Physiological Genomics

Gene expression profiling and phenotype analyses of S. cerevisiae in response to changing copper reveals six genes with new roles in copper and iron metabolism

Harm van Bakel, Eric Strengman, Cisca Wijmenga, Frank C. P. Holstege


Exhaustive microarray time course analyses of Saccharomyces cerevisiae during copper starvation and copper excess reveal new aspects of metal-induced gene regulation. Aside from identifying targets of established copper- and iron-responsive transcription factors, we find that genes encoding mitochondrial proteins are downregulated and that copper-independent iron transport genes are preferentially upregulated, both during prolonged copper deprivation. The experiments also suggest the presence of a small regulatory iron pool that links copper and iron responses. One hundred twenty-eight genes with putative roles in metal metabolism were further investigated by several systematic phenotype screens. Of the novel phenotypes uncovered, hsp12-Δ and arn1-Δ display increased sensitivity to copper, cyc1-Δ and crr1-Δ show resistance to high copper, vma13-Δ exhibits increased sensitivity to iron deprivation, and pep12-Δ results in reduced growth in high copper and low iron. Besides revealing new components of eukaryotic metal trafficking pathways, the results underscore the previously determined intimate links between iron and copper metabolism and mitochondrial and vacuolar function in metal trafficking. The analyses further suggest that copper starvation can specifically lead to downregulation of respiratory function to preserve iron and copper for other cellular processes.

  • DNA microarray
  • functional genomics
  • Saccharomyces cerevisiae
  • copper homeostasis

copper is an essential trace element for almost all organisms because of its ability to act as a cofactor in a variety of enzymes involved in electron transfer. However, the redox properties that make copper useful to cells can also lead to the generation of free radicals if left unchecked. An extremely tight regulation of copper homeostasis is therefore required, and this is reflected by the finding that free copper is restricted to less than one atom per cell in the yeast Saccharomyces cerevisiae (49). All organisms possess tightly regulated pathways for copper uptake and detoxification and these are highly conserved, even between eukaryotes and prokaryotes (41, 44, 62). Loss of appropriate copper trafficking in humans leads to disease (57), and more insight into copper and its regulation is required to understand such disorders. Copper homeostasis has been extensively characterized in S. cerevisiae (for review, see Refs. 7, 11, 44; see also Fig. 6). The high degree of conservation of the components of copper homeostatic pathways between different organisms, combined with the wealth of information on the yeast genome, makes S. cerevisiae an excellent model organism to gain insight in mammalian copper metabolism.

During copper starvation, copper is imported into yeast cells by the high-affinity transporters Ctr1p and Ctr3p (31). Copper can only be transported in the reduced Cu(I) state. Under normal conditions, this requires the action of the ferric/cupric reductases Fre1p and Fre2p (8, 9, 17, 18). The genes involved in high-affinity copper import are regulated by the transcription factor Mac1p (30, 70, 74). Copper excess results in the expression of the metallothioneins Cup1p and Crs5p, which can bind free copper in the cytoplasm, and the superoxide dismutase Sod1p, which is involved in free radical scavenging (21, 29, 67). The expression of these three genes is regulated by the transcription factor Ace1p (Cup2p) (5, 65).

There is a close link between copper and iron metabolism in S. cerevisiae because of the requirement of copper for high-affinity iron transport in the form of the cuproenzymes Fet3p and Fet5p. These paralogous proteins are multicopper oxidases that exhibit ferrous oxidase activity and form a high-affinity iron transport complex with the Ftr1p and Fth1p proteins, respectively (1, 59, 60). The genes involved in high-affinity iron transport are mainly regulated by Aft1p (Rcs1p), which activates transcription upon iron depletion (69). Additional genes that are regulated by Aft1p include a set of ferric reductases (FRE1FRE6), involved in reduction of iron before transport (9, 16, 37), and genes required for copper loading of Fet3p and Fet5p. The latter set includes the genes encoding the copper chaperone ATX1 and the copper-transporting P-type ATPase CCC2 (14, 47). S. cerevisiae also possesses a set of transporters encoded by ARN1–ARN4, which can acquire iron from siderophore-iron chelates in the medium (2426, 34, 73). The FIT1–FIT3 genes also play a role in siderophore-iron transport and encode cell wall proteins that may increase the amount of iron associated with the cell wall and periplasmic space (46). A second iron-responsive transcription factor, Aft2p, regulates a subset of Aft1p targets (4), but its role in iron homeostasis is less well understood.

Despite forming a relatively well-characterized system, novel insights into S. cerevisiae copper and iron homeostasis are still being generated (35, 50). It is therefore worthwhile to analyze copper homeostasis by systematic high-throughput approaches such as DNA microarray expression profiling and phenotype analysis. Previous studies of copper-related gene expression have focused on the role of the transcription factor Mac1p using single-point measurements under various conditions (12, 22). Here we have generated quadruplicate time course measurements of the changes in gene expression in S. cerevisiae in response to copper excess and deprivation over extended periods of time. High-throughput screening methods were applied to determine phenotypes of deletion mutants of genes identified in the microarray experiments. The results considerably extend our knowledge of copper homeostasis by uncovering new aspects of copper metabolism and provide an exhaustive analysis of the transcriptional response to environmental copper. The results also demonstrate the utility of combining systematic approaches, even for analyzing relatively well-characterized aspects of eukaryotic biology.


Microarray database accession numbers.

Minimal information about a microarray experiment (MIAME)-compliant microarray data in microarray gene expression markup language (MAGE-ML) and complete protocols have been deposited in the publicly available microarray database ArrayExpress (, with the following accession numbers: microarray layout, A-UMCU-4; data time course in copper starvation and excess, E-UMCU-15 and E-UMCU-16; data reference experiment, E-UMCU-17; protocols for growing time course samples, P-UMCU-29; and protocols for growing wild-type samples, P-UMCU-30. Array protocols are as follows: array layout, A-UMCU-4; total RNA isolation, P-UMCU-5; mRNA isolation, P-UMCU-6; labeling, P-UMCU-7; hybridization, P-UMCU-9; slide scanning, P-UMCU-10; and image analysis, P-UMCU-11.

Cultures for microarray experiment.

S. cerevisiae S288c (MATa, met15, ura3, his3 1, leu2) (Euroscarf, Frankfurt, Germany) was grown in synthetic complete (SC) medium supplemented with 2% glucose (QBiogene). All cultures for the microarray experiments were started at an optical density at 600 nm (OD600) of 0.025 from overnight (O/N) cultures and were harvested at midlog (OD600 = 0.5). At varying time points before midlog, either 100 μM bathocuproine disulphonate (BCS) or 8 μM CuSO4 was added to create conditions of copper starvation or excess, respectively. All samples were harvested at midlog by centrifugation at 2,000 g for 3 min, followed by snap-freezing in liquid nitrogen. An overview of the experimental setup can be found in Fig. 1.

Fig. 1.

Time course design. A: overnight (O/N) cultures were inoculated from 2 independent colonies and used the next day to start sampling cultures, reference cultures, and new O/N cultures, all at an optical density (OD) of 0.025. The new O/N cultures were used on day 2 to start sampling cultures for the 24-h time point, as well as a new set of O/N cultures for the 48-h time point on day 3. Bathocuproine disulphonate (BCS) or CuSO4 was added to the O/N cultures on day 1 after reaching an OD of 0.5. Sampling and reference cultures are indicated by gray boxes. B: schematic representation of growth curve for generating the 0.5- to 4-h time points, showing sample incubation times and point of harvest. BCS or CuSO4 was added to sample cultures at different times (*), which were then harvested after the time indicated by the arrows, resulting in harvesting of all samples at an identical OD of 0.5.

RNA isolation and microarray hybridization.

Yeast total RNA isolation, cDNA synthesis, labeling and microarray production, and hybridization were done as described previously (63). From each sample, 300 ng of cDNA (with a specific activity of 2–4% dye-labeled nucleosides) were hybridized for 16–20 h at 42°C. Microarray probes consisted of 70-mer oligonucleotides and included 3,000 control features and duplicate probes for 6,357 S. cerevisiae genes. Slides were scanned in an Agilent DNA Microarray Scanner (Agilent, model G2565BA). Spot quantification was carried out using Imagene 4.0 (Biodiscovery).

Microarray normalization and statistical analysis.

The microarray data were normalized by applying a Lowess function per subgrid on all gene spots (72), using the marrayNorm R package v.1.1.3 (15). After normalization, the data were variance stabilized with the VSN R package v.1.3.2 (28). The VSN function was calculated per array on a subset of genes only, where the spots with the lowest 2% intensity quantile were removed. After application of the VSN transformation to the full data set, the transformed data were used in an ANOVA analysis with the MAANOVA R package v.0.95-3 (68). The fixed-effect ANOVA model components included array, dye, spot, and sample effects (individual time points), which were estimated using restricted maximum likelihood. An F-statistic was used to test the sample effect for each gene by comparing the complete model to a null model without sample effects. P values were calculated by counting how often the maximum score in each of 5,000 randomized data sets was greater than or equal to the observed F-statistics. Permutation data sets were generated using unrestricted residual shuffling without replacement. Genes were considered significantly changed when the family-wise error rate corrected P value of the F2 test was ≤0.01. Clustering and preparation of graphs were done in Genespring 6.1 (Silicon Genetics). A set of 151 genes was removed before statistical analysis, as they displayed variable expression in the two reference pools, excluding genes with a 1.7-fold differential expression when compared in duplicate with a pool of 4 independently grown parent strains.

Phenotype screens.

Deletion mutants (MATa, met15, ura3, his3 1, leu2) were obtained from the Saccharomyces deletion project (Research Genetics) (19). Colonies were grown by transferring a small amount of the stock to single-well SC-agar plates (Omnitray: 242811) using a 96-pin tool (Biogene). The colony plates were subsequently used to start 300-μl liquid cultures in 96-well Nunc plates, which were grown O/N to an OD600 of ∼3. All screens were performed in duplicate from independent O/N plates.

For growth speed measurements in liquid SC medium, screen plates were prepared by diluting the O/N plates to a target start OD600 between 0.05 and 0.1. A GENIOS spectra fluor plus plate reader (Tecan) was then used to measure the OD600 at regular intervals. An R package (growth curve regression; GCR) was developed to identify the phase of logarithmic growth and subsequently determine the growth speed (OD/min) by linear regression. This package can be downloaded from Deletion mutant strains with a growth difference >20%, relative to growth in normal conditions, were selected after normalization. Normalization was done per screen plate, by adjusting the growth speeds of individual deletion strains by the median growth difference in SC and SC + stimulus medium of all strains in the plate. This was done to compensate for any growth differences of wild-type (WT) strains in these conditions.

For the 96-well spot assay to test growth in toxic copper conditions and on nonfermentable glycerol, O/N cultures were diluted to a target OD600 between 3.5 × 10−4 and 10.5 × 10−4 in 96-well plates. Cell counts of WT yeast indicated that this target range of OD600 corresponds to ∼20–60 yeast cells (H. van Bakel, data not shown) and resulted in spots where individual colonies could still be distinguished for the majority of deletion strains tested. Five microliters of diluted culture were spotted onto SC plates supplemented with 2% glycerol and onto SC plates supplemented with 2% glucose in the presence or absence of ferrozine and cupric sulfate. Differences in growth were scored based on the number and morphology of the resulting colonies.

Motif finding.

The MATCH 2.0 program in TRANSFAC professional 8.2 was used to screen a region of 500 nucleotides upstream of the translation start site of all significantly regulated genes for putative binding sites of transcription factors of interest. The size of the upstream region was selected on the basis that the vast majority of regulatory motifs in S. cerevisiae are located within 500 nucleotides upstream of the translation start site (23). Custom matrices for each transcription factor were created from a multiple alignment of known binding sites collected from literature and can be found at The cutoff for reporting putative binding motifs was set to minimize false positives according to TRANSFAC, after training the match algorithm on a set of exon sequences that were presumed to be free of transcription factor binding sites.

Atomic absorption spectrophotometry.

Parent strains were started at an OD600 of 0.025 in SC media and grown to an OD600 of 0.5. After the harvesting, cells were washed three times with milliQ demineralized water and then resuspended in 10 ml of milliQ demineralized water. An aliquot containing 20 OD units of the cell suspension was utilized for protein assay, and the remainder was digested in 40% nitric acid. Cell lysates were heated for 2 h at 55°C, diluted to 20% nitric acid, and assayed for copper and iron concentrations using a Varian SpectrAA 220-fast sequential atomic absorption spectrophotometer. For protein measurements, cells were broken in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton, 0.1% sodium deoxycholate, 0.1% SDS) by vortexing for 3 min with silica beads (0.5 mm zirconia). The protein concentration of the lysates was determined by a Bradford assay using BSA as standard.


The effects of prolonged changes in copper levels on gene expression in S. cerevisiae were examined in microarray experiments that covered extended periods of time. Copper starvation and excess were induced by adding 100 μM copper chelator BCS or 8 μM CuSO4, respectively. To avoid a large stress response confounding the results, these concentrations were chosen to ensure that growth rates during the course of the experiment were not affected (Supplemental Fig. S1; available at the Physiological Genomics web site).1 The time points for microarray sampling were selected to cover both short- (0–4 h) and long-term (24–48 h) effects. To prevent nutrient availability and cell density from affecting gene expression, all samples were harvested at the same growth stage and culture conditions (Fig. 1). For each time point, samples were taken from two biologically independent cultures and hybridized against an untreated reference sample in duplicate, swapping the label between the reference and sample. Each gene was represented twice on the microarrays used, resulting in eight measurements per gene for each time point. ANOVA was performed, and genes with a P value <0.01 after multiple testing correction were selected as significantly changed in expression between time points (see materials and methods). This yielded 128 genes, including 27 that have previously been validated as targets for copper- and iron-responsive transcription factors.

Temporal changes in copper and iron gene regulation.

A cluster diagram of the 128 genes with significantly altered mRNA expression in high- and low-copper conditions is shown in Fig. 2 (left). Four distinct clusters exhibit opposite regulation in copper-deprived vs. copper-excess conditions. Three of these clusters are highly enriched for known target genes of the copper- (6, 12, 18, 22, 33) and iron-responsive (4, 48, 51, 54, 71) transcription factors. DNA binding motif analysis of the upstream regions of all genes showed that binding sites for these transcription factors are predominantly present in the corresponding clusters. These three clusters therefore represent gene regulation by Aft1p and/or Aft2p, Mac1p, and Ace1p and are referred to by the names of these transcription factors. A fourth cluster was found to be highly enriched for genes encoding proteins localized in mitochondria. All the genes contained in these four clusters are listed in Table 1, along with their maximum fold change in copper-deprived and copper-excess conditions.

Fig. 2.

Genes responding to copper starvation and excess. Set of 128 genes with significant mRNA expression differences, as determined by MAANOVA analysis of the microarray time course, is shown after hierarchical clustering on both the low- and high-copper data sets using a Pearson correlation coefficient. Four clusters with genes involved in the response to low and high levels of copper or iron, as well as mitochondrial genes, are indicated in different colors, together with supporting evidence from literature and motif analysis. The results of the phenotype screens in liquid media and agar plates with varying concentrations of ferrozine (FZ) and CuSO4 are displayed at right. The levels of evidence in the agar plate screen were based on the amount of growth observed on plate. Genes with independently validated novel growth phenotypes in copper and iron metabolism are indicated in red.

View this table:
Table 1.

Overview of identified clusters

Genes in the Mac1p-regulated cluster (copper uptake) were among the first to show increased expression upon copper deprivation, with an initial response after 30 min and reaching a plateau within 2 h. Upregulation upon copper deprivation was also observed in the Aft1p/Aft2p cluster (iron uptake), but only at a later stage, starting at ∼1 h and reaching its maximum at 3–4 h (Fig. 2, left). This secondary response of iron transport genes reflects the interplay between copper and iron metabolism and is consistent with a model in which decreased cellular copper leads to a secondary iron starvation, as a result of diminished iron uptake through the copper-dependent reductive iron transport. Interestingly, the nonreductive (copper independent) siderophore-iron transport genes (FIT1–3, ARN1–4; Table 1 and Fig. 2) exhibited a strikingly higher induction than the genes required for reductive iron transport (FTR1, FTH1, FET3, FET4; Table 1 and Fig. 2). Preferential use of nonreductive siderophore-iron transport under limiting copper conditions has not been reported before and is interesting, as it suggests the presence of a regulatory mechanism capable of discriminating between copper-dependent and copper-independent iron transport.

To further investigate how upregulation of iron-uptake genes in the Aft1p/Aft2p cluster is achieved, total cellular copper and iron levels were determined by atomic absorption spectroscopy. No decrease in global cellular iron levels was observed after 4 h of copper starvation (Fig. 3). It is unlikely that reduced copper levels directly regulate iron transport; therefore, together these findings suggest the presence of a much smaller regulatory iron pool that directly influences Aft1p/Aft2p activity.

Fig. 3.

Cellular copper and iron levels. Parent strains were cultured in standard medium and subjected to copper depletion (100 μM BCS), copper excess (8 μM CuSO4), and iron depletion (200 μM FZ) for a period of 4 h. Sample growth and harvesting was performed according to Fig. 1B. Metal and protein levels were measured as described in materials and methods section. Assays were performed in 3 independent samples and averaged. Error bars indicate the standard deviation. SC medium, synthetic complete medium.

The third transcription factor-associated cluster contains genes under control of Ace1p (Fig. 2). This cluster behaves as expected given its role in copper detoxification (5, 65), with opposite regulation compared with the Mac1p and Aft1p/Aft2p clusters as well as a rapid induction.

A fourth cluster of genes showing low magnitude but reciprocal regulation under low and high copper contains 16 genes (Fig. 2, left). The majority of these encode proteins that have a mitochondrial localization or function, including key components of the respiratory chain (CYC1, CYT1, RIP1, ATP7, SDH4; Table 1) (3, 40, 42, 52). Altered expression of genes with mitochondrial function under changed environmental copper levels has not been described before. The downregulation of this cluster is gradual and may indicate a response to copper depletion by redistributing copper and/or iron away from the respiratory chain, which is not required under the conditions of fermentative growth assayed here. This hypothesis is further tested by the phenotype analyses described below.

Phenotype analyses of genes involved in copper and iron metabolism.

One of the goals of this study was to uncover novel genes involved in copper homeostasis. The majority of the 128 genes with altered mRNA expression have not been associated with a role in copper or iron metabolism or identified as targets of its regulators before. Altered mRNA expression under the conditions assayed here is a first indication of a putative role in metal metabolism. To further investigate roles in different aspects of metal metabolism, 111 of the 128 genes were screened for phenotypes upon gene deletion. Deletion strains for the other seventeen mutants were either not viable under standard growth conditions or unavailable in the collection used here (19).

Seven different conditions of growth were chosen for determining phenotypes. These were aimed at discovering defects in copper and iron import, defective copper detoxification, defective respiration, and increased viability under toxic copper levels (Table 2). All screens were performed in duplicate, and the 777 combinations of deletion mutants and conditions resulted in 99 phenotypes (Fig. 2, right, and Table 2 for overview). These phenotypes were found in 75 unique deletion mutants, several of which displayed phenotypes in multiple conditions.

View this table:
Table 2.

No. of phenotypes reported in screens

Quantification of phenotypes is challenging for growth-based agar plate assays. A high-throughput screening method was therefore developed for analysis of liquid media cultures, to allow more sensitive and quantitative determination of growth rates. Cultures were monitored in 96-well plates (Fig. 4), and growth rate was determined by linear regression, using the statistical R-package GCR developed for this purpose. All mutants were tested for growth in excess copper or low iron by adding either 0.3 mM CuSO4 or 0.3 mM iron chelator ferrozine (FZ). Screening in low iron was preferred to low copper, because phenotypes of deletion strains of known copper metabolism genes were more clearly observed in liquid medium or agar plates that were depleted of iron, rather than copper. This can be explained by the fact that defects in copper trafficking lead to impaired iron uptake by copper-dependent transporters. Iron is involved in a wider range of cellular processes than copper, and defects in copper metabolism may therefore be more apparent in low-iron rather than low-copper conditions. In addition, many of the differentially expressed genes identified in the microarray study are also iron responsive.

Fig. 4.

Phenotype screen in liquid medium for 111 deletion mutant strains in high copper and low iron. A: 96-well plate growth curves of wild-type (WT) strains in standard medium (black) and excess copper medium containing 300 μM CuSO4 (gray). Regression lines were fitted to points corresponding to logarithmic growth (solid circles), after excluding measurements during lag phase and diauxic shift (open circles). The average difference in slope of the regression lines in 2 independent growth curve measurements was used as a measure of differential growth. Growth rates of 111 deletion mutant strains for the genes identified in the microarray experiment were determined identically. B: an example of the deletion mutant growth curves (arn1-Δ) in standard medium (black) and with excess copper (gray) after normalization for growth differences of WT strains. C: scatter plot of growth rates in excess copper medium containing 0.3 mM CuSO4 relative to growth on standard medium. The solid diagonal line indicates no growth difference relative to WT strains, while the inner and outer dashed lines signify a growth difference of 10 and 20% relative to WT strains, respectively. aft1-Δ and cup2-Δ were included as controls (gray). D: result of screen as in C but for low-iron conditions (0.3 mM FZ). The screen was performed after an O/N preincubation with 300 μM FZ for 16 h to chelate intracellular iron stores. atx1-Δ, mac1-Δ, and aft1-Δ were included as controls (gray).

Results of the high-throughput liquid media assays were considered significant if a >20% growth reduction was observed in deletion mutant strains compared with WT, taking into account growth differences under standard conditions (Fig. 4, A and B; see also materials and methods). These criteria are based on the growth rates of deletion strains with previously established phenotypes (Fig. 4, C and D) and lead to identification of mutants with reduced growth in excess copper (arn1-Δ), low iron (mrs4-Δ, ctr1-Δ, ccc2, vma13-Δ), or both excess copper and low iron (fet3-Δ, ftr1-Δ, pep12-Δ) (Fig. 4, C and D). The growth phenotypes of ctr1-Δ, ccc2-Δ, fet3-Δ, and ftr1-Δ agree with previous studies (1, 14, 31, 60) and demonstrate the validity of the 96-well screening method. Reduced viability of mrs4-Δ under low iron has not been reported before but is in agreement with the fact that MRS4 encodes a mitochondrial iron transporter, expressed under iron-limiting conditions (12, 39, 66). The phenotypes of arn1, vma13-Δ, and pep12-Δ indicate novel roles for these proteins in copper and iron homeostasis.

Aside from screening in liquid cultures, several conditions were also screened using spot assays on agar plates. These included two conditions not suitable for screening in liquid media due to dramatically reduced growth rates of WT and mutant strains (glycerol and toxic copper). Excess copper (1 and 2.5 mM CuSO4) and low iron (300 μM FZ) were also included as agar plate screens, for comparison with the 96-well liquid medium screens.

Screening on agar plates with glycerol provides an alternative method to search for genes involved in copper and iron trafficking. Copper and iron are essential cofactors for several respiratory chain complexes, and impaired delivery of copper and iron to these complexes leads to dysfunction, resulting in an inability of cells to grow on nonfermentable carbon sources such as glycerol. Nine deletion mutants were identified with defective growth on glycerol (Fig. 2, right), three of which correspond to genes known to be involved in copper and iron metabolism (CTR1, FET3, FTR1) (1, 9, 60). The other six strains contain gene deletions for known components of the respiratory chain and one novel gene of unknown function (NRP1). In contrast to the glycerol phenotypes involving the three previously known copper and iron genes, addition of 2 mM CuSO4, 2 mM FeCl, or a combination of both failed to rescue the respiratory deficiency in deletion strains of the latter six genes (H. van Bakel, data not shown). This indicates that the glycerol phenotypes are a consequence of their role in energy metabolism, rather than copper or iron trafficking. This finding agrees with the hypothesis proposed above, that the mitochondrial cluster is downregulated when cellular copper is low during fermentative growth to preserve copper and iron for other cellular processes.

Toxic copper conditions were evaluated to identify genes that confer protection against high levels of copper when deleted. cyc1-Δ, fre1-Δ, ycr102c-Δ, and arn4-Δ display increased resistance to toxic copper levels relative to the parental strain (Fig. 2). FRE1 is involved in reduction of copper before uptake, and its deletion protects against excessive copper by lowering copper import (56). The phenotype of increased resistance to toxic copper upon deletion of CYC1, ARN4, and YCR102C has not been described before and suggests a novel function for YCR102C in copper metabolism.

The screen on agar plates with 2.5 mM CuSO4 revealed 66 gene deletion strains with increased sensitivity to excess copper compared with WT. Although the results support the phenotypes found in the other excess and toxic copper conditions (Fig. 2, right), the large number may reflect the fact that the concentration used is close to toxic levels. Although the genes were selected based on differential mRNA expression in either high or low copper, it is not excluded that any defect, rather than a specific copper defect, may be contributing to lethality under these conditions. These phenotypes were therefore not subsequently analyzed.

The performance of the low-iron and excess copper screen in liquid medium vs. agar plates was compared for the ability to identify phenotypes. In total, 19 phenotypes were reported when the screens in these conditions are combined (Table 2, first 4 rows), excluding 2.5 mM CuSO4 on agar plates. Of these phenotypes, 6 novel phenotypes could be validated in independent assays (described below) and 10 corresponded to known copper and iron phenotypes. The liquid screen performed better, reporting 10 validated or known phenotypes out of a total of 11, compared with 7 of 9 for the agar plate screen. The results of the liquid screen include four validated or known phenotypes not found in the agar plate screen. One validated deletion strain, hsp12-Δ, only shows increased sensitivity to high copper levels in the agar plate screen. Increased sensitivity of hsp12-Δ to excess copper was, however, also observed in liquid medium but failed to reach the significance threshold of 20% growth difference we imposed by 4% (Fig. 4).

Novel genes with roles in copper and or iron metabolism.

Together, the phenotype analyses revealed 20 deletion strains with one or more phenotypes, excluding the 2.5 mM CuSO4 screen on agar plates. Six strains only display respiratory phenotypes, not directly related to copper or iron metabolism. The other 14 strains contain gene deletions for 5 known genes of copper and iron metabolism and 9 novel genes. Phenotypes observed upon deletion of the nine novel genes were validated in independent serial dilutions on agar plates, confirming five phenotypes observed for hsp12-Δ, arn1-Δ, pep12, vma13-Δ, and cyc1-Δ (Fig. 5). The absence of a phenotype for mrs4-Δ on low-iron agar plates can be explained by the small growth difference of 24% in liquid culture, which may be missed on agar plates. One deletion mutant, crr1-Δ, that did not display a phenotype in the high-throughput screens was reevaluated in a serial dilution assay because it contains multiple Mac1p binding sites. crr1-Δ was found to provide increased protection against toxic copper levels (Fig. 5). This increased resistance became apparent at higher cell densities than those used in the spot screen on agar plates and explains why this effect was not initially observed. The six validated genes encode proteins with a wide range of functions, and their proposed novel roles in copper and iron metabolism are discussed further below and presented in Fig. 6.

Fig. 5.

Validation of phenotypes identified in high-throughput screens. Six 5-fold serial dilutions were plated on standard SC medium and SC medium with high and low copper or iron to verify the growth phenotypes observed in the high-throughput screens. Positive (+) and negative (−) controls are indicated. Plates were photographed after 2 days of growth. A: crr1-Δ and cyc1-Δ spotted on normal medium containing 3.0 mM CuSO4 to test the resistance of the two mutants against toxic copper levels. B: sensitivity of pep12-Δ and vma13-Δ to low iron was tested on normal medium containing 0.5 mM FZ. C: increased sensitivity of fet3, pep12, hsp12-Δ, and arn1-Δ to high-copper conditions was tested on medium with 0.6 or 1.0 mM added CuSO4.

Fig. 6.

Model of novel genes in existing copper pathways. The figure shows pathways involved in copper uptake and trafficking (green), iron trafficking (dark blue), and protection against toxic copper levels (light blue). The route from the cell membrane to the mitochondrion is indicated with a dotted line, since the proposed role of Cox17 as a copper chaperone has recently been questioned (38). Four transcription factors are known to regulate gene expression in these pathways and are indicated in yellow. The novel genes with validated growth phenotypes in varying conditions of environmental copper and/or iron are colored red. See text for more details.


Aside from revealing previously uncharacterized roles for genes in copper and iron homeostasis, the combined approach of mRNA expression profiling and phenotype analyses provides new insights into how copper and iron metabolism is regulated. These results are discussed first, followed by a presentation of the novel gene functions and their relation to copper and iron trafficking in general.

Copper and iron regulatory responses.

Microarray gene expression profiling revealed 128 genes with significant changes in mRNA expression in response to altered environmental copper levels (Fig. 2). These include 27 genes known to be regulated by copper- and/or iron-dependent transcription factors, representing almost the complete set of previously proposed targets for these factors. Cluster analysis grouped these genes into three clusters (Fig. 2), corresponding to the copper- and iron-responsive transcription factors. Delayed upregulation of iron import genes under low copper is consistent with the intimate link between copper and iron pathways (12, 22). Because no reduction in global cellular iron levels was observed to coincide with the upregulation of the iron regulon (Fig. 3), it is possible that this effect is mediated by a much smaller regulatory iron pool. The presence of such a pool would allow cells to respond to changes in environmental iron before depletion of iron affects essential cellular processes. An additional finding from the microarray time course is the differential regulation of the copper-dependent and -independent iron transport systems, suggesting a mechanism that favors upregulation of copper-independent iron transport in situations of copper starvation.

Twenty putative novel targets of the three copper- and iron-responsive transcription factors were identified, based on their cooccurrence in these transcriptional clusters. Additional evidence in the form of binding motifs for the corresponding transcription factors was found for five of these genes. These include FMP23 (YBR047W), VMR1 (YHL035C), and YLR047C, which are coregulated in the Aft1p/Aft2p cluster, and CRR1 and AQY2, which are found in the Mac1p cluster. FMP23, YBR047W, and VMR1 have previously been identified to be differentially regulated in mac1-Δ mutants (12). Both FMP23 and VMR1 localize to the mitochondrion (58) and could be involved in mitochondrial iron homeostasis. FMP23 encodes a protein of unknown function, whereas VMR1 is a member of the ATP binding cassette (ABC) family (10). YLR047C shows sequence similarity to the ferric/cupric reductases Fre1p–Fre7p and is therefore likely to play a role in the reduction of iron before transport across the plasma membrane.

The presence of copper-responsive elements in the promoters of CRR1 and AQY2 combined with their occurrence in the Mac1p cluster point to a role in copper import. This role is supported by increased resistance of crr1-Δ mutants to toxic copper conditions, which is discussed in more detail below. AQY2 encodes a water channel and is disrupted by a stop codon in many strains, including the strain used here, indicating that it is not essential for normal copper homeostasis.

A novel cluster, which contains 16 genes encoding mitochondrial proteins, was found to be downregulated after extended copper depletion (Fig. 2). Many mitochondrial proteins use copper and/or iron as a cofactor or structural component. Downregulation of these genes, which are not required during fermentative growth, could therefore represent a mechanism by which the cell redistributes copper and/or iron to more essential cellular processes. This agrees with our finding that the respiratory growth defects of genes in this cluster are not rescued by increased metal concentrations (Fig. 2; H. van Bakel, data not shown). The onset of regulatory changes in this cluster coincides with the upregulation of iron import, suggesting that limited availability of iron rather than copper plays a role in the regulation of this cluster. It was recently found that Cth2p plays a role in the remodeling of the cellular metabolism in response to iron deprivation by targeted mRNA degradation (48). This degradation is mediated by AU-rich elements in the 3′-UTR. CTH2 mRNA levels are induced as part of the iron regulon, in an opposite fashion to the mitochondrial-enriched cluster. A search of the 3′-UTRs of the genes in the mitochondrial-enriched cluster revealed the presence of AU-rich elements, indicating that Cth2p could play a role in their regulation (48).

The genes in the four transcriptional clusters that were identified reached their maximal up- or downregulation within 2–4 h and then continued to be expressed at this level for up to 72 h after addition of either CuSO4 or BCS. This indicates that prolonged changes in S. cerevisiae copper and iron metabolism gene expression, rather than transient responses, are needed to cope with lasting changes in environmental copper levels.

Genes with a novel role in copper and/or iron metabolism.

Genes identified by microarray expression profiling can be assigned specific cellular roles based on “guilt by association” with genes of established function. Systematic phenotype screening of gene deletion mutants provides additional evidence for novel roles in copper and or iron metabolism for six genes (Fig. 6). On the basis of their gene expression profiles, phenotypes, and previous reports of function, these can be divided into three groups involved in protection against copper toxicity (HSP12, ARN1), copper import or regulation (CYC1, CRR1), and vacuolar function (PEP12, VMA13).

HSP12 encodes a late embryonic abundant-like heat-shock protein involved in cellular responses against conditions of heat shock, oxidative stress, and osmotic stress (53, 64). HSP12 is thought to help maintain the stability of the cell membrane during stress conditions (53), and a likely function for the protein in excess copper conditions is to protect the membrane against oxidative damage. This is in agreement with the observed upregulation of HSP12 in excess copper (Fig. 2). De Freitas et al. (12) did not report reduced growth of HSP12 null mutants with high copper levels (1 mM CuSO4). This screen was performed with diploid strains in yeast-peptone-glucose (YPG) medium. The absence of a phenotype may be explained by the double gene dosage of metallothioneins in combination with an increased buffer capacity for copper in YPG medium, which may mask the HSP12 defect in these conditions. ARN1 is a known target of Aft1p/Aft2p and encodes a siderophore-iron transporter involved in the uptake of ferrichromes (25). The increased sensitivity of arn1-Δ to excess copper is not explained by the role of ARN1 as a siderophore-iron transporter (25), and further studies are required to elucidate its role in copper homeostasis.

Two strains, crr1-Δ and cyc1-Δ, exhibit increased resistance to toxic copper levels (Fig. 5). CRR1 is upregulated in low copper and has previously been reported to be a putative target gene of Mac1p, although no growth phenotype was reported (22). CRR1 encodes a putative glycosidase of the cell wall that is required for proper spore wall assembly (20). It is located divergently next to FRE1 in the genome, both genes sharing the same 644 nucleotide region upstream of their open reading frame, which contains Mac1p binding sites. Because fre1-Δ strains are known to display an increased resistance to toxic copper (Ref. 56; see also Fig. 5), the phenotype observed for crr1-Δ could conceivably result from an indirect reduction of FRE1 levels. However, FRE1 expression was actually found to be consistently increased in crr1-Δ relative to WT strains (Supplemental Fig. S2). Copper-dependent regulation of FRE1 gene expression was unaffected. Increased survival of crr1-Δ strains in toxic copper and similarity to cell wall glycosidases suggest that CRR1 functions to facilitate copper import, possibly by enriching copper in the cell wall and thus locally increasing the amount available for uptake (Fig. 6). The observed rise in FRE1 expression in crr1-Δ strains provides further support for this model, as cells would need to increase copper uptake to compensate for the loss of CRR1 function.

CYC1 is part of the mitochondrial cluster and encodes a protein of the respiratory chain responsible for the transfer of electrons from ubiquinone-cytochrome c oxidoreductase to cytochrome c oxidase (55). CYC1 also interacts with Ccp1p, a cytochrome c peroxidase involved in destruction of toxic radicals in the cell (32, 45). The increased viability of cyc1-Δ in toxic copper may be due to increased availability of Ccp1p for protection against free radicals caused by excessive copper.

It has recently become clear that the yeast vacuole plays an important role in copper and iron homeostasis through the actions of the copper transporter Ctr2p and the iron transporting p-type ATPase Ccc1p (35, 50) (Fig. 6). The functions of PEP12 and VMA13 are both linked to the yeast vacuole. PEP12 encodes a soluble N-ethylmaleimide-sensitive factor attachment protein receptor protein that is involved in endosome-to-vacuole fusion (2). VMA13 encodes a subunit of the vacuolar H+-ATPase that is needed for vacuolar acidification (13, 61). The growth defects of the pep12-Δ and vma13-Δ mutants in low iron and excess copper (Fig. 5) are likely to be the result of abnormal vacuole function. Endosomes are responsible for delivery of proteins to the vacuole (43). Impaired vesicle fusion in pep12-Δ strains may therefore disrupt both copper and iron homeostasis through defective delivery of metal transport proteins. VMA13 is an essential subunit of the vacuolar H+-ATPase, which is involved in acidification of the yeast vacuole (27). Proton gradients are used in many processes to drive the transport of ligands across membranes, and proton-coupled import of zinc into the yeast vacuole has previously been described (36). A similar mechanism may operate to release copper and/or iron from the vacuole when the availability of these metals is low. Alternatively, defects in vacuolar acidification could affect trafficking to the yeast vacuole and impair protein function, disrupting normal copper and iron homeostatic mechanisms.

S. cerevisiae copper and iron pathways have already been well characterized and thus serve as an excellent system to validate and test high-throughput approaches to determining gene function. It is encouraging that the combined approach of expression profiling and systematic phenotype screening employed here revealed novel aspects of gene regulation and novel genes as well as the majority of genes previously known to be involved in copper homeostasis. This may be attributed to the exhaustive nature of this study, encompassing quadruplicate microarray time courses with both low- and high-copper levels as well as seven different phenotype screens. The more quantitative liquid culture screens using growth curves compare favorably to the agar plate-based spot assays and can replace such screens in future systematic studies of gene function. Further experiments are required to test some of the specific aspects of copper and iron regulation and trafficking put forward in this study. These are worth pursuing, because at least three of the genes found here (PEP12, VMA13, and CYC1) have counterparts in higher eukaryotes.


This work was supported by The Netherlands Organization for Scientific Research (NWO) Grant 901-04-219.


We thank Jackie Senior for editing the manuscript and Dr. Leo Klomp for helpful comments during the preparation of the manuscript.


  • * C. Wijmenga and F. C. P. Holstege contributed equally to this work.

  • 1 The Supplemental Material for this article (Supplemental Figs. S1 and S2) is available online at

  • Address for reprint requests and other correspondence: C. Wijmenga, Complex Genetics Group, Dept. of Biomedical Genetics, Stratenum 2.117, Univ. Medical Center Utrecht, PO Box 85060, 3508 AB Utrecht, The Netherlands (e-mail: T.N.Wijmenga{at}



View Abstract