|Home | About | Journals | Submit | Contact Us | Français|
We used genome-wide expression analysis in Saccharomyces cerevisiae to explore whether and how the expression of protein-coding, RNA polymerase (Pol) II-transcribed genes is influenced by a decrease in RNA Pol III-dependent transcription. The Pol II transcriptome was characterized in four thermosensitive, slow-growth mutants affected in different components of the RNA Pol III transcription machinery. Unexpectedly, we found only a modest correlation between altered expression of Pol II-transcribed genes and their proximity to class III genes, a result also confirmed by the analysis of single tRNA gene deletants. Instead, the transcriptome of all of the four mutants was characterized by increased expression of genes known to be under the control of the Gcn4p transcriptional activator. Indeed, GCN4 was found to be translationally induced in the mutants, and deleting the GCN4 gene eliminated the response. The Gcn4p-dependent expression changes did not require the Gcn2 protein kinase and could be specifically counteracted by an increased gene dosage of initiator tRNAMet. Initiator tRNAMet depletion thus triggers a GCN4-dependent reprogramming of genome expression in response to decreased Pol III transcription. Such an effect might represent a key element in the coordinated transcriptional response of yeast cells to environmental changes.
In eukaryotic organisms, three forms of RNA polymerase (Pol) are involved in nuclear DNA transcription (48). Pol I transcribes a single essential gene to generate the rRNA precursor that is processed into the three largest rRNAs. Pol II synthesizes pre-mRNAs and a number of small RNAs involved in pre-mRNA splicing and rRNA modification. Pol III is devoted to the synthesis of tRNAs, 5S rRNA and a few other nontranslated RNAs participating in basic cellular functions such as pre-mRNA processing, tRNA maturation, or protein translocation. The triplication of the transcriptional machinery may provide a selective advantage to the eukaryotic cell, compared to the unique bacterial and archaeal RNA polymerases, by allowing independent or coordinated regulations of gene expression in response to environmental changes. Previous studies both in yeast and in higher eukaryotes have shown that Pol III transcription is coordinately regulated with Pol I transcription under many circumstances, e.g., upon nitrogen starvation or in response to a nutritional upshift (6, 9, 53, 58). Under other conditions, the regulatory pathways of Pol I and Pol III transcription are uncoupled. Upon amino acid starvation, for example, Pol I transcription was reported to diminish by 80%, while tRNA synthesis was only modestly affected (50). A coordinated repression of rRNA, tRNA, and ribosomal protein synthesis was observed in secretion-defective yeast cells or in response to rapamycin treatment (34, 41, 42). It is not known, however, whether such a coordinate regulation involves any direct cross talk between the transcription systems. In particular, the extent to which defects in Pol III transcription affect Pol II transcription has not been studied on a genomic scale yet. This issue deserves detailed investigation for two main reasons. First, the cellular levels of Pol III-synthesized RNAs might in principle exert a wide influence on genome expression, as recently demonstrated for the SINE-encoded mouse B2 RNA (1). Second, local Pol III-Pol II transcriptional interference phenomena have been described in Saccharomyces cerevisiae. An actively transcribed tRNA gene lying close to the silent HMR locus has been shown to serve as a barrier to heterochromatin spread and to prevent the silencing of the adjacent GIT1 gene (15). Similarly, the TRT2 tRNAThr gene upstream of STE6 has recently been shown to act as a barrier to repression in MATα yeast cells (51). Actively transcribed tRNA genes can also negatively regulate adjacent Pol II-transcribed promoters (27, 31). This effect, also called tRNA gene-mediated silencing, has been observed mostly with selected artificial constructions, in which Pol II-transcribed reporter genes were found to be inhibited 2- to 60-fold by a neighboring tRNA gene (27, 31). Only two cases of negative tDNA position effects have been reported to operate at native chromosomal loci (4, 51). At this time, it is not known whether such a silencing effect operates on other Pol II-transcribed genes among those neighboring the 274 tRNA loci that are dispersed in the yeast genome, nor whether local silencing effects are limited to the tRNA class of Pol III-transcribed genes. Several features of the RNA Pol III machinery might indeed favor positional effects on neighboring Pol II-transcribed genes. The core transcription apparatus acting on yeast class III genes comprises the 17-subunit RNA polymerase together with its transcription factors TFIIIA, TFIIIB, and TFIIIC (20). TFIIIA, a multi-zinc finger protein, is a 5S RNA gene-specific assembly factor for TFIIIC. TFIIIC is composed of six subunits, named τ138, τ131, τ95, τ91, τ60, and τ55. TFIIIC recognizes the promoter sequences of Pol III-transcribed genes and, once bound to DNA, acts as an assembly factor for TFIIIB, which is responsible for directing Pol III to its target genes. TFIIIB is composed of the TATA box-binding protein, the TFIIB-related component Brf1, and the Bdp1 protein. Using chromatin immunoprecipitation followed by microarray hybridization, recent studies have defined a complete inventory of the genes transcribed by Pol III in the yeast Saccharomyces cerevisiae and have strengthened the notion that these genes are persistently occupied by the transcription machinery during active growth (24, 38, 46). Such a tight occupancy state of class III transcriptional units, that are generally unclustered and thus interspersed throughout the chromosomes, has the potential to exert a genome-wide influence on the Pol II transcriptome through local positional effects. To address the issues of both Pol III-Pol II cross talk and positional interference, we analyzed by DNA microarray hybridization the genome-wide consequences, on Pol II transcription, of defects in Pol III transcription associated with mutations in four components of the Pol III machinery. We found very limited alteration in the expression of Pol II-transcribed genes lying close to class III genes, as if Pol II-transcribed genes were generally refractory to transcriptional interference by neighboring transcription units. Instead, we consistently found, as a consequence of defective Pol III transcription, a derepression of genes under the control of the Gcn4 transcription factor, and we demonstrate that such a response depends on depletion of initiator methionine tRNA.
The Pol III transcription mutant strains used in the present study are listed in Table Table1.1. Well-characterized thermosensitive mutants were chosen. The rpc160-112 mutant strain (YPH500 background) is affected in the largest subunit of Pol III (6, 13). The brf1-II.6 strain harbors a double aspartate to alanine substitution in Brf1, one of the components of TFIIIB (2, 10). The tfc3-G349E strain (YPH500 background) corresponds to a point mutation in τ138, the largest subunit of TFIIIC, located in the τB domain and involved in DNA-binding activity (3, 33). The tfc7-ΔN1strain (YNN282 background) harbors a N-terminal deleted version of the τ55 subunit of TFIIIC (37). The four strains present a strong temperature-sensitive growth defect at 37°C. The tfc3-G349EΔgcn2 strain was constructed by a one-step replacement strategy (35) integrating kanMX in the GCN2 locus in the strain tfc3-G349E. The Δgcn2 control strain (AG207) was a gift from R. Serrano (22). The tfc3-G349EΔgcn4, brf1-II.6Δgcn4, and YPH500Δgcn4 strains were constructed by one-step gene replacement integrating kanMX in the GCN4 locus. To this end, a gene disruption cassette was PCR amplified from the Δgcn4 strain of the yeast gene knockout collection (Open Biosystems, Huntsville, AL) (55).
Strains transformed with p180 (pGCN4-LacZ URA3 CEN), a gift from A. Hinnebusch (39), were grown at permissive temperatures (24 or 30°C) to an A600 of 0.5 under repressive conditions in standard rich medium (yeast extract-peptone-dextrose [YPD]) or under derepressive conditions (growth in YPD then shift to minimal medium with no amino acids for 3 h). The cells were collected and protein extracts were prepared with a glass bead procedure, and the β-galactosidase activity in the whole-cell extracts was measured as previously described (54).
Overnight cultures of cells grown in YPD at the permissive temperatures (24 or 30°C) were diluted in fresh medium (A600 = 0.1) and incubated with shaking, either at the permissive temperature or at 37°C, to an A600 of 0.4 to 0.6. The cells were collected and total RNA (from 50-ml cultures) was extracted with the RNeasy Midi Kit (QIAGEN) according to the manufacturer's instructions.
The DNA microarrays used in the present study were manufactured by the Service de Génomique Fonctionnelle (CEA/Evry, France) as described by Fauchon et al. (17). Labeled cDNAs were synthesized by using an indirect labeling protocol adapted from P. Brown (http://cmgm.stanford.edu/pbrown/protocols/aadUTPCouplingProcedure.htm). Prehybridization and hybridization conditions were those described by P. Brown and coworkers (http://brownlab.stanford.edu/protocols.html). Hybridized arrays were scanned by using a GenePix 4000A scanner (Axon Instruments, Inc.), and fluorescence ratio measurements were determined with the GenePix Pro 4.0 software (Axon Instruments, Inc.). Spots were considered for analysis if more than 70% of the pixels fluoresced with intensity above the median fluorescence intensity of the background plus two standard deviations. The data were normalized assuming that the arithmetic median of the ratios for every considered spot is equal to 1. The data analysis was performed by using GeneSpring 6.0 software (Silicon Genetics) and Microsoft Excel. The results were averaged from at least three independent experiments with two batches of RNAs from wild-type or mutant strains. Each experiment consisted of two different hybridizations in which the fluorochromes were swapped in order to minimize changes due to technical variability. The complete datasets are available upon request.
Total RNA (10 μg) was treated with DNase I (as described in the RiboPure-Yeast protocol; Ambion), followed by acid phenol-chloroform extraction and RNA precipitation. First-strand cDNA for each sample was synthesized by reverse transcription of total RNA using Superscript III reverse transcriptase (Invitrogen). The reaction was carried out at 50°C for 120 min, followed by thermal inactivation of the reverse transcriptase (70°C for 10 min). The reaction mixture (20-μl final volume) contained 1× reverse transcription buffer, 0.5 mM deoxynucleoside triphosphates, 1 μl of oligo(dT)20 (50 mM), 10 mM dithiothreitol, 20 U of SUPERase-In (Ambion), 200 U of Superscript III, and 2 μg of total RNA.
Real-time PCR was performed with a TaqMan ABI Prism 7000 Sequence Detector System (PE Applied Biosystems) according to the manufacturer's instructions. Reaction mixtures (25-μl final volume) were assembled with the following components: 2.5 μl of 10-fold serial dilutions of cDNAs, optimized amounts of each primer set (100 nM for all amplifications), 2× SYBR Green PCR Master Mix (PE Applied Biosystems). The housekeeping β-actin (ACT1) mRNA served as an independent internal standard. All primer pairs produced only one amplification band (ranging from 95 to 110 bp) when tested by conventional reverse transcription-PCR. The primer sequences are available upon request.
The specificity of individual real-time PCR products was assessed by melting-curve analysis (45) carried out immediately after PCR completion. Melting curves for individual PCR products displayed a single peak. The amplified products were also examined by agarose gel electrophoresis. Melting temperatures (Tm) were determined with the Dissociation Curve software (PE Applied Biosystems). All sets of reactions were conducted in triplicate, and each included a nontemplate control. The threshold cycle (CT) was used to calculate relative gene expression levels using the “Comparative CT Method” (PE Applied Biosystems, user bulletin number 2), which is based on the assumption that the efficiency of amplification of target and reference cDNAs are approximately the same. The validity of the above assumption was verified for each primer set using serial cDNA dilutions.
The IMT1 and EMT3 genes, coding for initiator and elongator tRNAMet, respectively, were PCR-amplified (together with ~70 bp of 5′-flanking and 30 bp of 3′-flanking regions) from yeast genomic DNA with two pairs of oligonucleotide primers—5′-TCCAAGATGAGAATTTTAAGTTTATGG (IMT1_fw) and 5′-CATTTCATTCTATGTATTAACAATAATAG (IMT1_rev) for IMT1 and 5′-TGCTATATCCTTTAATAACCATGG (EMT3_fw) and 5′-ATCTCAAAACATGAATGTTGAGC (EMT3_rev) for EMT3—and inserted into the SmaI site of YEp352 multicopy vector. The recombinant constructs (pIMT1 and pEMT3) were then transformed into the BRF1 and the brf1-II.6 strains (see Table Table1).1). The transformants were selected on the appropriate synthetic medium (lacking uracil). Total RNAs from transformed and untransformed strains were extracted from cells grown in YPD medium at the permissive temperature (retention of the YEp352-derived plasmids by the transformed strains was verified by replating on selective medium).
DNA fragments of approximately 250 to 500 bp flanking each tDNA were amplified from genomic DNA by PCR and directionally cloned into pBluescript SK(+) to flank the URA3 gene. These tdnaΔ::URA3 constructs were transformed in the parent strains (S. cerevisiae W303 strain for AMD2, ACO1, YEL033W, and YJL200C, S288C strain for ARO8 and POR1) and plated on minimal medium lacking uracil. Ura+ transformants were screened by PCR for properly integrated constructs. To create strains containing specific deletions of the tDNAs, fragments of approximately 700 to 1,000 bp flanking the tDNAs were amplified by PCR and subcloned into pCR2.1-TOPO (Invitrogen). The tDNA sequences were deleted by site-directed mutagenesis (Stratagene QuikChange kit [oligonucleotide sequences used for deletions available on request]), and each tdnaΔ construct was transformed into its corresponding tdnaΔ::URA3 strain and plated on 5-fluoorotic acid (5-FOA) media. FOAr, uracil auxotrophic transformants were analyzed by PCR to verify the desired tdnaΔ loci.
To study the impact on Pol II-dependent gene expression of defects in Pol III transcription, we decided to focus on four well-characterized thermosensitive strains bearing mutations located in the C160 subunit of Pol III (rpc160-112 mutant ) in the Brf1 component of TFIIIB (brf1-II.6 mutant ) and in two different subunits of TFIIIC, τ138 (tfc3-G349E mutant ) and τ55 (tfc7-ΔN1 mutant ). The four strains presented a strong growth defect at 37°C. Even at the permissive temperature (24°C for the rpc160-112 and brf1-II.6 strains and 30°C for the tfc3-G349E and tfc7-ΔN1 strains), the mutant strains have a doubling time longer than that of the wild-type strains (150 to 190 min instead of 100 to 120 min), most likely reflecting defects in Pol III transcription. Indeed, we verified by real-time PCR amplification that, at the permissive temperature, the four mutants display a two- to threefold reduction of the steady-state levels of tRNALeu(CAA) or tRNAThr(AGT) compared to wild type (data not shown; see also reference 6). A newly discovered, unexpected phenotype of these mutants was their ability to grow at the restrictive temperatures when the medium was supplemented with 1 M sorbitol, an osmotic stabilizer (data not shown). The mutants also grew at the restrictive temperatures in the presence of 500 mM KCl, suggesting that the rescue of growth was due to an osmotic effect rather than to a specific chemical property of the solute. An osmoremedial growth defect is often correlated with a weakened cell wall structure and has been found to be associated with several mutations in components of the Pkc1 mitogen-activated protein (MAP) kinase pathway, which mediates maintenance of cell integrity in yeast (23, 25). The mutants in the Pol III machinery, especially tfc7-ΔN1, displayed other phenotypes also reported for mutants in the Pkc1 MAP kinase pathway, such as sensitivity to caffeine and to Calcofluor White (data not shown) (36). These observations made it clear that defects in Pol III-mediated transcription can influence cell physiology in unpredictable ways and further motivated us to analyze on a genome-wide scale, by DNA microarray technology, the transcriptional response of S. cerevisiae to defective class III gene transcription. Yeast cultures of the mutants and of the corresponding background strains were exponentially grown in rich media at the permissive temperature. RNAs were extracted and fluorescently labeled cDNAs, synthesized from wild-type and mutant sample RNA by using an indirect labeling protocol, were hybridized to DNA microarrays that bear 6,144 yeast open reading frames (ORFs) (17). At least three independent experiments (technical replicates) were performed with two different batches of RNA from each strain (biological replicates). After spot quantification and normalization, the data sets were analyzed by using GeneSpring 6.0 software (Silicon Genetics). Differentially expressed genes were defined as at least twofold up- or downregulated in the mutant compared to the wild type (datasets available upon request). According to this criterion, more than 4% of the analyzed genes displayed significant changes in expression levels in each mutant, thus indicating that a significant remodeling of genome expression is elicited to allow yeast cells to adapt to defects in class III gene transcription. In all cases, the percentage of upregulated genes was significantly higher than the percentage of downregulated genes. The rpc160-112, tfc3-G349E, tfc7-ΔN1, and brf1-II.6 mutants showed an at least threefold upregulation for 38, 78, 80, and 31 genes, respectively. Among all differentially expressed genes, only 2.5% increased their transcript levels by >10-fold, with a maximum of 30-fold. A large fraction of differentially expressed genes (including all downregulated genes) were specifically deregulated in some mutants and not in others. Only a small, but highly significant set of genes were coinduced by more than twofold in all of the four mutants (see below). We also conducted genome-wide expression studies after shift of the yeast cultures to the nonpermissive temperature (37°C). In this case, however, datasets were more difficult to interpret, likely due to the complicating effects of cell cycle arrest. Only data obtained at the permissive temperature were thus considered for the analyses described below.
The genes differentially expressed in each mutant were first categorized based on the FunCat annotation scheme, version 2.0 (47), available at the MIPS Web site (http://mips.gsf.de/projects/funcat). The number of genes in each category was expressed as a percentage of the total number of differentially expressed genes from each of the mutants. As shown in Fig. Fig.1,1, Pol III transcriptional defects bring about a significant enrichment in transcripts of genes whose products are involved in central metabolism, generation of energy, and cell rescue/defense and, to a lesser extent, in interactions with the cellular environment. This reshaping of functional categories associated with defects in Pol III transcription is very similar to the remodeling of genome expression reported in previous studies analyzing the global gene expression changes in response to environmental stress, including for example, changes in temperature, osmolarity or salinity (8, 19, 44, 57). A similar functional remodeling has also been reported in a study analyzing the global expression changes induced by defects in the cell wall construction (32). Since we observed that mutants in the Pol III transcriptional machinery could share growth phenotypes with mutants of the Pkc1-MAP kinase pathway (see above), we were interested in comparing our data with global analysis of gene expression changes constitutively induced by cell wall mutations, which have been shown to trigger three major transcriptional responses: the Pkc1-MAP kinase pathway, the global stress system mediated by Msn2/Msn4 and Hsf1 transcription factors, and the Ca2+/calcineurin-mediated pathway (32). A significant overlap was found between the gene expression changes described previously (32) and those observed in the tfc7-ΔN1 mutant of the Pol III transcriptional machinery. For example, 26 of the 79 genes coinduced (at least twofold) in five mutants of cell wall construction were also more than twofold upregulated in the tfc7-ΔN1 mutant. Nine (ASP3D, CWP1, ECM4, PIR3, PST1, SED1, SPI1, YGP1, and YNL158c) of the thirty-three genes implicated in cell wall biogenesis whose expression was increased at least twofold in response to cell wall mutations were also upregulated more than twofold in tfc7-ΔN1. These results may explain why the tfc7-ΔN1 strain was the most sensitive to low concentrations of Calcofluor White and caffeine, two cell wall-interfering drugs. In contrast, the expression of most of these genes was not significantly deregulated in the three other mutants of the Pol III transcriptional machinery. The statistical thresholds used in our data analyses did not allow the identification, in these three mutants, of genes whose deregulation could be correlated with the phenotypes they shared with mutants of the Pkc1-MAP kinase pathway, such as their osmotic remediability. The reasons for the peculiar deregulation profile of tfc7-ΔN1 are unknown, but they might be related to the complex roles of the τ55 subunit of TFIIIC (encoded by TFC7) in the interplay between Pol III transcription and cell metabolism (37).
To better understand the biological meaning of gene expression remodeling associated with Pol III transcriptional defects, we used the GO::TermFinder software (5) available at the SGD website (http://db.yeastgenome.org/cgi-bin/GO/goTermFinder) to search for significantly enriched gene ontology (GO) terms associated with the four datasets of induced genes. Table Table22 shows the results of such analysis. The listed GO terms are those considered to be significantly over-represented (P < 10−4) in the four lists of upregulated genes (the same analysis conducted on the four lists of downregulated genes did not reveal any significantly enriched GO term). The most evident feature emerging from Table Table22 is that a significant fraction of upregulated genes encodes proteins involved in amino acid metabolism. Importantly, such an enrichment is common to upregulated genes in all of the four mutants, and almost all of the genes that are coinduced in the four mutants (listed in Table Table3)3) fall into this functional category. Previous studies on a genomic scale (28, 40) have shown that the expression of most of these genes is under the control of the Gcn4p transcriptional activator. Starvation for amino acids, purine, and glucose limitation stimulate GCN4 mRNA translation and thus induce the synthesis of Gcn4p, which in turn controls the expression of at least 500 genes in yeast cells, including genes involved in amino acid biosynthetic pathways, cofactor biosynthesis, organelle biogenesis, mitochondrial transport, autophagy, and glycogen homeostasis (40). We found that more than 20% of all upregulated genes in each of the Pol III transcriptional machinery mutants corresponded to genes identified by Natarajan et al. (40) as being under the control of Gcn4p. Table Table33 lists the genes that are consistently upmodulated in the Pol III transcription mutants. Among them, 14 genes were induced more than twofold in all of the four mutants (ARG1, ARG3, ARG4, ARG5-6, CPA2, ECM40, HIS4, HIS5, HOM2, LEU4, MTG1, SNO1, SNZ1, and TMT1, all reported to be Gcn4p regulated). The cells used for microarray analyses were grown in rich medium, i.e., under conditions in which GCN4 mRNA translation is expected to be repressed (26). Since the levels of GCN4 mRNA were found to be substantially unchanged in the different Pol III mutants with respect to control strains, the activation of a large number of Gcn4p-responsive genes suggested that Pol III transcriptional defects result in constitutive derepression of GCN4 mRNA translation.
To further investigate this possibility, we first verified that the set of genes defined as induced by Gcn4p by Natarajan et al. (40) were also upregulated upon GCN4 derepression in our background strains. To do this, we transformed one of our control strains (YNN282) with a plasmid harboring a modified version of GCN4 that is expressed constitutively when the cells are grown in a rich medium [p238(GCN4c) plasmid] (16). The genome-wide analysis of the expression changes induced by Gcn4p in our background (data not shown but available upon request) revealed a significant overlap (~65%) with the analysis by Natarajan et al. (40) with respect to upregulated genes. Importantly, as shown in Table Table3,3, most of the genes found to be consistently upmodulated in the four Pol III transcription mutants were also responsive to constitutive GCN4 activation in our background strain. We next directly measured the translational derepression of GCN4 in the Pol III transcription mutants by using a β-galactosidase colorimetric assay. A GCN4-lacZ reporter plasmid, coding for a fusion transcript whose expression is under the translational control typical of GCN4 mRNA (39), was introduced into the wild-type strain or in the rpc160-112, tfc3-G349E, tfc7-ΔN1, and brfI-II.6 mutants. As shown in Table Table4,4, a 4- to 5.5-fold induction of GCN4-lacZ expression, under nonstarvation conditions, was measured in the Pol III machinery mutants with respect to the wild type, indicating that GCN4 translational induction was constitutively operating in the mutants. The derepression of GCN4 in the mutants under nonstarvation conditions was, however, not at its maximum level since a further ~2-fold increase of GCN4-lacZ expression was observed when the cells were starved for amino acids for 3 h (Table (Table4).4). To conclusively prove that the induction of GCN4 target genes in the Pol III mutants is dependent on Gcn4p, we sought to determine whether deleting GCN4 in the mutants could eliminate the response. The GCN4 gene was deleted in the tfc3-G349E and brf1-II.6 mutant strains and in the YPH500 control strain, and the levels of several mRNAs were measured by real-time PCR. As shown in Table Table5,5, the pronounced activation observed in the mutant strains for two GCN4 target genes, HIS4 and ARG1 genes, was completely abolished upon deletion of GCN4, thus demonstrating that GCN4 is necessary and sufficient for their upmodulation in the Pol III transcription mutants. A different behavior was observed for PIC2, a gene that we found only weakly (1.8-fold) upregulated in the presence of constitutive GCN4 activation in the YNN282 strain (the same gene is completely unresponsive to GCN4 in the analysis by Natarajan et al. ). PIC2 was 2.2- and 2.9-fold upmodulated in the tfc3-G349E and brf1-II.6 mutant strains, respectively. As shown in Table Table5,5, GCN4 deletion did not affect PIC2 activation in the brf1-II.6 mutant, whereas it abolished the response in the tfc3-G349E context, thus suggesting that PIC2 upmodulation depends on different pathways in the two mutants. The modulation of AVO2, another gene that is not responsive to constitutive GCN4 activation, also displays a complex, mutant-specific pattern. This gene is clearly upmodulated in the brf1-II.6 mutant, whereas its expression is unchanged in the tfc3-G349E strain. Unexpectedly, the deletion of GCN4 abolishes the response in brf1-II.6. Such a behavior might be explained by assuming that GCN4 is necessary but not sufficient for AVO2 activation in brf1-II.6. Reported in Table Table55 are also the consequences of GCN4 deletion on the response of three genes that are downmodulated in the Pol III mutants: PHO11, HXT2, and GAL1. PHO11 and HXT2 are unresponsive to constitutive GCN4 activation and, accordingly, their downmodulation in the mutants was left unchanged by GCN4 deletion. A different behavior was observed for GAL1. On the basis of our analysis and that of Natarajan et al. (40), GAL1 is also unresponsive to constitutive GCN4 activation. However, its marked downmodulation in the two Pol III mutants was abolished upon deletion of GCN4. Again, this result might be explained by assuming that GAL1 repression in the mutants depends on the action of both GCN4 and another regulator(s).
We finally wondered whether GCN4 constitutive derepression is a general feature of the mutants in the Pol III transcriptional machinery. To this purpose, GCN4-lacZ expression was measured in six other thermosensitive mutants grown under nonstarvation conditions at the permissive temperature, and it was found to be increased in all mutants with the exception of the tfc1-E447K mutant (29) and a brf1-HA-tagged strain (24) (data not shown).
Previous studies showed that GCN4 expression is stimulated under conditions producing a decrease in the levels of the ternary complex composed of eIF2, GTP, and Met- (26). In response to amino acid or purine starvation, the reduced formation of the ternary complex is controlled by the Gcn2 protein kinase. There are also several examples where GCN4 is derepressed in a manner dependent on the reduced levels of the ternary complex but independent of Gcn2 (26, 43). To test the GCN2-dependence of GCN4 derepression in the Pol III transcription mutants, the GCN2 gene was deleted in tfc3-G349E strain, and GCN4-lacZ expression was measured in tfc3-G349E and tfc3-G349EΔgcn2 strains and a Δgcn2 control strain (22). As shown in Table Table4,4, similar low levels of β-galactosidase activity were obtained when the Δgcn2 cells were grown under repressing or derepressing conditions, as expected. The ~4-fold derepression of GCN4-lacZ in tfc3-G349E versus wild-type cells grown in rich medium was also observed in tfc3-G349EΔgcn2 cells, thus indicating that GCN4 derepression in the Pol III transcription mutant is independent of GCN2. In contrast, the 2.5-fold increase in GCN4-lacZ expression observed in the mutant upon amino acid starvation is GCN2-dependent.
A reduced gene dosage of initiator tRNAMet, or defects in its maturation, have previously been shown to increase Gcn4p expression by reducing the concentration of eIF2 · GTP · Met-ternary complex (7, 11). By real-time PCR analysis, the steady-state levels of were found to be ~2-fold reduced in the four Pol III mutant strains with respect to control strains at the permissive temperature (data not shown). To see whether increased GCN4 expression under conditions of defective Pol III transcription could be due to such a reduction of levels, the brf1-II.6 mutant strain was transformed with multicopy plasmids carrying either the IMT1 gene, coding for the initiator tRNAMet, or the EMT3 gene, coding for the elongator tRNAMet. We verified by both real-time PCR and Northern analysis that overexpression of IMT1 in the mutant restored levels of initiator tRNAMet that were ~1.5-fold higher than in the wild type; we also observed that the growth defect of brf1-II.6 mutant was partially corrected by increased IMT1, but not EMT3, gene dosage (data not shown). The expression levels of ARG1 and HIS4, two Gcn4p-responsive genes whose mRNAs were found to be increased ~5- and ~2-fold, respectively, in brf1-II.6 by microarray analysis, were analyzed in nontransformed and transformed brf1-II.6 and wt cells by real-time PCR. As shown in Table Table6,6, the upregulation of both ARG1 and HIS4 genes typically observed in brf1-II.6 was abolished by increased IMT1 gene dosage, whereas it was unaffected (or even slightly exacerbated) by increased EMT3 dosage. Initiator tRNAMet depletion is thus specifically responsible for the Gcn4p-dependent transcriptional response in the brf1-II.6 mutant.
It has previously been proposed that the expression of a number of chromosomal, Pol II-transcribed genes might be influenced by tDNA position effects (4). Based on computational evidence, such effects would be mainly inhibitory, and two cases have been reported of chromosomal genes whose expression is slightly increased when the neighboring tRNA genes are transcriptionally inactivated or deleted (4, 51). Cases of tRNA genes with a protective function against transcriptional repression have also been reported (15, 51). To establish how widespread are the tDNA proximity effects on Pol II-directed transcription in the yeast genome, we compared the average expression changes, at the permissive temperature, of the subset of Pol II-transcribed genes that are nearest neighbors of tDNAs with those of the other genes (the complete data set for modulation of the tDNA nearest-neighbor ORFs is available in the supplemental material). We considered as significantly modulated the genes whose expression varied at least twofold in the mutant with respect to control strains. As shown in Fig. Fig.2,2, the tDNA-flanked genes appear to be slightly more modulated than the other genes in the tfc7-DN1 and brf1-II.6 mutants, but not in the case of the tfc3-G349E and rpc160-112 mutants. The data in Fig. Fig.22 also take into account the distance between the tDNA and its nearest neighbor, since it is not known how far tDNA positioneffects may reach. In the case of brf1-II.6, the tDNA-flanked genes located at less than 1,000 bp from the tDNA were significantly more deregulated than those located at larger distances. The situation was reversed in the case of the two TFIIIC mutants, for which the genes located at more that 1,000 bp from the tDNA were more deregulated. The modulation of tDNA neighbor genes was also separately analyzed for the subset of genes that are not regulated by GCN4. The results of such analysis were almost identical to those in Fig. Fig.2.2. These results argue in favor of the existence of some transcriptional interference between tDNAs and neighboring Pol II-transcribed genes (see Discussion). However, it is also evident that the expression of most (>90%) of the tDNA-flanked genes was not affected significantly in the Pol III transcriptional mutants. To gain more insight into tDNA position effects, we thus decided to undertake a tDNA deletion approach by focusing on a small number of loci in which the expression of a tDNA-flanked gene was found to be consistently modulated in at least three of four Pol III transcriptional mutants. Of the ORFs considered for this analysis (Fig. (Fig.3),3), five were upmodulated >1.5-fold in at least three of the four mutants, whereas only one, YEL033W, was consistently downmodulated in the mutants. The corresponding tDNAs were removed by a two-step homologous recombination strategy, first replacing each tDNA with URA3 and then transforming the tdnaΔ::URA3 strains with genomic fragments lacking the tDNA and selecting them on 5-FOA media. As shown in Fig. Fig.3,3, only two of the five genes upregulated in the Pol III mutant microarray analysis showed a consistent increase in transcript level by Northern blot analysis upon deletion of the adjacent tDNA (ACO1 and ARO8). The observed upregulation was modest, however, at best a 1.4-fold increase upon deletion of the tDNA. The other three genes (AMD2, POR1, and YJL200C) showed no or inconsistent changes in transcript levels upon tDNA deletion compared to their respective parent strains. Interestingly, deletion of the tDNA immediately downstream of YEL033W, a downregulated gene in the microarray analysis, resulted in a significant downregulation of YEL033W transcript level. Both the weak upregulation of ACO1 and ARO8 and the downregulation of YEL033W upon deletion of the nearest tDNA were not a result of GCN4 derepression, possibly triggered by tDNA deletion, because the expression of the GCN4-responsive ARG1 and HIS4 genes was not affected in the tDNA deletants, as revealed by Northern blot analysis (data not shown). A conclusion of this analysis is that the upmodulation of ACO1, AMD2, ARO8, POR1, and YJL200C, observed in the Pol III transcription mutants, cannot be accounted for by a tDNA proximity effect (upmodulation ACO1 and YJL200C might instead be attributed to the GCN4 response, because both of these genes were more than twofold activated in response to constitutive GCN4 derepression).
The interplay between the three nuclear RNA polymerase systems in eukaryotes is presumably a key aspect of growth control, yet it remains largely uncharacterized. In the present study, we have addressed this issue by analyzing, on a genome-wide scale, the influence of RNA Pol III-directed transcription on the expression of RNA Pol II-transcribed genes in S. cerevisiae. The problem of interplay between Pol III and Pol II systems is twofold. First, class III genes are characterized by an extremely high rate of transcription and transcription factor occupancy during active cell growth (14, 18, 24), a feature that might favor repressive or antirepressive position effects on neighboring RNA Pol II promoters. Second, Pol III synthesizes very abundant, essential RNAs whose levels might influence in a complex way the expression pattern of protein-coding genes. The results of our study provide insight into both of these aspects of Pol III-Pol II interplay. As for the positional interference exerted by class III genes on neighboring Pol II promoters, we have shown that it operates only to a limited extent and thus probably does not represent a key feature of yeast genome expression. As for the effects on the Pol II transcriptome of reduced levels of class III gene products, we have revealed an extensive, GCN4-mediated reprogramming of yeast genome expression that entirely depends on initiator tRNAMet depletion.
The general lack of positional effects exerted by tDNAs and the other class III genes was somehow unexpected, because actively transcribed class III genes have been shown to strongly repress adjacent Pol II promoters in some cases (27, 30, 31) and to protect neighboring genes from the propagation of repressive chromatin in other cases (15, 51). Although information on tRNA gene-mediated repression mostly comes from studies in which tRNA genes and Pol II reporter genes were artificially juxtaposed in plasmid constructs, repressive tDNA positional effects have also been reported to operate at native chromosomal loci (4, 27, 51). The general lack of such effects emerging from our microarray data might be due to the fact that the transcriptional defects associated with the Pol III system mutations are not sufficiently strong to cause a loss of the positional effect. Transcriptional interference might require nothing more than the presence of TFIIIB and/or TFIIIC on class III genes, regardless of their actual transcription. In the Pol III catalytic mutant rpc160-112, for example, TFIIIC and TFIIIB are likely to be assembled on tDNAs as they are in the wild type and thus to produce the same level of transcriptional interference. On the other hand, tDNA occupancy by TFIIIC and/or TFIIIB is expected to be decreased in the other three mutants with respect to wild type. In these mutants, however, reduced transcription complex assembly might not be sufficient to impair transcriptional interference. These considerations prompted us to test the Pol III-Pol II interference at selected loci by a tDNA deletion approach. With the interesting exception of YEL033w, whose expression was clearly reduced upon deletion of the adjacent S(AGA)E tDNA, the results of this analysis confirmed the conclusion that positional interference produces small magnitude effects and probably operates on a very small set of Pol II-transcribed genes. This in turn implies that the majority of Pol II transcription units in S. cerevisiae are well insulated and thus refractory to in cis perturbation by nearby tDNAs. The observation that, in the brf1-II.6 mutant, the tDNA-neighbor genes closer to tDNAs were more deregulated than those located at larger distances argues in favor of the presence of tDNA position effect at the subset of loci very close to tRNA genes, in agreement with the conclusion of a previous bioinformatic analysis of the tRNA gene loci database (4). The reversal of the distance dependence of modulation in the two TFIIIC mutants, however, suggests the existence of more subtle and less predictable effects of tDNA transcription on neighboring genes, perhaps related to the extensive presence of Ty elements around tRNA genes. Given this complexity, we believe that the systematic deletion of all yeast tDNAs will be required in order to conclude about the extent of tDNA position effects in yeast.
Based on transcriptomic data, the GCN4-dependent transcriptional response is the major consequence of defects in Pol III transcription. The activation of several Gcn4p-regulated genes was invariably observed in all of the Pol III mutants analyzed, thus suggesting that a reduced level of Pol III transcription products is the primary intracellular signal eliciting such a response. The molecular pathway leading from Pol III transcript depletion to GCN4 derepression turned out to be unexpectedly straightforward. An increase in initiator tRNAMet gene copy number was indeed found to be sufficient to abolish GCN4 derepression in the Pol III mutant context, thus indicating that depletion is the main (if not the sole) factor responsible for eliciting the Gcn4p-mediated response. It has previously been shown that GCN4 expression can be translationally induced by reducing the tRNAiMet gene dosage, in agreement with a model inversely coupling GCN4 translation to the level of eIF2 · GTP · Met-ternary complex (11). What our data add to this picture is that a reduction in ternary complex levels can occur as a direct (and perhaps, for cell physiology, the most important) consequence of a decrease in Pol III transcription. Interestingly, a genome expression remodeling reminiscent of the general amino acid control response has recently been observed in response to treatment of yeast cells with a small molecule inhibitor (UK-118005) of RNA Pol III (56). Such a response was also found to be GCN2 independent; however, in disagreement with our observations, GCN4 mRNA translation was not found to be increased in UK-118005-treated cells. The reasons for such a discrepancy are presently unclear but might include side effects of the drug on other cellular processes.
Recent genome-wide studies have shown that Gcn4p induces an unexpectedly large set of genes, encompassing as much as 1/10 or more of the yeast genome, and that it does so in response not only to amino acid deprivation but also to purine and glucose starvation, high salinity, and treatment with UV, methyl methanesulfonate, and rapamycin (reviewed in reference 26). Gcn4p thus appears to be a key regulatory protein in the global remodeling of RNA Pol II-dependent genome expression elicited by diverse signals of starvation and stress. Several environmentally stressful conditions, such as nutrient limitation, secretion defects, and treatment with DNA-damaging agents, are known to severely downregulate RNA Pol III-dependent transcription in yeast (9, 12, 21, 34, 49, 52, 58). The reduction in the level of initiator tRNAMet resulting from such perturbations might contribute to GCN4 translational activation and the consequent, genome-wide reprogramming of RNA Pol II-dependent transcription. Together with the still largely unknown molecular strategies allowing for coregulation of Pol I and Pol III transcription, the -mediated cross talk between Pol III and Pol II transcription might represent a key element of the coordinated regulation of the three eukaryotic transcription systems in response to environmental changes.
The involvement of other regulatory pathways, besides GCN4 activation, in the genome-wide response to defects in the Pol III machinery is made evident by the high number of genes that are significantly (at least twofold) up- or downregulated in the mutants but that are not responsive to constitutive GCN4 activation. These genes represent ~80% of all modulated genes in each of the mutant strains. It is important to note, however, that the up- or downregulation of most of these genes was not found to be conserved in the different mutants. As already mentioned, the general remodeling of genome expression in the Pol III mutants, illustrated in Fig. Fig.1,1, is reminiscent of the previously described common environmental response (8) and environmental stress response (19). For each of the mutants, however, a search for significantly enriched GO terms associated with the sets of modulated, GCN4-independent genes did not produce any reliable result. The complexity and heterogeneity of responses in the different mutants is probably related to the wide impact exerted on cell physiology by a reduced rate of protein synthesis (consequent to reduced levels of class III gene products) and to the fact that the different mutations in the Pol III system might affect the rate of protein synthesis to different extents.
We are grateful to Michel Werner for support and critical comments on the manuscript. We thank A. Hinnebusch, Steven Hahn, and Ramon Serrano for plasmids and strains; Véronique Bordas-Le Floch and Olivier Harismendy for help with microarray analysis; Tauro Maria Neri for the use of real-time PCR instrumentation; and Milena Preti for help with plasmid construction.
This study was supported by a grant from the Human Frontier Science Program Organization (to G.D. and D.D.); by the Italian Ministry of Education, University and Research; by the National Science Foundation (MCB-0342113 to D.D.); and by the Association pour la Recherche Contre le Cancer (Villejuif, France).
†Supplemental material for this article may be found at http://mcb.asm.org/.