Tài liệu Báo cáo khoa học: Protein kinase CK2 activates the atypical Rio1p kinase and promotes its cell-cycle phase-dependent degradation in yeast pdf

Replacement of all six serine residues by aspartate, mimicking constitutive phosphorylation, stimulates Rio1p kinase activity about twofold in vitro compared with wild-type or the corres

and promotes its cell-cycle phase-dependent degradation

in yeast

Michaela Angermayr1,*, Elisabeth Hochleitner2,†, Friedrich Lottspeich2and Wolfhard Bandlow1

1 Department Biologie I, Bereich Genetik, Ludwig-Maximilians-Universita¨t Mu¨nchen, Germany

2 Max-Planck-Institut fu¨r Biochemie, Martinsried, Germany

The protein kinase casein kinase 2 (CK2) is ubiquitous

in eukaryotes and is responsible for the Ser⁄ Thr

phos-phorylation of a large number of protein substrates

[1–3] The active holoenzyme is most often a

hetero-tetramer composed of two catalytic a subunits, a

(encoded by CKA1) and a¢ (encoded by CKA2), and

two regulatory b subunits, b and b¢ in Saccharomyces

cerevisiae(CKB1 and CKB2) The enzyme occurs in all

possible combinations of a and b subunits [4,5] In

yeast, deletion of the gene for one of the two catalytic

subunits has little effect, but deletion of both

homolo-gous genes results in loss of viability [6] To date, more

than 300 endogenous CK2 substrates are known to be

involved in quite diverse processes, e.g cell proliferation,

signal transduction, transcriptional regulation, transla-tion and metabolism [3] Despite this eminent role in strictly regulated cellular processes and although CK2

is indispensable for cell life, CK2 activity by itself is apparently unregulated [4,5], although some fluctua-tion in activity in correlafluctua-tion with cell-cycle progres-sion has been seen in cultured mammalian cells [7,8] The physiological effect of substrate phosphorylation

is surprisingly low with almost all targets described to date [9,10] In S cerevisiae, it has been shown using temperature-sensitive CKA2 alleles that protein kinase CK2 is required for passage across the G2⁄ M bound-ary and for cell-cycle progression through the G1 phase [11]

Keywords

atypical protein kinase; casein kinase 2

substrate; cell-cycle phase-dependent

degradation; protein–protein interactions;

Rio1 protein kinase

2

1 Correspondence

M Angermayr

E-mail: M.Angermayr@lrz.uni-muenchen.de

Present address

*ac-Pharma AG, Oberhaching, Germany

†Wacker Chemie AG, Burghausen, Germany

(Received 16 May 2007, revised 11 July

2007, accepted 16 July 2007)

doi:10.1111/j.1742-4658.2007.05993.x

Using co-immunoprecipitation combined with MS analysis, we identified the a¢ subunit of casein kinase 2 (CK2) as an interaction partner of the atypical Rio1 protein kinase in yeast Co-purification of Rio1p with CK2 from Dcka1 or Dcka2 mutant extracts shows that Rio1p preferentially interacts with Cka2p in vitro The C-terminal domain of Rio1p is essential and sufficient for this interaction Six C-terminally located clustered serines were identified as the only CK2 sites present in Rio1p Replacement of all six serine residues by aspartate, mimicking constitutive phosphorylation, stimulates Rio1p kinase activity about twofold in vitro compared with wild-type or the corresponding (S > A)6 mutant proteins Both mutant alleles (S > A)6 or (S > D)6 complement in vivo, however, growth of the RIO1(S > A)6 mutant is greatly retarded and shows a cell-cycle pheno-type, whereas the behaviour of the RIO1 (S > D)6mutant is indistinguish-able from wild-type This suggests that phosphorylation by protein kinase CK2 leads to moderate activation of Rio1p in vivo and promotes cell pro-liferation Physiological studies indicate that phosphorylation by CK2 ren-ders the Rio1 protein kinase susceptible to proteolytic degradation at the

G1⁄ S transition in the cell-division cycle, whereas the non-phosphorylated version is resistant

Abbreviations

Aky2p, adenylate kinase 2; CK2, casein kinase 2; GST, glutathione S-transferase.

Rio1p from yeast has been identified as the

found-ing member of a novel family of atypical protein

serine kinases [12–14] It is essential in yeast and is

only distantly related to previously characterized

pro-tein kinases The RIO1 gene is transcribed

constitu-tively at an extremely low level [15,16], and Rio1p is

a very low abundant protein [12] Yeast has only one

such gene, whereas at least two orthologues of Rio1p

are present in higher eukaryotes The primary

struc-ture of the catalytic domain and the N-terminal

so-called Rio1-family domain are highly conserved

from archaea to man, whereas a great deal of

sequence variation resides in the extreme N- and

C-terminal portions [13]

We found that Rio1p plays a role in cell-cycle

pro-gression Yeast cells deprived of RIO1 lose

minichro-mosomes at an increased rate relative to wild-type and

arrest either as large G1cells (i.e late in the G1 phase

of the cell-division cycle) or as large-budded M cells

with a single DNA mass at the bud neck and short

spindles This indicates that Rio1p is simultaneously

required in the G1phase and for the onset of anaphase

(and⁄ or nuclear division and chromosome segregation)

[12] Vanrobays et al [17] obtained evidence from a

synthetic lethal screen with GAR1, an essential gene

required for 18S rRNA maturation, that Rio1p might

be involved in ribosome biogenesis However, it is

fea-sible that Rio1p has more than one target and plays a

role in several pathways in yeast (as may be deduced

from the fact that two orthologues occur in higher

eukaryotes) [13]

The biological role of Rio1p or even the pathways

in which the Rio1 protein kinase is involved are far from being understood Targets or interaction partners have not been identified as yet We report here first attempts to identify interaction partners and found that the activity and cellular concentration of Rio1p are regulated by phosphorylation through CK2 in a cell-cycle-dependent fashion

Results

Rio1p interacts with Cka2p

In an effort to identify the interaction partners of the essential Rio1p kinase, we performed co-purification experiments after overexpression of an N-terminally myc3-tagged version of Rio1p from yeast extracts Sub-sequently, proteins were identified by mass spectrome-try We found several presumptive Rio1p interaction partners which co-purified exclusively with full-length Rio1p but not with a C-terminally truncated version (M Angermayr, unpublished) Among them we identi-fied Cka2p, one of the two a subunits of protein kinase CK2 using this approach, however, we did not recover the other a subunit (Cka1p) or any of the b subunits (Ckb1p or Ckb2p)

To verify interactions between Rio1p and Cka2p, RIO1 was fused, at its 5¢-end, to a myc3 tag-encoding sequence (the RIO1 deletion mutants used are compiled

in Fig 1), and CKA2 was equipped, at its 5¢-end, with

a nucleotide sequence encoding an HA3-tag, both

Fig 1 Rio1 protein kinase (A) Domains of the Rio1p kinase are indicated (according to the classification of Hanks et al [45]); candidates for CK2 phosphorylation are

7 indicated by an asterisk (*); pos., positions relative to the translational start codon A, N-Terminal domain; B, Rio1 family-specific domain (R-domain); CK, serine-rich acidic domain (pos 402–435, CK2 domain); and K, lysine-rich C-terminal domain (K domain, pos 436–484) (B) N- and C-terminally truncated versions of the Rio1 protein used in this study; amino acid sequences expressed are indicated (C) CK2 phosphorylation sites in the C-terminal domain (CK2 domain) of Rio1p; the respective S residues are emphasized and positions are indicated (pos 402–435); the respective CK2 phosphorylation sites are numbered consecutively (1–6).

transcribed from the inducible GAL10 promoter Using

yeast extracts, co-immunoprecipitations were performed

once with anti-(myc agarose) to purifiy Rio1p and once

with HA antibodies

Cka2p or Rio1p was subsequently detected by western

analysis with HA or myc antibodies, respectively

(Fig 2) Because we presumed that the C-terminal

por-tion of Rio1p was involved in protein–protein

inter-actions and might serve as a substrate for CK2, we also used a C-terminally truncated version of Rio1p, 1–408 [12] as a control (Fig 1B) In addition, we investigated whether a catalytically inactive allele of RIO1, Rio1-D244N [12] interacts with HA3–Cka2p as well (inactive RIO1 alleles were rescued by the, untagged, genomic copy of RIO1) When Cka2p was immunoprecipitated with HA antibodies, we detected the active or inactive versions of Rio1p in a subsequent western blot by using myc antibodies, indicating that both active and inactive Rio1 proteins interact with Cka2p (Fig 2A,B) This was also true, when anti-(myc agarose) was applied to pre-cipitate Rio1p (Fig 2C,D) The interaction of Rio1p with CK2 is extremely stable and resistant to extensive washing (not shown), whereas the C-terminally trun-cated Rio1p (1–408) displays only weak interactions with Cka2p (Fig 2A,C; in C, only a faint signal was detected)

The above results indicated that the C-terminus of Rio1p might play a role in the interaction with Cka2p

in yeast cells To determine domains of Rio1p which are necessary and sufficient to interact with Cka2p, we pro-duced a series of RIO1 truncations Products containing only amino acids 1–46 or 1–76 turned out to be unstable

in yeast In order to test the possible importance of the N–terminus in the interaction with CK2 we could there-fore use only N-terminal truncations (46–484, 76–484; Fig 1) in this experiment We made another C-terminal truncation (1–402; in this construct an additional pre-sumptive CK2 phosphorylation site has been deleted in comparison with 1–408) In a complementary experi-ment, we used a construct containing exclusively the C-terminal part of Rio1p (starting immediately C-termi-nal adjacent to domain XI, amino acids 335–484) Co-purification was performed using anti-(myc aga-rose), and co-purified HA3–Cka2p was subsequently identified by western blot analysis using HA antibodies (Fig 2E,F) Interactions between Rio1p and Cka2p were abolished completely as soon as the 82 C-terminal amino acids of Rio1p were deleted (constructs 1–402, 46–402 or 76–402) Conversely, Cka2p co-purified with the C-terminal fragment of Rio1p (amino acids 335– 484) Thus, the C-terminus is necessary and sufficient to establish this interaction N)Terminal truncations behave like full-length Rio1p in this respect and do not affect Rio1p–Cka2p complex formation, demonstrating that the N-terminus of Rio1p has no bearing on the interactions between Rio1p and Cka2p

Rio1p is a target of CK2 The above results indicated that Rio1p and Cka2p interact with one another and that the interaction

Fig 2 Co-immunopurification experiments and identification of

domains essential for Rio1p–Cka2p interactions (A) HA 3 -tagged

Cka2p was immunoprecipitated using HA antibodies and protein

A–Sepharose; co-purified myc3-tagged Rio1p was subsequently

detected by immunodetection with myc antibodies (B) Control

detection of total immunoprecipitated HA3-tagged Cka2p using HA

antibodies (C) Myc3-tagged Rio1 proteins were immunoprecipitated

using anti-(myc agarose) Co-purified HA 3 -tagged Cka2p was

subse-quently detected by immunodecoration with HA antibodies (D)

Control detection of total immunoprecipitated myc3-tagged Rio1

proteins using myc antibodies (E) Myc 3 -tagged Rio1 proteins were

immunoprecipitated with anti-(myc agarose) Co-purified Cka2p was

subsequently detected by HA antibodies in a western blot (F)

Immunoprecipitation efficiencies of the Rio1 proteins were

con-trolled by immunodetection with myc antibodies Cka2p, HA 3 -Cka2p

which was still immunoreactive during the second detection on the

same blot in (F) p1, p2, control plasmids containing exclusively the

GAL10 promoter and the myc 3 - or HA 3 -tag, respectively fl,

full-length Rio1p; 1–408p, C-terminally truncated Rio1p.

domain might involve the C-terminal segment of

Rio1p To analyse whether Rio1p and Cka2p interact

directly, i.e without bridging factors from yeast, we

purified recombinant glutathione S-transferase (GST)–

Rio1p from Escherichia coli Rio1 wild-type protein

has autophosphorylation activity, but GST-fused

Rio1p is enzymatically inactive, as observed with

sev-eral kinases Therefore, this fusion protein is a suitable

substrate to unambiguously test whether Rio1p is a

target of CK2 GST–Rio1p was incubated without and

with recombinant human CK2 holoenzyme in the

pres-ence of [32P]ATP[cP] (Fig 3A) GST served as an

additional negative control in the kinase assays No

autophosphorylation (absence of CK2) was detected

corroborating that GST–Rio1p is inactive GST–Rio1p

phosphorylation signals were detected only after

incu-bation with CK2 C-Terminally truncated GST–Rio1p

(1–408p) was poorly phosphorylated by CK2 when the

signal strengths of precipitated GST–Rio1p and

GST-1–408p were compared (Coomassie Brilliant

Blue-stained gel, Fig 3B) No phosphate incorporation was

detected when amino acids 1–402 of Rio1p served as a

substrate for CK2 (Fig 3C), suggesting the absence of

CK2 sites N-terminal of position 402 in the catalytic

domain and the presence of several phosphorylation

sites for CK2 in the C-terminal portion of Rio1p, one

(or more) of them in the segment between positions

402 and 408 In the complementary experiment, the

C-terminal fragment of Rio1p (335–484p) was heavily phosphorylated (Fig 3C)

These results show that: (a) Rio1p and CK2 interact directly in vitro, because both proteins are of recombi-nant origin; (b) recombirecombi-nant CK2 holoenzyme has the capacity to phosphorylate Rio1p; and (c) the CK2 phosphorylation sites of Rio1p lie within a region between amino acid 402 and the C-terminus at position 484

To provide evidence that Rio1p is also a target of CK2 in vivo, we examined the extent of Rio1p phos-phorylation from extracts of a Dcka1 or Dcka2 yeast mutant, respectively As controls we used an inactive allele of RIO1, Rio1-D244N, and the truncated Rio1(1– 402-D244N) mutant, the latter is both inactive and not phosphorylated by Cka2p (Fig 4) Using yeast extracts obtained from cells wild-type for CKA1 and CKA2, tagged versions of Rio1p or Rio1-D244Np were heavily phosphorylated, whereas only a weak signal was detected with Rio1(1–402-D244N)p (Fig 4A) (This residual phosphorylation is independent of both Rio1p and CK2 kinase activities and, likely attributable to the action of still another protein kinase; M Angermayr, unpublished) However, in the Dcka1 or Dcka2 genetic backgrounds, phosphate incorporation dropped to

 40% (Dcka1) or 25% (Dcka2) with either Rio1p or Rio1(D244N)p (Fig 4C) indicating that Rio1p is phos-phorylated mainly by a heterotetramer containing both

Fig 3 Rio1p is a target of CK2 (A) Affinity-purified recombinant inactive GST-tagged Rio1 proteins were incubated with (+) or without (–) recombinant CK2 in the presence of [ 32 P]ATP[cP], electrophoresed and autoradiographed (B) Coomassie Brilliant Blue-staining of the respec-tive gel GST served as a negarespec-tive control in in vitro kinase assays *, degradation product of the recombinant full-length version of Rio1p (present in lanes 2 and 3) (C) Different recombinant GST-tagged Rio1 protein full-length or truncated versions were affinity-purified, incu-bated with (+) or without (–) recombinant CK2 in the presence of [ 32 P]ATP[cP], and detected by autoradiography after SDS ⁄ PAGE * denotes degradation products of full-length Rio1p and the C-terminal portion (amino acids 335–484) (D) Coomassie Brilliant Blue-stained gel of (C).

Cka1p and Cka2p less efficiently by a heterotetramer

composed of two Cka2p subunits, and only poorly by

the two Cka1p catalytic subunits-comprising

holoen-zyme The observed differences in phosphate

incorpora-tion ( 20%) between Rio1 wild-type and mutant

(D244N) proteins are likely attributable to

simulta-neous autophosphorylation of Rio1p and⁄ or the action

of still another protein kinase (M Angermayr,

unpublished observations)

The above results indicate that Rio1p–CK2

inter-actions are not restricted to Cka2p, but might be

exerted via Cka1p as well To test the capacity of

tagged versions of Cka1p or Cka2p to compete with

the respective residual version of CKA in either a

Dcka1 or Dcka2 genetic background, we performed

co-immunoprecipitation experiments using an HA3

-tagged version of Cka1p (Fig 5) Cka1p interacts with

Rio1p, although to a much lesser extent than Cka2p

Quantification of co-immunoprecipitates in the Dcka1

or Dcka2 genetic background showed that Rio1p binds with higher affinity to Cka2p, corroborating the results obtained with in vitro phosphorylation experiments

Protein kinase CK2 phosphorylates six clustered serine residues of Rio1p

Computational analyses (http://scansite.mit.edu/ motifscan_seq.phtml) indicated that several (four to six, depending on the stringency set) high- and low-affinity

Fig 4 Phosphorylation of Rio1p by CK2 after Co-purification from

yeast cellular extracts (A) Myc 3 -tagged Rio1 proteins were

immu-noprecipitated with anti-(myc agarose) using yeast extracts from

wild-type-, Dcka1-, and Dcka2 yeast strains; immunoprecipitates

were incubated in the presence of [32P]ATP[cP], and detected by

autoradiography after SDS ⁄ PAGE fl, full-length Rio1p (B)

Coomas-sie Brilliant Blue-stained gel; IgG, antibody heavy chain; fl,

full-length proteins (C) Quantitative evaluation of phosphate

incorpora-tion into the respective Rio1 proteins with respect to the different

genetic backgrounds (wild-type, Dcka1, Dcka2).

Fig 5 Interaction of Rio1p with Cka1p or Cka2p (A) Myc3-Rio1p [myc3-RIO1] was immunoprecipitated from yeast extracts from dif-ferent genetic backgrounds (wild-type, Dcka1, Dcka2), coexpressing either HA3-Cka1p, [HA3-CKA1], or HA3-Cka2p, [HA3-CKA2], respec-tively Co-purified HA 3 -Cka1p or HA 3 -Cka2p was subsequently iden-tified by immunodetection in a western blot using HA antibodies IgG, antibody heavy chain (B) Control of immunoprecipitation effi-ciencies of myc3-Rio1p by western blot (C) Quantitative evaluation (normalized to the strongest signal in 5B) of relative interaction effi-ciencies between Rio1p and Cka1p or Rio1p and Cka2p, respec-tively, in yeast strains disrupted for either CKA1 or CKA2 Genotypes are indicated below (C), and alleles in brackets denote the tagged (and immunoprecipitated) isozyme of CK2 (Quantitative evaluation is only shown for the respective Dcka1 or Dcka2 genetic background, respectively).

phosphorylation sites for CK2 might exist exclusively

in the C-terminal part of Rio1p To determine whether

these sites are functional, we changed the candidate Ser

residues one by one to Ala using site-directed in vitro

mutagenesis In vitro kinase assays with recombinant

CK2 holoenzyme and the respective (enzymatically

inactive) recombinant GST-fused Rio1 mutant proteins

as substrates revealed a total of six tightly clustered

ser-ine residues as CK2 phosphorylation sites [S402 (S1),

S403 (S2), S409 (S3), S416 (S4), S417 (S5), S419 (S6)]

consecutively numbered 1–6; cf Fig 1C (Fig 6) The

total number of CK2 phosphorylation sites of Rio1p

was deduced from experiments with several single,

dou-ble, and triple S to A mutations; not all combinations

are shown Recombinant (inactive) GST-fused RIO1

mutant alleles in which all six presumptive

phosphory-lation sites for CK2 had been mutated exhibited no

residual phosphorylation signal at all after incubation

with recombinant CK2 proving that all CK2

recogni-tion sites within the Rio1p kinase had been destroyed

Quantitative evaluation of phosphate incorporation indicated that CK2 displays different affinities towards the respective serine residues (Fig 6C)

Phosphorylation by CK2 stimulates Rio1p kinase activity

To investigate the possible physiological importance of Rio1p phosphorylation by CK2, we changed all six CK2 phosphorylation sites from S to either A (S > A)6,

or D (S > D)6, respectively N-Terminally His6-tagged wild-type and mutant proteins were purified from E coli and analysed in vitro by kinase assays using histone H2B [12] as a substrate (Fig 7) Quantification revealed comparable rates of phosphate incorporation into the heterologous substrate by the Rio1 wild-type and (S > A)6mutant proteins, as expected for recombinant proteins lacking modifications (e.g unphosphorylated

by CK2) However, when the CK2 sites were changed to

D (S > D)6 (mimicking permanently CK2-phosphory-lated Rio1p), Rio1 mutant protein kinase activity was stimulated approximately twofold

To prove the functionality of the respective CK2 phosphorylation sites in yeast, we incubated immuno-precipitated myc3-tagged Rio1 wild-type, (S > A)6, or (S > D)6mutant proteins from yeast in the presence of [32P]ATP[cP] (Fig 8) Quantification of phosphate incorporation into the respective Rio1p versions showed that the Rio1 wild-type protein was heavily phosphorylated (Fig 8C) However, when the six CK2 phosphorylation sites were mutated to either alanine or aspartate, phosphate incorporation dropped to 20 or 40%, respectively, which reflects autophosphorylation

of Rio1p and⁄ or the presence of a site for another as yet unidentified kinase which co-immunoprecipitated together with Rio1p in addition to CK2

Biological implications of phosphorylation of Rio1p by CK2

In order to examine the possible biological importance

of Rio1p phosphorylation in vivo, we tested whether substitution of all six CK2 phosphorylation sites in Rio1p by either A or D (mimicking unphosphorylated

or permanently phosphorylated Rio1p, respectively) has any consequences on yeast viability or growth rate For this purpose the respective mutant alleles were brought into the genuine genomic context (i.e at the RIO1 locus) Gene-shuffling experiments showed that the (S > A)6as well as the (S > D)6mutant alleles comple-mented the deletion of the RIO1 wild-type copy How-ever, growth rates of yeast cells harbouring the (S > A)6 mutant allele were significantly reduced

Fig 6 Mutational analyses of the CK2-sites of Rio1p (A)

Recombi-nant GST-fused Rio1 mutant proteins were incubated without (–) or

with (+) recombinant CK2 in the presence of [32P]ATP[cP],

sepa-rated by SDS⁄ PAGE and autoradiographed (B) Coomassie Brilliant

Blue-stained gel (C) Quantification of phosphate incorporation

(relative to GST–Rio1p precipitated amounts) into the respective

Rio1p mutant proteins by protein kinase CK2; * denotes

degrada-tion products.

(Fig 9A), indicating that the respective serine residues must be phosphorylated in vivo to give a biologically fully active Rio1p kinase The yeast strain carrying the mutant (S > D)6allele behaved indistinguishably from wild-type, suggesting that the Rio1p kinase phosphory-lated by CK2 is the fully functional form of the enzyme

in vivo These findings obtained with the respective RIO1 mutant alleles corroborate the results obtained

in vitro, i.e that the Rio1p kinase is moderately activated by CK2 phosphorylation also in vivo in the wild-type and that this activation accelerates cell proliferation These observations imply that lack of phosphorylation is disadvantageous for cell prolifera-tion

One possible reason for the slow growth of the non-phosphorylatable (S > A)6 mutant could be that these cells are impeded in entering or exiting from a certain

Fig 8 Functionality of the CK2 phosphorylation sites in yeast (A) Myc 3 -tagged Rio1 wild-type or mutant proteins were immuno-precipitated from yeast extracts with anti-(myc agarose) and incu-bated in the presence of [ 32 P]ATP[cP], separated by SDS ⁄ PAGE and autoradiographed (B) Coomassie Brilliant Blue-stained gel (C) Quantitative evaluation of phosphate incorporation into the respec-tive Rio1 proteins Values represent the average of three indepen-dent experiments, SD bars are given in the figure.

Fig 7 CK2 phosphorylation stimulates Rio1p kinase activity.

(A) In vitro kinase assays were performed using recombinant

affin-ity-purified His6-tagged wild-type or mutant Rio1p proteins (S > A)6

or (S > D)6, with histone H2B as a substrate (autoradiograph) (B)

Coomassie Brilliant Blue stain of Rio1p input (C) Quantitative

evalu-ation of phosphate incorporevalu-ation into H2B (normalized to H2B input

and to input of Rio1 proteins) Values represent the average of

three independent experiments (three different purifications from

E coli), SD bars are given in the figure.

phase of the cell-division cycle Therefore, we tested,

using light microscopy and after DAPI staining, whether

the slow growth yields a cell-cycle phenotype Cells of

the logarithmically growing wild-type or the (S > A)6

mutant strain were photographed and cells assigned to

the respective phase of the cell-division cycle (similarly

as described previously) [12] In the (S > A)6 mutant,

the number of cells in the S phase having a small bud

was drastically diminished to almost one third relative

to wild-type or the (S > D)6 mutant, i.e 8% in the

(S > A)6mutant versus 22% in the wild-type, and the

number of G1cells was increased accordingly – 39% in

the (S > A)6 mutant versus 29% in the wild-type By contrast, G2plus M phase cells were not affected signifi-cantly – 53% in the (S > A)6mutant versus 49% in the wild-type (Fig 9B) However, we found a slight imbal-ance with respect to the distribution of G2⁄ M cells: the number of metaphase cells with a single DNA mass at the bud neck and the number of anaphase cells were increased slightly in the mutant (metaphase cells: 30.2%

in the mutant versus 24% in the wild-type; anaphase cells: 4.5% in the mutant versus 3.2% in the wild-type), whereas the number of telophase cells was decreased slightly (18.3% in the mutant versus 21.8% in the wild-type) These slight imbalances are considered insignificant, in contrast to the differences observed with the distribution of G1 and S phase cells These obser-vations suggest that (S > A)6 mutant cells, that fail

to become phosphorylated, mainly have difficulties entering the S phase

The reason for the accumulation of cells in the

G1phase may be due to the slightly lower kinase activity

of the (S > A)6 mutant Rio1p kinase during the G1 phase relative to wild-type or the (S > D)6 mutant, thereby retarding entry into the S phase However, we obtained direct evidence that a different mechanism may play a role A first hint to this mechanism came from quantitative analysis of Rio1 protein concentra-tions in the two mutants In logarithmically growing cells of the (S > A)6 mutant the concentration of the respective Rio1 protein generally exceeded (approxi-mately fivefold) that of the wild-type or the (S > D)6 mutant proteins (see Fig 10; log¼ cycling cells) One possible explanation for this finding, which was further

Fig 9 Growth rates of RIO1 wild-type and mutant yeast strains

and cell-cycle phase distribution (A) A yeast strain carrying a single

genomic copy of the RIO1 (S > A) 6 mutant allele is hampered in

growth rate; the respective (S > D) 6 mutant yeast strain is

indistin-guishable from the wild-type Two independent clones were tested

in all cases Maximum deviations are indicated by bars (B)

Loga-rithmically growing cells were stained with DAPI, photographed

and evaluated according to the stages of the cell-division cycle as

indicated below the diagram A total of 546 wild-type cells or 239

mutant (S > A) 6 mutant cells have been analysed.

Fig 10 Phosphorylation of Rio1p by CK2 renders the protein sus-ceptible to proteolysis.RIO1 wild-type, (S > A)6, or (S > D)6mutant cells were arrested with a-factor, hydroxyurea (HU), or nocodazole (Noc) or left untreated (log) as a control in the presence of galactose

as a carbon source (for details please refer to Experimental proce-dures) Cellular extracts were separated by SDS ⁄ PAGE, and the respective proteins were detected by western blotting (A) Immuno-precipitated myc3-tagged versions of Rio1p were detected by myc antibodies after SDS ⁄ PAGE in a western blot (B) Loading control The stable protein Aky2p served as an input control and was analo-gously detected by Aky2p antibodies derived from hen egg yolk.

pursued, is that the unphosphorylated version is stable,

whereas the (S > D)6 mutant protein, mimicking the

phosphorylated form (and obviously also wild-type

Rio1p) is proteolytically degraded Generally increased

proteolytic instability of Rio1p wild-type and the

(S > D)6 mutant proteins relative to the (S > A)6

mutant was, however, not observed in unsynchronized

cells (not shown) Therefore, we tested whether Rio1p

and Rio1(S > D)6 mutant proteins are degraded at a

certain stage of the cell cycle Such temporary instability

may explain the lower steady-state concentration of

the Rio1 mutant protein in the (S > D)6p version and

wild-type Rio1p relative to (S > A)6p in asynchronous

cells To test whether phosphorylated Rio1p is, in fact,

degraded at a certain stage of the cell-division cycle,

whereas the unphosphorylated form is not, we

per-formed cell-cycle arrest experiments (see Experimental

procedures) Logarithmically growing cells were induced

by galactose (RIO1 alleles under the control of the

GAL10promoter) and simultaneously arrested by

treat-ment with either a-factor (arrest before the G1⁄ S

transi-tion), hydroxyurea (S phase), or nocodazole (before

onset of anaphase), respectively Cellular concentrations

of Rio1p were measured relative to adenylate kinase 2

(Aky2p), which is a constitutively expressed, stable

pro-tein [18] as a loading control (see Experimental

proce-dures) Our results indicate that in the RIO1 wild-type

and the (S > D)6 mutant the level of Rio1 protein is

low to undetectable in the S phase but normal in G1and

during mitosis compared with cycling cells (Fig 10;

traces of material detected after arrest with hydroxyurea

may be attributable to nonarrested cells; 10–15%)

However, the level of Rio1p is surprisingly high and

constant in the (S > A)6mutant and, most notably, not

at all affected by the stage of the cell-division cycle

(Fig 10), thus displaying significant resistance to

pro-teolytic degradation These findings demonstrate that

Rio1p and the (S > D)6 mutant proteins, mimicking

the phosphorylated version, are degraded at the G1⁄ S

transition, whereas the nonphosphorylatable (S > A)6

version is not

Discussion

The Rio1p kinase from yeast is essential and highly

conserved from archaea to man and, thus, likely to

serve an evolutionarily ancient, highly important, as yet

unknown function [12–14,17,19] Therefore, it is of

gen-eral interest to identify interaction partners or

sub-strates and, as a consequence, the pathway(s) this

kinase is involved in We have found that the catalytic

a subunits of protein kinase CK2, Cka2p and to a

les-ser extent Cka1p, specifically interact with Rio1p We

have shown that the C-terminus of Rio1p is essential and sufficient for this interaction We have also shown that Rio1p is a substrate for CK2 holoenzyme in vitro and, most likely, also in vivo

Several combinations of serine mutations in the C-terminal portion of Rio1p proved the presence of six clustered CK2 phosphorylation sites with different affinities Neither the (S > A)6 mutant nor the 1–402 C-terminally truncated protein served as a substrate for CK2 (excluding the presence of additional CK2 sites N-terminal of position 402), whereas the C-terminal fragment Rio1-335–484p alone displayed strong wild-type-like phosphorylation by CK2

The C-terminal portion of yeast Rio1p displays a striking two-partite primary structure The part C-ter-minally adjacent to the catalytic domain is rich in serines and acidic residues (referred to as CK2 domain, positions 402–435), whereas the extreme C-terminus (positions 436–484) lacks serines and is highly posi-tively charged (mainly lysines, referred to as K-domain, Fig 1A)

It is noteworthy that the C-terminal domain of Rio1p is least conserved in evolution Archaea, that do not have CK2, lack the CK2- and K-domains com-pletely, but also among higher eukaryotic Rio1p ortho-logues high sequence divergence is observed in the C-terminal part Higher eukaryotes harbour two ortho-logues of Rio1p named the SUDD-type and the

ad034-type according to their first identification [13] The SUDD proteins have only a short stretch of basic amino acid sequences lacking C-terminal CK2 sites

ad034 proteins are more closely related to yeast Rio1p and have both a CK2- and a K-domain, although the direct sequence similarity is low We have cloned both types of human cDNAs, ad 034 and SUDD, as myc3 -tagged version and expressed them in yeast and E coli Neither, alone or together, complements RIO1 defi-ciency in yeast Nevertheless, both recombinant orthol-ogous proteins are heavily phosphorylated by CK2

in vitro(M Angermayr, unpublished results)

In vitro kinase assays with recombinant Rio1 pro-teins from yeast have revealed that the (S > D)6 muta-tion, mimicking permanent phosphorylation by CK2, stimulates Rio1p kinase activity about twofold with H2B as a heterologous substrate The same is true when the extent of Rio1p autophosphorylation of the respective mutant proteins isolated from yeast is com-pared, suggesting that phosphorylation of Rio1p by CK2 has only a minor, presumably modulating effect

on the kinase activity of Rio1p

Also with other substrates of CK2 that have been described in detail in the literature, very small physio-logical effects of phosphorylation by CK2 have been

reported Many of the targets of CK2 have been found

to serve essential functions For example, CK2

increases the efficiency of transcription of the tRNA

and 5S rRNA genes by RNA polymerase III due to

phosphorylation of TBP within TFIIIB [20]

Eukary-otic topoisomerase II is another target of protein

kinase CK2 which is essential for viability However,

the importance of CK2 phosporylation of

topoisomer-ase II is not well understood, because mutation of the

respective CK2 phosphorylation sites does not cause

an obvious phenotype in yeast [21,22]

Cyclin-depen-dent protein kinase from S cerevisiae, Cdc28p, is

phosphorylated at a single serine residue by protein

kinase CK2 [23] Lack of phosphorylation at this site

affects neither Cdc28p kinase activity in vitro nor yeast

growth rate, but leads to a slightly decreased cell size

during the G1phase [9,23]; by contrast, S > E

muta-tion of this residue stimulates Cdc28p twofold, at least

in vitro [9] Sic1p, the cyclin⁄ CDC28 cell-cycle kinase

inhibitor that prevents premature entrance into the

S phase, is another interesting substrate of CK2, but

phosphorylation by CK2 has little influence on its

physiological function [10,24,25] The essential

transla-tional initiation factors, eIF2a (encoded by SUI2) [26]

and eIF5 (encoded by TIF5) [27], are additional

tar-gets of CK2, but phosphorylation by CK2 of either

eIF2a or eIF5 by CK2 is not essential for their

respec-tive functions A seeming exception of a low effect of

CK2 site mutation is constituted by Cdc37p, a

kinase-associated molecular chaperone required in concert

with Hsp90p in the regulation of the activity of several

signalling protein kinases Mutation of the single CK2

site on Cdc37p is not lethal but severely impedes

growth, presumably because of the additive negative

effects on several important protein kinases [28,29]

Taken together, there are many examples of proteins

which serve important functions that are substrates of

CK2, but mutational alteration of the respective CK2

phosphorylation sites has little effect or, at least, is not

deleterious to cell viability This is more surprising as

deletion of CK2 (in yeast the double deletion of CKA1

and CKA2) leads to inviability [6] However, the

pleio-tropy of CK2 may explain its indispensability Failure

in CK2 deletion mutants of modulating a plethora of

processes, although within narrow limits, is likely to be

detrimental to cell life Thus, moderate activation of the

Rio1p kinase upon phosphorylation by CK2 is in the

same range as observed with most other substrates

We have shown that mutant RIO1 alleles, in which

all six CK2 phosphorylation sites have been mutated to

alanine or aspartate, respectively, sustain yeast viability

The (S > D)6 mutant behaves indistinguishably from

wild-type indicating that Rio1p is heavily

phosphory-lated by CK2 in vivo In contrast, yeast cells harbouring exclusively the (S > A)6mutant allele, a substrate that does not become phosphorylated by CK2, are severely hampered in growth rate In addition, we have observed

a cell-cycle phenotype with this mutant Cytological approaches investigating cell-cycle stages reveal that this mutant yeast strain accumulates G1 cells, whereas the number of S phase cells is drastically diminished It may be noteworthy that we detected a slight imbalance

of metaphase, anaphase and telophase cells as well Most significantly, this is in accordance with our previ-ous observation that either depletion of Rio1p in the cell or the use of a weak D244E mutant allele leads to increased loss of minichromosomes and to the accumu-lation of both, large-budded M cells with a single DNA mass at the bud neck and large G1cells, indicating that Rio1p is required for exit from mitosis and during G1 phase, but obviously not during S phase [12] However,

in contrast to our previous Rio1p depletion experiments

or the use of the weak active site mutant of Rio1 (D244E) in which inhibition of the entrance of ana-phase was the most significant effect, we describe here with the nonphosphorylatable (S > A)6 mutant that the exit from the G1phase is more pronouncedly retarded than the arrest in mitosis These findings indi-cate that Rio1p phosphorylation by CK2 mainly plays

a role in the G1phase conceivably by slightly increasing Rio1p kinase activity, but is less (or not) important during mitosis

What might be the physiological basis of the moder-ate G1arrest phenotype? By comparing the cellular con-centrations of Rio1p and Rio1 mutant proteins, we were able to exclude that Rio1p or Rio1 (S > D)6p are pro-teolytically less stable per se than the (S > A)6mutant protein This means that phosphorylation by CK2 causes no general signal for the degradation of Rio1p Rather, we observed that the cellular concentration of Rio1p and (S > D)6p is extremely low to undetectable

in the S phase, indicating that phosphorylated Rio1p and the (S > D)6 protein are destined for degradation

at the G1to S boundary, whereas the (S > A)6protein

is not In this context, it may be relevant that the CK2 phosphorylation sites of Rio1p overlap a bona fide destruction box, an amino acid sequence rich in P, E, S and T residues which has been implied to be involved in the degradation of the respective protein in a ubiquitin-and proteasome-dependent manner [30] In the case

of Rio1pk, this potential degradation signal may be activated upon phosphorylation to act as a phospho-degron [31] and to effect cell-cycle phase-dependent proteolysis of Rio1p before entrance into the S phase This presumably occurs at the same time when other important G1-specific proteins (e.g Sic1p, Cln1p,