Transcription of an important number of divergent genes of Saccharomyces cerevisiae is controlled by intergenic regions, which constitute factual bidirectional promoters. However, few of such promoters have been characterized in detail. The analysis of the UGA3-GLT1 intergenic region has provided an interesting model to study the joint action of two global transcriptional activators that had been considered to act independently. Our results show that Gln3p and Gcn4p exert their effect upon cis-acting elements, which are shared in a bidirectional promoter. Accordingly, when yeast is grown on a low-quality nitrogen source, or under amino acid deprivation, the expression of both UGA3 and GLT1 is induced through the action of both these global transcriptional modulators that bind to a region of the bidirectional promoter. In addition, we demonstrate that chromatin organization plays a major role in the bidirectional properties of the UGA3-GLT1 promoter, through the action of an upstream Abf1p-binding consensus sequence and a polydAdTtract. Mutations in these cis-elements differentially affect transcription of UGA3 and GLT1, and thus alter the overall relative expression. This is the first example of an intergenic region constituting a promoter whose bidirectional character is determined by chromatin organization.
It has long been recognized that the genomes of a variety of organisms contain open reading frames that are transcribed divergently from closely spaced intergenic regions (Beck and Warren, 1988). Whole genome sequencing of Saccharomyces cerevisiae has extended this observation allowing the identification of a large proportion of gene pairs whose expression could be regulated from presumed bidirectional promoters, indicating that this organization is widely represented in the yeast genome. In regard to the physiological meaning of this organization, it has been proposed that bidirectional promoters constitute a regulatory instance allowing a tightly controlled co-ordinate expression of genes pertaining to the same or related metabolic pathways (Zhang and Smith, 1998). Thus, the expression of both genes depends on the transcriptional activation mediated by shared activators responding to a specific environmental signal.
The S. cerevisiae GLT1 gene encodes the enzyme glutamate synthase (GOGAT), involved in glutamate biosynthesis (Cogoni et al., 1995), and is located in opposite orientation, next to the UGA3 gene, which codes for a transcriptional activator of the genes involved in γ-aminobutyrate (GABA) catabolism (Andre, 1990). Thus, the intergenic sequence located between GLT1 and UGA3 could afford binding sites for trans-acting regulatory elements determining the expression of either one of the two divergent genes. The products of these two genes do not participate in the same metabolic pathway; nevertheless, when yeast is grown on media containing glucose and GABA as carbon and nitrogen sources, the catabolism of the latter compound is the only nitrogen-providing pathway, and glutamate is preferentially synthesized through the concerted action of the GLN1-encoded glutamine synthetase and GOGAT, because under this condition the transcription of GDH3 encoding NADP+-dependent glutamate dehydrogenase 3 is glucose-repressed and the expression of GDH1 encoding NADP+-dependent glutamate dehydrogenase 1 is decreased as compared with that found on ammonium (DeLuna et al., 2001; Riego et al., 2002). These facts suggest that the expression of UGA3 and GLT1 should be co-ordinately controlled in GABA-grown cells.
It had long been considered that genes like UGA3, which are involved in nitrogen catabolism, were primarily regulated by Gln3p and/or Gat1p/Nil1p (Minehart and Magasanik, 1991; Stanbrough et al., 1995; Coffman et al., 1996), while those involved in amino acid biosynthesis, like GLT1, were Gcn4p-dependent (Hinnebusch, 1985). However, analysis of the transcriptional activation profile of yeast cells treated with rapamycin revealed that the inhibition of the target of rapamycin (TOR) cascade elicited the transcriptional activation of genes involved in the catabolism or transport of nitrogenous compounds, through the action of both Gln3p and Gcn4p (Valenzuela et al., 1998; Cardenas et al., 1999). These studies showed that the expression of catabolic genes could be achieved through the combined action of Gln3p and Gcn4p. In regard to biosynthetic genes, it has been found that the expression of GDH1 and GLN1 is determined by the action of Gln3p and Gcn4p (Mitchell and Magasanik, 1984; Riego et al., 2002), indicating that Gln3p-determined transcriptional activation is not exclusive of catabolic genes, and that the expression of biosynthetic genes can also be influenced by this activator.
Although gene activation is achieved through transcriptional modulators, the action of the cis- and trans-acting factors on gene expression is hindered or favoured by the specific status of chromatin organization of a given promoter under specific physiological conditions. The role of chromatin organization in the expression of bidirectional promoters has only been thoroughly studied for the GAL1-GAL10 promoter of S. cerevisiae (Bash and Lohr, 2001) and the prnD-prnB and niiA-niiD of Aspergillus nidulans (Muro-Pastor et al., 1999; Garcia et al., 2004). The parallel activation of the GAL1 and GAL10 genes is induced by galactose, while there is little transcription in non-inducing carbon sources. Full repression correlates with correct positioning of a nucleosome, which is lost upon induction (Bash and Lohr, 2001). Induced expression of both prnD-prnB and niiA-niiD requires chromatin remodelling through the removal of the nucleosomes, which are positioned along the intergenic region under repressive conditions (Muro-Pastor et al., 1999; Garcia et al., 2004). Thus, nucleosomal phasing can repress the expression of two divergent genes, and induction can be achieved after chromatin remodelling.
We undertook the study of the UGA3-GLT1 intergenic region as a model to analyse the concerted action of Gln3p and Gcn4p, two global transcriptional activators that respond to different metabolic signals. To understand the role of chromatin remodelling on the bidirectional action of this intergenic region, we also determined the chromatin organization of the UGA3-GLT1 promoter in the wild type and mutants altered in relevant cis-acting elements. Our results indicate that the UGA3-GLT1 intergenic region constitutes a genuine bidirectional promoter, and that transcriptional activation in response to the quality of the nitrogen source and to amino acid availability is co-ordinately determined through the combined action of both Gln3p and Gcn4p. The expression of UGA3 and GLT1 is asymmetrically dependent upon chromatin remodelling through the combined action of an Abf1p binding site and a polydAdTtract, which when impaired, exclusively hinder GLT1 expression. These results suggest that the wild-type chromatin organization facilitates the action of the pertinent cis- and trans-acting elements on GLT1 expression, although these are located more than 500 bp away from the GLT1 translation initiation site.
Conserved cis-acting elements in the UGA3-GLT1 intergenic region determine the expression of the two divergent genes
The intergenic sequence located between the divergently transcribed UGA3 and GLT1 genes could provide binding sites for trans-acting regulatory elements determining their expression. Previous observations had already indicated that the cis-acting elements present in a 132 bp region of the UGA3-GLT1 intergenic sequence, between −489 and −357 with respect to the GLT1 transcription initiation site (Fig. 1), play a major role in GLT1 expression (Valenzuela et al., 1998). These sites are the canonical gcn4box, palbox and gatabox (Fig. 1) presumably binding Gcn4p (Hinnebusch, 1985), a member of the Zn-Cys binuclear family of activators (Akache et al., 2001), and Gln3p or Gat1p/Nil1p (Minehart and Magasanik, 1991; Blinder and Magasanik, 1995; Stanbrough et al., 1995; Coffman et al., 1997). Figure 1 also shows the UGA3 transcription initiation site that was determined by primer extension analysis (Supplementary material Fig. S1), a consensus TATA box was not found for UGA3. The previously determined transcription initiation site and the presumed GLT1 tatabox (TATTTA) are also depicted in Fig. 1 (Valenzuela et al., 1998).
The UGA3-GLT1 intergenic sequences from five different Saccharomyces species were aligned using clustalw. The GLT1 tatabox and the above-mentioned cis-elements contained in the 132 bp region are conserved (Fig. 1). In addition, sequence comparison analysis allowed the identification of an Abf1 binding element and a polydAdT-rich region contiguously located to the 132 bp cluster (Fig. 1, Region 1).
To determine whether the cis-acting elements located in the 132 bp region affected UGA3 and GLT1 expression, plasmids independently carrying the full wild-type intergenic sequence or a deletion construct lacking the 132 bp tract (Fig. 1) were fused to the Escherichia coliβ-galactosidase (β-gal) coding region, oriented to either UGA3 or GLT1 (pUGA3-1, pUGA3-0, pGLT1-1 and pGLT1-0) as stated in Experimental procedures (Tables 1 and 2). Wild-type CLA1 strain was independently transformed with each one of the four plasmids, transformants were recovered and grown on minimal medium (MM) with GABA as sole nitrogen source. Extracts were prepared and β-gal activity was determined. The results indicated that the 132 bp promoter region houses the relevant cis-acting elements needed for both UGA3 and GLT1 expression, because the deletion constructs showed an important decrease in β-gal activity as compared with those determined for plasmids harbouring the wild-type promoter (238 versus 20 nmol min−1 mg−1 for pUGA3 series; 1250 versus 40 nmol min−1 mg−1 for pGLT1 series). In addition, these results show that, as has been found for other bidirectional promoters (Johnston and Davis, 1984; van der Merwe et al., 2001), one of the two genes, GLT1, displayed a considerable higher expression as compared with that detected for its partner, UGA3. Mutation of the presumed GLT1 tatabox resulted in null β-gal activity towards GLT1, while UGA3 showed wild-type expression (238 versus 250 nmol min−1 mg−1), indicating that GLT1 transcription does not negatively affect UGA3 expression.
Table 1. Oligonucleotides used in this work.
Bold letters indicate restriction site used for generate 132 bp deletion, lower case lettering indicates the mutations generated.
Point-substitution mutations of the three cis-acting elements located in the 132 bp cluster impair UGA3 and GLT1 expression, while mutations of the abf1box and polydAdTtract exclusively affect GLT1 expression
To analyse the role of each one of the above-mentioned cis-acting elements and that of the abf1box and the polydAdTtract, lacZ fusion plasmids bearing point-substitution mutations for each element were constructed (pGLT1-1 through 8 and pUGA3-1 through 8, Tables 1 and 2). These plasmids were independently introduced in a wild-type strain by transformation, and β-gal activities were determined in extracts obtained from cells grown on MM with GABA as sole nitrogen source. GCN4 binding site (gcn4box) was modified according to Sellers et al. (1990). This mutation diminished 1.5- and 1.7-fold GLT1 and UGA3 expression, respectively, as compared with that found in the wild-type strain (Fig. 2). Mutation of the gatabox, resulted in a 2.8-fold reduced GLT1 expression, while it had a less severe effect on UGA3, which only showed a 1.6-fold reduced expression (Fig. 2). The palindrome (palbox) was modified by altering the sequence of the second inverted CGG half site to atG; a twofold or threefold reduced expression for UGA3 or GLT1, respectively, was observed in this mutant. The double mutant carrying the palbox and gatabox mutations, showed 3.5-fold decreased activity for either UGA3 or GLT1 expression. As mentioned earlier, the palindromic sequence of the palbox suggests that it could bind a member of the Zn-Cys binuclear family of activators (Klug and Rhodes, 1987; Siddiqui and Brandriss, 1989), while the gatabox could constitute an UASNTR (UAS nitrogen regulated) element determining UGA3 and GLT1 nitrogen controlled response, which could bind Gln3p (Blinder and Magasanik, 1995). As Fig. 2 shows, the simultaneous mutation of the palbox and gatabox elements resulted in a negative additive effect, decreasing β-gal activity towards either UGA3 or GLT1 and suggesting that each one of these sequences could act independently. Mutations in either the abf1box or the polydAdTtract (Fig. 2) showed a twofold reduced β-gal activity towards GLT1, while UGA3 transcriptional activity was not affected and the wild-type phenotype was observed. Furthermore, in an abf1box polydAdTtract double mutant, GLT1 expression was sixfold reduced; while UGA3 maintained wild-type β-gal activities (Fig. 2). The above presented results confirm that the UGA3-GLT1 intergenic region determines UGA3 and GLT1 expression through the bidirectional action of a cluster of three cis-acting elements, while the abf1box and polydAdTtract exclusively determine GLT1 expression. Thus, the bidirectional capacity of the UGA3-GLT1 intergenic region is asymmetrically altered when expression is fostered by a fusion plasmid carrying impaired abf1box and polydAdTtract elements. An intriguing observation was the fact that although the relevant cluster of three cis-acting elements was found to be located 500 bp away from the GLT1 protein-coding sequence, and 100 bp away from the UGA3 protein-coding sequence, their effect was more profoundly exerted on GLT1 transcription. These results suggested that the asymmetric effect of the abf1box and polydAdTtract on UGA3 and GLT1 expression could constitute a regulatory instance, determining a chromatin organization of this intergenic region that could provide a means to overcome a possible restrain on GLT1 transcription by narrowing the physical distance between GLT1 tatabox and the DNA tract containing the cluster of cis-acting elements.
The abf1box and polydAdTtract determine chromatin organization of the UGA3-GLT1 promoter
Although it has been shown that ABF1 encodes a multifunctional essential protein that plays an important role in transcriptional activation and repression, gene silencing, recombination and telomere structure (Halfter et al., 1989; Rhode et al., 1989; Yarragudi et al., 2004), our results suggest that UGA3-GLT1 promoter-driven expression is influenced by the combined action of the abf1box and the polydAdTtract. In this regard, the most plausible mechanism that could involve the two elements could reside in chromatin organization (Lascaris et al., 2000). Thus, in order to get a deeper insight of the nature of the discriminative effect of these elements on UGA3 and GLT1 expression, the nucleosomal organization of the UGA3-GLT1 intergenic region was analysed. Chromatin was prepared as described in Experimental procedures and nucleosomal phasing of the chromosomal UGA3-GLT1 intergenic region was determined. As Fig. 3 shows, nucleosomes were preferentially positioned from the border of the polydAdTtract towards the GLT1 coding region, indicating a closed chromatin organization of this region. However, the promoter sequence containing the gcn4box, pal box, gata box, polydAdTtract and abf1box showed an open configuration, indicating that this region was free to interact with the pertinent trans-acting elements. A similar chromatin organization was observed when it was determined in the wild-type strain harbouring plasmid pGLT1-1. Three nucleosomes could be positioned (A, B and C, in Fig. 4A), while the rest of the promoter showed a more fluid organization and nucleosomes did not exhibit a preferred positioning (Fig. 4B). Thus, nucleosomal phasing indicated that the region in which the presumed cis-acting elements and the UGA3 transcription initiation site were located is accessible to factors determining transcriptional activation. Thus, it seems likely that this peculiar nucleosomal arrangement facilitates transcription, although it cannot be excluded whether transcriptional activation is the driving force determining nucleosomal phasing.
Next, chromatin organization was determined in a wild-type strain harbouring plasmids with mutations in the abf1box or the polydAdTtract. A shift of the position of nucleosome C was observed in the abf1box mutant, while the mutation in the polydAdTtract resulted in a more fluid organization reducing the preferential positioning of the three nucleosomes (Fig. 4). As expected, the combined effect of both mutations was observed in the abf1box polydAdTtract double mutant (Fig. 4).
The fact that mutants altered in the abf1box and polydAdTtract exclusively affected GLT1 expression (Fig. 2), suggests that changes in chromatin organization caused by these mutations could not result in the occlusion of the cis-acting elements located in the 132 bp cluster, which are also needed to sustain UGA3 transcription. High-resolution analysis of wild-type chromatin organization was performed in order to obtain a precise map of nucleosomes in the wild type and double abf1box polydAdTtract mutant. Figure 5 shows that, in both strains, the relevant cis-acting sites were not covered by nucleosomes and thus were accessible to the transcriptional machinery.
Gln3p and Gcn4p determine nitrogen-dependent expression of UGA3
Previous studies have shown that GLT1 expression is determined by the nature of the nitrogen source. Transcription is repressed when cells are grown on glutamate-rich nitrogen sources and activated in GABA-grown cultures through the action of the GLN3 and GCN4-encoded transcriptional regulators (Valenzuela et al., 1998). GLT1 was the first studied example in which the nitrogen-determined transcriptional response depended on Gln3p and Gcn4p, suggesting a physiological interaction between these two activators. In order to establish whether these modulators also determined nitrogen-dependent expression of the divergent UGA3 gene, transcriptional regulation was analysed by determining β-gal activities in a wild-type strain harbouring plasmid pUGA3-1. As Table 3 shows, β-gal activities were higher in the wild-type strain grown on GABA than on glutamate, glutamine or asparagine, showing that, as well as for GLT1, UGA3 expression was nitrogen-regulated, glutamate-repressed and activated when cells were grown on poor nitrogen sources. It is worth mentioning that chromatin prepared from cells grown on glutamate or GABA showed the wild-type pattern described above, indicating that nucleosome remodelling did not play a role in determining nitrogen-dependent response (Supplementary material Fig. S2).
Table 3. β-Gal activity of wild type, gat1Δ, gln3Δ and gcn4Δ harbouring plasmid pUGA3-1 strains grown on different nitrogen sources.
. Values are presented as means from three independent experiments ± SD.
Cells were grown on MM and harvested during exponential growth (Abs600 = 0.5–0.6).
238 ± 40
238 ± 40
100 ± 10
80 ± 10
161 ± 20
150 ± 20
70 ± 10
50 ± 4
126 ± 17
130 ± 10
29 ± 5
40 ± 3
85 ± 5
80 ± 8
54 ± 6
40 ± 3
To analyse whether Gln3p, Gat1p/Nil1p or Gcn4p had a role in UGA3 expression, plasmid pUGA3-1 was transformed into gat1Δ, gln3Δ and gcn4Δ mutant strains (Table 3). It was found that on GABA, glutamate, glutamine or asparagine, lack of either Gln3p or Gcn4p diminished UGA3 expression, indicating that both Gln3p and Gcn4p were needed to sustain UGA3 expression in all tested nitrogen sources; Gat1p/Nil1p had no effect. The effect of Gcn4p and Gln3p on UGA3 expression was confirmed by Northern analysis using total RNA samples extracted from the wild-type CLA strain and the pertinent mutant strains grown on GABA as sole nitrogen source (Fig. 6). To asses if GCN4 mRNA translation was increased in the presence of low-quality nitrogen sources like GABA, β-gal activity was determined in the CLA1 strain harbouring plasmid p180 (GCN4-lacZ CEN4 ARS1 URA3) bearing the translational GCN4–lacZ fusion (Valenzuela et al., 2001). It was found that β-gal activity fostered by plasmid p180 was equivalent in extracts prepared from ammonium or GABA (45 versus 50 nmol min−1 mg−1 respectively), showing that GCN4 mRNA was similarly translated under both nitrogenous conditions. Consequently, the fact that nitrogen-determined response of UGA3 and GLT1 was Gcn4p-dependent could not be attributed to increased GCN4 mRNA translation.
To confirm that GLT1 and UGA3 transcriptional activation was determined by the action of Gln3p and Gcn4p, a gln3Δgcn4Δ double mutant was transformed with plasmids pUGA3-1 and pGLT1-1, colonies were recovered and grown on GABA as sole nitrogen source. Extracts were prepared and β-gal activity was determined. A sixfold or 11-fold lower activity was detected for UGA3 or GLT1 as compared with that found in the wild-type strain (238 versus 44 and 1240 versus 105 nmol min−1 mg−1 respectively), indicating that Gcn4p and Gln3p co-ordinately determined expression of these genes. Furthermore, chromatin immunoprecipitation (ChIP) experiments, carried out on Gln3-myc13- and Gcn4-myc13-tagged strains, showed that Gln3p and Gcn4p acted upon the sequences located in the 132 bp cluster (Fig. 7).
The above presented results show that UGA3 and GLT1 expression is determined by the nature of the nitrogen source through the action of Gcn4p and Gln3p, which bind cis-acting elements present in the UGA3-GLT1 intergenic region.
Gcn4p and Gln3p determine UGA3 and GLT1 expression during amino acid deprivation in the presence of GABA as sole nitrogen source
Considering that Gcn4p and Gln3p determined UGA3 and GLT1 transcription and that it has been shown that GLT1 transcriptional activation is increased after 3-aminotriazole (3-AT) treatment (Valenzuela et al., 1998), we decided to analyse whether UGA3 expression was increased during amino acid deprivation. As Table 4 shows, the expression of UGA3 and GLT1 was twofold increased during amino acid deprivation in a Gcn4p-dependent manner, when ammonium was used as nitrogen source. Under these conditions, 3-AT-dependent transcriptional activation was also observed in a gln3Δ mutant, showing that UGA3 and GLT1 transcriptional activation in response to amino acid deprivation was exclusively Gcn4p-dependent. However, when GABA was used as nitrogen source, 3-AT response was not elicited on either a gcn4Δ or gln3Δ genetic background, indicating that on a low-quality nitrogen source like GABA, lack of either one of these activators hindered 3-AT-induced response (Table 4), these observations were confirmed by Northern analysis (Fig. 6). These results add to previous observations suggesting a physiological link between Gcn4p and Gln3p (Mitchell and Magasanik, 1984). In order to further analyse the role of Gln3p on 3-AT-mediated response, the wild-type strain carrying plasmids pUGA3-4 or pGLT1-4 bearing a mutation in the gatabox were grown on GABA and treated with 3-AT. It was found that these strains showed wild-type 3-AT response indicating that impairment of the Gln3p binding site did not affect response to amino acid deprivation (Table 4). Chromatin organization was determined after 3-AT treatment and it was found to be identical to that shown in Fig. 3 for GABA-grown cells, indicating that 3-AT-induced expression was not determined by a particular nucleosome organization (Supplementary material Fig. S2).
Table 4. β-Gal specific activitya of wild type, gcn4Δ and gln3Δ mutants grown on ammonium or GABA with or without 3-AT.
. Mean of three independent experiments. Variations were ≤15%.
Cells were grown on MM with ammonium or GABA as nitrogen source. 3-AT treatment is described in Experimental procedures. ND, not determined.
CLA 100–0 gcn4Δ/pGLT1-1
CLA 100–0 gcn4Δ/pUGA3-1
CLA 302–0 gln3Δ/pGLT1-1
CLA 302–0 gln3Δ/pUGA3-1
We have undertaken the analysis of the UGA3-GLT1 intergenic region, addressing two issues: (i) the combined role of Gln3p and Gcn4p transcriptional activators to co-ordinately support expression of either biosynthetic or catabolic genes and (ii) the role of chromatin organization on the bidirectional capacity of a promoter.
UGA3-GLT1 transcriptional response to the quality of the nitrogen source and to amino acid deprivation is determined by the combined action of both Gln3p and Gcn4p
When the yeast S. cerevisiae is grown in the presence of poor nitrogen sources, such as GABA, genes coding for the enzymes involved in the catabolism of these compounds are highly expressed. Conversely, in the presence of high-quality nitrogen sources, such as glutamine or asparagine, a decrease in the levels of catabolic enzymes is observed. This response is brought about through the action of a regulatory system known as nitrogen catabolite repression (NCR) (Coffman et al., 1996; Magasanik and Kaiser, 2002). NCR operation relies upon two transcriptional activators, Gln3p and Gat1p/Nil1p, each containing a GATA-binding zinc finger motif (Stanbrough et al., 1995). The positive cis-acting transcriptional element common to Gln3p, Gat1p/Nil1p-regulated genes, has been considered to be a bipartite UASNTR (gatabox) consisting of two separated dodecanucleotides each containing the sequence 5′-GATAA-3′ at their core (Rai et al., 1995). However, analysis of the nitrogen-dependent regulation of GLN1 and the proline dehydrogenase-encoding PUT1 gene showed that a single UASNTR site was able to combine with another unrelated cis-acting element to mediate transcription, which resulted in hybrid regulatory responses that did not correspond to a clear NCR-response. The nature of the trans-acting elements acting on the unrelated site was not analysed (Rai et al., 1995). Studies on GLT1 transcriptional regulation showed that Gln3p atypically activated GLT1 expression on glutamine-grown cultures; it was thus considered that Gln3p could act in combination with an unrelated cis-acting element (Valenzuela et al., 1998). Results presented in this paper further confirm this observation and show that as well as GLT1, UGA3 expression on glutamine and asparagine is Gln3p-dependent and that this response is elicited through the combined action of Gln3p and Gcn4p. This unveils the possibility that the non-related cis-acting element functioning in combination with the UASNTR present in the UGA3-GLT1 intergenic region is constituted by the Gcn4p-binding box, suggesting that Gln3p and Gcn4p synergistically determine UGA3 and GLT1 nitrogen-dependent expression. In this regard, the fact that mutation of either GLN3 or the gatabox did not result in a dramatic decrease of either GLT1 (Valenzuela et al., 1998) or UGA3 expression indicates that the nitrogen-elicited transcriptional response of GLT1 and UGA3 cannot be regarded as classical NCR, probably because of the fact that Gcn4p is also contributing to this response.
On GABA-grown cells, translation of GCN4 mRNA was not enhanced, showing that the effect of Gcn4p was not due to an increase of Gcn4p concentration. Furthermore, this fact suggests the possibility that on GABA-grown cells, when Gln3p is maximally nuclearly localized (Beck and Hall, 1999), this transcriptional activator could enhance or determine Gcn4p capacity to stimulate UGA3-GLT1 expression. This proposition is also supported by the fact that on GABA-grown cells, 3-AT-dependent transcriptional activation was determined by the simultaneous action of Gcn4p and Gln3p. In addition, the observation that mutation of the gatabox determining Gln3p-dependent transcriptional activation did not impair amino acid deprivation response argues in favour of the proposal that there exists a physiological interaction between Gln3p and Gcn4p, which allows increased UGA3 and GLT1 transcription on 3-AT-treated GABA cultures.
A physiological interaction between Gln3p and Gcn4p was first observed by Mitchell and Magasanik (1984). These authors found that GLN1 3-AT-Gcn4p-dependent expression was influenced by Gln3p, thus the 29-fold higher GLN1 expression found on 3-AT as compared with that found in its absence, was decreased to only fourfold in a gln3 genetic background. Our results show that lack of either Gcn4p or Gln3p completely abolished GLT1 and UGA3 amino acid deprivation response.
Our results thus suggest that Gln3p and Gcn4p could synergistically determine expression of genes whose promoters simultaneously bind these two global activators. In this regard, our results support the proposition that indicates that certain regulator pairs (co-ocurring regulators), which occur more frequently within the same promoter regions than would be expected by chance, could either interact physically or have related functions at multiple genes (Harbison et al., 2004). Most interesting was the fact that the Gln3p-Gcn4p pair was classified by Harbison et al. (2004) as a co-ocurring regulator pair.
Chromatin organization influences bidirectionality of the UGA3-GLT1 intergenic region
The fact that chromatin organization was not modified when yeasts were grown on either nitrogen repressive or non-repressive conditions indicated that UGA3 and GLT1 GABA-induced expression was not the result of a chromatin-dependent improved accessibility of the relevant cis-acting elements under these physiological conditions. In addition, the fact that mutations of the abf1box and polydAdTtract altered chromatin organization and GLT1 expression without hindering UGA3 expression indicated that chromatin remodelling played a role in the capacity of the UGA3-GLT1 intergenic region to promote transcription, which did not reside in the occlusion or exposure of the relevant cis-acting elements. As has been previously put forward (Iyer and Struhl, 1995; Beck and Hall, 1999), the abf1box and polydAdTtract could function as insulator elements establishing boundaries between regions of more compact or relaxed chromatin organization, which could influence the physical properties of flexibility or curvature that promote contact between the GLT1 tatabox and the relevant cis-acting elements (Parvin et al., 1995), and thus curtailing the distance between the activating sequences and the GLT1 tatabox. The RPS28A-GLN4 bidirectional promoter has a similar organization to that herein described for the UGA3-GLT1 promoter, bearing an Abf1p binding site and a T-rich element that act synergistically to create a nucleosome free-region (Lascaris et al., 2000). However, the question of whether these elements play a role in the bidirectional capacity of the RPS28A-GLN4 promoter was not addressed. Conversely, the GAL1-GAL10 and the GCY1-RIO1 regions have been shown to, respectively, bear a Reb1 site and a polydAdT site, which neither affected nucleosome positioning nor functioned as insulator elements (Angermayr and Bandlow, 1997; Reagan and Majors, 1998). Thus it could be considered that abf1box and polydAdT-like sites could play a role determining expression of certain bidirectional promoters.
Bidirectional promoters are a common feature of genome organization in eukaryotic organisms (Trinklein et al., 2004). The analysis of the UGA3-GLT1 promoter has put forward the possibility that bidirectional organization of genes whose products must respond to various environmental conditions could facilitate the interaction of trans-acting factors eliciting hybrid responses that more accurately could allow adaptation to a changing environment. Analysis of chromatin organization showed that it plays an important role in the bidirectional activity of the UGA3-GLT1 promoter, giving relevance to the fact that chromatin organization not only determines the accessibility of a DNA sequence to the transcriptional action of the activators but that it could also determine the intrinsic architecture of a given regulatory sequence, influencing its ability to promote interactions between the tatabox and the trans-acting element. It might be the case that an abf1box combined with a polydAdTtract could constitute a remodelling barrier similar to that herein described in those intergenic regions in which the cis-acting elements would be asymmetrically localized in the control region, as was the case for the UGA3-GLT1 promoter. However, this consideration will only be amenable for analysis when more bidirectional promoters are experimentally studied.
Strains and plasmids
The strains CLA1 (MATαura3 leu2), CLA-100-0 (MATαgcn4Δ::URA3 leu2), CLA-302-0 (MATαura3 leu2 gln3Δ::kanMX), CLA-303-0 (MATαgcn4Δ::URA3 leu2 gln3Δ::kanMX) and CLA-102-0 (MATαgat1Δ::URA3 leu2) (Valenzuela et al., 1998; Valenzuela et al., 2001) were transformed according to the method described by Ito et al. (1983), with either pUGA3-1 (UGA3-lacZ 2µLEU2) or pGLT1-1 (GLT1-lacZ 2µLEU2).
A collection of plasmids derived from pGLT1-1 (GLT1-lacZ 2µLEU2) (Valenzuela et al., 1998) or pUGA3-1 (UGA3-lacZ 2µLEU2) bearing various mutations affecting cis-acting elements were constructed as described in Tables 1 and 2. These plasmids were independently introduced in the wild-type strain CLA1 (Valenzuela et al., 1998). GLN3 and GCN4 were tagged in the carboxy-terminal end using the 13myc-kanMX6 modules described by Longtine et al. (1988). Two pairs of deoxyoligonucleotides (GCN4F2/GCN4R1 and GLN3F2/GLN3R1) were designed based on the GLN3 and GCN4 coding sequence and that of pFA6a-13myc-kanMX6 multiple cloning site (Table 5).
Table 5. Oligonucleotides used for protein tagging and PCR analysis of ChIP.
Lower case lettering indicates sequence of the multiple cloning site of plasmid pFA6a-13Myc-KanMX6.
Strains were grown on MM containing salts, trace elements and vitamins following the formula of yeast nitrogen base (Difco). Filter-sterilized glucose (2% w/v) was used as the carbon source and 0.2% (w/v) (NH4)2SO4 or 0.1% (w/v) glutamate, glutamine, asparagine, or γ-aminobutyrate (GABA) were used as nitrogen sources. Amino acids needed to satisfy auxotrophic requirements were added at 0.01% (w/v). Cells were incubated at 30°C with shaking (250 r.p.m.). For amino acid deprivation experiments, CLA1/pGLT1-1 or pUGA3-1, CLA100/pGLT1-1 or pUGA3-1 and CLA302/pGLT1-1 or pUGA3-1 were treated as previously described (Valenzuela et al., 1998).
Determination of β-gal activities
Soluble extracts were prepared by suspending whole cells in the corresponding extraction buffer (Cogoni et al., 1995) and grinding them with glass beads. β-gal activities were determined as previously described (Valenzuela et al., 1998). Specific activity was expressed as nmoles of ο-nitrophenol produced per minute per milligram of protein. Protein was measured by the method of Lowry et al. (1951), with bovine serum albumin as a standard.
Plasmid construction and polymerase chain reaction (PCR)-mediated mutagenesis
PCR-based methods were used to create deletion and substitution mutations in the UGA3-GLT1 promoter region. Primers MK1 and NQ13-2 (Table 1) were used to synthesize a fragment encompassing nucleotides +30 of UGA3 to +107 of GLT1 allowing in frame fusion of the GLT1 ATG to the lacZ gene of YEp363 plasmid (Myers et al., 1986). Primers MK2 and NQ14 (Table 1) were used to synthesize a fragment covering nucleotides from +12 UGA3 to +120 GLT1 allowing in frame fusion of UGA3 ATG to the lacZ gene of YEp363 plasmid. Site-directed mutagenesis of the tatabox, gcn4box, palbox, gatabox, palbox gatabox, abf1box, polydAdTtract and abf1box polydAdTtract were constructed and designed as previously reported (Ho et al., 1989; Sellers et al., 1990). Primer combinations used and the plasmids created are all presented in Tables 1 and 2. Expand High Fidelity PCR System polymerase (Roche) was used in all PCR. All PCR products were digested with BamHI and SalI and cloned into YEp363 (2 µm LEU2) (Myers et al., 1986). All fusion plasmids were sequenced, using as primers the deoxyoligonucleotides 1291 and 1163 (Table 1).
Pertinent S. cerevisiae strains were transformed with the lacZ fusion plasmids by the method described by Ito et al. (1983). Transformants were selected for leucine prototrophy on MM supplemented with auxotrophic requirements as needed.
Micrococcal nuclease (MNase) treatment
Cell cultures grown on MM with 0.1% GABA (200 ml at OD600 of 0.4) were pelleted and suspended in 10 ml of buffer 1 (1 M Sorbitol, 50 mM Tris-HCl, 10 mM β-mercaptoethanol) and incubated with 20 units of lyticase for 30 min at 30°C. The pelleted cells were washed with 20 ml of 1 M sorbitol and suspended in 5 ml of Nystatin buffer (1 M sorbitol, 20 mM Tris-HCl pH 8.0, 50 mM NaCl, 1.5 mM CaCl2 and 10% Nystatin); incubation was carried out for 5 min at room temperature (Venditti and Camilloni, 1994). Permeabilized cells were then subjected to micrococcal nuclease (MNase) digestion with increasing nuclease concentration (0–3 U ml−1) and the reaction was stopped with 1% SDS and 50 mM EDTA (final concentration). Proteinase K (40 µg per sample) was added and the samples were incubated at 56°C for 45 min. The DNA was then purified by three phenol-chloroform extractions, followed by an ethanol, 3 M NaOAc precipitation. The pellet was suspended in 200 µl of TE buffer and treated with RNAse (150 µg per sample). Naked DNA was prepared with the same treatment of nystatin-permeabilized spheroplasts, extracted and digested in vitro with MNase (0.01–0.1 U ml−1), purified by three phenol-chloroform extractions followed by an ethanol, 3 M NaOAc precipitation. RNAse treatment was also performed. All samples were suspended in 50 µl of distilled water and a 2 µl aliquot was electrophoresed on a 1% agarose-TAE 1× gel to determine the extent of digestion.
Spheroplasts prepared from CLA1 strain carrying multicopy (2µ) plasmids pGLT1-1, pGLT1-6, pGLT1-7 and pGLT1-8 were permeabilized with nystatin and treated with MNase. Purified DNA was digested with BamHI and SalI and subjected to indirect end-labelling using a probe flanking BamHI and SalI restriction sites. As control, naked DNA from pGLT1-1 was digested with MNase; a 50 bp ladder was used as molecular weight marker. After addition of loading buffer, DNA samples were electrophoresed on a 1.5% agarose-TBE 1× gel containing 0.1 µg ml−1 ethidium bromide at 30–40 V for 15–18 h at room temperature. DNA was transferred to Hybond-N paper by the alkaline Southern blotting method. The 172 bp BamHI-probe flanking the restriction site of the lacZ reporter gene was generated by PCR with primers MK1 and 1435 (Table 1). The 155 bp SalI probe flanking the restriction site of the lacZ reporter gene was generated by PCR with primers NQ13-2 and 845 (Table 1).
Primers 957 and 985 (Table 1) were 5′ end labelled (Venema, 1998) and purified using Quiaquick kit (Qiagen). Primer extensions on MNase digests of episomal (2µ) constructs were performed according to Zhu and Thiele (1996). Samples were precipitated and electrophoresed on a 6% polyacrilamide-urea gel. Dried gels were exposed to a Phosphor Imaging screen (Molecular Dynamics). For sequencing, a cycle-sequencing kit (Amersham) was used with 5′ end-labelled primers 957 and 985 (Table 1).
Formaldehyde-crosslinking and immunoprecipitations were carried out by the procedure described by Hecht et al. (1995). Yeast cells (200 ml of OD600 of 0.4) were cross-linked with 1% formaldehyde for 60 min at room temperature. After addition of 125 mM glycine and incubation for 5 min, cells were harvested and washed with saline Tris-buffer. Pellet cells were suspended in spheroplast buffer (1 M sorbitol, 50 mM Tris-Cl, 25 mM HEPES) with 600 U ml−1 of lyticase and incubated at 30°C for 30 min. Spheroplasts were harvested, washed and suspended in RIPA buffer (50 mM HEPES, 137 mM KCl, 1 mM EDTA, 0.1% Na deoxycholate, 0.5% Triton X-100, 0.1% SDS) with protease inhibitor cocktail (Complete Mini, Roche). Spheroplasts were sonicated to produce chromatin fragments of 200–1000 bp, with average size of ∼500 bp. ChIP was conducted with 1 µg of antic-myc antibody (9E 10, Santa Cruz Biotechnology). The primers sets used for PCR analysis 1227 and 1508 are listed in Table 5. PCR products corresponding to the region from −331 to −589, which contains 132 bp sequence harbouring the relevant cis-acting elements, were resolved on a 1.5% agarose gel stained with ethidium bromide.
Northern analysis was carried out as previously described by Gonzalez et al. (1992). Total RNA was prepared according to Struhl and Davis (1981) from 200 ml cultures of MM with GABA as nitrogen source, with or without 3-AT as described in growth conditions. Prehybridization and hybridization conditions were settled as previously described (Valenzuela et al., 2001). Three sets of deoxyoligonucleotides were used to PCR-amplify three fragments of 1574 bp (FwUga3/RvUga3), 781 bp (FwGlt1/RvGlt1) and 877 bp (FwAct1/RvAct1) that were used as probes for sequential hybridization of UGA3, GLTI and ACT1 (Table 1). Blots were scanned using the ImageQuant 5.2 (Molecular Dynamics) software.
This work was supported in part by the Dirección General de Asuntos del Personal Académico, Universidad Nacional Autónoma de México (IN221103-2), and by the Consejo Nacional de Ciencia y Tecnología (U40506Q). We acknowledge L. Alba-Lois, P. Ballario and J. M. Cantú for useful discussions and critical review of the manuscript. We are grateful to Sergio Zonszein and Benedetto Grimaldi for technical assistance and to A. Hinnebusch for kindly providing plasmid p180. We also acknowledge L. Ongay, G. Codiz and M. Mora (Unidad de Biología Molecular, Instituto de Fisiología Celular, UNAM) for DNA sequencing and synthesis of deoxyoligonucleotides. C.I. received a grant (register number 119309) from the Consejo Nacional de Ciencia y Tecnología.