To determine the effect of overexpression of HAP4 on mitochondria, several different experimental approaches were used to quantify mitochondrial components and mitochondrial structures in the cell.

First, the quantity and quality of the respiratory-chain complexes were analyzed by subjecting mitochondrial extracts to two-dimensional gel electrophoresis. As shown in Figure 1, in the glucose-grown wild-type strain (WT YPD in Figure 1) most of the various respiratory chain or OXPHOS (oxidative phosphorylation) complexes are undetectable or barely detected as a result of the glucose-mediated repression of respiratory function. Only complex V (the ATP synthase) can be clearly distinguished, in particular an intense spot that consists of the α- and β-subunits of the F1 subcomplex. Glucose-grown cells that overexpress HAP4, on the other hand, show very distinct OXPHOS complexes (see HAP4 YPD in Figure 1). The pattern of spots of the glucose-grown HAP4 overproducer is similar to wild-type cells that grow in fully respiratory mode on a non-fermentable carbon source (compare HAP4 YPD and WT YPEG in Figure 1). The increased abundance of OXPHOS complexes in the glucose-grown HAP4 overproducer is further shown by the fact that mitochondrial suspensions from the HAP4 overexpression strain contain approximately twofold more protein per gram wet weight compared to its wild-type counterpart (Figure 2a).

The cellular content of cytochromes c + c₁ and b is also markedly increased, 50% and 65% respectively, when the HAP4 gene is overexpressed in glucose-grown cells (Figure 2b). The cytochrome aa₃ content is also increased in the HAP4 strain, although the corresponding part of the difference spectra does not allow proper quantification.

The mitochondrial content of cells was further analyzed by fluorescence microscopy using the fluorescent dye DASPMI, which specifically stains mitochondria [5]. To compare mitochondrial DASPMI fluorescence between strains directly, glucose-grown wild-type and HAP4-overexpressing cells were mixed. Wild-type and HAP4 cells were identified in situ by also staining one of the strains with the cell-wall dye CalcoFluorWhite (CFW [6]) before mixing the cells. As shown in Figure 3, HAP4-overexpressing cells contain more mitochondrial structures per cell (HAP4 cells are identified by CFW fluorescence in the left panels and by the absence of CFW fluorescence in the right panels). This is consistent with the fact that more membranous vesicles are present in mitochondrial suspensions extracted from HAP4-overexpressing cells (data not shown). Interestingly, HAP4-overexpressing cells also exhibit increased DASPMI fluorescence. Although this may suggest an increased membrane potential over the mitochondrial inner membrane [7], we favor the idea that the inner membrane has a more compact structure in these mitochondria [5].

Nuclear DNA encodes the majority of mitochondrial proteins. Several mitochondrial components, such as a number of subunits of enzymes of the respiratory chain, are encoded by the mitochondrial genome. Figure 4 shows that overexpression of HAP4 in glucose-growth conditions not only results in increased transcription of nuclear-encoded mitochondrial protein genes ([2], and see below) but also leads to enhanced transcription of all mitochondrial genes tested. In particular, COX2 mRNA was induced to high levels. It should be noted that the increase in transcription in the mitochondrion is not accompanied by strong alterations in the copy number of mitochondrial DNA (mtDNA) (data not shown), indicating that the increase in mitochondrial transcription is mainly due to increased activity of the mitochondrial transcription machinery.

The strong increase in OXPHOS complex abundance and in mitochondrial protein and cytochrome content per cell, together with the evidence that shows that HAP4 cells contain more mitochondrial structures per cell, conclusively shows that overexpression of HAP4 in glucose-growth conditions results in increased mitochondrial biogenesis.

Genome-wide expression was monitored with Affymetrix GeneChips using RNA from wild-type and HAP4 cells grown aerobically on minimal glucose medium (YNB) in a batch fermenter. For each strain, two separate hybridizations were carried out and the log₂-expression ratios or fold-changes were calculated for HAP4 overexpression versus wild type.

Because the HAP4-overexpression system can induce a physiological change that resembles that of the diauxic shift [2], we first analyzed whether the state of the regulatory activity network in the HAP4 overexpressor is similar to one or more of the transient states of the diauxic shift [1]. For this, the regulatory activity network was fingerprinted using the Quontology algorithm (see Materials and methods). This algorithm enables the identification of functional groups of genes that have changed in expression as a result of increased expression of HAP4. A Z-score is calculated for the functional categories from the MIPS database [8], which is a measure of the changes that occur in the regulatory activity network. Categories with a Z-score larger than 4 can be considered to be induced or repressed significantly. Figure 5 shows all functional categories that were induced or repressed significantly in the HAP4-overexpressing strain or in one of the expression profiles from the diauxic shift.

From Figure 5 it is evident that overexpression of HAP4 results in alterations in the regulatory network that resembles the time points of the diauxic shift with respect to several important functional categories (see categories marked with a filled circle).

The alterations in the regulatory activity network of the diauxic shift mainly occur in three phases, as indicated in Figure 5: high glucose concentrations; strongly decreasing glucose concentration; and glucose depletion where cells grow in respiratory mode on the remaining ethanol. The HAP4 overexpression profile is similar in a number of respects to the profiles of cells growing in decreasing glucose concentrations (13.5 and 15.5 hours; see Figure 5). However, in this case also, the changes that occur in the diauxic shift are much more extensive, in particular with respect to the categories 'metabolism of energy reserves' and 'rRNA transcription'. Hence, the regulatory activity network of the HAP4-overexpressing strain has moved to a state that resembles one of the transition states of the diauxic shift, but at the same time differs in a number of important aspects.

All functional categories that are significantly increased (Z-score > 4) in the HAP4-overexpressing strain are related to mitochondrial function (see Figure 5), including the main category 'protein synthesis' which consists largely of mitochondrial protein genes. Consistently, the vast majority of the 139 nuclear genes that are induced more than twofold in the HAP4-overexpressing strain (see genes listed in Figure 6 and Materials and methods) encode proteins that localize to mitochondria. It is striking that if the expression of these genes is compared to the expression in respiratory-state cells (columns 2-8 in Figure 6) or in glucose-repressed cells (column 9), it can be observed in the right part of Figure 6 that there are a large number of genes that are strongly upregulated in the HAP4 overproducer but not upregulated in respiring cells. As these genes also encode mitochondrial proteins, this suggests that overexpression of HAP4 leads to a novel state in which the expression of mitochondrial proteins goes beyond levels that are required in normal physiological conditions.

The two clusters of genes in Figure 6 fall into different functional categories. The group of genes that is responsive to changes in the respiro-fermentative growth mode consists mainly of genes encoding proteins that are directly involved in respiration (see genes listed in the left part of Figure 6). The second, larger and non-responsive group mainly consists of genes that encode proteins involved in mitochondrial protein synthesis (see genes listed in the right part of Figure 6). Apparently, overexpression of HAP4 not only results in increased expression of genes encoding proteins required for mitochondrial energy conservation (or generation) but also of genes that are required for the synthesis of functional mitochondria.

The upstream activation sequences (UAS) of the genes of the non-responsive group - but not those of the responsive group - generally all lack the typical Hap2/3/4/5p-binding element CCAATCA. This is interesting, as this particular element strongly correlates with the changes in expression that occur as a result of overexpression of HAP4 ([4], REDUCE: p < 10^-6). Furthermore, with a more in-depth computational search for potential Hap2/3/4/5p sites using a nucleotide-distribution matrix, no difference could be observed between promoters of the non-responsive group and random-sequence DNA (Figure 7a). In contrast, in the gene promoters of the responsive gene cluster a large number of potential Hap2/3/4/5p sites can be detected that are distributed over the promoter according to a typical yeast UAS (100-400 bp upstream of the translational start codon ATG). These data strongly suggest that the Hap2/3/4/5p complex does not directly regulate the vast majority of genes of the non-responsive gene cluster, namely genes important for mitochondrial proliferation.

In addition to the above-mentioned genes, which are upregulated and lead to increased mitochondrial biogenesis, it should be noted that the expression of 23 genes is reduced by overexpression of HAP4. These genes are RPR1 (-2.3), EUG1 (-2.3), ZRT1 (-17.0), ADH4 (-7.9), YGL258W (-3.9), YHR048W (-2.5), HXT4 (-2.7), TDH1 (-3.0), ZAP1 (-6.8), CWP1 (-3.0), ZRT3 (-3.4), ZRT2 (-2.9), TFS1 (-2.4), THI7 (-2.9), CAR2 (-2.7), SNO1 (-3.8), SNZ1 (-2.9), FET4 (-2.7), YNL254C (-5.3), FRE4 (-3.0), COS10 (-2.4), YOL154W (-27.3), YOR387C (-18.4). Interestingly, a large number of these genes are involved in zinc metabolism. Genes indicated in bold are most strongly affected by overexpression of HAP4 (<-3) and have been proposed to be targeted by zinc-specific transcription factor Zap1p [9]. In fact, the expression of ZAP1 itself is strongly repressed by overexpression of HAP4 and, furthermore, the repressive motif ACC-5-GGT, the DNA-binding site of Zap1p [10], was identified by the REDUCE algorithm with high significance, p < 10^-12. Hence, we conclude that the main repressive effect of overexpression of HAP4 is mediated through decreased expression of the Zap1p transcription factor. Genes that are typically not altered in expression on zinc depletion or zap1-deletion [9] are given in non-italic type above, and are suggested to be repressed through a different mechanism, possibly through the same mechanism that initially represses the ZAP1 gene.

The wild-type strain CEN.PK113-7D and its HAP4-overproducing counterpart 436 GH [16], in which the HAP4 gene is overexpressed more than fivefold in glucose-growth conditions, were used. The latter contains a genomically integrated HAP4-overproduction cassette in which the TDH3 promoter regulates the HAP4 gene. Cells were grown aerobically in YPD (1% yeast extract, 1% Bacto-peptone and 2 or 3% D-glucose) or YPEG (YP containing 2% ethanol and 2% glycerol). In all experiments cells were harvested at early logarithmic phase.

Yeast cells were grown in 100 ml YPD or YPEG medium until OD₆₀₀ of 0.7-1.0 and mitochondria were isolated according to [17]. Protein concentration was determined using Bradford. Two-dimensional gel electrophoresis was carried out according to [18]: the first dimension consisted of blue native, 5-13% gradient polyacrylamide gel electrophoresis (PAGE), and the second of denaturing PAGE on gels containing 10% SDS. Protein was visualized by Coomassie staining of the gels. The identity of a constitutive spot at the upper-left part of the gels was determined using mass spectrometry [17].

Cells were grown to OD₆₀₀ = 1.0-1.2, frozen in liquid nitrogen and stored in -20°C. The cell pellet was resuspended in a final volume of 2 ml using a buffer containing 100 mM KPO₄ pH 7.3, 250 mM sucrose and 0.5% Na-cholate. Difference spectra were derived from dithionite-reduced cells minus ferricyanide-oxidized cells. Spectral measurements were carried out at room temperature in a DW2000 dual-beam spectrophotometer. Concentrations of cytochromes were determined essentially as in [19].

Mitochondrial biogenesis in living cells was analyzed by DASPMI staining of cells grown logarithmically in liquid medium (YPD). YPD culture samples of 5 ml were incubated at 28°C in the presence of 10 μl of 1 mg/ml CalcoFluorWhite [6], centrifuged, resuspended in 20 μl YPD. CFW stained cells (2 μl) and 2 μl unstained cells were mixed, 1 μl of 1 mg/ml DASPMI and 15 μl 0.1 M Tris-Cl pH 8.0 were added. After 5 min, cells were washed in 0.1 M Tris-Cl pH 8.0, centrifuged and resuspended in 0.1 M Tris-Cl pH 8.0. After addition of 1 volume YPD, cells were aerated by pipeting and analyzed microscopically.

RNA isolation, northern blotting and hybridizations were carried out essentially as in [20]. Probes for mitochondrial transcripts COX1, COX2 and SCEI and the loading control PDA1 were obtained by [³²P]ATP labeling of PCR-generated DNA fragments. For microarray analyses, a pre-culture was grown overnight in YNB medium (0.67% yeast nitrogen base) containing 2% raffinose. The main culture was inoculated at OD₆₀₀ = 0.1 in a batch fermenter containing 500 ml YNB medium buffered at pH 5.0 with 100 mM sodium phthalate and supplemented with 2% glucose, stirred at 500 rpm and aerated at 1 vv (vessel volume per minute). Both strains grew equally well, having growth rates of μ = 0.36/h and 0.34/h for wild-type and HAP4, respectively. The respiratory quotient (RQ), decreased from RQ = 3.5 in wild-type cells to 2.6 in the HAP4-overexpressing strain. This shows that the latter strain grows in a more respiratory mode, similar to [2], which is accompanied by a decrease in the glucose consumption and ethanol production rates: from q_glucose = 15.6 mmol/g/h and q_ethanol = 19.8 mmol/g/h for the wild-type strain, to q_glucose = 11.1 mmol/g/h and q_ethanol = 15.6 mmol/g/h for the HAP4-overexpressing strain. The cultures were grown until OD₆₀₀ = 1.0 (approximately 8 h), cells were chilled rapidly by addition of ice, centrifuged and stored at -70°C. RNA was extracted using hot phenol according to [21] and poly(A)⁺ mRNA was isolated using Oligotex (Qiagen). Labeling and hybridization of GeneChip Yeast Genome S98 Arrays were carried out according to Affymetrix.

The average ratio was calculated from log₂ expression ratios that derive from two independent experiments of the HAP4 overexpressor relative to two independent wild-type experiments. ORFs for which the expression ratios diverged twofold in duplicate experiments were excluded from the analysis. Using interpolated variance analysis, 110 genes were calculated to be regulated significantly (p ≤ 0.01; data not shown). In this analysis genes are selected that show a twofold change, k-means clustering was carried out using Cluster [22], corrected for a minor bug that inappropriately displays the gene of the second cluster as the final gene of the first cluster. Matrix searches were carried out using the Hap2/3/4/5/p nucleotide-distribution matrix from the TransFac database [23] and the MatInspector program [24]. Matches to the Hap2/3/4/5/p matrix were analyzed in random-sequence DNA (A/T 31%, C/G 19% [25]) that contained the same number of promoters of the non-responsive gene cluster.

Several publications have used genome-wide expression patterns to score functional gene categories, by considering the overlap between each category and the set of genes induced or repressed above a certain threshold. We take a very different approach that is essentially identical to the REDUCE algorithm for scoring promoter elements [4] but with manually defined gene categories from an ontology replacing the set of genes whose promoter contains a particular DNA motif. Again, a Z-score can be calculated for each category that measures the deviation of the average log-ratio for genes in the category from the genome-wide average, in units of the standard deviation. A similar method has recently been discussed by Pavlidis et al. [26].