Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/cfg.254 Conference Paper

We have compared Saccharomyces cerevisiae global gene expression in wild-type and mutants (Δhap2 and Δhap4) of the HAP transcriptional complex, which has been shown to be necessary for growth on respiratory substrates. Several hundred ORFs are under positive or negative control of this complex and we analyse here in detail the effect of HAP on mitochondria. We found that most of the genes upregulated in the wild-type strain were involved in organelle functions, but practically none of the downregulated ones. Nuclear genes encoding the different subunits of the respiratory chain complexes figure in the genes more expressed in the wild-type than in the mutants, as expected, but in this group we also found key components of the mitochondrial translation apparatus. This control of mitochondrial translation may be one of the means of coordinating mitochondrial and nuclear gene expression in elaborating the respiratory chain. In addition, HAP controls the nuclear genes involved in several other mitochondrial processes (import, mitochondrial division) that define the metabolic state of the cell, but not mitochondrial DNA replication and transcription. In most cases, a putative CCAAT-binding site is present upstream of the ORF, while in others no such sites are present, suggesting the control to be indirect. The large number of genes regulated by the HAP complex, as well as the fact that HAP also regulates some putative transcriptional activators of unknown function, place this complex at a hierarchically high position in the global transcriptional regulation of the cell.


Introduction
The yeast Saccharomyces cerevisiae has a predominantly fermentative metabolism. When grown on media containing glucose as carbon source, yeast cells repress their respiratory metabolism up to the point where all glucose has been consumed, leaving only ethanol as carbon source. In order to use ethanol, the cell has to reprogram its metabolism, a phase called 'diauxic shift'. Previous data have shown that the greater part of this reprogramming is under the control of the HAP complex.
The HAP complex is a heteromeric transcriptional regulator composed of four proteins. Hap2p, Hap3p and Hap5p associate to form the DNAbinding part, while Hap4p contains the activation domain (review in Forsburg and Guarente, 1989a). While HAP2, HAP3, HAP5 are constitutively expressed, transcription of HAP4 is induced during the diauxic shift (De Risi et al., 1997). The

S. Buschlen et al.
HAP complex was originally identified as upregulating the expression of cytochrome c (Forsburg and Guarente, 1989a) and later on, of several genes encoding TCA cycle and respiratory chain enzymes (for review, see De Winde and Grivell, 1993).
During the global transcriptome analysis of yeast genes in relation to growth phase on glucose, De Risi et al. (1997) clustered genes according to their expression pattern and found that nuclear-encoded genes playing a role in respiratory chain function and TCA cycle were also transcriptionally induced at the diauxic shift, leading to the hypothesis that HAP4 was responsible for this induction. The fact that HAP4 plays a role in programming a respiratory metabolism is further confirmed by its metabolic effect when overexpressed (Blom et al., 2000) and by its induction when yeast cells are placed in conditions of adaptative evolution under glucose-limiting conditions (Ferea et al., 1999).
Despite these convergent observations, no precise identification of HAP4 target genes has yet been made. We decided to address this question by comparing the transcriptome of three strains (a wild-type strain, a hap2 deletion and a hap4 deletion) during growth on galactose medium. Since the hap2 and hap4 deletion mutants are not able to grow on non-fermentable carbon sources, their effects on gene expression could only be examined on fermentable carbon sources. Galactose was chosen as carbon source to avoid glucose repression, which would have flattened out any possible effects. We report here the systematic analysis of genes related to mitochondrial function and biogenesis and show that the HAP complex positively controls the expression of genes involved in mitochondrial respiration and mitochondrial translation, thus allowing the coordinated expression of both genetic compartments.

RNA isolation
RNA extraction was done mechanically in liquid nitrogen as described by Hauser et al. (1998) from cultures grown in complete galactose medium [2.3% BactoAgar, 1% BactoPeptone, 1% Yeast Extract (Difco) and 2% Galactose]. Yeast cells (200 ml) grown to OD 600 0.8-1.0 were pelleted. The pellet was then stirred a few seconds on a vortex and released as individual drops directly into liquid nitrogen, kept in the small Teflon vessel of a micro-dismembrator. After adding a 7 mm tungsten carbide bead, frozen drops were mechanically broken for 2 min at 3000 rpm. The frozen powder was instantly taken up in Trizol reagent (Invitrogen) and treated as instructed.
Three independent RNA preparations and labelling were made from three different cultures, for each of the three strains (WT, hap2 and hap4) in parallel.

Probe generation and GeneChip hybridization
Experimental procedures for the GeneChip were performed as recommended by Affymetrix. Briefly, mRNA was isolated from total RNA with the Oligotex mRNA kit, as outlined by the manufacturer (Qiagen). Double-stranded cDNA was synthesized from mRNA with the SuperScript Choice system (Life Technologies) and a T7-(dT) 24 (GENSET) primer. In vitro transcription was performed on the cDNA to produce biotin-labelled cRNA with an Enzo Transcription Kit (Enzo). The biotinylated cRNA was cleaned with an RNeasy Mini Kit (Qiagen), fragmented to 50-200 nucleotide lengths, and then hybridized to Affymetrix YGS98 arrays. The arrays were then processed on the Affymetrix fluidics station and scanned on an HP GeneArray scanner.

Data acquisition and analysis
Acquisition and quantification of array images were performed with Affymetrix Microarray Suite 4.0. RNA preparations from the wild-type and the deleted strains hap2 and hap4 were done three times, on different days and therefore from different cultures. All the data were normalized at the 75th percentile. The expression ratios were obtained using both normalized intensity and noise data through the PFOLD algorithm (Theilhaber et al., 2002). Briefly, components of noise in the Array analysis of yeast HAP complex 39 intensities are identified and assumed to have normal distributions for both scans. An analytical expression is derived for the probability distribution function of intensities, using Bayes' theorem and the above assumptions about noise distribution. A joint probability distribution (JPD) of the fold change is derived using the above distributions. The JPD is used to calculate the median, confidence limits and p values for the estimated fold-change. The fold change of a particular gene corresponds to the median of the fold changes deduced from the three biological replicates.

Results and discussion
Transcriptome analysis reveals a large number of genes whose expression is dependent upon the presence of the HAP complex The data point to the conclusion that the expression of several hundred genes is controlled directly or indirectly by the HAP complex. This large number of targets places this complex as an important and hierarchically highly-placed transactivator. About 230 genes are positively regulated (higher expression in the wild-type than in the hap2 and hap4 mutants), while about 240 seem to be negatively regulated. A closer examination, focused on respiratory metabolism, identifies the great majority of those genes upregulated in the wild-type as being directly involved in mitochondrial function and TCA cycle. Among the negatively regulated genes, the situation is quite different, since we found only two nuclear-encoded mitochondrial genes downregulated (negative effect of the Hap2p/Hap4p proteins), which are isoforms of positively-regulated ones. We will focus in this report only on those genes involved in mitochondrial function and biogenesis.
The HAP complex controls the expression of genes coding for all the respiratory chain complexes and the enzymes of the TCA cycle When De Risi et al. (1997) examined yeast global expression in relation to the growth phase, they observed around the diauxic shift an increase in expression of genes encoding enzymes of the TCA cycle and of the pathways that fuel substrates to it. This increase could not be directly attributed to HAP4 as a general rule but this was strongly suspected, since HAP4 itself was induced and since we and others have previously shown some of these genes to be regulated by the HAP complex (De Winde and Grivell, 1993;Dang et al., 1994). In the experiments reported here, we found all the genes identified by De Risi et al. (1997) to be upregulated in the presence of the HAP complex (they showed two-fold higher expression in the wild-type than in the hap2 and hap4 mutants). These were CIT2, CIT1, ACO1, MDH2, MLS1, ICL1, IDP2, MDH1, FUM1, SDH2, SDH3, SDH4, KGD1, KGD2, LPD1, IDH1, as well as PCK1, PYC1, PYC2, ALD2 and ACS1 (Table 2, partA). It was expected that IDH2 would be regulated, but we did not observe a significant effect of HAP. Another contradiction was the differential expression (ratio, hap2 : wt = 0.5, where wt = wildtype) observed for IDP1, which was described as not carbon source-responsive by Haselbeck and McAlister-Henn (1993) and was not regulated in the De Risi et al. (1997) experiment. Except for these two cases, which have to be verified by other methods, it is now clear that the HAP complex controls the complete TCA cycle and related pathways. A coordinated increase in expression of about twoor three-fold in the wild-type as compared to the mutants was also observed for nuclear genes that code for the different components of the five respiratory chain complexes in yeast and some proteins (not all) involved in their assembly (Table 1). This co-response of different elements of the same functional families reinforces the interpretation that it reveals a mechanism allowing a good coordination of all transcripts when necessary. Related to this observation, two genes involved in heme biosynthesis, HEM1 and HEM2, which had already been described as being controlled by HAP2 (Keng et al, 1992;Keng and Guarente, 1987), were also found to be upregulated by a factor of 2.
Table 1 also shows that in statistically significant cases the regulatory effect is the same in both hap2 and hap4 mutants. In cases where isoenzymes exist, the HAP complex can exert different effects on their gene expression, probably in relation to the metabolic needs of the cell, e.g. the HAP genes were first isolated as a regulator of CYC1 (iso1-cytochrome c), which is indeed upregulated in the wild-type strain in our transcriptome data. Conversely, CYC7 (iso2-cytochrome c) is found Table 1. Effect of the HAP2 and HAP4 deletion on the expression of nuclear genes encoding components and associated proteins of the five respiratory chain complexes of S. cerevisiae. In the column 'CCAAT site?'; 'yes' indicates the presence of a CCAAT sequence defined arbitrarily between −150 and −800 from the ATG and whatever the orientation. In the delta hap2/wt and delta hap4/wt columns, the ratio indicated is the median of the expression in the mutant divided by the median of the expression in the wild-type, these median coming from three independent experiments, made with different RNA preparations. SD, standard deviation. We considered data as significant only when p < 0.001. If this was not the case, the value is provided but is followed by an asterisk to be downregulated (2.3-fold less expression in the wild-type than in the mutants), as would be expected for a gene which is expressed in anaerobiosis. MDH1 and MDH2, encoding respectively the cytoplasmic and mitochondrial forms of malate dehydrogenases, are both regulated by HAP (ratio of 0.4 and 0.5, respectively) while MDH3, the peroxisomal isoform, does not seem to be regulated (0.8), an observation perhaps reflecting the need to co-regulate the two elements of the redox-malate shuttle. Transcriptome experiments do not tell us whether the regulatory effect is direct (binding of the transactivator upstream of the gene) or indirect (through a cascade). A 'CCAAT' binding site (arbitrarily chosen between −150 and −800 from the translation start) is present upstream of practically all the genes described in this table, but this does not necessary mean that this site is functional.

Mitochondrial gene expression also seems to vary in function of the HAP complex
The Affymetrix yeast DNA chips also contain oligomers from mitochondrially-encoded elements of the respiratory chain. We found some of them to be also upregulated by a factor of 2-3. Two facts argue against a direct control of HAP on the mitochondrial genes. Transactivators associated to pol II are located in the nucleus and can be translocated into the cytoplasm but there is no evidence so far for mitochondrial localization. Only one transcription factor, XTC1, has a dual localization, but its mitochondrial function is different (Traven et al., 2002). Furthermore, direct mitochondrial transcriptional regulation has not been observed in yeast, while information pointing to translational control does exist, particularly  Table 1. Only data with the hap2/wt ratio are provided, since no statistical differences were observed in hap4/wt as compared to hap2/wt. SD, standard deviation. We considered the data as significant only when p < 0.001. If this was not the case, the value is provided but is followed by an asterisk through translational activators that seem to be limiting (Green-Willms et al., 2001), and a coupling between mRNA synthesis and/or processing and mRNA maturation via the SLS1 gene (Bryan et al., 2001). If a (indirect) regulatory effect of HAP exists on mitochondrial gene expression, the simplest explanation would be that HAP partly controls the expression of chromosomal genes involved in mitochondrial translation. The observed increase in mRNAs would reflect their possible stabilization (e.g. as proposed above via the SLS1 gene). A closer examination does indeed favour this hypothesis, since the key translation factors (genes encoding initiation factor TUF1, elongation factors MEF1 and MEF2, termination factor MRF1 ) are all upregulated by a factor of 2-4. The SLS1 gene, probably involved in the coupling between transcription and translation, could also be regulated, but the p value was not robust enough to conclude this with confidence. Several other genes controlling tRNA maturation (RPM2, upregulated four times) or mitochondrial aminoacyl tRNA synthetases are in the same situation. We examined genes encoding mitochondrial ribosomal proteins and found some of them to be also upregulated by a factor of 2 (MRPL6, MRPL19, MRPS5, MRPL7, PET123, MRPL11, YMR188c and NAM9 ), but it remains to be understood why other ribosomal genes are not regulated similarly.
Other mitochondrial genes under control of the HAP complex Table 2 provides the list of all S. cerevisiae ORFs found to be regulated by HAP in these studies and involved in mitochondrial biogenesis and function. All these genes are regulated by a factor of 2-3, and controlled in the same way by the Hap2p (Table 2) and the Hap4p (data not shown) proteins. Apart from genes coding for elements of the mitochondrial translation apparatus (see above), other groups of genes are also informative on the role of this transactivator. The HAP complex controls elements of the mitochondrial import pathway, focusing essentially on chaperones (probably reacting to the increased amount of nuclear proteins that have to be imported for respiratory functions) and especially the TIM44/SSE2 complex, regarded as a central motor unit of the mitochondrial import machinery (reviewed in Rassow et al., 1995). A few genes related to mitochondrial shape and movement but not to mitochondrial inheritance in general are also under the control of HAP. This result is consistent with the differences observed for mitochondrial shape in relation to glucose repression (discussed in Church and Poynton, 1998). Finally, in addition to some genes for which functions are not clearly defined, or which are homologous to components of the respiratory chain, a few mitochondrial carriers (for citrate, oxodicarboxylate and ATP/ADP) are also regulated. The case of ATP/ADP carriers again reflects the differential expression of the three isoforms, as previously described. It was shown that AAC2, the main carrier, was regulated by HAP (Betina et al., 1995); AAC1 was regulated by heme and oxygen (Gavurnikova et al, 1996); and AAC3 was regulated by carbon source but not by the HAP complex (Sokolikova et al., 2000). In our work, we indeed found that AAC1 was not regulated, while AAC3 was downregulated by HAP2. This last effect is certainly indirect, since Sokolikova et al. (2000) have shown that the responsive sequence of the repressor is not the HAP binding site and may be mediated by other transactivators.
The expression of some putative transactivators of unknown function is regulated by HAP With respect to the last point discussed in the previous paragraph, it is interesting to note that several putative transactivators were found to be controlled by HAP2, some of them possessing a degenerate 'CCAAT box'. This effect remains to be confirmed, especially since for some of them the standard deviation is quite large. Transcriptome analysis of these functionally uncharacterized transactivators in simple ( transactivator) or double ( transactivator hap2) mutants should help to understand their exact place in these regulatory cascades.
What is the function of HAP4 in evolution?
HAP4 mRNA is strongly induced at the diauxic shift, practically undetectable in glucose and expressed in glycerol and galactose. As already shown by others (Forsburg and Guarente, 1989b) and confirmed by our data, Hap4p is the moiety required for reprogramming of the cell from fermentation to respiration. Because this role is linked to the characteristic metabolism of S. cerevisiae, and because HAP4 orthologues had not yet been identified in other species, it could even be hypothesized that HAP4 would be unique to this yeast. More recently, we isolated KlHAP4, the HAP4 gene of the yeast Kluyveromyces lactis (Bourgarel et al., 1999), a yeast which has a respiratory metabolism and which is poorly glucose-repressed (Ferrero et al., 1978;Breunig, 1989). The other components of the complex (KlHAP2, KlHAP3, KLHAP5 ) have also been characterized (Nguyen et al., 1994;Mulder et al., 1994;Bourgarel, 2000). The question of the function of the HAP complex in K. lactis (and possibly in other yeasts) is therefore opened. In K. lactis, the nuclear genes involved in mitochondrial function (such as KlCYC1, KlQCR7 or KlQCR8 ) are constitutively expressed (Freire-Picos et al., 1995) and, up to now, no genes, with