The COP9 signalosome (CSN) is a highly conserved eukaryotic protein complex which regulates the cullin RING family of ubiquitin ligases and carries out a deneddylase activity that resides in subunit 5 (CSN5). Whereas CSN activity is essential for the development of higher eukaryotes, several unicellular fungi including the budding yeast Saccharomyces cerevisiae can survive without a functional CSN. Nevertheless, the budding yeast CSN is biochemically active and deletion mutants of each of its subunits exhibit deficiency in cullins deneddylation, although the biological context of this activity is still unknown in this organism. To further characterize CSN function in budding yeast, we present here a transcriptomic and proteomic analysis of a S. cerevisiae strain deleted in the CSN5/RRI1 gene (hereafter referred to as CSN5), coding for the only canonical subunit of the complex. We show that Csn5 is involved in modulation of the genes controlling amino acid and lipid metabolism and especially ergosterol biosynthesis. These alterations in gene expression correlate with the lower ergosterol levels and increased intracellular zinc content which we observed in csn5 null mutant cells. We show that some of these regulatory effects of Csn5, in particular the control of isoprenoid biosynthesis, are conserved through evolution, since similar transcriptomic and/or proteomic effects of csn5 mutation were previously observed in other eukaryotic organisms such as Aspergillus nidulans, Arabidopsis thaliana and Drosophila melanogaster. Our results suggest that the diverged budding yeast CSN is more conserved than was previously thought.
chromatin immunoprecipitation associated with microarray technology
cullin RING ubiquitin E3 ligases
filter-aided sample preparation
false discovery rate
limiting zinc medium
Mpr1-Pad1 N terminal
proteasome, COP9 and initiation factor 3
Skp1,Cdc53/Cullin, F-box ubiquitin ligase
vacuolar transporter chaperone
The COP9 signalosome (CSN) is a highly conserved eukaryotic complex with a canonical composition of eight subunits that was initially discovered as a repressor of light-regulated development in Arabidopsis thaliana . Subsequently it was found that the CSN is involved in diverse cellular and developmental processes, and is a key regulator of the ubiquitin proteasome system . A canonical CSN is composed of eight subunits (Csn1–8), six subunits (Csn1–4; Csn7–8) harbouring a PCI domain (proteasome, COP9 and initiation factor 3) and two subunits (Csn1–6) harbouring an MPN domain (Mpr1-Pad1 N terminal) [1-4].
The best studied function of the CSN is the enzymatic regulation of cullin RING ubiquitin E3 ligases (CRLs). CRLs are active when covalently attached to the ubiquitin-related protein Nedd8 (known as Rub1 in plants and budding yeast). Neddylation/deneddylation cycles are required for the regulation of CRL function, and the CSN mediates deneddylation. Deneddylation activity resides in CSN subunit CSN5, which catalyzes the deconjugation of the small peptide Nedd8 from the cullin subunit of CRL . Csn5 is a highly conserved subunit that harbours a unique catalytically active metal binding MPN+/JAMM metalloprotease motif responsible for an isopeptidase activity that is dependent on fully formed CSN complex . The opposite event – conjugation of Nedd8 to cullins – stimulates CRL activity [7-10]. The observed positive role of CSN in regulating CRL activity can be explained by the finding that CSN-mediated inactivation of CRLs counteracts autocatalytic breakdown of CRL substrate receptor subunits [11-13]. Recent data, however, suggest that CSN regulation of CRLs can also occur in a non-enzymatic fashion, suggesting that CSN regulates CRL activity by multiple mechanisms [14-17].
While CSN-deneddylating activity is highly conserved, its subunit composition can be different in unicellular organisms. Csn5 is the most conserved subunit and the only canonical subunit within the budding yeast CSN complex , and has been maintained throughout evolution probably because of its essential role in deneddylation [1-3, 6, 19, 20]. The budding yeast Saccharomyces cerevisiae, for instance, seems to have evolved a particularly divergent CSN. The ScCSN (S. cerevisiae CSN) is composed of six subunits, four PCI subunits (Csn9, Csn10/Rri2, Csn11/Pci8, Rpn5), one MPN subunit (Csn5/Rri1) and Csi1, a diverged homologue of Csn6 which does not contain recognizable domains [18-22].Whereas CSN is essential for the development of organisms as different as plants (A. thaliana) and animals (Drosophila melanogaster; Caenorhabditis elegans; mouse and human), budding yeast can survive without a CSN. Indeed, deletions of CSN5 and CSN9 genes show some increase in pheromone sensitivity and mating efficiency but do not affect yeast viability . More recently, neddylation of the budding yeast Cdc53/Cul1 (aka Cdc53) has been shown to be tightly regulated upon carbon source shifting with possible metabolic implications . These phenotypic effects, again, do not compromise budding yeast vitality, thus making it a suitable model to study whether CSN cellular functions have diverged in evolution. In order to further investigate the cellular functions of the ScCSN, we have conducted a transcriptomic analysis of an S. cerevisiae null mutant for CSN subunit 5 (hereafter referred to as ∆csn5). The preliminary results  showed a strong involvement of Csn5 in transcriptome modulation. Here we validated the main transcriptomic observations by completing the analysis with a genome-wide proteomic screening and phenotypic assessments. We show that Csn5 is necessary for efficient ergosterol biosynthesis and proper zinc uptake.
The ∆csn5 mutant has defects in deneddylation and this phenotype can be rescued by exogenous Csn5
Null mutants of the budding yeast CSN, including the null mutant of the enzymatic subunit ∆csn5, have been studied previously in the parental BY4741 genetic background, and the lack of cullin deneddylation has been described . For this study, we generated a new ∆csn5 deletion mutant in the W303 parental background. To confirm that this newly constructed mutant was also affected in deneddylation, we conducted an immunoblot analysis of total protein extracts from the mutants, using a commercial antibody for Cdc53, which is one of the three cullins encoded by the S. cerevisiae genome . Only the neddylated form of Cdc53 was observed in the lane corresponding to ∆csn5, indeed indicating a deneddylation deficiency in this mutant (Fig. S1A, lane 3). To further confirm that this was indeed due to absence of the CSN5 gene, we monitored Cdc53 neddylation status also in ∆csn5 cells transformed with a plasmid bearing wild-type CSN5 with its natural promoter. This strain showed a complete rescue of the deneddylation activity (Fig. S1A, lanes 4 and 5). When the same experiment was conducted using a plasmid encoding a Csn5 mutated in the catalytic JAMM domain, CSN deneddylating activity could not be rescued (Fig. S1A, lane 6). In addition, the lack of Csn5 did not affect either the levels of Csn9, genomically tagged by 13 repeats of Myc (Fig. S1B), or the levels of Rpn5, which is a shared subunit between the proteasome lid and the CSN complexes. Furthermore, the lack of Csn5 did not lead to significant changes in the transcript expression of other CSN subunits, compared with their relative wild-type strain, suggesting that absence of Csn5 does not lead to changes in the levels of other CSN subunits (Fig. S1C).
Csn5 affects transcription of specific categories of genes
In order to gain a general view on the transcriptomic effects due to the absence of CSN5, we performed a genome-wide analysis using DNA microarrays containing the entire repertoire of S. cerevisiae ORFs. The mRNA profile of the ∆csn5 strain was compared with its isogenic wild-type strain W303. Two independent cell cultures were analysed for each strain. The microarray data derived from two biological replicates revealed that the expression levels of several S. cerevisiae genes were altered in the mutant compared with its isogenic wild-type (Table S1). We identified 2386 ORFs which gave reliable signals in both biological replicas for both strains (see 'Materials and methods' for filtering criteria). These data were subsequently used for functional analysis by the t-profiler tool  (Table 1). This method scores changes in the average activity of predefined groups of genes and it does not require a pre-selected modulation cut-off in order to obtain a general view of transcriptomic regulation. Surprisingly, the results revealed that ‘Structural constituent of ribosome’ was the most upregulated functional category in the ∆csn5 strain, 123 genes of this category being around 1.5-fold upregulated compared with the wild-type (average log2 ratio 0.67 and E value < 1.0 × 10−15). Additionally, we found that 90 genes involved in the lipid metabolism category were downregulated (average log2 ratio −0.43 and E value < 3.42 × 10−5). In particular, downregulated genes within the last category included most of the genes involved in the biosynthesis of fungal sterol ergosterol (Fig. 1).
Table 1. Functional categories which are globally modulated in the Δcsn5 strain compared with the isogenic wild-type strain. T value and E value were calculated as described in . T value > 1 indicates the gene categories that are upregulated; T value < 1 indicates gene categories that are downregulated. We selected 2386 ORFs, mean 0.0, standard deviation 0.8
Mean log2 ratio Δcsn5/W303
No. of ORFs
Structural constituent of ribosome
Cytosolic ribosome (sensu Eukarya)
Structural molecule activity
Cytosolic large ribosomal subunit
Large ribosomal subunit
Small ribosomal subunit
Carbohydrate transporter activity
Response to inorganic substance
Fructose transporter activity
We used the same tool also to score groups of genes whose promoters had been previously shown to bind particular transcription factors by chromatin immunoprecipitation associated with microarray technology (ChIP on chip) analysis  and which were also showing a significant average upregulation or downregulation (Table 2). Our analysis revealed that approximately 100 genes which bind Rap1 and/or Fhl1 in vivo were on average 1.5-fold upregulated in the ∆csn5 compared with wild-type (E value < 2.12 × 10−6 and E value < 1.0 × 10−15, respectively), while 71 Hap1 target genes were on average 1.35-fold downregulated (P <1.21 × 10−4). Downregulation of Hap1 binding genes is particularly intriguing, since this transcription factor is known to be activated by a high concentration of heme which shares with the ergosterol biosynthesis pathway farnesyl pyrophosphate (FPP) as the precursor .
Table 2. Number of ORFs found modulated in ∆csn5 and whose promoters are known to be bound in vivo by specific transcription factors. List of known in vivo targets of the transcription factors Flh1, Rap 1 and Hap1 found also to be significantly modulated in ∆csn5 cells, compared with the isogenic wild-type. T value and E value were calculated as described in . (For ChIP on chip data, see .)
No. of ORFs
Several gene modulations were chosen to be further validated by real-time RT-PCR (Figs 2 and S2). We selected two genes involved in ergosterol biosynthesis (ERG2 and ERG28), one gene involved in lipid biosynthesis (FAA4), a gene known as a modulator of cullin activity (SCM4) , and a series of genes involved in zinc metabolism (ZRT1, ZRT2, IZH1, IZH2 and IZH4), which were all downregulated in ∆csn5 cells. The genes involved in zinc metabolism were previously shown to be under the control of transcription factor Zap1 . We could not obtain reliable data for ZAP1 mRNA accumulation by DNA microarrays due to its low level of expression but real-time RT-PCR showed a significant reduction in the Δcsn5 strain (Fig. S2A). Furthermore, we also tested a couple of induced genes involved in the autophagy pathway (ATG16 and ATG33). As shown in Fig. 2, there is an excellent correlation between the real-time RT-PCR and the microarray data (Pearson's r correlation 0.92; P =1.6 × 10−5). Next, we tested the expression of selected genes in strains deleted in other subunits of the Sc CSN complex (Csn9, Csn10/Rri2, and Csn11/Pci8 and Csi1; Figs 3 and S2). Similar modulations of transcript levels were also observed in these strains, suggesting that the changes in gene expression observed in ∆csn5 were caused by a defect in the function of the whole complex rather than being linked to a missing function of the single Csn5 protein.
The proteome of the ∆csn5 strain partly recapitulates the transcriptome
Since the transcriptomic alterations observed in the ∆csn5 mutant appeared reproducible and coherent, we decided to verify if they corresponded to similar variations in protein abundance by analysing the genome-wide proteome of the Δcsn5 strain using a label-free shotgun approach (see 'Materials and methods'). We analysed total protein extracts purified from four independent cultures for each strain, run simultaneously. Samples from the same strains exhibited a high level of correlation, thus excluding that the observed modulations could be the consequence of technical artifacts (Fig. S3). We obtained reliable data for approximately 750 proteins (see 'Materials and methods'). Among them, 106 showed quantitative and reproducible changes in their levels [false discovery rate (FDR) cut-off of 20%; see Table 3]. An analysis of these modulated proteins showed a correlation between the results of our transcriptomic and proteomic analysis for enzymes involved in ergosterol biosynthesis, such as Erg2, Erg6, Erg9, Erg11 and Erg20. These enzymes are downregulated at the mRNA level (see Fig. 1), and they are also less abundant at the protein level (Table 3). However, a positive correlation between the two sets of data was not always found (Table S2). As an example, proteins involved in amino acid metabolism, such as Aro9 and Aro10, are more abundant in the Δcsn5 mutant, while their respective mRNAs are modestly downregulated. Similarly, the thiol peroxidase Hyr1 is more abundant at the protein level, while the respective mRNA level does not change in Δcsn5 compared with the wild-type. On the other hand, similarly to what was observed at the RNA level, the fraction of proteins whose abundance is lower in ∆csn5 is highly enriched in proteins encoded by genes known to be bound by the Hap1 transcription factor in cells grown in rich medium (Table S3). An important complementation of this study would be the analysis of the half-life of these proteins, in the context of a Csn5 deficiency.
Table 3. Modulated proteins in the Δcsn5 strain compared with the isogenic wild-type strain. Values represent the average of four independent extracts for each strain. Values with FDR < 0.20 are reported and values with FDR < 0.05 and FDR < 0.10 are marked with ‘+’
log2 ratio ∆csn5/W303
FDR < 0.05
FDR < 0.10
The ∆csn5 mutant has reduced ergosterol content
Comparative analysis of the transcriptome and proteome suggested a significant impact of Csn5 on ergosterol biosynthesis. Because absence of Csn5 leads to downregulation of several enzymes of the ergosterol biosynthesis regulon, it can be predicted that, as a consequence, the overall ergosterol biosynthesis may be affected. To verify this hypothesis, we measured the ergosterol content in the ∆csn5 mutant cells by GC analysis. Table 4 shows that the ∆csn5 strain shows a significant (25%) reduction in ergosterol content compared with the isogenic wild-type. We further confirmed this finding by assessing the ability of the ∆csn5 mutant to grow in the presence of nystatin and ketoconazole. Nystatin targets ergosterol in the cell membrane, while ketoconazole interferes with ergosterol biosynthesis . As expected, the results shown in Table 4 indicate that ∆csn5 cells are indeed more sensitive to ketoconazole and more resistant to nystatin compared with the wild-type, again confirming an alteration in ergosterol biosynthesis. The ergosterol reduction observed in the ∆csn5 strain is within the range (12–51%) observed in yeast cells treated with different azole derivatives at sub-inhibitory concentrations (0.01–1 μm·L−1) [32, 33]. At higher azole concentrations (3–10 μg·mL−1), a further decrease in ergosterol is observed in these cells, accompanied by a decrease in chitin levels and by an evident cytostatic activity. This explains why the Csn5 deleted strain shows higher sensitivity to ketoconazole, although it can still grow efficiently in normal conditions.
Table 4. Ergosterol content and related phenotypes for the Δcsn5 strain compared with the isogenic W303 strain
3.27 mg·(mg dried weight)−1
Sensitive to 6 U·mL−1
Sensitive to 4 mg·mL−1
2.43 mg·(mg dried weight)−1
Sensitive to 24 U·mL−1
Sensitive to 1 mg·mL−1
∆csn5 mutants have alterations in transition metal uptake
A reduced content of ergosterol in the cell membrane has previously been linked with an increased permeability to monovalent and divalent cations [34, 35] in general, and to metal cations in particular . Indeed, as mentioned earlier, our transcriptomic analysis showed repression of several genes coding for key regulators of zinc metabolism in YPD medium (Fig. 2), such as ZRT1, ZRT3, IZH1, IZH2, IZH4, and of ZAP1, which encodes their transcription activator (for a schematic representation, see Fig. S4). This alteration in gene expression could be caused by an increased content of zinc in the cytoplasm. To verify whether the low expression of zinc metabolism genes observed in ∆csn5 cells could be correlated to an alteration in zinc content, we checked zinc intracellular levels on cells grown in YPD using the fluorescent dye zinquin ethyl ester (see 'Materials and methods'). This zinc-specific dye has already been used successfully to measure relative zinc levels in S. cerevisiae [36, 37]. While poor zinquin staining was observed in wild-type cells grown under normal conditions, addition of 500 μm ZnSO4 to the medium resulted in the formation of small, bright fluorescent granules (zincosomes) (Fig. 4). Several zincosomes were instead already visible in the cytoplasm of the ∆csn5 strain grown in normal conditions and they further increased upon addition of 500 μm ZnSO4 to the medium. The zincosomes appear mainly localized around the vacuoles suggesting that, besides an increased passive import in the cell, the deleted strain shows also a defective zinc import and storage in the vacuole. A similar increased permeability to zinc is observed in a strain deleted in ERG6 (Fig. 4), confirming a direct relationship between ergosterol deficit and passive zinc income.
We also tested the growth yield of the ∆csn5 strain and of its isogenic wild-type strain in limiting zinc medium (LZM) containing different zinc concentrations (see 'Materials and methods'). Indeed, the growth of the ∆csn5 is slower when the cells are transferred in a medium with low zinc concentration compared with its isogenic wild-type (Fig. 5A). This growth defect was rescued only at concentrations above 500 μm, a zinc concentration at which active import is probably less stringent. Ectopic expression of the CSN5 gene was sufficient to restore efficient growth at a low zinc concentration (Fig. 5B).
We extended our analysis also to other transition metals and found a reduced viability of the ∆csn5 strain when Ni and Cd were added to a half synthetic defined medium depleted of zinc (Fig. S5). These results could reflect an increased passive influx of these divalent cations at the concentrations used, due to lower ergosterol concentration in the cell membrane. Another explanation could be related to a specific effect of Cd on the function of the F-box protein Met30. F-box proteins are the substrate adaptors of the Cul1 containing SCF (Skp1,Cdc53/Cullin1, F-box) ubiquitin ligase. In the presence of Cd, Met30 is detached from the SCF, becoming inactive, and cells may lose their viability . Indeed a role of ubiquitin-dependent proteolysis on the resistance to cadmium was previously proposed . In this scenario, lack of the regulatory role of the CSN on one or more SCFs could enhance metal toxicity.
In this study, we have deciphered the transcriptomic and proteomic modulations of the budding yeast ∆csn5 in the W303 genetic context. We show that the deletion of CSN5 in budding yeast does not change the mRNA level of other CSN subunits, nor the protein level of Csn9 (Fig. S1). On the other hand, our results demonstrate that Csn5 is able to regulate, directly or indirectly, several sets of genes and proteins.
First, we observed a general upregulation of genes coding for components of the protein biosynthetic apparatus (Table 1), including genes coding for at least 26 ribosomal proteins (RPs). These genes are known to be mainly controlled by the Rap1 and Fhl1 transcription factors. Interestingly, similar genes were also found to be regulated by CSN in other eukaryotic organisms. For example, in A. thaliana, Ma et al.  previously found a general upregulation of genes coding for at least 29 RP gene mRNAs in dark-grown csn8 (cop9-1) mutant seedlings. The same genes were even more strongly upregulated in dark-grown cop10-1 (a mutant in a ubiquitin E2 variant) seedlings, suggesting that the regulation of their expression might involve CSN and its ubiquitin ligase-mediated regulation of specific transcription factors or of proteins involved in mRNA stability. Indeed, the results from our proteomic analysis show that in S. cerevisiae the CSN-mediated regulation of these genes remains substantially confined at the mRNA level, since only four of the transcriptionally upregulated RPs (RPL3, RPL42B, RPS24A and RPS30A) are found to be significantly increased also at the protein level while three of them (RPP2B, RPL14B and RPL16B) are instead found to be significantly decreased at the protein level (Table S2). It is likely that, in the ∆csn5 mutant, a translational negative control partly compensates the RP mRNA accumulation. It has been previously shown that several mammalian RPs are selectively neddylated  and therefore potentially less stable in the ∆csn5; yet, so far, the link between this neddylation and the CSN complex has not been confirmed. In this case, this transcriptional upregulation could represent a feedback compensation mechanism.
Second, most of the genes that are involved in lipid biosynthesis are downregulated in ∆csn5 cells (Table 1). This is reflected at the phenotypic level by a reduction of the ergosterol content and by an increased sensitivity and resistance to ketoconazole and to nystatin, respectively. Interestingly, the group of downregulated genes includes about 70 genes which are known to be bound and regulated by the transcription factor Hap1. This is particularly intriguing, since this transcription factor is known to be activated by a high concentration of heme which shares FPP with the ergosterol biosynthesis pathway as precursor (Fig. 1).
Third, there is a significant downregulation of sugar transporter genes. This could be related to the carbon source regulation which has recently been shown for Cdc53 neddylation . This downregulation appears to be limited to the RNA level.
Furthermore, the gene ontology analysis of the proteins differentially modulated in Δcsn5 cells shows a highly significant enrichment of proteins involved in amino acid biosynthesis and metabolism and oxidation–reduction processes. Indeed, genes belonging to these ontology categories were previously found to be modulated at the mRNA level in other csn mutants from D. melanogaster, A. thaliana and Aspergillus nidulans [40, 42, 43]. Of the five proteins found to have significantly increased levels in the As. nidulans csnE mutant (the ScCsn5 orthologue), two (Sam2 and Ilv5) were enzymes involved in amino acid metabolism . Strikingly, their orthologues in Sc∆csn5 are downregulated (Table 3), suggesting that their modulation by CSN is conserved through evolution. The fact that the two proteins are not regulated at the RNA level either in As. nidulans or in S. cerevisiae suggests a direct post-transcriptional regulation by the CSN.
The mis-regulation of Sam2 and Ilv5 in the ∆csn5 strain could be responsible for the observed alterations in the expression of several other genes and proteins involved in amino acid biosynthesis and metabolism. In fact, several proteins involved in lysine biosynthesis are also downregulated in Sc∆csn5 cells. Among them, we found the homocitrate synthase Lys20, which catalyzes the acetyl CoA-α-chetoglutarate condensation required for lysine biosynthesis. The A. thaliana orthologue gene, IMS2, is also strongly downregulated in a csn5 null mutant, suggesting that this can be a crucial step in the CSN control on lysine biosynthesis ( and Table S4). With regard to genes involved in oxidation–reduction processes there is a striking parallelism in the regulation of the thiol peroxidase and hydroperoxide receptor Hyr1, which is one of the most overexpressed proteins in the Sc∆csn5 strain (Table 3), and its A. thaliana orthologue At4g11600 is strongly induced in a csn5 null mutant ( and Table S4). Since this protein is an important sensor of lipid oxidation and is involved in the cellular response to oxidative stress, we think that this is a noteworthy feature of the CSN regulatory action.
A third protein which is mis-regulated both in As. nidulans and in S. cerevisiae is 3-hydroxy-3-methyl-glutaryl-CoA (HMG-CoA) synthase (Erg13). Its repression could provide an explanation for the downregulation of the ergosterol pathway observed in S. cerevisiae ∆csn5 cells. Indeed, downregulation of Erg13 could result in a decrease of HMG-CoA synthesis, a crucial step in the isoprenoid biosynthetic pathway (Fig. 1). This is consistent with various observations. The first observation is a reported positive genetic interaction between CSN5 and FMS1, a gene coding for a key enzyme in the pantothenic acid synthesis pathway . The second observation is a general downregulation of ACS2 both at the RNA and protein level. ACS2 is one of the two genes coding for the acetyl-CoA synthetase which utilizes acetate to produce acetyl-CoA, the substrate utilized for HMG-CoA synthesis. Indeed, it was recently shown that overexpression of ACS2 leads to a consistent increase in acetyl-CoA and upregulates seven key genes in the mevalonate pathway . Similarly, repression of the mevalonate pathway could lead to a reduction in ACS2 expression. Moreover, this enzyme has a key role in determining the extent of histone acetylation and its regulation could have secondary effects on the Δcsn5 transcriptome . The transcriptional downregulation of the ergosterol biosynthetic pathway could therefore be the consequence of the decreased availability of HMG-CoA and subsequently of FPP, which is the key precursor in the ergosterol biosynthetic pathway (Fig. 1). Likewise, a decrease in FPP levels could be the cause of a reduction in heme biosynthesis, and therefore of the observed downregulation of the Hap1 regulon. To this purpose, it should be noted that we observed downregulation both of the aerobic and anaerobic Hap1 targets, which is an effect typically observed in the presence of a reduction in heme biosynthesis under aerobic conditions .
Downregulation both at the RNA and protein level of ERG20 which controls the synthesis of geranyl-P-P, the precursor of FPP and of ERG9, which in turn controls FPP conversion to squalene on the ergosterol pathway, underscores the relevance of Csn5 control in the isoprenoid biosynthesis (Fig. 1). In yeast, as well as in mammals, there is a coordinated regulation of fatty acids and cholesterol . In the ∆csn5 strain we indeed observed downregulation of OLE1 at the RNA level and of FAS1 at the protein level, which are two key enzymes in the fatty acids biosynthetic pathway. Correspondingly, Chamovitz and co-workers observed mis-regulation of genes coding for sterol-acyl-transferase and acetyl-CoA acyl-transferase (orthologue of FAS1) in Drosophila csn5 null mutants .
It will be interesting in the future to assess whether HMG-CoA synthesis can be regulated by CSN also in Drosophila, plants and mammals. From this point of view it is noteworthy that, in A. thaliana, the orthologue of S. cerevisiae acyl-CoA reductase (HMG1/HMG2) is strongly downregulated in csn5 null mutants ( and Table S4), confirming that the control of mevalonate synthesis, a rate-limiting step in isoprenoid and sterol biosynthesis, by the CSN is indeed conserved through evolution (Fig. 1). On the other hand, the orthologue of the acetyl-CoA carboxylase (ACC1), another key regulator of histone acetylation , is upregulated in A. thaliana csn5 null mutants, confirming that a defect in CSN levels could have important effects in histone modification.
S. cerevisiae is an excellent model organism for studying the regulation of lipid metabolism in eukaryotes, as most of the regulatory mechanisms are conserved between yeast and mammals  and there are good chances that this CSN regulatory role could be conserved also in mammals.
Our transcriptomic analysis also points to a CSN function in regulating, directly or indirectly, zinc metabolism and uptake. The genes coding for the zinc high affinity transporters Zrt1 and Zrt3, as well as the Zn/Fe transporter Fet4 to a lesser extent, and the zinc transcription factor Zap1, are all downregulated at the transcript level. As a result, ∆csn5 cells are sensitive to low Zn, Ni and Cd concentrations in the medium. It is possible that, similarly to what we suggested for the RP and ergosterol genes, CSN might control the stability of some transcription factors involved in the regulation of the metal transporter genes through the ubiquitin ligase activity. Alternatively, the differential transcription of these genes in the ∆csn5 might be simply the consequence of the reduced ergosterol content in the membrane. Membranes with lower ergosterol content are known to be more permeable to cations . As a result, zinc, when present at high concentration in the medium, could easily accumulate in the cytoplasm and reduce transport gene expression by negative feedback regulation. The positive regulatory effect of zinc and copper on ergosterol biosynthesis which was recently demonstrated  could also be explained by a feedback regulation of divalent cation influx. Indeed, in support of the hypothesis that the zinc-related phenotypes of the ΔCSN5 mutant are due to its lower ergosterol content, we show here that also a strain deleted in Erg6 accumulates zinc in the cytoplasm (Fig. 4). At present we cannot rule out that zinc storage in the vacuole is also defective in the Δcsn5 strain and that this can contribute to the increased zinc accumulation in the cytoplasm. Indeed, the ∆csn5 mutant has a negative genetic interaction with mutants deleted in ZRC1 , which encodes the main zinc importer in the vacuole. Moreover, among the downregulated proteins in ∆csn5 (Table 3) we found Vtc4, which is a subunit of the vacuolar transporter chaperone (VTC) complex that is involved in synthesis and transport of polyphosphate to the vacuole . This complex regulates membrane trafficking and zinc storage in the vacuole and deletion of VTC4 has been previously shown to decrease resistance to zinc . Imbalance in zinc or cadmium uptake has previously been shown to cause alterations in free amino acid pools  and the expression of some of the proteins which we found significantly downregulated in ∆csn5 such as Mae1, Leu1, Bat1, Aro10 and Met6 are directly responding to zinc levels through a complex transcriptional and post-transcriptional regulatory system . Further work will be required in order to establish a defined cause–effect relationship between metal influx deregulation and proteomic alterations in ∆csn5.
Taken together, our results suggest that Csn5 has a pervasive effect on budding yeast metabolism. The effects of Csn5 on yeast metabolism seem to be a hallmark not only of Csn5 but of the whole CSN holocomplex, since a panel of selected genes were found by us to be regulated in a very similar fashion in mutants of four other CSN subunits. It is still not clear, however, whether their regulation is neddylation/deneddylation dependent or independent and additional studies are required. Future studies on a double ∆csn5/rub1 mutant may help to solve this issue. A recent report  indicates that SCF complexes from csn mutants in budding yeast fail to release substrate adaptors, which remain bound to the complex even in the absence of the substrate and delay the formation of new SCF complexes. A similar scenario could be envisaged for the csn mutants from S. cerevisiae, where an altered stability of specific transcription factors and of their cognate substrate adaptors might be responsible for the observed defects. Clearly, further work lies ahead to shed light on this hypothesis.
Materials and methods
Yeast strains and growth conditions
The yeast strains and plasmids used throughout this work are described in Table 5. All yeast strains were grown in YP medium supplemented with 2% glucose (YPD). Limiting zinc medium (LZM) was prepared in the same manner as low-iron medium (LIM)  except that ZnSO4 in LIM was replaced with 10 μm FeCl3 in LZM. Cell numbers in liquid cultures were determined by measuring the optical density at 600 nm (A600). For zinquin staining experiments cells were grown in YDP* (YPD with 1.5 mm triptophane, 80 mm adenine hemisulfate, 0.6 mm methionine, 10 mm succinic acid and 15 mm potassium bicarbonate) (Sigma-Aldrich, St Louis, MO, USA) as described previously .
RNA was extracted from 20 mL of S. cerevisiae cell cultures at A600 = 1 (strains were grown in YPD medium at 30 °C). Cells were suspended in 1 mL of AE buffer (50 mm sodium acetate pH 5, 10 mm EDTA) (AppliChem GmbH, Darmstadt, Germany), centrifuged and suspended in 0.4 mL of AE buffer plus 1% (w/v) SDS (AppliChem GmbH). Cells were lysed with phenol : chloroform (5 : 1, pH 4.7, Sigma-Aldrich), heated at 65 °C for 10 min, transferred at −80 °C for 10 min and the aqueous phase was separated by centrifugation. After a second extraction with phenol : chloroform (24 : 1, pH 5.2, Sigma-Aldrich), RNA was precipitated with ethanol, dried and suspended in sterile water. RNA quantity and purity were assessed with a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA) at 260 nm and at 260/230, 260/280 nm ratios, respectively. RNA integrity was assessed by electrophoresis on ethidium bromide stained 1% agarose-formaldehyde gels.
20 μg of total RNA extracted from cell culture as described above was mixed with 2 μg of 16mers oligo dT and incubated at 70 °C for 10 min. cDNA was synthesized in a final volume of 40 μL with 25 mm each of dATP, dCTP and dGTP, 15 mm of dTTP, 10 mm of aminoallyl-dUTP, 10 mm dithiothreitol and 400 U of SuperScript III reverse transcriptase (Life Technologies, Carlsbad, CA, USA) in 1× reaction buffer.
The samples were incubated for 2 h at 42 °C. RNA was hydrolyzed with 0.1 m NaOH, incubated for 10 min at 70 °C and subsequently neutralized.
cDNA was labeled with Cy3 and HyPer5 Post-Labeling Reactive Dye (GE Healthcare cod. 28-9224-19, Little Chalfont, UK) according to the manufacturer's protocol. The labeled cDNA was purified using QIAquick® PCR Purification Kit (Qiagen, Venlo, The Netherlands) and eluted with 30 μL of double distilled H2O. The NanoDrop 1000 spectrophotometer was used to quantify Cy3 and HyPer5 incorporation.
Hybridization and image acquisition of microarray
The microarrays used for analysis were the cDNA Microarray Yeast 6.4k ver. 7 purchased from UHN Microarray Centre (Toronto, Canada, http://data.microarrays.ca/arrays/index.htm). Slides were pre-hybridized at 42 °C for at least 45 min in a solution containing 5 × saline sodium citrate buffer (NaCl/Cit), 0.1% SDS and 0.1% BSA. The labeled cDNAs (Cy3 sample and HyPer5 sample mixed) were added to an equal volume of hybridization buffer containing 50% formamide, 10 × NaCl/Cit and 0.2% SDS pre-heated at 70 °C for 3 min.
Hybridization was carried out for 16 h at 42 °C and unbound DNA was washed off using three steps with solutions containing 1 × NaCl/Cit, 0.2% SDS pre-heated at 42 °C; 0.1 × NaCl/Cit, 0.2% SDS; two times 0.1 × NaCl/Cit. A PerkinElmer ScanArray Gx Plus Microarray Scanner (PerkinElmer, Waltham, MA, USA) was used to acquire images, and genepix pro 6.1 software and scanarray express software were used to quantify hybridization signals. Absent and marginal spots were flagged automatically by the software and subsequently each slide was inspected manually.
Microarray data analysis
We filtered the data to exclude artifacts, saturated spots and low signal spots. Assuming that most of the genes have unchanged expression, the Cy3/HyPer5 ratios were normalized using Goulphar script  (http://transcriptome.ens.fr/goulphar/index.php) running on r software using a Global Lowess Normalization. The parameters used for the hierarchical clustering were the Euclidean distance and the average linkage method.
1 μg of total RNA extracted from cell cultures was reverse transcribed using 200 ng of 16mers oligo dT (Life Technologies) with SuperScript III First-Strand Synthesis System for RT-PCR (Life Technologies), according to the manufacturer's instructions. cDNA served as template for subsequent real-time PCR reactions that were set up in duplicate for each sample using the SensiMix SYBR Mix (Bioline, London, UK) and an Applied Biosystems Prism 7300 Sequence Detector. The reaction mixtures were kept at 95 °C for 10 min, followed by 40 cycles at 95 °C for 15 s and 60 °C for 1 min. The threshold cycle (Ct) was calculated using the Sequence Detector Systems version 1.2.2 (Life Technologies) by determining the cycle number at which the change in the fluorescence of the reporter dye (ΔRn) crossed the threshold. To synchronize each experiment, the baseline was set automatically by the software. Relative quantification was carried out with the 2−ΔΔCt method , using the abundance of actin transcript as endogenous housekeeping control. Data were statistically analysed by Student's t test.
Proteins were extracted from 20 mL of S. cerevisiae cell cultures at A600 = 1 (strains were grown in YPD medium at 30 °C).
The yeast culture was centrifuged at 1200 g for 3 min at 4 °C. The cell pellet was washed with 10 mL of 1 × Tris buffered saline (NaCl/Tris) (Sigma-Aldrich) and centrifuged as described above. Following centrifugation, the cell pellet was suspended in 500 μL lysis buffer containing 50 mm Tris/HCl (pH 6.8) (Applichem), 1% (w/v) SDS (Applichem), 1% (v/v) 2-mercaptoethanol (Sigma-Aldrich) and protease inhibitor cocktail (2.5 μg·mL−1 aprotinin, 2.5 μg·mL−1 chymostatin, 2.5 μg·mL−1 leupeptin, 0.5 μg·mL−1 pepstain A) (Sigma-Aldrich).
The cells were lysed in the presence of 200 μL of glass beads (Sigma-Aldrich) (425–600 μm, acid washed) by vortexing in MultiVortex at 4 °C for 30–40 min and centrifuged at 16 000 g for 10 min at 4 °C.
The supernatant containing the proteins was collected, drop-frozen in liquid nitrogen and stored at −20 °C. The protein concentration of yeast lysate was determined by Bradford assay (Bio-Rad Laboratories, Hercules, CA, USA) using BSA (Sigma-Aldrich) as a standard.
Sample preparation for label-free proteomics
Proteins were digested using the filter-aided sample preparation (FASP) method . Briefly, 65 μg of protein was loaded on the filter and washed twice with buffer containing 8 m urea. The proteins were then alkylated using iodoacetamide, and the excess reagent was washed through the filters. The reduced and alkylated proteins were digested overnight in a wet chamber at 25 °C using endoproteinase LysC, which cleaves at the C terminus of lysine residues, with an enzyme to protein ratio of 1 : 50. Then trypsin was added with an enzyme to protein ratio of 1 : 100 and digestion was stopped after 4 h. Peptides obtained by FASP were micro strong anion exchange fractionated into six fractions and finally loaded onto C18 StageTips.
Mass spectrometric analysis
LC-MS/MS experiments were performed on a nano-HPLC system Ultimate 3000 connected to an Orbitrap XL Discovery equipped with a nanoelectrospray source (Thermo Fisher Scientific). Each peptide sample was auto-sampled and separated in a 10 cm analytical column (75 μm inner diameter) in-house packed with 3 μm C18 beads (Magic C18AQ 200 Å, Michrom Bioresources Inc., Auburn, CA, USA) with a 160 min gradient from 5% to 80% acetonitrile in 0.1% formic acid. The effluent from the HPLC was directly electrosprayed into the mass spectrometer.
The Orbitrap MS instrument was operated in data-dependent mode to automatically switch between one full-scan MS and five MS/MS acquisitions. Survey full-scan MS spectra (from m/z 300 to 2000) were acquired in the Orbitrap detector with resolution R = 30 000 at m/z 400. The five most intense peptide ions with charge states ≥ 2 were sequentially isolated with an isolation window of 2 Th to a maximum target value of 500 000 using automatic gain control and fragmented by collision induced dissociation in the linear trap using a normalized collision energy of 35 and activation time of 30 ms. Dynamic exclusion was used to minimize the extent of repeat sequencing of the peptides, and singly charged peptides were excluded from sequencing throughout the run.
Standard mass spectrometric conditions for all experiments were spray voltage 1.9 kV; no sheath and auxiliary gas flow; heated capillary temperature 275 °C.
All raw data analysis was performed with maxquant computational proteomics platform (www.maxquant.org)  version 18.104.22.168 supported by Andromeda (www.andromeda-search.org)  as the database search engine for peptide identifications. Mass tolerance for searches was set to an initial precursor mass window of 6 ppm and a fragment mass window of 0.5 Th. Data were searched with carbamidomethylation as a fixed modification and protein N-terminal acetylation, methionine oxidation and lysine di-glycine modification as variable modifications. A maximum of two mis-cleavages was allowed while protease specificity was set to trypsin. We used Andromeda to search the data against a concatenated target/decoy (forward and reversed) version of the Yeast ORF database containing 6301 protein entries. The cut-off false discovery rate for proteins and peptides was set to 0.01, and peptides with a minimum of seven amino acids were considered for identification.
Label-free quantification and statistical analysis
Label-free quantification was performed in maxquant. This included quantification of peptides recognized on the basis of mass and retention time but identified in other LC–MS/MS runs (‘match between the runs’ option in maxquant). Feature matching between raw files was enabled, using a retention time window of 2 min. maxquant data were filtered for reverse identifications (false positives), contaminants and ‘only identified by site’. Only proteins that had a label-free quantification intensity in at least three of the four biological replicates of each strain were included for statistical and clustering analysis.
An example of the data quality is given in Table S5 for Erg2, Erg11 and Erg13 proteins but similar data for all the other proteins included in Table 3 are available on request.
Data were evaluated and statistics calculated using the perseus software (version 1.2, Max Planck Institute of Biochemistry, Martinsried, Germany).
Saponification and gas chromatography analysis
Yeast cells were grown to stationary phase at 30 °C. Cells were pelleted in 50 mL conical tubes for 5 min at 5000 g. Cells were then resuspended in 3 mL alcoholic KOH (25% w/v) and transferred to glass tubes for refluxing at 87 °C for 2 h. After cooling to room temperature 3 mL of n-heptane and 1 mL sterile H2O were added to extract the non-saponifiable lipid fraction. Gas chromatography was routinely carried out on an HP5890 series II utilizing a fused silica DB5-MS capillary column (15 m × 0.25 mm × 0.25 μm film thickness), with nitrogen as carrier gas using the Hewlett Packard chemstation software for quantitation .
Samples were run in a semi-splitless mode with a starting temperature of 195 °C for 1 min increasing to 240 °C in 20 °C·min−1 increments and then 2 °C·min−1 increments to a final temperature of 280 °C which was held for an additional 5 min.
Nystatin and ketoconazole assay
The same amount of cells (104·mL−1) for each sample were spotted in a maxi-well plate with different nystatin and ketoconazole (Sigma-Aldrich) concentrations and incubated at 30 °C overnight. Cell number in liquid cultures was determined by measuring the optical density of cell suspensions at 600 nm (A600).
Yeast strains were grown in YPD* with 500 μm of ZnSO4 and without zinc as control to an optical density of A600 = 1. Cells was pelleted from liquid culture (23 °C, 1000 g, 1 min). The cells were immediately resuspended and washed twice with 1 mL of a buffer solution of 50 mm Tris base and 1 mm sodium azide. Cells to be stained were suspended in 50 μL buffer, diluted with a pre-mix of 1 μL of dimethyl sulfoxide and 1 μL of zinquin ethyl ester (Sigma-Aldrich), mixed gently, and incubated at 23 °C for 60 min with occasional mixing. The cells were then washed with buffer, resuspended in three pellet volumes of supernatant, and prepared for microscopy (Zeiss Motorized Axio Imager Z1 Fluorescence Microscope, Zeiss, Oberkochen, Germany). Zinquin stock solutions were stored in the dark at 23 °C.
Yeast strains were grown in LZM medium plus different ZnSO4 concentrations (0–600 μm) for 16 h at 30 °C. Cell number in liquid cultures was determined by measuring the optical density of cell suspensions at 600 nm (A600).
This work was supported by FIRB 2011-2013 (grant no. RBIN06E9Z8) ‘Molecular Bases of Diseases’, PRIN 2009 ‘Role of S. cerevisiae General Regulatory Factors (GRF) in Chromatin Organization and Dynamics’ and Progetti di Ricerca di Ateneo (grant no. C26A1139XY) to RN; Israel Ministry of Science and Technology (MOST) – Italy Ministry of Foreign Affairs (MAE) grant 3-9022 to RN, TR, GS and EP; and Israel Science Foundation grant (EP355/10) for EP; Progetto di Ricerca C26A1089CJ (2010), Sapienza Universita' di Roma, to RN, TR and GS. VL's fellowship is supported by a grant from Regione Lazio.