The CreA Repressor Is the Sole DNA-binding Protein Responsible for Carbon Catabolite Repression of the alcA Gene inAspergillus nidulans via Its Binding to a Couple of Specific Sites*

Carbon catabolite repression is mediated inAspergillus nidulans by the negative acting protein CreA. The CreA repressor plays a major role in the control of the expression of the alc regulon, encoding proteins required for the ethanol utilization pathway. It represses directly, at the transcriptional level, the specific transacting gene alcR, the two structural genes alcA and aldA, and other alc genes in all physiological growth conditions. Among the seven putative CreA sites identified in the alcApromoter region, we have determined the CreA functional targets in AlcR constitutive and derepressed genetic backgrounds. Two different divergent CreA sites, of which one overlaps a functional AlcR inverted repeat site, are largely responsible for alcA repression. Totally derepressed alcA expression is achieved when these two CreA sites are disrupted in addition to another single site, which overlaps the functional palindromic induction target. The fact that derepression is always associated with alcA overexpression is consistent with a competition model between AlcR and CreA for their cognate targets in the same region of the alcApromoter. Our results also indicate that the CreA repressor is necessary and sufficient for the total repression of the alcA gene.

Carbon catabolite repression is mediated in Aspergillus nidulans by the negative acting protein CreA. The CreA repressor plays a major role in the control of the expression of the alc regulon, encoding proteins required for the ethanol utilization pathway. It represses directly, at the transcriptional level, the specific transacting gene alcR, the two structural genes alcA and aldA, and other alc genes in all physiological growth conditions. Among the seven putative CreA sites identified in the alcA promoter region, we have determined the CreA functional targets in AlcR constitutive and derepressed genetic backgrounds. Two different divergent CreA sites, of which one overlaps a functional AlcR inverted repeat site, are largely responsible for alcA repression. Totally derepressed alcA expression is achieved when these two CreA sites are disrupted in addition to another single site, which overlaps the functional palindromic induction target. The fact that derepression is always associated with alcA overexpression is consistent with a competition model between AlcR and CreA for their cognate targets in the same region of the alcA promoter.
Our results also indicate that the CreA repressor is necessary and sufficient for the total repression of the alcA gene.
Aspergillus nidulans is a versatile organism capable of growing on a wide variety of nutrients by expressing enzymes and permeases involved in specific utilization pathways. When ethanol or a related carbon source is added to a growing culture of the fungus A. nidulans, the alc genes are expressed. They comprise not only the genes necessary for ethanol degradation, namely alcA and aldA, encoding alcohol dehydrogenase I (ADHI) 1 and aldehyde dehydrogenase, respectively (1,2), but at least four other alc genes (alcO, -M, -S, and -U), recently identified (3), which are clustered in the same locus in chromosome VII, as are alcA and alcR. This latter gene encodes the specific transactivator of the ethanol utilization pathway, AlcR, a zinc binuclear cluster protein of the C6 class, with features that differentiate it from the other proteins of this class (4,5).
The AlcR protein is able to bind in vitro to specific DNA single sites occurring as either direct or inverted repeats (6,7). An interesting aspect of AlcR activity is that in vivo, only repeated sites mediate specific transcriptional induction (6,8). The alcR gene itself is subjected to a positive feedback control (9) via the binding of the AlcR protein to targets in the alcR promoter (6). The strong inducible alcA promoter encompasses three clustered AlcR targets (10), which are responsible for the synergistic activation of the alcA gene (8).
When a favored carbon source such as glucose is added to the culture medium, transcription of the alc regulon is severely reduced (11). This phenomenon is mediated by CreA, the repressor responsible for carbon catabolite repression in A. nidulans (12). CreA contains two Cys 2 -His 2 fingers and is related to MIG1, the repressor responsible for glucose repression of several genes in Saccharomyces cerevisiae, and to the mammalian Krox20/Egr and Wilm's tumor proteins (13). CreA counterparts have also been found in other fungal species such as Aspergillus niger and Trichoderma reesei (14,15). All of these proteins bind to a GC-rich motif. Binding sites for CreA, whose consensus sequence is 5Ј-(C/G)YGGRG-3Ј, have been identified in the alc regulon (11,16,17) and in the proline cluster (18). The repressor CreA exerts transcriptional repression on the transacting gene alcR and independently on structural genes alcA and aldA (17). A great number of CreA binding sites have been identified in the upstream regions of the alcR and alcA genes, and these sites are localized in the same region as the specific AlcR binding sites (16). In the alcR promoter region, it has been shown that for at least one of the CreA binding sites overlapping the AlcR inverted repeat target, direct competition between the two regulators occurs both in vitro and in vivo. More generally, under nonrepressing physiological growth conditions (induced and noninduced), it was shown that both regulators control the expression of the alc genes (19,20). The interplay between these two circuits therefore regulates the expression of the ethanol regulon genes.
In the alcA promoter, several CreA binding sites have been identified by sequence analysis. Three are close to or overlap AlcR binding sites. The alcA promoter is one of the strongest inducible genes in A. nidulans. Synergistic transcriptional activation by AlcR is mediated by three sets of AlcR repeat sites, which were shown to be functional in vivo and represent alc upstream activating sequences (8). The transcriptional induction of the alcA gene is absolutely dependent upon the presence of AlcR. In glucose-repressed conditions, the alcR gene is totally repressed, and as a consequence (by a cascade mechanism), alcA is not expressed (9).
The aim of this work was to understand the mechanism of carbon catabolite repression in the alcA promoter. We have addressed several questions: (i) which CreA binding sites are functional in vivo; (ii) whether CreA is the only repressor responsible for alcA glucose repression; and (iii) whether there is competition between the activator AlcR and the repressor CreA for the same region in the alcA promoter in A. nidulans.

EXPERIMENTAL PROCEDURES
Strains, Media, and Growth Conditions-The A. nidulans strain used as the host for transformation was alcA Ϫ argB Ϫ uaZ Ϫ (alcA4951, argB2, pabaA1, yA2, uaZ11). Media and supplements were as described by Cove (21). The mycelia for Northern analyses and ADHI activities were grown for 7 h at 37°C on 0.1% fructose as the sole carbon source. Induction was achieved by adding the gratuitous inducer ethyl methyl ketone (50 mM) or ethanol (50 mM). Cells were harvested after a further 2.5 h (induced conditions). Under repressed conditions, 1% glucose was added simultaneously with the inducer.
Plasmid Construction and Transformant Selection-The constitutive and derepressed pgpdA:alcR:argB plasmid (bAN8) was constructed as follows. The alcR coding sequence was amplified by polymerase chain reaction from the SalI-SalI (Ϫ1064 to ϩ2848) fragment of the alcR gene, introducing an EcoRI site at the 5Ј-end and an NcoI site at the ATG. The EcoRI-SalI (Ϫ5 to ϩ2848) fragment was cloned into the EcoRI-SalI sites of pBluescript SKϩ, resulting in plasmid bAN6. The EcoRI-NcoI fragment corresponding to the gpdA promoter was cloned into the EcoRI-NcoI sites of bAN6, resulting in bAN7. The XbaI-XbaI fragment (made blunt end), of the argB gene was cloned into the EcoRI site (made blunt end) of plasmid bAN7, resulting in plasmid bAN8.
The pyrG:alcR:argB (bAN10) plasmid was constructed as follows. The pyrG promoter was amplified by polymerase chain reaction introducing an EcoRI at the 5Ј-end and an NcoI site at the 3Ј-end. The EcoRI-NcoI fragment was cloned in the EcoRI-NcoI sites of bAN6, resulting in plasmid bAN9. The argB gene was then cloned, as for bAN8, resulting in plasmid bAN10.
The alcA Ϫ argB Ϫ uaZ Ϫ strain was transformed with bAN8 or bAN10. Southern blot analysis of the selected argB ϩ transformants showed that the plasmids were integrated at the argB locus (data not shown).
A plasmid deleted for the CreA binding sites C, D 1 , D 2 , and E was obtained by deleting the HindIII-PvuII (Ϫ738 to Ϫ423) fragment from plasmid bAuaZ, which contains the alcA gene (HindIII-KpnI (Ϫ738 to Ϫ2024) fragment) and the 3Ј ClaI-XhoI fragment of the uaZ gene able to complement by site-directed integration the uaZ11 translocation, resulting in plasmid bAN50.
Disruption of CreA binding sites (A, B 1 , and B 2 ) was achieved by site-directed mutagenesis carried out by the method of Kunkel et al. (22). A uracil-SK template (bANC), containing the alcA promoter from the MluI (Ϫ255) to ApaI (ϩ19) sites, was mutagenized. Mutations in CreA sites A, B 1 , B 2 , B 1 B 2 , and AB 1 B 2 (single, double, and triple mutations) were confirmed by sequencing. Cloning of these mutated sequences was achieved by substituting the ApaI-ApaI or the MluI-ApaI fragments of bAN50 with the corresponding mutated sequences, resulting in plasmids pUA (A mutated), pUB 1 (B 1 mutated), pUB 2 (B 2 mutated), pUB 1 B 2 (B 1 B 2 mutated), and pUAB 1 B 2 (AB 1 B 2 mutated).
All of the plasmids were used to transform the gpdA:alcR (Tg) strain, resulting in the TgmA strain transformed by pUA, the TgmB 1 strain transformed by pUB 1 , the TgmB 2 strain transformed by pUB 2 , the TgmB 1 B 2 strain transformed by pUB 1 B 2 , the TgmAB 1 B 2 strain transformed by pUAB 1 B 2 , and the Tg⌬CD 1 D 2 E strain transformed by bAN50. The pUB 1 B 2 and pUAB 1 B 2 were also used to transform the argB:pyrG:alcR (Tp) strain, resulting, respectively, in TpmB 1 B 2 and TpmAB 1 B 2 transformant strains. Southern blot analysis of the selected uaZ ϩ transformants showed that the plasmids were integrated in a single copy at the uaZ locus (data not shown).
Isolation of RNA and Quantitative Analyses-Total RNA was isolated from A. nidulans as described by Lockington et al. (23), separated on agarose gels after denaturation by glyoxal (24) and blotted on Hybond-N membranes (Amersham Corp.). Hybridization was carried out in 0.5 M sodium orthophosphate, pH 7, 1 mM EDTA, 7% SDS, and 1% bovine serum albumin at 65°C for at least 14 h. The probes used were the entire genes either of alcR or alcA, cloned into Bluescript plasmids (1,25) and labeled with [ 32 P]dCTP using random hexanucleotide primers (Amersham). The membranes were also hybridized with a BamHI-KpnI restriction fragment containing the actin gene, isolated from the pSF5 plasmid (26), as an internal control to monitor the amount of mRNA loaded. Autoradiograms were developed at various times to avoid saturation of the film. Densitometric scanning was performed with a system Biosoft-Orkis. Experiments were repeated at least three times. The amount of specific mRNA was calculated relative to the actin signal. Observed variations among the various hybridizations were 20 -30%.
Polyacrylamide Gel Electrophoresis and ADHI Activity-Gel electrophoresis and activity staining of ADHI were carried out according to the method described by Sealy-Lewis and Fairhurst (27). Protein concentration was measured according to Bradford (28). Concentrations of protein from different extracts were equalized before loading onto a gel and verified by Coomassie staining. Densitometric scanning was performed. Variations of 20 -30% were observed between experiments. Fig. 1 shows that seven CreA binding sites are localized in the alcA promoter. Two of them, A and B1, have been identified by footprint analysis (16), and the others, B 2 , C, D 1 , D 2 , and E, have been identified by sequence analysis through searching for the same consensus motif, 5Ј-(G/C)YGGRG-3Ј. We have shown previously that in vitro CreA is able to bind to any of these sites even when the flanking DNA sequence is completely unrelated to A. nidulans, such as the polylinker region of a Bluescript plasmid (16). This was confirmed by in vitro analysis of CreA binding to the penicillin biosynthetic gene in A. nidulans, which is not subject to CreA control in vivo (29) and was also suggested by Cubero and Scazzocchio (18)  prn cluster. The same conclusions were drawn after in vitro localization of MIG1 binding sites, the repressor responsible for glucose-repressed genes in S. cerevisiae (30).

Identification of CreA Binding Sites in the alcA Promoter-
It has been shown previously that the alcA structural gene is under the direct control of CreA (17). Therefore, it was important to determine which of the putative CreA sites are actually functional in the alcA promoter.
alcA Expression Is Dependent upon the Level of the AlcR Protein-The analysis of the mechanism of carbon catabolite repression of the alcA promoter has to be performed in an alcR-derepressed background. The reason already mentioned is the absolute dependence for alcA transcription on the presence of an active AlcR protein that is repressed in glucose growth medium. A constitutively derepressed alcR strain was constructed using the promoter of gpdA, encoding glyceraldehyde-3-phosphate dehydrogenase, upstream of the full-length alcR gene with its own transcriptional termination signal, unlike our previous construct in which the first six amino residues and the termination signal of AlcR were deleted (4,17).
The gpdA promoter is very strong and drives the expression of a high amount of alcR gene. Under repressing growth conditions, alcA transcription is substantially derepressed (50%) as a consequence of AlcR synthesis. Since the level of alcA expression is dependent on that of alcR (17), it was important to set up glucose-repressing conditions in which alcA expression could be monitored with suitable accuracy. It has been shown previously (31) that total alcohol dehydrogenase activity in crude extracts varies considerably, depending upon the alcohols and ketones used as inducers. The best inducer was shown to be the gratuitous inducer butan-2-one (also called ethyl methyl ketone (EMK)). Transcriptional analyses of alcR and alcA and semiquantitative measurements of ADHI activity, estimated by specific staining of gels after native electro-phoresis, of the gpdA:alcR strain were compared with the wild type strain after the addition of two different external inducers, ethyl methyl ketone and ethanol. Northern blots, presented in Fig. 2A, show that both these external inducers have about the same transcriptional inducing ability for alcR in the wild type strain and in the gpdA:alcR strain, in which a higher steady state amount of alcR mRNA is also observed. As a result, an increase in alcA transcription is observed in the gpdA:alcR strain compared with the wild type strain, which, in the presence of the inducer ethanol, is 4-fold higher and in the presence of EMK is 8-fold higher.
As expected with the constitutive and derepressed gpdA:alcR promoter, the same steady state amount of alcR mRNA is observed in mycelia grown in noninduced, induced, and glucose-repressed conditions. In agreement with our previous results (9,17), no alcA transcription is observed in the absence of an external co-inducer. Interestingly, in the gpdA:alcR strain, in the presence of the inducer ethanol, alcA mRNA derepression is weak (10%), whereas in the presence of EMK, it reaches 50% ( Fig. 2A). Semiquantitative evaluation of ADHI activity by native gel staining (Fig. 2B) shows that with ethanol, only a weak ADHI derepression is observed, whereas with EMK, it is much more efficient. Therefore, this gpdA:alcR strain, when induced with ethanol, is suitable to analyze the mechanism of alcA repression.
Among Seven CreA Binding Sites, a Pair of Sites Is Largely Responsible for alcA Repression-As seen in Fig. 1, seven putative CreA binding sites, sharing the consensus sequence, 5Ј-(C/G)YGGRG-3Ј, have been identified in the alcA promoter. Four of them, C, D 1 , D 2 , and E, localized in the 5Ј alcA promoter region, have been deleted. The resulting deleted alcA plasmid, containing the selectable uaZ gene marker, was used to transform the A. nidulans gpdA:alcR alcA Ϫ uaZ Ϫ strain. Transfor- FIG. 2. A, analysis of alcR and alcA transcription in the gpdA:alcR (Tg) strain grown under various physiological growth conditions. The construction of the gpdA: alcR strain and growth conditions were as described under "Experimental Procedures." NI, noninduced growth conditions; I 1 induction with 50 mM ethanol; I 2 , induction with 50 mM EMK; I 1 G or I 2 G, glucose-repressed conditions in the presence of the inducer (ethanol or EMK). The membranes were hybridized with various 32 P-labeled probes, as indicated to the left of the Northern blots. The actin signal reveals the amount of mRNA loaded in each lane. The scanning diagram was normalized to the value 10, representing the cognate mRNA level of the wild type (WT) strain under ethanol-induced growth conditions, and is the average of three independent experiments (Ϯ20%). B, differently induced ADHI activities in the gpdA:alcR (Tg) strain. Growth conditions were as described under "Experimental Procedures." Each track was loaded with 50 g of wild type strain protein and with 30 g of Tg strain protein. The amount of protein loaded per track was equalized and controlled by Coomassie staining. Independent experiments were performed at least twice. ADHI activity staining was according to Sealy-Lewis and Fairhurst (27). The scanning diagram was normalized to a value of 10, representing ADHI activity of the wild type strain under ethanol-induced growth conditions. The numbers reported are accurate to within 20 -30%. mants containing a single copy of the alcA gene, integrated in the uaZ locus, were selected. ADHI activity of the selected transformants (TA⌬CD 1 D 2 E) was estimated by native gel staining after noninduced, induced, and glucose-repressed conditions and compared with the control strain carrying the entire alcA gene in the gpdA:alcR background. Similar ADHI activities were obtained in all growth conditions (results not shown). Therefore, we can conclude that CreA binding sites C, D 1 , D 2 , and E do not contribute to alcA repression.
The involvement of CreA sites A, B 1 , and B 2 in alcA repression was tested by individual site-directed mutagenesis to leave intact all the AlcR induction sites and to retain the natural distance to the alcA transcription start. Fig. 3B shows that mutagenesis of the CreA A binding site, localized near the AlcR b target, results in an increase in ADHI activity and a slight derepression (10%) compared with the transformation control Tg. Fig. 3A shows that individual mutations in CreA binding site B 1 or B 2 , localized downstream of the AlcR c target, result in an increase in ADHI activities (12-fold) and a strong ADHI derepression (about 80%). Therefore, disruption of either CreA binding site, B 1 or B 2 , is sufficient to derepress the alcA promoter.
alcA Derepression following CreA Binding Site Disruption Is Independent of AlcR Level-To know if alcA expression could be totally derepressed, the three active CreA binding sites A, B 1 , and B 2 were disrupted. All mutagenized alcA transformants were selected as single copy integrants at the uaZ locus.
Interestingly, analysis of transformants in which B 1 and B 2 CreA binding sites are simultaneously mutagenized results in the same increase in induction and the same level of derepression as observed after the disruption of either B 1 or B 2 CreA binding sites (Fig. 4). Therefore, it seems that CreA sites B 1 and B 2 are working as a pair. When the three CreA binding sites A, B 1 , and B 2 are disrupted simultaneously, as in Tgm (AB 1 B 2 ), it is clear that there is a large increase in induced ADHI activity besides an almost totally derepressed ADHI activity (Fig. 4). To verify that this total derepression did not result from the high AlcR concentration driven by the strong gpdA promoter, AlcR was expressed under the control of the pyrG promoter (pyrG encodes orotidine-5Ј-phosphate decarboxylase). This promoter is not subject to carbon catabolite repression and is considered as a weak, noninducible promoter. A translational fusion between the pyrG promoter and the alcR coding sequence was constructed, and the argB auxotrophy gene was also inserted into the plasmid. A selected argB ϩ transformant, containing pyrG:alcR, was then used as the recipient strain (TpyrG alcA Ϫ uaZ Ϫ ) for plasmids carrying the mutagenized alcA gene. This strain contains the pyrG:alcR gene in addition to the alcR resident gene.
As shown in Fig. 5, under induced growth conditions in the pyrG:alcR strain (Tp), the ADHI activity is higher than in the wild type strain due to the presence of the pyrG:alcR gene and the chromosomal alcR gene. In this pyrG:alcR strain, when the three CreA binding sites A, B 1 , and B 2 in the alcA promoter are disrupted, ADHI derepression is observed along with increase in ADHI activity in induced growth conditions (Fig. 6). The ADHI derepression observed in this strain can only be accounted for by the pyrG:alcR derepressed gene, whereas in induced growth conditions, alcA expression is driven by two alcR genes as mentioned above. This ADHI derepression is complete with regard to the pyrG:alcR gene. We can conclude that mutagenesis of the three CreA binding sites A, B 1 , and B 2 , in a context in which AlcR is moderately expressed, results in a totally derepressed alcA promoter. In other words, disrupted CreA sites have a dominant effect on AlcR concentration in the cell. DISCUSSION We have previously shown that the repressor CreA, responsible for carbon catabolite repression in A. nidulans, acts at two different levels. First, it prevents completely the transcription of the trans-acting gene alcR, which could account for the drastic decrease of transcription of the two structural genes alcA and aldA (9) and also of the other clustered alc genes (3). Second, however, direct repression of the two structural genes alcA and aldA (17) and of other alc genes such as alcS and alcO (3) also occurs via the CreA binding sites localized in the promoter regions of these genes. This was shown in a partially derepressed alcR context, obtained either by mutagenesis of one CreA binding site in the alcR promoter or by driving alcR transcription from a constitutive (and derepressed) promoter (17). Results shown here demonstrate that among seven CreA binding sites identified in the alcA promoter, two sites (B 1 and B 2 ) are largely responsible for alcA repression, with site A making a minor contribution. These three functional CreA targets have completely different flanking regions (as seen in Fig. 1), and no AT-rich region is found. The importance of a flanking AT-rich region in site recognition by MIG1 was shown by Lundin et al. (30). Since the mode of recognition of MIG1 and CreA are similar, a similar role for the flanking regions may also have been expected. In vitro binding of CreA to sites in the ipnA promoter region is AT context-dependent (29). Functional CreA targets in the alcR promoter (17) and in the prnB promoter (18) show AT-rich regions upstream from the GC consensus sequence. However, it has to be pointed out that in some physiological CreA target, as in the alcR promoter, 2 and in known MIG1 sites (in GAL1 and GAL4) involved in glucose repression, no AT-rich flanking sequence have been found. In the ipnA promoter, the in vitro CreA binding sites are not functional in vivo (29). We have to mention that in vitro CreA binding sites have been localized using a dimeric GST-CreA fusion protein. We have recently shown in our laboratory (7) that the utilization of GST, as an AlcR GST fusion protein, can introduce a serious bias in AlcR DNA recognition of AlcR sites in the alcR, alcA, and aldA promoters, probably as a consequence of the normally monomeric structure of AlcR. It is also expected that CreA is monomeric in solution, as demonstrated for the related Zif268 protein by crystallography studies. It is therefore possible that the use of a GST-CreA fusion protein will also introduce DNA recognition artifacts. It was postulated by Espeso and Peñ alva (29) and by Cubero and Scazzocchio (18) that the dimeric nature of GST:CreA fusions could stabilize CreA binding to sites that do not fit the GC consensus site. In this regard, physiological results should also be used when defining CreA targets. Therefore, CreA targets in the alcA promoter, localized in vivo by site-directed mutagenesis, can be considered as upstream repressing sequences.
Interestingly, functional CreA binding sites are more often organized as pairs of sites directly repeated (e.g. the alcR promoter (16)) 2 or as inverted repeats (e.g. the alcA promoter (this work) and in the prnB promoter region (18)). The single CreA A site in the alcA promoter makes only a minor contribution to glucose repression. MIG1 sites are also often organized as repeats (e.g. the promoters of GAL1, SUC2 (32)). It is well established that MIG1 recruits TUP1 and SSN6, which are general repressors, to establish repression of glucose-repressed promoters. It is possible that CreA acts in a similar way to repress the transcriptional machinery and that two CreA binding sites are necessary.
Disruption of the three CreA targets, B 1 , B 2 , and A, results in totally derepressed alcA gene expression. This result indicates that CreA seems to be the only repressor acting directly on alcA during carbon catabolite repression. In S. cerevisiae, it has been shown recently that besides MIG1, another similar zinc finger repressor MIG2, which binds to the same binding sites as MIG1, is involved in glucose repression of SUC2 expression (33). In fact, disruption of MIG1 relieves most of the glucose repression of GAL1 and GAL4 expression, partially relieves SUC2 repression, and has no effect on the glucose repression of other genes (34). In A. nidulans, no carbon catabolite repressor, other than CreA, has been found. This is in agreement with our results presented here and with the complete glucose-derepressed prnB expression observed after the disruption of two CreA binding sites. 3 In this system, it was shown recently that CreA binding prevents the activity of a positive element on the prnB promoter (35).
In the alcA promoter, CreA sites are very close to, and even overlap with, the AlcR binding sites. It is noticeable that there is a correlation between the level of alcA derepression and that of overexpression. In agreement with our previous results (17,19), this correlation also indicates that CreA is active under derepressing conditions. The disruption of the minor CreA A site results in a weak increase in alcA expression, whereas disruption of the pair sites, B 1 and B 2 , playing the major role in repression, results in a drastic increase in alcA expression. The alcA promoter region encompassing the induction target c comprises direct repeats (c 1 -c 2 ) and an overlapping inverted repeat ( recently shown that to reach full alcA transcriptional activation, the three sites, c 1 , c 2 , and c 3 , are necessary (8). The CreA B 1 site overlaps the AlcR c 3 site, and the CreA B 2 site is localized only 17 nucleotides downstream from B 1 . The size of the AlcR protein is 92 kDa, and that of CreA is 47 kDa. It is possible, therefore, to imagine a steric hindrance between AlcR and CreA for their cognate targets localized in the same region of the alcA promoter. This steric competition could account for alcA overexpression when functional CreA binding sites are disrupted. The induction of alcA is correlated with the level of AlcR, as observed in contexts in which AlcR is highly (gpdA: alcR) or moderately (pyrG:alcR) expressed. Interestingly, when the functional CreA targets are disrupted in the alcA promoter, the level of derepression is independent of AlcR. This result is expected, since in the absence of CreA sites, AlcR can fully occupy its cognate targets. In the alcR promoter, which is positively autoregulated, the induction site overlaps a CreA site. Disruption of this CreA site results in an overexpression of alcR and only a partial derepression, since other CreA sites should be functional in the alcR promoter. The mechanism involved is a direct competition between the two regulators AlcR and CreA (17). Such a competition mechanism could control the expression of other clustered alc genes, such as alcS and alcO, where promoter regions contain both AlcR and CreA sites (20). These results are in agreement with a regulatory mechanism whereby the relative levels of these proteins govern the degree of alc regulon expression under different growth conditions.