Genome-wide expression screens indicate a global role for protein kinase CK2 in chromatin remodeling

Protein kinase CK2 (also known as casein kinase 2) is a serine/threonine phosphotransferase that occurs ubiquitously in eukaryotes. CK2 is a tetramer,composed of two catalytic subunits, α, and two regulatory subunits,β, formed through stable β-β dimer intermediates that complex with α to constitute the α₂β₂holoenzyme (Niefind et al.,2001). Isoforms of these subunits may exist and may play tetramer-independent roles in the cell. Humans have one type of regulatory subunit and two catalytic subunit isoforms (α and α′) and thus form α₂β_2,α′ ₂β₂ andαα′β ₂ holoenzymes(Bodenbach et al., 1994),whereas the budding yeast Saccharomyces cerevisiae has two catalytic(Cka1 and Cka2) and two regulatory (Ckb1 and Ckb2) subunits that appear to build an obligatory heterotetramer(Glover, 1998). In yeast,deletion of any of the four subunit genes (CKA1, CKA2, CKB1 and CKB2) individually, and CKB1 and CKB2 together, has no overt phenotypic effect in normal media, but such strains exhibit varying degrees of sodium sensitivity. Deletion of CKA1 and CKA2together is lethal, but can be rescued by catalytic subunits of other organisms such as fly and man (Padmanabha et al., 1990; Dotan et al.,2001), demonstrating that protein kinase CK2 is both vital for cells and conserved over large evolutionary distances.

Despite being one of the earliest protein kinases discovered(Burnett and Kennedy, 1954),the biological role of CK2 is still far from being completely characterized. This is mainly a consequence of CK2's marked pleiotropic nature. So far, more than 300 substrates have been identified(Pinna, 2002). The nature of these substrates – predominantly proteins related to transcription and transcription-directed signaling – has led to the assumption that CK2 plays a global role in cell regulation, with roles particularly in gene expression and signal transduction. Many CK2 functions appear to be individually dispensable, but their collective loss is not. On the basis of its vital importance and known interactions with components of stress signaling, growth signaling and survival signaling pathways, a role for CK2 as a survival factor has been proposed (Ahmed et al., 2002).

In S. cerevisiae, CK2 is required for G2/M and G1 phases of the cell cycle (Glover, 1998). Previously, we have shown that the cell cycle entry of human cells from G0 and the subsequent early progression through G1 also requires CK2; its perturbation in cultured human cells at either the nucleic acid level (by antisense oligonucleotides) or the protein level (by microinjection of antibodies or substrate peptide analogs) significantly inhibits this process(Pepperkok et al., 1993; Pepperkok et al., 1994; Lorenz et al., 1993; Lorenz et al., 1999). Concomitantly, expression of immediate early genes such as fos is suppressed (Pepperkok et al.,1993). The immediate early gene products, for their part, trigger waves of gene expression, reflecting phase-specific hierarchical transcription programs (Iyer et al., 1999). When cells re-enter the cell cycle, they either resume proliferation or go into apoptosis. Because cells are able to remain in G0 for any length of period, re-entry may, in a cell sociological context, have important effects,including cancer development. CK2 has been linked to such pathophysiological processes (ole-MoiYoi, 1995; Seldin and Leder, 1995; Landesmann-Bollag et al.,2001).

CK2 may carry out its global role in gene expression through common transcriptional features and/or through higher-order structural processes such as nucleosomal organization. The mechanism CK2 uses is not known, but it should be disclosed by comprehensive comparison of CK2-affected genes. We examined the situation at cell cycle entry by using a comparative genome-wide expression analysis of S. cerevisiae CK2 subunit deletion strains. The data not only provide a comprehensive overview of CK2-linked gene expression, but also reveal specific contributions of the individual CK2 subunits and identify functional connections. Several CK2-affected genes encode metabolic pathway and spindle pole body (SPB) components as well as cell cycle control proteins such as cyclins and are differentially affected by the respective CK2 subunits. Strikingly, most CK2-linked genes lack common control features, and a significant proportion of temporarily altered genes encode chromatin-remodeling and modification proteins. Together with available data on CK2's association with and phosphorylation of diverse chromatin components and modifiers, our results strongly suggest a global role for protein kinase CK2 in nucleosomal remodeling processes that are particularly important at transition points such as cell cycle (re-)entry.

The CK2 mutant strains ckb1Δ ckb2Δ[YAPB10-2c, MATaCKA1 CKA2 ckb1-Δ1::HIS3 ckb2-Δ1::LEU2 (Bidwai et al., 1995)], cka1Δ [JC4-1a, MATacka1-Δ1::HIS3 CKA2 CKB1 CKB2(Chen-Wu et al., 1988)] and cka2Δ [RPG22-1b, MATaCKA1Δ cka2::TRP1 CKB1 CKB2(Padmanabha et al., 1990)] as well as their respective wildtypes (YPH499, MATaCKA1 CKA2 CKB1 CKB2 and YPH250, MATaCKA1 CKA2 CKB1 CKB2)were generously provided by C. V. C. Glover (Biochemistry and Molecular Biology, University of Georgia, Athens, GA). All strains were grown in YPD medium at 30°C and shaken at 230 rpm.

Pheromone-based synchronization was performed as previously described(Spellman et al., 1998). Briefly, yeast cells were grown to early log phase (OD₆₀₀ 0.2) and arrested by the addition of α-factor (12 μg/ml), followed by 2 hours of cultivation. After centrifugation, arrest release was effected by resuspension of the cell pellet in fresh (pheromone-free) medium to an OD₆₀₀ of 0.18. During further cultivation samples were taken 0, 7 and 14 minutes after release by adding culture aliquots to sterile ice. Cells were immediately collected by centrifugation, and pellets were flash-frozen in liquid nitrogen. Synchrony was verified by FACS analysis as previously described (Nash et al.,1988).

The following procedures were performed as previously described(Ackermann et al., 2001). Briefly, total RNA was isolated by applying the hot phenol protocol followed by mRNA concentration using oligo(dT)-cellulose (Ambion). 3 μg of mRNA were reverse transcribed using SuperScript™ Choice System for first and second strand cDNA synthesis (Gibco BRL); 1 μg cDNA was transcribed in vitro using biotinylated ribonucleotides (ENZO BioArray™ High Yield™ RNA transcript labeling kit, ENZO Diagnostics). After fragmentation, 15 μg of biotin-labeled cRNA were taken for hybridization to oligonucleotide arrays (YG-S98 Arrays, Affymetrix) at 45°C, 60 rpm for 16 hours. Staining (performing the three-stain procedure including a signal amplification step) and washing steps (applying the protocol EukGE-WS2.v4)were carried out on a GeneChip Fluidics Station 400 (Affymetrix). Arrays were scanned on a HP GeneArray™ scanner (Affymetrix). The resulting image data were analyzed using the GeneChip Expression Analysis software (Micoarray Suite version 4.0; Affymetrix).

For data analysis, expression values corresponding to an `absent call' or to a negative value were set to one, followed by uploading data in M-CHIPS[multi-conditional hybridization intensity processing system(Fellenberg et al., 2002)]. This MATLAB-based tool was used for further analysis (normalization and filtering).

For normalization, loglinear regression accounting for affine-linear deviations among the different hybridizations(Beissbarth et al., 2000) was applied. Each hybridization experiment was normalized with respect to the gene-wise median of the control condition (time point 0 of the corresponding wildtype), which was subsequently referred to as the standard. To correct for distortions, the 5% quantile of each hybridization was subtracted initially. As the transcription levels of the majority of genes were unaltered in the different conditions, normalization factors were calculated on the basis of the majority of the spots. In contrast to Beissbarth et al.(Beissbarth et al., 2000),low-intensity signals were kept, and hybridizations were shifted additively back to a more natural range, that is, to the level of the standard(Fellenberg et al., 2001).

To visualize interdependencies among the high-dimensional data received in two-way contingency tables with rows representing genes and columns representing hybridizations, singular value decomposition(Alter et al., 2000) was used for reducing dimensionality, and the so-called chi-squared distance among the data points was approximated from below. Afterwards, genes and hybridizations were plotted together in a two-dimensional space, where expression specificities are indicated by the distances from the center.

In order to identify genes whose expression depends on or is affected by protein kinase CK2 at cell cycle (re-)entry, we performed a comparative genome-wide expression analysis of S. cerevisiae CK2 deletion strains and corresponding wildtypes. The deletions were a regulatory CK2 subunit double mutant, strain ckb1Δ ckb2Δ, and mutants of the individual catalytic subunits, strains cka1Δ and cka2Δ. All strains, grown to early log phase (OD₆₀₀0.2) in order to avoid metabolite and oxygen deficiency effects(DeRisi et al., 1997), were synchronized by α-pheromone treatment (verified by FACS analysis), and transcript profiles were determined at 0, 7 and 14 minutes after pheromone release on oligonucleotide arrays as previously described(Ackermann et al., 2001). According to Spellman et al., gene expressions peaking at 0 to 7 minutes after release are specific for M/G1 phase transition, and those at 14 minutes, for early G1 phase (Spellman et al.,1998). Data were uploaded in M-CHIPS [multi-conditional hybridization intensity processing system(Fellenberg et al., 2002)] and normalized with respect to the gene-wise median of the standard (time point 0 of the corresponding wildtype) in order to correct distortions caused by background and multiplicative factors so that the different data sets were comparable.

We found various interdependencies between gene expression profiles and CK2 perturbations. By applying correspondence analysis to the data(Fellenberg et al., 2001), the association of individual genes with a specific condition (yeast strain and time point of interest) was determined from their distance from the center and location; the further the distance from the center, the more pronounced the association of the gene with the considered condition(Fig. 1). For instance, SLK19, a G1 phase gene whose product is involved in spindle dynamics control, is modestly repressed in the ckb1Δ ckb2Δ strain but strongly repressed in strains cka1Δ and cka2Δ and thus only slightly(Fig. 1A) or significantly(Fig. 1B,C) shifted to wild-type areas, respectively. In addition to such strain-specific transcript deviations, time-point-dependent deviations are also indicated for each gene:transcript levels of YGP1, for example, a M/G1 phase gene encoding a starvation-induced glycoprotein, were elevated at all time points in strains ckb1Δ ckb2Δ and cka2Δ but not in cka1Δ so that the gene is localized at a significant distance from the center along the 7 minutes mutant axis(Fig. 1A,C) or within the center (Fig. 1B), respectively. By contrast, TSM1, a G2/M phase gene encoding a TATA-binding protein-associated factor (TAF150), increased transcription at the later time points in strains cka1Δ and cka2Δ but not ckb1Δ ckb2Δ, positioning TSM1 farther from the center, between the 7 and 14 minutes mutant axes(Fig. 1B,C), or within the center, respectively. In summary, correspondence analysis shows CK2-linked expression of various genes at cell cycle entry and indicates specific contributions by the individual CK2 subunits.

By considering the greater than two-fold transcript deviations in at least one of the CK2-deficient strains at one of the time points as significant, we identified 283 genes of the ∼900 characterized as cell-cycle-regulated in the literature (Spellman et al.,1998; Cho et al.,1998) as linked to CK2 in their expression(Fig. 2A). Most transcriptional alterations (140) were observed in the regulatory subunit double mutant. Eighty of these were ckb1Δ ckb2Δ specific, the rest was also altered in the catalytic subunit mutants. In the latter, fewer genes were concerned; 56 and 47 were cka1Δ and cka2Δ specific, respectively. Regarding time point distribution, the total number of affected genes were comparable at 0 and 7 minutes (157 and 155 genes, respectively) but lower at 14 minutes (128 genes),suggesting stronger CK2 perturbation effects at earlier cell cycle stages(Fig. 2B).

Surprisingly, the 283 genes affected by CK2 perturbation at cell cycle(re-)entry do not represent a collection of early genes as one might expect from observations with human cultured cells(Lorenz et al., 1993; Pepperkok et al., 1993; Pepperkok et al., 1994). Rather, according to their peak expression(Spellman et al., 1998; Cho et al., 1998), the genes can be assigned to various cell cycle stages. Aside from affected early genes(37 M/G1 phase and 95 G1 phase genes), genes ascribed to S phase (30 genes),S/G2 phase (44 genes), G2/M phase (71 genes) and diverse phases (6 genes) also exhibited altered transcriptional levels(Fig. 3A). For all of these groups, the number of genes deviating in expression remained more or less the same over time, except for G1 phase genes. Their number decreased steadily so that less than half of the alterations at 0 minutes were seen at 14 minutes(data not shown), suggesting particularly strong CK2 requirement for early G1-specific gene expression.

When the CK2 mutant strains were grown as asynchronous permanently cycling cultures (early log phase, OD₆₀₀ 0.46), the transcript levels of 106 cell-cycle-regulated genes were altered. This is a significantly smaller number than that obtained for the cell cycle (re-)entry cultures above. Strikingly, only 23 G1-specific genes were altered(Fig. 3B), indicating that CK2 perturbation in asynchronous cultures does not primarily affect G1-specific genes. G2/M-specific genes form the largest group of affected genes (28 genes); the number of M/G1 genes (21 genes) was similar to that of G1 genes,and 14 and 15 expression alterations were found for S- and S/G2-specific genes, respectively.

By comparing the results from cells synchronized at cell cycle entry with those in permanently cycling cultures, 58 were identified that were altered in both cultures. 40 of these were found in the regulatory CK2 subunit mutant, 16 being ckb1Δ ckb2Δ specific. In the catalytic CK2 subunit mutants, 29 and 25 genes were altered, and 5 and 6 were cka1Δ and cka2Δ specific, respectively(Fig. 4). Thus, deletion of the regulatory CK2 subunits not only affects expression of more cell cycle genes than deleting a catalytic subunit, it is also characterized by a significantly stronger persistence. The names, expression and cell cycle phase-association of these 58 persistently altered CK2-linked genes, together with the physiological roles of their respective gene products, are given in Table 1.

More than two thirds of the persistently altered genes exhibited the same deviation characteristics in terms of transcription elevation/repression owing to a certain CK2 mutation in both asynchronous and early cell cycle cultures. For instance, the expression of PCL2, a gene encoding a G1 cyclin,was elevated predominantly in the cka2Δ mutant, whereas MIF2, encoding a centromer protein (the human CENP-C homologue), was strongly repressed in the ckb1Δ ckb2Δ mutant. A particularly interesting situation was provided by the S phase genes. The gene products of all seven continuously altered S phase genes (MET6, MET10,MET14, STR3, ECM17, ICY2, SAM3) are involved in methionine biosynthesis and belong to a 20-gene `MET' cluster peaking coordinately during the cell cycle (Spellman et al.,1998). Remarkably, although repressed in the ckb1Δ ckb2Δ and cka2Δ mutants, significantly altered MET genes exhibited exclusively elevated transcript levels in the cka1Δ mutant. We also noticed this CK2-subunit- and isoform-dependence for the S/G2-specific MET cluster gene MET16. In summary, the MET genes represent an excellent example of subunit- and isoform-specific CK2 perturbation effects that are apparently not restricted to early cell cycle expression but rather persist throughout the cell cycle.

ECA39, the mammalian homologue of the S/G2 gene BAT1, is regulated by c-Myc and promotes apoptosis in murine cells(Eden and Benvenisty, 1999). Yeast BAT1, encoding a branched-chain amino acid transaminase, is probably involved in the regulation of the G1/S transition, as bat1Δ strains exhibit elevated growth rates owing to a shortened G1 phase and higher rates of UV-induced mutations(Schuldiner et al., 1996). The persistent BAT1 repression in the ckb1Δ ckb2Δ mutant suggests an apoptosis-inhibiting CK2 function. Evidence for a role of CK2 in apoptotic cellular processes not only via gene regulation but also at the protein level is increasing. For instance, Li et al. have recently shown that CK2-mediated phosphorylation is required for activation of a caspase-inhibiting protein in order to prevent apoptosis(2002).

Other S/G2 genes as well as G2/M and M/G1 genes with similar deviation characteristics in early cell cycle and asynchronous cultures are involved in amino acid metabolism and nutrition supply. These include genes for the amino acid permease Gap1, the aromatic amino acid aminotransferase Aro9, the acid phosphatase Pho5 and the propionate metabolism-associated protein Pdh1.

The remaining third of permanently altered genes exhibited diverse deviation characteristics. Several of the genes encode proteins involved in the pheromone signaling pathway, including FUS1, KAR4 and SST2, which encode a MAP kinase, a transcription factor, and a GTPase-activating protein, respectively. In the ckb1Δ ckb2Δ mutant, the genes were strongly repressed during the early cell cycle but elevated in asynchronous culture, whereas in the cka2Δ mutant they were unaltered in the early cell cycle and repressed in asynchronous culture.

In contrast to genes associated with the pheromone pathway, MET cluster and nutrition supply, various gene groups that constitute considerable portions of the CK2 perturbation-affected genes at cell cycle (re-)entry (see below) were significantly underrepresented among the genes deviating in permanently cycling CK2 mutant cultures. For example, of ∼30 genes linked to chromatin remodeling (CHR) and/or spindle pole body (SPB) organization that exhibited expression deviations at cell cycle (re-)entry, only two genes (MIF2and ICY2) were altered in asynchronous cultures. Similarly, of at least 15 genes deviating at cell cycle (re-)entry whose products are involved in DNA replication and cell cycle control, only PCL2 and NDD1 showed altered transcript levels in asynchronous cultures. This contrast clearly indicates that, although CK2 perturbation considerably affects early cell cycle expression of genes crucial for the (re-)entry, yeast cells are obviously able to compensate for these transcriptional deviations during the course of cell cycle progression.

After subtraction of the 58 permanently altered genes, the remaining 225 genes should more closely reflect CK2-linked genes specifically related to cell cycle (re-)entry. Below, we present all the genes falling into this category, grouped according to when their expression peaked during the cell cycle. These data and the physiological roles of their respective gene products are presented in Table 2,Table 2,Table 2,Table 2.

84 cell-cycle-regulated genes are significantly affected by CK2 perturbation at cell cycle (re-)entry peak during G1. Generally, if deviation tendencies are also considered, we found that 90% of altered G1 genes were altered in at least two of the three analyzed CK2 subunit deletion strains. About two thirds of these 84 genes exhibited similar deviation characteristics(repression, elevation) in the CK2 mutants. The rest included genes, such as HSN1 or YJR154W, whose expression was oppositely affected,providing further evidence for a functional specialisation of the catalytic CK2 isoforms and autonomous roles for the regulatory subunits, as already noted for mammalian cells (Lorenz et al.,1999; Vilk et al.,1999). Most of the significantly altered G1-specific genes were repressed. One group of such genes is represented by four members of the 11-gene TOS family (TOS6, 7, 10, 11). TOS originally comes from a multicopy suppressor of DNA topoisomerase II (TopII) mutants, named TOS1 (Thomas et al.,1991). TopII is regulated by CK2-mediated phosphorylation(Alghisi et al., 1994);however, as long as the biological function of the TOS gene family is not fully understood, a possible connection to CK2 remains speculative.

Because of their temporal coregulation with CLN2 transcription, 76 G1-specific genes peaking at mid-G1 have been categorized into a CLN2 cluster(Spellman et al., 1998), which contains many genes involved in DNA replication and other important cell cycle functions. Collectively, we found 14 altered CLN2 cluster genes, including CDC45, RNR1 and CLN2 itself. Several CLN2 cluster genes are induced by the transcription factors MBF and/or SBF(Spellman et al., 1998). Six of the 21 known MBF-regulated genes and 10 of the 24 SBF-regulated genes, as determined by Simon et al. (Simon et al.,2001), were altered; however, we did not observe a common deviation pattern. The related heterodimers SBF (Swi4/Swi6) and MBF(Mbp1/Swi6) provide crucial G1-specific transcriptional regulation in the yeast cell cycle. One of their main functions is the transcriptional activation of the G1/S cyclin genes CLN1 and CLN2. SBF and MBF clearly do not work alone; the list of factors that influence SBF and MBF activities or cooperate with them is growing. Cdc68, a general transcriptional regulator, is an activator of their components Swi4 and Swi6(Lycan et al., 1994) and has been identified by Glover as a high probability CK2 substrate relevant to cell cycle progression (Glover,1998). Recently, Cdc68 has been found to be associated with all four CK2 subunits in yeast (Gavin et al.,2002; Ho et al.,2002). Possible phosphorylation of Cdc68 by protein kinase CK2 might explain the polymorphic deviation patterns for SBF/MBF-regulated genes in the CK2 mutants, because mutations in CDC68 can repress or activate the regulation of transcription (Wittmeyer et al., 1997).

During G1, the SPB replicates and expression of several SPB components consistently peaks at this stage (Spellman et al., 1998). A remarkably high number of G1 genes affected in our study encode proteins required for SPB assembly or spindle movement control: the SPB component genes NUF1 and CNM67 exhibited divergent expressional alteration characteristics in the diverse CK2 mutants,whereas HCM1 and PAC11, whose gene products are needed for SPB assembly, were primarily repressed. Repression was also observed for SLK19, IPL1 and KAR3, which encode proteins participating in spindle dynamic control. By contrast, transcript levels of SMC1 and CSM2, whose products are also directly or indirectly involved in chromosome segregation processes (CSM2 in meiosis), were exclusively elevated. Thus, in the early cell cycle, CK2 perturbation causes complex expressional changes in spindle-function-associated genes.

Three of the affected G1 genes are involved in chromatin remodeling: CAC2, encoding a chromatin assembly factor subunit, had elevated expression in the cka2Δ strain, whereas ESC4 and ASF1, encoding proteins with chromatin silencing and anti-silencing functions, respectively, exhibited divergent expression deviations depending on the CK2 subunit(s) deleted. Recently, ASF1 deletion in yeast has been reported to induce cell death with apoptotic features(Yamaki et al., 2001). An interpretation of the ASF1 transcript profiles (slightly repressed in ckb1Δ ckb2Δ; significantly elevated in cka1Δ) in this context seems difficult.

In common with the G1 genes, we found that most of the 23 S phase genes that show significant deviation in expression are repressed. In the ckb1Δ ckb2Δ strain, only a single gene, STU2, whose product may play a role in organization of microtubule ends at SPBs, exhibited significantly elevated transcript levels. In the catalytic subunit mutants, the repression:elevation distribution was more balanced. In addition to STU2, another microtubule/SPB-associated S phase gene, DAM1, which encodes a protein required for spindle integrity and kinetochore function, was altered (elevated at 0 minutes in both cka1Δ and cka2Δ strains). This is consistent with the diverse expression deviations of spindle-function-linked G1 genes(see above). In common with the G1 group, certain S phase genes that are significantly affected by CK2 perturbation, such as SWI1 and TEL2, are associated with chromatin-remodeling processes. In addition to the eight MET cluster genes also altered in asynchronous cultures (see above), we identified another three MET genes (MET13, SER33 and MET28) exhibiting virtually the same isoform/subunit-specific expression deviations (repression in ckb1Δ ckb2Δand cka2Δ, elevation in cka1Δ). For these genes,the corresponding deviation tendencies were also observed in the asynchronous culture (data not shown).

35 S/G2-specific genes exhibited CK2-perturbation-linked expression deviations at cell cycle (re-)entry and again repression was the major deviation. As before, genes linked to SPB/microtubule organization, such as KIP3 (encoding a kinesin-related protein) and CIK1 (encoding a SPB-associated protein), and genes whose products are involved in chromatin remodeling, such as HOS3 (encoding a histone deacetylase) and CSE4 (encoding a CENP-A homolog), exhibited diverse deviation characteristics. In this context, expression of FKH1, encoding a forkhead transcription factor involved in cell cycle control, was continuously elevated in the ckb1Δ ckb2Δ strain, but unaltered in cka1Δ and slightly elevated (only at 0 minutes) in cka2Δ strains, respectively. Fkh1 target gene expression is high in virtually all cell cycle phases but primarily in G2/M(Simon et al., 2001). Interestingly, Fkh1 binds to promoters of various chromatin-structure-associated genes, including the S-specific TEL2gene (see above) and HOS3 (Simon et al., 2001). Deviation characteristics of these genes do not allow us to ascribe these changes directly to altered FKH1expression; however, the association of Fkh1 with all four CK2 subunits(Ho et al., 2002) suggests that its transcriptional activity might be affected by CK2.

Genes that peak at G2/M phase form the second largest group of altered genes. Transcription of 54 genes was affected, with repression and elevation occurring equally frequently. Again, chromatin-structure-linked genes, such as SRI1 (repressed), whose product interacts with the Swi/SNF and RSC remodeling complex, and AHC1 (elevated), encoding a Ada histone acetyltransferase complex component, exhibited expression deviations. Remarkably, transcript levels of the G1 cyclin Cln3, whose cooperation with Cdc28 (also named Cdk1) triggers the first step of a new cell cycle in yeast,was repressed at the early time points in the catalytic subunit mutants,suggesting a possible cell cycle (re-)entry delay compared with the wildtype.

Spellman et al. defined a MCM cluster comprising 34 genes peaking at about the M/G1 boundary that is primarily involved in initiation of DNA replication(Spellman et al., 1998). Transcriptional levels of 15 MCM cluster genes classified as G2/M specific were altered in one or more of the CK2 mutants. Two interesting MCM cluster genes exhibiting significantly different deviation characteristics between the regulatory subunit double mutant and the catalytic subunit deletion strains are represented by SPO12 and TSM1, which encode a putative positive regulator of mitosis and TAF150 (see Fig. 1), respectively. TSM1 expression was significantly elevated at all time points in the catalytic subunit mutants (except cka1Δ at 0 minutes) but only showed an elevation in the ckb1Δ ckb2Δ strain at 7 minutes. Conversely, SPO12 was found continuously repressed in the regulatory subunit double mutant but virtually unaltered in the catalytic subunit deletion strains. Most MCM cluster genes contain binding sites for the transcription factor Mcm1 in their promoter regions(Spellman et al., 1998). We detected expression deviations at cell cycle (re-)entry for 10 out of the 27 genes found inside and outside of the MCM cluster identified as Mcm1-regulated by Simon et al. (Simon et al.,2001). Mcm1 is an essential, multifunctional cell cycle regulator that is highly homologous to human serum response factor(Kuo et al., 1997). Differential phosphorylation of Mcm1 may modulate its transcriptional activity on specific cofactors. Kuo et al. have identified more than eight differently phosphorylated Mcm1 isoforms (Kuo et al.,1997), including one that is necessary for the osmotic stress response. There is strong evidence that this specific Mcm1 phosphorylation is provided by protein kinase CK2, which has also been demonstrated to be involved in salt tolerance pathways(Glover, 1998). It is conceivable that CK2-mediated phosphorylation of Mcm1 – possibly at additional sites – causes expression deviations of Mcm1-regulated genes related to further physiological and cell cycle functions.

Transcription levels of 26 M/G1-specific genes were significantly affected by CK2 perturbation at cell cycle entry. Generally, repression was dominant(it was observed for two thirds of the altered M/G1 genes) and was most prominent in the regulatory subunit deletion strain. Several primarily repressed genes are associated with metabolic pathways and nutritional supply. Examples include the phosphate transporter gene PHO89 and the long-chain fatty acid CoA-ligase gene FAA1, which exhibit various degrees of repression in the three CK2 mutants. The gene products of FAA1, encoded on chromosome 15, and its putative duplication on chromosome 8, FAA4, which, by contrast, showed elevated expression in the catalytic CK2 subunit mutants, are involved in fatty acid import(Faergeman et al., 2001). Remarkably, YGP1, which encodes a glycoprotein synthesized in response to starvation conditions, exhibited elevated transcription levels in the ckb1Δ ckb2Δ and cka2Δstrains, suggesting an impaired nutrient supply owing to CK2 perturbation.

Six of the altered M/G1 genes belong to a SIC1 cluster, which comprises 27 genes peaking in late M or at the M/G1 boundary(Spellman et al., 1998). There was no common expressional deviation pattern for these genes; instead,characteristics were quite divergent with respect to the direction of alteration (repression or elevation) and to the CK2 mutants concerned. A further four SIC1 cluster genes (CTS1, YER124C, YHR143W and SCW11) affected were classified by Spellman et al. as G1 specific(see above) (Spellman et al.,1998). Similarly, two altered MCM cluster genes (HST4 and YGP1) have been characterized as M/G1 specific(Spellman et al., 1998). The fact that temporally coregulated genes are categorized into the same cluster but assigned to different cell cycle stages demonstrates that phasing of cell-cycle-regulated genes can only be crude(Spellman et al., 1998) and includes many borderline cases.

Another three cell-cycle-regulated genes that peak at multiple phases(Cho et al., 1998) exhibited significant deviations in expression. Among these genes whose transcript levels were primarily altered in the regulatory subunit double mutant were the DNA helicase gene RRM3 (repressed) and BUD2 (elevated),which encodes a GTPase-activating protein involved in bud site selection that is necessary for viability in the absence of the G1 cyclin Cln3(Benton et al., 1993).

We carried out our search for the global roles of protein kinase CK2 using genome-wide expression profiling as this type of analysis provides the necessary global data. It allows a comprehensive comparison of genes whose expression levels are affected by perturbation of CK2 and discloses specific effects of the individual CK2 subunits. Genome-wide analysis was expected to reveal whether CK2 acts through defined transcription control features common to diverse genes or through higher-order structural processes such as chromatin remodeling. In the latter case, protein kinase CK2 would act as a global (unspecific) effector of gene expression. We present here for the first time a comprehensive dataset addressing this issue (Tables 1 and 2,2,2,2).

Our analysis shows that the expression levels of a relatively high proportion of genes, roughly a quarter of the 900 estimated to be cell-cycle-regulated in yeast (Spellman et al., 1998; Cho et al.,1998), are affected by protein kinase CK2 and are related to cell cycle (re-)entry from a G0-like state (pheromone arrest). In addition, 58 genes are also altered in permanently cycling cells (asynchronous cultures),and these represent rather persistent perturbation effects. According to the expression peaks, both groups comprise genes from various cell cycle phases. While not surprising for permanently cycling cells, it is quite remarkable for cells (re-)entering the cell cycle as these cells were expected to depend on early genes. Genes with persistently altered expression are primarily involved in metabolic pathways, nutritional supply and the pheromone response. This group, therefore, is not specific for cell cycle entry per se. These genes have various expression features in common. For instance, all significantly affected MET genes, members of a gene cluster responsible for methionine synthesis, exhibit the same CK2 subunit- and isoform-specific expression alteration, that is, repression in ckb1Δ ckb2Δand cka2Δ strains but elevation in the cka1Δstrain. Similarly, transcription of the PHO genes, a gene cluster essential for phosphate supply, is strongly repressed in the ckb1Δ ckb2 Δ strain, and the same is true for their common transcriptional activator Pho4. As this activator is subject to the control of a cell-cycle-linked regulatory complex whose components are CK2-independently expressed, CK2 perturbation uncouples PHO gene expression from its regulator (Barz et al., 2003). It should be noted that as phosphate supply is vital for all cells, this observation supports the survival factor hypothesis of CK2(Ahmed et al., 2002).

The most interesting genes related to cell cycle (re-)entry fall into two groups. The first group comprises, predominantly, genes linked to cell cycle progression and exit, including genes associated with cell cycle engine control as well as cell division and apoptotic machineries(Fig. 5). The genes linked to cell cycle engine control include cyclin-encoding genes and genes encoding cyclin destruction proteins. For instance, CLN3, which encodes the cell cycle start cyclin in S. cerevisiae, is repressed in ckaΔ strains, indicating a requirement for CK2 phosphorylation function for expression. CLN2, which encodes a cyclin active at G1/S transition and is induced by the Cln3-avtivated central yeast CDK Cdc28, is repressed in ckbΔ strains, indicating a requirement for the regulatory CK2 subunits. Further, we find a Cka1-linked increase in expression in CDC20, which encodes a subunit of the anaphase-promoting complex,which is involved in proteolysis of M phase cyclins. Another exit-related gene is SPO12, which encodes a putative positive exit regulator that requires Ckb1/Ckb2 for expression. Prominent among the genes related to the cell division machinery are genes encoding SPB components and SPB interaction partners. These include HCM1, which encodes a transcription factor involved in the regulation of SPB assembly, the SPB component gene NUF1, and CIK1, which encodes another SPB-associated protein. Remarkably, these SPB genes exhibit high CK2-subunit- and isoform-specific expression dependencies.

The second group comprises genes with a striking link to chromatin remodeling and modification. Chromatin-modifying complexes are divided into two main groups: ATP-dependent chromatin-remodeling complexes that use energy for nucleosome sliding along the DNA fibre and thus disrupt histone-DNA contacts (Gasser, 2002; Becker, 2002); and histone-modifying complexes, which add or remove covalent modifications from histone tails that influence histone-DNA binding and thus affect gene activation and silencing (Cosma,2002). An exactly coordinated cooperation between chromatin-remodeling complexes, histone modifyers, general transcription factors and specific activators is required for successful transcription of a certain gene. For each cell cycle phase, we find altered expression of genes encoding proteins involved in this important process (Fig. 6). They are involved in silencing mechanisms (ESC4 and TEL2) and nucleosome assembly (ASF1, CAC2 and CSE4) or represent chromatin-remodeling components (SWI1 and SRI1), a DNA helicase (MCM6) and histone (de)acetylases (HOS3, AHC1,HST4). In common with the SPB genes, their expression dependence on different CK2 subunits and isoforms is highly heterogenous, and – in contrast to the persistent alterations – shared deviation patterns for functionally related and/or commonly regulated genes (e.g. by identical transcription factors) could not be detected. Conversely, our promoter analyses of genes assigned to one cluster according to similar deviation characteristics did not reveal any potential upstream regulatory sequences that might indicate shared transcriptional regulation (data not shown). The indicated participation of CK2 in chromatin remodeling might explain this apparently `undirected' effect of CK2 perturbation at cell cycle(re-)entry.

In addition to the indicated involvement of CK2 in the expression of chromatin-remodeling genes at cell cycle (re-)entry, evidence for CK2 implications in remodeling processes at the protein level is increasing. In yeast, all CK2 subunits have been discovered in complex with histones and general chromosomal remodeling factors(Gavin et al., 2002; Ho et al., 2002; Krogan et al., 2002). Phosphorylation of the nucleosome assembly proteins NAP-1 and -2 by CK2 is supposed to control nuclear-cytoplasmic histone translocation in fly and man,respectively (Li et al., 1999; Rodriguez et al., 2000). Furthermore, architectural non-histone chromatin components, such as the high mobility group proteins (HMGs) and the heterochromatin-associated protein 1(HP1), are phosphorylated by CK2, affecting their DNA binding and gene silencing activities as well as their interaction with specific transcription factors (Wisniewski et al.,1999; Zhao et al.,2001; Krohn et al.,2002). It has also been shown that CK2 interacts with the basic leucine-zipper domains of several transcription factors such as ATF1, CREB,c-Fos or c-Jun, and in this way indirectly binds to DNA so that it can phosphorylate not only its immediate interacting partner but also proteins bound to nearby elements (Yamaguchi et al., 1998). The yeast ATF1/CREB homologue Sko1 regulates transcription of hyperosmotic-stress-induced genes such as the sodium pump gene ENA1, whose expression is defective in ckbΔstrains (Glover, 1998). Interestingly, Sko1 is important for recruitment of SWI/SNF nucleosome remodeling complexes and SAGA histone acetylase to osmotic-inducible promoters(Proft and Struhl, 2002). CK2-mediated phosphorylation of human histone deacetylases 1 and 2 (HDAC1/2)has been reported to promote complex formation and enzymatic activity(Pflum et al., 2001; Tsai and Seto, 2002; Sun et al., 2002). Moreover,Guo et al. found that CK2-nucleosome association and transcription initiation coincide at previously inactive chromatin regions, suggesting that CK2 participates in promoting the transition to a transcriptionally active conformation (Guo et al.,1998). Consistent with this, evidence for a differential CK2-mediated phosphorylation of nucleosomal proteins depending on the state of transcriptional activity has been presented(Guo et al., 1999).

Collectively the data strongly suggest that, independent from CK2 involvement in the regulation of specific gene groups (such as the PHO or MET genes), there is a global role for protein kinase CK2 in gene expression,which is based on its functions in essential chromatin-remodeling processes. If this was the case, a perturbation of CK2 would have a general, far-reaching and partly undirected effect on gene expression that is particularly pronounced at transition points such as cell cycle (re-)entry, where switching from one gene expression program to another requires significant chromatin reorganization. That this effect is particularly reflected by altered transcript levels of remodeling-associated genes might be a simple result of their preferred induction at cell cycle entry where their gene products are particularly needed.

We thank C. V. C. Glover for generously providing the yeast strains and critical reading of the manuscript. We also thank M. Hergenhahn for assistance in the Affymetrix GeneChip system handling, Kurt Fellenberg and Christian Busold for help with M-CHIPS and the members of the Biochemical Cell Physiology group for discussion. The work was supported by the Strategiefonds III Program of the Helmholtz Gesellschaft Deutscher Forschungszentren. This article is dedicated to Harald zur Hausen on occasion of his retirement as head of the Deutsches Krebsforschungszentrum with appreciation for two decades of leadership.