that while in a ras2 haploid invasive growth is prevented , in a ras2/ras2 diploid starved for nitrogen PH growth still occurs, although the pseudohyphae are composed of cells that are round instead of being elongated .
2.1The MAP kinase cascade
The MAP kinase pathway involved in PH and invasive growth includes a large number of elements. The available evidence supports the scheme shown in Fig. 2, which operates as follows. Upon receipt of an adequate signal (see below) Ras2 is activated and activates the guanine nucleotide exchange factor Cdc24. Activated Cdc24, in turn, activates Cdc42 by facilitating the formation of its GTP-bound form . Activated Cdc42 interacts with the protein kinase Ste20 [14,15], and displaces the negative regulator Hsl7 . In this way Ste20 can activate the MAP kinase cascade formed by Ste11, Ste7 and Kss1 [17–19]. An important function of the interaction of Cdc42 with Ste20 appears to be the targeting of Ste20 towards the site of growth, a process required for a pseudohyphal growth pattern [14,15]. The MAP kinase Kss1 in its unphosphorylated form is able to interact with the transcription factor Ste12 and with the negative regulators Dig1(Rst1) and Dig2(Rst2), thereby potentiating the Dig-mediated block of Ste12 activity [20,21]. When Kss1 is phosphorylated by Ste7, both on a threonine and on a tyrosine residue, it is able to phosphorylate Ste12 and Dig1/2; the Dig proteins dissociate and this allows derepression of the target genes to which Ste12 may bind . Elements called FRE (for Filamentation and invasive Response Element) have been identified in the promoters of several genes . These are composite elements with two adjacent binding sites for the transcription factors Ste12 and Tec1 which bind cooperatively to the consensus sequences TGAAACA and CATTCT/C respectively. Dig1 and Dig2 interact with Tec1 in a two-hybrid assay  but there is no evidence that Kss1 phosphorylates Tec1. On the other hand, there is a FRE element in the TEC1 promoter itself, which is required for PH growth . The MAP kinase cascade appears therefore to regulate both TEC1 expression through the FRE element and Ste12 activity through its phosphorylation.
Further proteins which participate in the modulation of this MAP kinase cascade are Bmh1, Bmh2 and Ste50. The proteins Bmh1 and Bmh2 appear to play a redundant role, they have been found to associate with Ste20 and are critical for the expression of a reporter gene under the control of a FRE element  These proteins are not involved in invasive growth but are required for cell elongation. Since bmh1 mutants have been isolated which cause a reduced expression of a FRE-lacZ reporter gene but allow normal elongation, Bmh proteins may play a role in elongation, independent of the MAP kinase cascade . Deletion of STE50 leads to defects in PH growth and since Ste50 associates with Ste11 it has been suggested that Ste50 is necessary for Ste11 activity .
The GTPase activating proteins Bem3 and Rga1 appear to stimulate the hydrolysis of GTP on Cdc42 [25,26] and a rga1bem3 double mutant has an aberrant, elongated cell morphology . It is not known, however, whether pseudohyphal or invasive growth are stimulated in this double mutant. There is also no information on possible mechanisms of control for the proteins Bem3 and Rga1 which may operate when morphogenesis is induced.
Some of the elements of this MAP kinase pathway are shared with other regulatory pathways: pheromones activate the Ste20/Ste11/Ste7 cascade and a high osmolarity causes activation of Ste11 (reviewed in ). It is, therefore, important to prevent inappropriate cross-talk between the pathways. This is achieved through the specificity of the MAP kinases themselves, Kss1 involved in PH growth and invasiveness, Hog1 responding to osmotic stress, and Fus3 being responsible for the pheromone response. In the absence of the Fus3 protein, however, Kss1 can be recruited in response to pheromones and then activate both the mating response and invasive growth . Similarly, in a pbs2-3 strain, defective in the HOG pathway, Kss1 is phosphorylated in response to hyperosmotic shock and this leads to the production of long projections by the yeast . It can be noted that Hog1 has some inhibitory effect on the STE signaling pathway and that hog1 strains are consistently more filamentous than wild-type strains .
2.2The cAMP-dependent pathway
Activated Ras2 can interact with Cyr1, adenylate cyclase ; this results in an increase in the intracellular concentration of cAMP which in turn activates the cAMP-dependent protein kinases (Fig. 3). Although any of the three cAMP-dependent protein kinases from S. cerevisiae is sufficient for cell viability, they play different roles in PH differentiation [29,30]. Deletion of TPK2 blocks PH growth, while deletion of TPK3 stimulates it. For TPK1 there is a lack of agreement between different groups, while Robertson and Fink  did not observe any effect when the gene was interrupted, Pan and Heitman  reported that Tpk1 had a repressing role on PH growth. In two-hybrid assays only Tpk2 was shown to interact with Sfl1 and with Mga1 . The proteins Sfl1 and Mga1 contain a type of helix-turn-helix DNA binding-motif first described in the heat-shock transcription factor Hsf1  and consensus Tpk phosphorylation sites (5 in Sfl1 and 2 in Mga1). Sfl1 acts as a repressor for the SUC2 gene  and its deletion enhances PH growth . The deletion of MGA1 did not interfere with the formation of elongated cells during nitrogen starvation but caused a large increase in the proportion of cells budding randomly ; the defect in pseudohyphal growth was reported as strong  or ‘not dramatic’.
Two other proteins with the same helix-turn-helix motif, Skn7 and Hsm2, could also be involved in PH growth since a deletion of either of the corresponding genes causes a moderate defect in this type of growth . This points to Sfl1 and other proteins of the family as possible substrates for Tpk2, which would play opposing roles: Sfl1 would act as a repressor and Mga1, Skn7 and Hsm2 as (redundant?) activators. The downstream target(s) of these proteins in the regulation of pseudohyphal growth have not been yet identified but the gene MUC1/FLO11 (see below) could be a good candidate, since MUC1/FLO11 mRNA levels are increased in a sfl1 mutant and sfl1muc1/flo11 homozygous diploids are defective in filamentation. The fact that filamentation is not completely abolished in the double mutant, as it is in a muc1/flo11 mutant, suggests that Sfl1 is able to repress other genes involved in pseudohyphal development . In a bcy1 mutant where the Tpks are constitutively activated, there is an increase in MUC1/FLO11 expression which is strictly dependent on FLO8, a gene involved in flocculation  and required for PH growth  (see below); this could indicate that FLO8 is also a target of the cAMP-dependent protein kinases . The fact that a double mutant sfl1tpk2 shows no defect in filamentation , suggests that the activation of FLO8 is not strictly dependent on Tpk2 and can be performed also by Tpk1 and/or Tpk3.
A further role of the Ras2/cAMP pathway in stimulating invasive growth may be the suppression of the activity of the stress-responsive transcription factors Msn2 and Msn4. It has been shown that invasive growth does not take place in an haploid ras2Δ mutant, but occurs in ras2Δmsn2Δmsn4Δ or ras2Δyap1Δ mutants . This suggests that the product of some gene(s) activated by the transcription factors Msn2/4 and yAP-1 has a negative effect on invasiveness.
The importance of the cAMP-dependent pathway for morphogenesis is highlighted by the fact that external cAMP stimulates PH growth [13,36] while this form of growth is blocked by overexpression of the phosphodiesterase Pde2 .
2.3Interaction between the pathways
The two pathways activated by Ras2 converge to regulate the expression of the same gene, MUC1/FLO11, which encodes a cell-surface flocculin required for PH growth of diploid cells and invasive growth of haploids [38,39]. The MUC1/FLO11 promoter is much larger than most yeast promoters, cis-acting elements being present at least at 2.8 kb from the first coding triplet . It is also extremely complex, as it appears to contain at least four upstream activating sequences and nine upstream repressing sequences. MUC1/FLO11 expression requires Tec1, Ste12 and Flo8, but the effect of ste12 and flo8 mutations can be suppressed by overexpression of FLO8 and STE12, respectively . However, Flo8 is specifically required for the cAMP-mediated increase in MUC1/FLO11 transcription . While it has been found that the upstream region of the MUC1/FLO11 promoter contains a FRE element with a Ste12-binding site and a Tec1-binding site in close proximity , it is not yet established whether the nuclear protein Flo8 binds to the MUC1/FLO11 promoter or plays a more indirect role. It has also been suggested that Ste12 may act at some other site of the MUC1/FLO11 promoter, independently of Tec1, interacting perhaps with a different activator .
Although Flo8 may also facilitate an increased transcription of FLO1, a gene encoding a cell wall protein involved in flocculation , neither Flo1 nor the related protein Flo5 appear required for PH growth .
Further elements involved in the transcriptional activation of MUC1/FLO11 have been characterized recently. Msn1, a nuclear protein with capacity for transcriptional activation , and Mss11 are required for invasive growth; they may act upstream of MUC1/FLO11 since their absence may be partially compensated by expression of MUC1/FLO11 under the control of a strong constitutive promoter . On the other hand, MUC1/FLO11 cannot be the only target for Mss1, since overexpression of MSS11, but not of MSN1, allows weak invasive growth in a muc1/flo11 strain. Whereas Mss11 appears to act downstream of Ste12 and of Msn1, Msn1 and Ste12 would act independently in parallel pathways, which are both Ras2-dependent . There is also evidence that Mss11 acts downstream of Flo8 to trigger invasive growth and that Flo8 and Msn1 act independently of each other . A tentative model is shown in Fig. 4.
Another mechanism which would allow the two Ras2-dependent pathways to interact is the operation of the Tpks on the Ste12 and/or Tec1 transcription factors. This is suggested by the observation that in ras2 haploid cells, overexpression of TPK1, TPK2 or TPK3 stimulates the expression of a FRE-lacZ reporter gene .
The observation that Σ1278b strains are more competent than most laboratory strains for filamentous growth has been related with the fact that the usual laboratory strains have a defect in the FLO8 gene . It has also been reported that the Ras2/cAMP pathway is overactive in the Σ1278 background . It may be concluded therefore that the unique behavior of Σ1278b strains is due to a combination of several peculiarities of its genetic background.
A variety of conditions promote PH growth, among them nitrogen limitation , a poor carbon source in the medium [8,38], or some possible stress conditions [8,44].
How these situations trigger the operation of the MAP kinase cascade and of the cAMP-dependent protein kinase pathway is not well understood. There are, however, some elements involved in signaling which have already been identified and which will be discussed in this section (cf. Fig. 5).
Surprisingly, although Ras2 has been placed upstream of both the MAPK and the cAMP-dependent pathways, there is at present no information on the mechanisms by which Ras2 may be activated in conditions leading to PH and invasive growth.
In SLAD medium, a medium with only 50 μM ammonium , Sho1, a putative membrane-spanning protein , is required for PH growth of a diploid . It has been therefore proposed that Sho1 may respond not only to high osmolarity  but also to a ‘low nitrogen’ signal and relay it to Ste20 and the MAPK cascade which activates Ste12 and Tec1 .
PH growth requires also the G protein Gpa2 [13,36] and its coupled receptor Gpr1 [47–49]. It has been suggested that Gpr1 plays a role in nitrogen detection and, in fact, GPR1 mRNA was strongly induced in response to starvation for nitrogen but not for a carbon source . However, there is also recent evidence that Gpr1 can monitor glucose [49,51,52] and it has been proposed that both low nitrogen and a high level of glucose or a related sugar are required to trigger PH growth .
There is, in addition, a high affinity ammonium permease, Mep2, also needed for PH growth of a diploid in SLAD medium  which suggests that Mep2 could act as a nitrogen sensor. There is no evidence, however, for a direct interaction between Mep2 and Gpa2. It is noteworthy that another ammonium permease, Mep1, may inhibit pseudohyphal differentiation under certain conditions .
The elements Gpr1, Gpa2 or Mep2 appear to work through the cAMP pathway, since the defects in filamentation of a gpr1, a gpa2 or a mep2 mutant are suppressed by a RAS2Val19 mutation or by external cAMP [13,36,48,49,53]; the defect in filamentation in a gpr1 mutant is also suppressed by a bcy1 mutation .
The phosphatidylinositol-specific phospholipase C, Plc1, is required for the interaction between Gpr1 and Gpa2. Two-hybrid assays and coimmune precipitation of epitope-tagged proteins demonstrated that Plc1 can bind both Gpr1 and Gpa2, while the association between Gpr1 and Gpa2 required the presence of Pcl1 . This explains why is Plc1 necessary for PH growth upon nitrogen starvation . On the other hand the fact that the mutation RAS2Val19 suppresses the filamentation defects of gpr1 and gpa2 strains, but not those of a plc1 strain, suggests that Plc1 has a further function downstream of Ras2, perhaps controlling the MAP kinase pathway via Cdc42 . Although it had been proposed that Gpa2 could act in a cAMP-independent pathway that activates the protein kinase Sch9 , it was later shown that Sch9 is not absolutely required for PH growth and could even play a negative role on invasive growth and on MUC1/FLO11 expression in haploid cells .
A further element which may play a role in signaling the availability of nitrogen is the glutamine tRNACUG molecule. A mutation which affects this tRNA allows PH growth in a nitrogen-rich solid complex medium where wild-type cells grow exclusively in the yeast form . It has also been found that the MAP kinase pathway and the Ste12 transcription factor are dispensable for the PH growth of the tRNACUG mutants , an observation which suggests that tRNACUG operates through the cAMP-dependent signaling pathway. However, the putative regulatory protein interacting with tRNACUG has not been yet identified.
It has been reported recently that in a Σ1278b diploid strain a mild thermal stress or the presence of aliphatic alcohols (ethanol, propanol or butanol) induces invasive growth on YPglucose; the alcohols trigger also the formation of pseudohyphae  In pde2 diploid cells, which do not share the Σ1278b genetic background, alcohols or a high salt concentration induce PH growth, but only in the presence of cAMP in the medium . Since Sho1 activates Ste11 in response to osmotic stress , Sho1 is likely to act as the sensor for high salt which triggers morphogenesis. On the other hand, information is lacking on the target(s) for thermal stress and aliphatic alcohols, although they may act by modifying the structure of the plasma membrane.
3Control of specific morphological changes
The switch between growth in the yeast form and PH growth in S. cerevisiae involves important changes in the pattern of progression through the cell cycle. For cells in the yeast form, the bud grows first predominantly at its apex; later on growth becomes more uniform over its whole surface, and finally the daughter cell separates from the mother before reaching the size of the latter. As a consequence, the smaller new cell has to spend some time growing in G1 before reaching the critical size for progression to S phase at Start. In contrast, during PH growth there is an extended G2/M period and when the septum separating mother and daughter is formed the daughter and the mother cells are the same size, so that both cells can start a new cycle synchronously .
Cell cycle progression is mediated by the association of the protein kinase Cdc28 with distinct groups of cyclins, mainly Cln1, Cln2 or Cln3 during the G1 phase and Clb1 or Clb2 before mitosis . This suggests that Cdc28 and/or the Cln and Clb cyclins may be targets for the signal transduction pathways leading to filamentation and there is recent evidence for a lengthening of the G2/M growth period induced by an activation of the MAP kinase pathway which inhibits mitotic cyclin/Cdc28 complexes .
Filamentous growth involves changes in the budding pattern, in elongation and in invasiveness. Although these cellular processes are interconnected, they are not always controlled by the same elements . For instance, while both Ste12 and Tpk2 are required for agar invasion, Ste12 has little influence in the pattern of budding and Tpk2 does not affect cell elongation . Therefore the different processes leading to PH growth will be considered separately.
3.1Regulation of the budding pattern
Pseudofilamentation is absolutely dependent on a polar budding pattern, the normal pattern of growth in diploid cells. The mode of budding is controlled by a large number of genes which are listed in Table 1 and which will be discussed below.
Table 1. S. cerevisiae genes implicated in the control of the budding pattern and in elongation
Mutations in BUD1, BUD2 or BUD5 which cause a random distribution in the localization of new buds do not allow the formation of pseudohyphae in diploids [3,59] or haploids . Similarly a bud8/bud8 diploid mutant which buds with high frequency at the birth end of the cell is also defective in PH growth . Interestingly neither Bud1 nor Bud8 are required for cell elongation or agar invasion [4,5].
Other mutations have been described which facilitate filamentation by changing the budding pattern. A bud4 mutation, which causes polar budding in haploid cells allows a moderate increase in filamentation [60,61]. Similarly, overexpression of the WHI2 gene triggers polar budding in haploids, hinders the completion of cytokinesis, and causes hyperpolarized growth which results in the production of cells with an elongated shape; nevertheless, these cells do not invade the agar .
The mutation cdc28-127 in the cyclin-dependent kinase Cdc28 replaces also the axial budding pattern of an haploid strain by an unipolar or bipolar budding which causes PH growth, even during exponential growth in liquid cultures of YPD . It should be noted that cdc28-127 does not cause delayed cell separation, a further indication that the different processes contributing to pseudofilamentation may be regulated by different elements.
It has been reported that deletion of SEM1 allows pseudohyphae formation upon nitrogen starvation in a flo8 strain and it was suggested that Sem1 interferes with the shift from bipolar to unipolar distal budding which is required for PH growth . Although the mode of action of Sem1 is not known, this small acidic protein appears to play a negative regulatory role in exocyst function, and this would support the idea that there is a link between exocytosis and polarization of growth, involving the actin cytoskeleton . Unipolar bud selection in pseudohyphal cells is also dependent on actin function .
3.2Regulation of elongation
Mutations in a large number of genes affect cell elongation (see Table 1). It is important to note that in some cases the cells may become elongated due to a delay in mitosis and a defect in the switch from apical to isotropic growth. Nevertheless, it is still debated whether this is the mechanism underlying the change in cell shape observed in pseudohyphae. Among the genes related with elongation, those encoding cyclins, or proteins able to regulate cyclins or the cyclin-dependent kinase Cdc28 itself, are prominent. At least one of the G1 cyclins, Cln1 or Cln2, is required for PH growth and haploid invasive growth [67–69], while overexpression of the cyclins Clb1 or Clb2 blocks PH growth triggered by low nitrogen, even in the presence of the dominant active mutations RAS2Val19 or STE11-4, and suppresses the elongation phenotype of a cdc28-127 mutant . Since deletion of CLB2 produces an elongation phenotype, even in rich medium [63,67], it appears that the complexes Cdc28-Clb and Cdc28-Cln play opposite roles in the morphogenic process . The negative role of Clb2 would be dependent on the Cln3 cyclin, since both a cln3 and a clb2 mutation bypass the requirement of Cln1 for elongation . This indicates that the MAP kinase cascade activates at least two parallel pathways with some degree of redundancy. It is interesting to note that, although expression of CLN1 is under the control of Tec1 and Ste12 , overexpression of STE20 or the dominant mutation STE11-4 can partially suppress defects in PH growth of a cln1/cln1 cln2/cln2 mutant . It has been suggested that the complex Cdc28-Cln1/2 would promote vegetative morphogenic functions of Ste20 while inhibiting functions related with pheromone signaling . The observation that a grr1 mutation promotes PH growth  may in turn be related with the fact that Grr1 participates in the degradation of Cln1 and Cln2 . Grr1, however, should control at least another protein involved in elongation, since a grr1/grr1 cln1/cln1 cln2/cln2 mutant forms elongated cells even in a rich medium . The crucial role of Cdc28 in controlling S. cerevisiae morphogenesis is also shown by the fact that PH growth is facilitated by inhibition of Cdc28 by Swe1 (which phosphorylates Tyr-19) and impaired by Hsl1, a protein kinase which inhibits Swe1  or by Mih1, a protein phosphatase which acts as antagonist of Swe1 .
A variety of mutants which present an elongated morphology have been called elm[70,72]; other mutants such as bni1, spa2, tpm1, pea2/dfg9, dfg5 and dfg10 are defective in both cell polarity and cell elongation, but still invade the agar . ELM1 encodes a serin–threonine protein kinase [72,73] which appears located mainly at the bud neck . It may activate the protein kinase Hsl1  to allow cytokinesis and mitosis to proceed. The elm1 mutation would then cause a delay in cytokinesis which results in elongation of the cells and eventually PH growth. Another elm mutation has been characterized as cdc12 and its phenotype can be related to the fact that CDC12 encodes one of the septin proteins essential for cytokinesis. The role played by Cdc55, a regulatory subunit of yeast PPase 2A, is less clear. Whereas a diploid with a single functional copy of the CDC55 gene shows an elongated morphology , a diploid with the homozygous cdc55-100 mutation is impaired in cell elongation .
While Dfg5 and Dfg10 have not been characterized further, Spa2, Pea2, Bni1 and another protein, Bud6, appear to function together in a complex to promote polarized cell growth . These proteins, and also Tpm1, a major form of tropomyosin, could interact with actin and thereby direct the actin cytoskeleton to specified growth sites in the yeast cells. It should be noted that Spa2 is also able to interact with Ste11 and Ste7, two elements of the MAP kinase pathway which controls PH growth . Some mutations in actin interfere with cell elongation suggesting that the polarization of actin patches at the tip of the cell is important for elongation .
It has been reported that strains with a deletion of the ASH1 gene are impaired in PH growth . At the end of mitosis there is an asymmetric accumulation of the repressor Ash1 in daughter cells due to an asymmetric distribution of ASH1 mRNA which is dependent on proteins such as Act1, Tpm1 or Bni1 which regulate the actin cytoskeleton . It would be interesting to check whether the defect in pseudofilamentation observed in a ash1Δ mutant is actually due to the lack of the Ash1 protein or may be a secondary effect of the lack of ASH1 mRNA which could distort the structure of the actin cytoskeleton required for polarized growth. Although this mechanism may seem rather unusual, it has recently been reported that the association between the complex She3–Myo4 and the She2 protein is dependent on the presence of Ash1 mRNA .
3.3Regulation of agar invasion
In a general screen to isolate genes involved in filamentous growth, in which different classes of mutants were identified, only one mutant, dfg16, was obtained which was defective in invasion but not affected in cell polarity or filament formation . DFG16 encodes a protein predicted to contain six membrane-spanning domains, which could play a role in cell wall synthesis , but has not been further characterized. A cdc 55-100 mutant was impaired both in cell elongation and in invasiveness .
In addition, some actin alleles cause a marked defect in agar invasiveness; a possible reason for this would be the participation of actin in the secretion of putative hydrolases able to facilitate growth through the agar .
It should be noted that not every gene which could have been identified was revealed in the screening described in . The MUC1/FLO11 gene discussed in Section 2.2 was not found, for example, although it has been reported to be required for agar invasion and cell–cell adhesion but not for mediating cell elongation or the budding pattern switch [30,67].
4Other genes involved in PH growth and invasive growth
Overexpression of PHD1 induces vigorous PH growth of diploid cells on rich medium  and PHD1 expression is strongly induced in a medium with low nitrogen . Nevertheless, deletion of PHD1 does not interfere with the filamentation response to nitrogen starvation, and overexpression of PHD1 does not induce PH growth in wild-type haploid strains, although it does in a bud4 mutant strain, which has the polar budding pattern . Phd1 has been shown to be localized in the nucleus and to contain a region similar to one present in different fungal transcriptional regulators . This region includes a positively charged stretch followed by a helix-turn-helix motif potentially able to bind DNA; there is no information, however, on the possible target(s) of Phd1. In any case the enhancement of filamentous growth by Phd1 is independent of Tpk2 , Tec1 , Ste12 or Ash1 . Although this enhancement is prevented in a gpa2Δ mutant , it takes place in the diploid double mutants gpa2mep1 or mep1mep2. A protein with significant homology to Phd1 in its putative DNA-binding region, Sok2, has been found to play a negative role on PH growth since sok2/sok2 diploid strains form pseudohyphae at an accelerated rate . Here again the target(s) for Sok2 are not known but it should be noted that the positive effect of a SOK2 deletion on filamentation requires the presence of PHD1, although sok2phd1 diploids still exhibit PH growth when starved for nitrogen.
Pseudohyphae formation may require an enhanced production of chitin, since pseudohyphae have been shown to accumulate chitin on their whole surface and not only at the septa . This is consistent with the observation that Σ1278b-derived strains which form pseudohyphae easily  have a 30-fold lower chitinase activity and a level of chitin at least 70% higher than strains unable to undergo PH growth . It can be noted also that in yeasts with an impairment in PH growth caused by a non-functional FLO8 allele, disruption of ACE2, which encodes a positive regulator of the chitinase gene CTS1, results in the production of pseudohyphae. Since disruption of the CTS1 gene itself in a flo8 background also allows PH growth, but at a reduced level , Ace2 may control another gene involved in PH growth.
So many genes have been described as involved in the control of PH growth that it becomes difficult to distinguish between direct and indirect effects. For instance, the impairment of dom34 mutants in forming pseudohyphae appears related to a decrease in bulk protein translation, which, in turn, alters the cell cycle . The observation that shr3/shr3 diploids show enhanced filamentous growth can be connected with a role of Shr3 in the secretory pathway specifically required for the localization of amino acid permeases in the plasma membrane . In the absence of a functional Shr3, the yeast exhibits starvation responses, even when amino acids are present in the medium. MSN5 is one of the genes identified in different screenings, for which no role has been suggested. Overexpression of the gene restores filamentation in a mep1mep2 mutant, and its deletion in a wild-type strain causes a moderate defect in PH growth . Since Msn5 is involved in the nuclear export of phosphorylated proteins  it could facilitate the removal from the nucleus of some regulatory protein which represses PH growth.
A form of the RNA polymerase II holoenzyme includes a negative regulator formed by Srb10, a cyclin-dependent kinase, and several associated proteins. This Srb–CDK complex, which would function through phosphorylation of the RNA polymerase carboxy-terminal domain, appears to down-regulate a small fraction of S. cerevisiae genes, many of them derepressed during nutrient deprivation . It has been found that diploid cells lacking Srb10 exhibit increased PH growth and this could be related with the observation that in a srb10Δ mutant growing in YPD, MUC1/FLO11 is induced 15-fold and FLO1 up to 100-fold . Since the lack of Srb10 did not affect significantly the mRNA levels of either STE12, TEC1, FLO8 or SFL1, Srb10 could control directly the transcription of the FLO1 and MUC1/FLO11 genes. As cells enter the diauxic shift Srb10 levels decrease strongly and other control mechanisms should become operative.
Recently, targets for the MAP kinase Kss1 have been looked for by performing transcript profiling experiments [68,87]. Although dozens of genes regulated by the Kss1 signaling pathway have been identified, most of them are not required for invasive growth in haploid cells . This could indicate that many of these genes play a redundant role, but also that some of them participate in a process related to filamentous growth but not required for morphogenesis. A clear-cut example is the PGU1 gene, which encodes a secreted enzyme that hydrolyzes polygalacturonic acid, a component of the plant polysaccharide pectin. This enzyme is not required for invasiveness but would facilitate the access of S. cerevisiae to substrates such as rotting fruits . Profiling experiments have also confirmed that there is some functional overlap between the filamentous growth and mating responses .
5Differences in the behavior of haploids and diploids
PH growth of diploid cells and invasiveness of haploids present very similar requirements. Both depend on the MAPK cascade, starting with activation of Cdc42, including the Ste20, Ste11, Ste7 and Kss1 proteins and their downstream targets, the transcription factors Ste12 and Tec1 [4,11,12,17]. Flo8  and Muc1/Flo11  are also required for the morphogenetic change of both haploids and diploids. Finally the PKA pathway plays a role in haploids  as it does in diploids.
However, the regulation of these pathways is different in the two types of cells. Whereas, in an haploid strain, MUC1/FLO11 mRNA levels increased steadily during exponential growth in YPD and decreased to very low levels at later phases of growth, in diploid cells growing in YPD MUC1/FLO11 mRNA levels always remained low . In SLAD medium, on the other hand, while MUC1/FLO11 mRNA levels increased steadily with growth in diploids, in haploids there was some increase in the first hour of growth and then a subsequent decrease . Expression of a reporter gene depending on the transcription factors Ste12 and Tec1 was 10-fold higher in haploids than in diploids in a complete mineral medium (SC). The rate of transcription of the reporter gene was increased in SLAD medium, which has a low level of ammonium, the effect being more marked in the case of diploid cells (nine-fold vs. three-fold for haploids) .
It has been stated  that while in haploids an increase in protein kinase A activity stimulated the expression of a reporter gene controlled by Ste12 and Tec1, in diploids a very high protein kinase A activity inhibited the transcription of this reporter gene. In an haploid strain in SLAD medium external cAMP repressed the expression of the reporter gene up to 50-fold, while in SC medium cAMP had no effect . In the same SC medium the overexpression of TPK1, TPK2 or TPK3 increased the expression of the reporter gene in a ras2 haploid three to five-fold . For diploid strains growing on SLAD, overexpression of TPK1 increased the expression of the reporter gene about two-fold . Looking at the effects of an activation of the PKA pathway on the expression of the reporter gene controlled by Ste12 and Tec1, no clear difference emerges between haploids and diploids. On the other hand, the effects are strongly dependent on the metabolic conditions of the cell.
As discussed in Section 2.1, regulation of cyclin activity plays a role in the control of morphogenesis. There are, however, some differences in the behavior of haploid and diploid cells. Whereas single knockouts of CLN1 or CLN2 do not block invasive growth in haploids , a cln1/cln1 diploid strain, but not a cln2/cln2 strain, is unable to induce pseudohyphae upon stimulation of the MAP kinase pathway [67,68].
The most striking difference between the two types of cells appeared to be that haploids are able to invade agar in the presence of a good nitrogen source, while diploids are not . However, in a diploid growing on a rich medium (YP), filamentation and invasiveness may be induced by cAMP together with a stressful condition such as a poor carbon source or the presence of aliphatic alcohols .
Although TEC1 transcription is higher in haploid than in diploid cells, there is little correlation between TEC1 transcription and haploid invasive growth, an observation which suggests that a low level of TEC1 transcription is sufficient for invasive growth . It has been shown that invasive growth in haploids does not take place in rim mutants, while in homozygous diploid rim mutants there is no defect in PH growth . The exact role played by the Rim proteins is not yet known, but from the characteristics of these proteins (yeast protein databases) several points emerge. Rim101 (earlier called Rim1) is a Zn finger protein with similarity to PacC from Aspergillus. It is activated by limited proteolysis which, in turn, requires the genes RIM8, RIM9 and RIM13. RIM8 has not been cloned, Rim9 may acetylate Rim13 and Rim13, similar to PalB, a calpain-like protein from Aspergillus nidulans, may be the protease which cleaves Rim101.
6Regulation of dimorphism in other fungi
Although dimorphism has been observed in a large number of fungi, most studies on this process have been carried out in two main species, S. cerevisiae and the pathogenic fungus Candida albicans (see  for a recent review). C. albicans can grow as a budding yeast or as pseudohyphae, and may also develop germ tubes and true hyphae, in which growth takes place by apical extension and division occurs without leaving constrictions at the septa. Pseudohyphae formation is induced by growth on a variety of solid media (Spider, milk-Tween, etc.) while true hyphae are formed in liquid media containing serum or in Lee's medium at pH 6.7.
The role of the MAP kinase pathway in the transition from yeast to pseudohypha in C. albicans has been well established. Functional homologues of the S. cerevisiae STE20, STE7 and KSS1 genes have been identified; these are CST20[91,92]HST7[91–93] and CEK1 respectively, and interruption of these genes causes lack of pseudofilamentation in solid media but no defect in the formation of hyphae in response to serum. In contrast another Ste20-like protein, Cla4, appears to be absolutely required for hyphal growth; homozygous cla4 mutants show an aberrant morphology which may indicate a role for Cla4 in cytokinesis . In this respect it can be noted that there is evidence that Cla4 is important for the formation of the septin ring , although a requirement for Cla4 for pseudohyphal growth in S. cerevisiae has not been directly tested. The downstream target of the MAP kinase pathway in C. albicans is Cph1, the homologue of S. cerevisiae Ste12, and homozygous cph1 mutants show also a defect in pseudohyphae formation on a solid medium, such as Spider medium, but are able to form hyphae in liquid Lee's medium and to induce germ tubes in response to serum . Although an homologue of the S. cerevisiae TEC1 gene has been identified by DNA sequencing, its role on morphogenesis has not been yet studied. On the other hand, while PH growth is constitutive in C. albicans cpp1 mutants, lacking a protein phosphatase which acts on phosphorylated Cek1 , there is no information on the effects on PH growth of a MSG5 deletion in S. cerevisiae, although recent evidence suggests that the protein phosphatase Msg5 acts on Kss1 .
Targets for Cph1 have not been identified but INT1, a gene encoding a surface protein with a cytoplasmic tail, may be a candidate. Disruption of INT1 suppressed hyphal growth in Spider or milk-Tween solid media but not on Lee's medium or in the presence of serum . Besides, overexpression of INT1 led to the production of germ tubes in a haploid S. cerevisiae strain and even in the corresponding ste12 mutant .
Morphogenesis in C. albicans has been shown to respond also to Efg1, a homologue of S. cerevisiae Phd1. Reduced expression or deletion of EFG1 caused an inability to form germ tubes and true hyphae in the presence of serum, but elongated pseudohyphal cells could be observed in certain solid media [101,102]. While overexpression of EFG1-enhanced pseudohyphal growth , a double mutant cph1/cph1 efg1/efg1 was completely defective in filamentous growth . It has been postulated that Efg1 lies in a Ras-activated, cAMP-dependent pathway , and recent results indicate that mutation of the single possible PKA phosphorylation site in Efg1 to alanine does not allow filamentation, whereas a change to glutamate results in hyperfilamentation (D. Bockmühl and J.F. Ernst, personal communication). Evidence is also accumulating that a cAMP-dependent protein kinase is involved in the dimorphic transition in C. albicans. Dibutyryl-cAMP enhanced germ-tube formation in exponentially growing cells , inhibitors of protein kinase A blocked the germ-tube formation induced by N-acetyl glucosamine, although not that induced by serum , and Catpk2 mutants are impaired in hyphal development unless EFG1 is overexpressed . One of the targets of the cAMP-activated protein kinase may be the glucosamine-6-phosphate synthase required to provide the substrate for chitin synthesis, which is more abundant in mycelium than in yeast cells .
Very recently the RAS1 gene from C. albicans has been isolated. No other close homologue of RAS seems to be present in this species and it has been shown that a ras1Δ/ras1Δ homozygous strain could no longer form germ tubes and hyphae, although it still formed some pseudohyphae when exposed to serum . A C. albicans chaperonin, Cct8, appears to play a role in morphogenesis. Filamentation is defective both in a cct8/CCT8 strain and in a strain overexpressing CCT8. In S. cerevisiae, Cct8 from C. albicans can block pseudohyphal growth induced by RAS2Val19, an observation which suggests that an excess of chaperonin subunits interferes with the Ras signal transduction pathway .
No homologue for Flo8, the target for the cAMP pathway in S. cerevisiae, has been reported in C. albicans and this is also the case for the MUC1/FLO11 gene, which plays a central role in S. cerevisiae.
Hyphal development in C. albicans is under both positive and negative control. Deletion of the TUP1 or RBF1 genes stimulates filamentous growth. Homozygous tup1 mutants display filamentation in a great variety of media but no germ tube formation progressing to true hyphae in the presence of serum. A double mutant tup1cph1 behaves as a tup1 mutant, and it has been suggested that repression by Tup1 is relieved via Cph1 . However, as discussed below, Tup1 and Cph1 are probably acting in different pathways. It can be noted that in S. cerevisiae the lack of Tup1 has a different effect: it reduces PH growth in diploids and invasiveness in haploids  and it induces flocculation .
A further gene involved in C. albicans morphogenesis is HWP1, a hypha-specific gene, which encodes a cell surface protein involved in host attachment but of unknown function in hyphal development. In hwp1 mutants the ability to form hyphae on solid medium is lost and in the presence of serum there are reduced levels of peripheral hyphae . The ability to invade the agar beneath the colony is maintained as well as the capacity to form germ tubes in liquid media. Expression of HWP1 is dependent on EFG1 but is not dependent upon CPH1. HWP1 is expressed at a reduced level in a strain with a deletion in RBF1 and expression can take place in non-inducing conditions in a tup1 null mutant . Since the expression of HWP1 is repressed by Tup1 but does not require Cph1, the suggestion that Cph1 acts by counteracting the action of Tup1 remains unproven. Although HWP1 can be situated downstream of EFG1, the inability of constitutively expressed HWP1 to suppress an efg1 null mutation indicates that Efg1 regulates additional genes required for hyphal development . Recent evidence also suggests that Cph1, Efg1 and Tup1 control different pathways which make additive contributions to filamentous growth and that a fourth pathway may still operate when the others are blocked . A tentative model integrating the data available is shown in Fig. 6. It should also be noted that, as reported for S. cerevisiae, G1 cyclins are involved in morphogenesis. In a Cacln1/Cacln1 mutant in liquid Lee's medium the HWP1 gene is not induced and hyphal growth does not take place . In a serum-containing medium, on the other hand, the cln1 mutation has only a slight effect on the induction of HWP1 and does not interfere with germ tube formation or hyphal growth .
Finally, the interruption in C. albicans of HRM101, the homologue of S. cerevisiae RIM101, blocks filamentation on Spider plates . This is in contrast with the situation in S. cerevisiae, where PH growth is not impaired in homozygous diploids, although rim101 haploids are defective in invasive growth .
As happens with S. cerevisiae, other genes which have been implicated in morphogenesis may affect differentiation in an indirect way. For instance, disruption of the C. albicans TPS1 gene encoding trehalose-6-phosphate synthase impaired the formation of hyphae during incubation with serum at 37°C, in the presence of glucose, or during growth on galactose or glycerol at 42°C . In the first case the impairment could be due to alterations in the metabolism of glucose, whereas in the second it may reflect a need for trehalose for the maintenance of the hyphae at a high temperature.
In Yarrowia lipolytica the protein Mhy1, which is able to bind to a STRE sequence, is required for the yeast-to-hypha transition . This is in contrast with the situation in S. cerevisiae, where the absence of either or both the transcription factors Msn2 and Msn4, which bind to the STRE sequences has no effect in pseudohyphae formation , and can even facilitate the morphogenetic switch in some genetic backgrounds . During the yeast-to-hypha transition the transcription of MHY1 is strongly increased and Mhy1 is concentrated in the nuclei of actively growing cells found at the hyphal tip. Mhy1 transcription is unaffected by thermal stress and decreases upon carbon source starvation and osmotic or oxidative shock .
It is interesting to note that while C. albicans hyphal development is activated by a variety of stresses such as nitrogen starvation or an increased temperature or pH, nitrogen starvation inhibits hyphae formation in liquid cultures of Y. lipolytica. A further peculiarity of Y. lipolytica is that the YlSTE7 gene, an homologue of the STE7 genes from C. albicans and S. cerevisiae, is not required for hyphal development (A. Domínguez, personal communication). On the other hand, deletion of YlTUP1 stimulates hyphal growth as it does in C. albicans (A. Domínguez, personal communication).
XPR2 which encodes an alkaline extracellular proteinase from Y. lipolytica appears to be involved in regulation of dimorphism ; the expression of XPR2, in turn, is controlled by Rim101, homologous to S.c. Rim101 [119,120], while the processing of the native Xpr2 is carried out by a dibasic endoprotease Xpr6 . This regulatory pathway would be equivalent to the pathway controlled by the RIM genes in S. cerevisiae.
The product of the homeogene HOY1, which shows some regions of homology with the S. cerevisiae transcriptional activator Pho2, is required for hyphal formation in Y. lipolytica. The expression of HOY1 increases when the yeast cells are induced to form hyphae but there is not yet information on the genes which could be controlled by Hoy1.
Even the fission yeast Schizosaccharomyces pombe may undergo pseudohyphal growth in certain stress conditions. Although very little is known on the mechanisms underlying the morphological transition, there have been reports that a defect in a MAP kinase, Pmk1/Spm1 [123,124] or in a MEK kinase, Mkh1  interferes with the completion of cytokinesis and results in growth of filaments with multiple septa and nuclei. The absence of a coiled-coil protein, Alm1, with similarity to myosin and other fibrous proteins, results in the production of a high proportion of elongated mononucleate cells . Since Alm1 is associated with the medial region of the cells during nuclei separation at anaphase, it has been suggested that Alm1 plays a role in initiating cell division.
The fungal pathogens Cryptococcus neoformans and Ustilago maydis show some characteristics which set them apart from the fungi we have considered until now. Under suitable environmental conditions haploid, budding cells mate and generate a dikaryotic filamentous cell type [9,127]. However, a true dimorphic transition from a haploid yeast phase to a hyphal phase has also been observed for both species [128,129]. In the case of C. neoformans this transition, called haploid fruiting, is stimulated in the presence of a dominant RAS1 mutant protein  and requires a STE12 homologue [130,131]. STE12 in turn is under the control of the G-protein β subunit GBP1, which has been proposed to activate a MAP kinase pathway . Although cAMP has been shown to suppress the mating defect of a gpa1 mutant lacking a G-protein α subunit homologue, there is no evidence that cAMP affects haploid fruiting .
For U. maydis, a disruption of the genes encoding adenylate cyclase (UAC1) or the major catalytic subunit of cAMP-dependent protein kinase (ADR1) resulted in a constitutive filamentous phenotype [129,134]. Filamentous growth in a uac1 mutant was suppressed by the lack of the MAP kinase kinase, Fuz7, or of the MAP kinase, Ubc3 . These observations support the idea that morphogenesis in U. maydis is controlled by both cAMP and MAP kinase signal transduction pathways. However, while in the well studied cases of S. cerevisiae and C. albicans, cAMP stimulates filamentation, in the case of U. maydis cAMP is required for budding growth. It may be noted, too, that for Mucor rouxii, cAMP levels were found to decrease before the yeast-to-hypha transition induced by exposure to air and that this transition was blocked by addition of dibutyryl-cAMP .
7Conclusions and perspectives
The role played in morphogenesis by a MAP kinase cascade and by a cAMP-dependent pathway has been clearly established in S. cerevisiae. The MAP kinase cascade is required for cell elongation while the cAMP-dependent protein kinase Tpk2 controls the switch to unipolar budding . For invasiveness both pathways appear redundant to some extent since a defect in one of the pathways may be compensated by an increased activity in the other .
However, large gaps in knowledge remain, which concern different aspects of the morphogenetic process. Without being exhaustive, the following points can be listed:
1How do external signals trigger the operation of the regulatory pathways?The receptor Gpr1 and its coupled G protein, Gpa2, appear to be important in the response to the nutritional situation, and a role has also been suggested for the membrane protein Sho1. However, due to the fact that Ras2 is situated upstream of the two well-known regulatory pathways, it should be expected that some signal is transmitted to the Ras2 protein itself and this point has not yet been addressed.
2What are the final targets for the different regulatory pathways?Actin appears to play a very important role in determining the budding pattern and the morphology of the cells (round vs. elongated), but it is not yet clear how Tpk2 and the MAP kinase cascade influence the structure of the actin cytoskeleton.It is also clear that the formation of chains of cells as well as invasiveness would require specific cell-surface proteins. The only such protein identified up to now is Muc1/Flo11 but other proteins may be involved and it could be worth looking more closely at the putative membrane protein Dfg16 required for agar invasion.
3What are the roles for a number of elements involved in PH growth but which have not been yet connected with the main regulatory pathways?One of these elements is the cyclin-dependent kinase Cdc28. It has been suggested that the MAP kinase cascade which leads to the activation of Ste12 could modulate the cyclin/Cdc28 complexes either by promoting hyperphosphorylation of Cdc28  or by inhibiting the mitotic Clb/Cdc28 complexes . However, the evidence for this is still rather indirect and further research will be necessary to test these suggestions.Another protein which stimulates PH growth, but for which no mechanism of action has been described is Phd1. It has been shown that Phd1 acts independently of Tpk2 , but Phd1 could be a substrate for Tpk1 or Tpk3. In this context, it can be noted that the Phd1 homologue in C. albicans (Efg1) is likely to be controlled through phosphorylation by a cAMP-dependent protein kinase (D. Bockmühl and J.F. Ernst, personal communication).
An intriguing point which has not received much attention is the fact that invasiveness of haploid strains grown in liquid medium and spotted in test plates is strongly dependent on the growth phase at which the cells were sampled from the liquid cultures. Late-exponential phase yeast cells showed a stronger invasive phenotype. It has been suggested that the phase of growth could even affect the results of epistasis analysis . Differences due to the growth state of the cells used in different laboratories may explain some of the contradictions between reports from different groups.
When the factors which regulate morphogenesis in different yeasts are compared, it is plain that no unified picture emerges. Nitrogen starvation stimulates filamentous growth in both S. cerevisiae and C. albicans, but inhibits hyphae formation in cultures of Y. lipolytica. With respect to cAMP, it has a positive effect on filamentation in S. cerevisiae and C. albicans, but a negative one in M. rouxii or U. maydis. There are also marked differences between S. cerevisiae and C. albicans: disruption of TUP1 stimulates hyphal growth in Candida and reduces PH growth in Saccharomyces diploids, whereas disruption of RIM101 has no effect in a Saccharomyces diploid, but disruption of its homologue, HRM101, impairs filamentation in Candida. On the other hand some signaling pathways, such as the MAP kinase cascade leading to phosphorylation of Ste12, the cAMP-dependent pathway, or the pathway involving Phd1 in S. cerevisiae and its homologue Efg1 in C. albicans are found in the two yeasts although they could differ in the regulatory details. Therefore, even in the absence of extensive studies in species other than S. cerevisiae and C. albicans, it could be concluded that the fungal signal transduction pathways have been only partially conserved during evolution. Depending on the ecological niche where each particular species has evolved, diversity has been generated, through selection of the mechanisms which allowed a better adaptation to the features of each particular environment.
I am grateful to all the colleagues who communicated to me unpublished results. I thank Carlos Gancedo and Oscar Zaragoza for valuable comments on the manuscript, and O. Zaragoza for the pictures shown in Fig. 1. Work in my laboratory has been supported by Grant PB97-1213-CO2-01 from the Dirección General de Investigación Científica y Técnica.