Transferable Domain in the G₁ Cyclin Cln2 Sufficient To Switch Degradation of Sic1 from the E3 Ubiquitin Ligase SCF^Cdc4 to SCF^Grr1

Abstract

Degradation of Saccharomyces cerevisiae G₁ cyclins Cln1 and Cln2 is mediated by the ubiquitin-proteasome pathway and involves the SCF E3 ubiquitin-ligase complex containing the F-box protein Grr1 (SCF^Grr1). Here we identify the domain of Cln2 that confers instability and describe the signals in Cln2 that result in binding to Grr1 and rapid degradation. We demonstrate that mutants of Cln2 that lack a cluster of four Cdc28 consensus phosphorylation sites are highly stabilized and fail to interact with Grr1 in vivo. Since one of the phosphorylation sites lies within the Cln2 PEST motif, a sequence rich in proline, aspartate or glutamate, serine, and threonine residues found in many unstable proteins, we fused various Cln2 C-terminal domains containing combinations of the PEST and the phosphoacceptor motifs to stable reporter proteins. We show that fusion of the Cln2 domain to a stabilized form of the cyclin-dependent kinase inhibitor Sic1 (ΔN-Sic1), a substrate of SCF^Cdc4, results in degradation in a phosphorylation-dependent manner. Fusion of Cln2 degradation domains to ΔN-Sic1 switches degradation of Sic1 from SCF^Cdc4 to SCF^Grr1. ΔN-Sic1 fused with a Cln2 domain containing the PEST motif and four phosphorylation sites binds to Grr1 and is unstable and ubiquitinated in vivo. Interestingly, the phosphoacceptor domain of Cln2 binds to Grr1 but is not ubiquitinated and is stable. In summary, we have identified a small transferable domain in Cln2 that can redirect a stabilized SCF^Cdc4 target for SCF^Grr1-mediated degradation by the ubiquitin-proteasome pathway.

The precise temporal and spatial control of key biological regulatory proteins via degradation by the ubiquitin-proteasome machinery is a critical regulatory process in eukaryotes. Due to its flexibility and specificity, the ubiquitination machinery controls many diverse regulatory pathways, including the cell cycle, cellular stress responses, intracellular signaling, and development. Flexibility is brought about by the modular nature of the ubiquitination process, involving three classes of enzymes, termed E1, E2, and E3 (reviewed in reference 16). A single ubiquitin-activating enzyme, E1, activates the 76-amino-acid (76-aa) polypeptide ubiquitin via formation of a high-energy thiolester bond. Activated ubiquitin is then transferred to one of a large family of E2 ubiquitin-conjugating enzymes; the relatively unspecific E2 enzymes are targeted to an appropriate substrate by a class of substrate receptor complexes termed E3 ubiquitin ligases. Specificity of the ubiquitin pathways is conferred by the nature and activity of the E3 complexes. Together, the E2 and E3 enzymes catalyze the formation of isopeptide bonds between ubiquitin and lysine residues on the target proteins. Additional ubiquitin molecules are added to form multiubiquitin chains, and the multiubiquitin proteins are recognized and rapidly degraded into short peptides by the 26S proteasome. One key to understanding the regulation of ubiquitin-mediated degradation is to elucidate the molecular nature of substrate recognition. Here we describe the identification of a transferable domain in a model substrate, the G₁ cyclin Cln2. We show that this domain, when activated by a phosphorylation signal, can interact with a specific E3 enzyme, the SCF^Grr1 complex, resulting in rapid ubiquitin-mediated degradation.

Two E3 ubiquitin ligase complexes, the anaphase-promoting complex (APC) and the SCF (Skp1/Cdc53/F-box protein) complex, have attracted considerable attention in recent years due to their prominent role in controlling the cell cycle. Both factors are highly conserved among eukaryotes and function in a modular fashion to recognize different classes of substrates (reviewed in reference 16). However, the two complexes differ in their mode of regulation. The APC appears to be activated by posttranslational mechanisms and by association with two specific adapter molecules. In contrast, the SCF complex is proposed to be constitutively active; substrate phosphorylation promotes recognition by the SCF complex, which is followed by ubiquitination and degradation of the substrate. Target proteins that bind to the SCF complex only in their phosphorylated state include the G₁ cyclins Cln1 and Cln2 (24, 31, 39, 45), the cyclin-dependent kinase (CDK) inhibitors Sic1 (11, 39) and p27 (5, 41), and the transcription factor NF-κB inhibitor IκBα (reviewed in reference 21). Using in vivo mutational analysis and in vitro ubiquitination assays, degradation of each of these substrates has been demonstrated to depend both on a specific F-box protein and on the phosphorylation state of the substrate. Thus, ubiquitination of specific targets not only depends upon the identity of that target but also upon its state.

Functional SCF complexes, composed of Skp1, Cdc53/cullin, Rbx1/Roc1, and an F-box protein, associate with the E2 enzyme Ccd34 to mediate targeting and ubiquitination of specific substrate proteins (20, 30, 39, 40, 42). The SCF complex is a critical regulator of the G₁/S transition, and subunit composition and function are highly conserved among eukaryotes (1, 20, 26, 28, 32, 48). A variety of SCF complexes exist, each with a different F-box protein. A large number of F-box proteins have been identified; the Saccharomyces cerevisiae genome contains at least 17 proteins, and database searches and two-hybrid screens have revealed more than 50 F-box proteins in plants and animal cells and more than 60 in the nematode Caenorhabditis elegans (1, 32, 46). This suggests that the SCF system controls a large number of regulatory processes in eukaryotes. However, the vast majority of F-box proteins have yet to be assigned a function. Three F-box proteins in yeast cells (Grr1, Cdc4, and Met30) and two in mammalian cells (Skp2 and βTrCP) have been studied in detail. SCF^Cdc4 is required for degradation of the CDK inhibitors Sic1p (11, 39) and Far1 (15), the replication factor Cdc6 (9), and the transcription factor Gcn4 (23). SCF^Grr1 targets the G₁ cyclins Cln1 and Cln2 and the putative Cdc42 effector Gic2 (18) for ubiquitination and degradation (2, 24, 31, 39).

The transition from G₁ to S phase controls entry into and exit from the cell cycle and is, therefore, a crucial regulatory point. In the yeast S. cerevisiae, this transition requires the activity of the CDK Cdc28 complexed with the G₁-specific cyclins, Cln1, Cln2, and Cln3. The cyclic accumulation of Cln1 and Cln2 in G₁ phase and rapid decay once cells enter S phase are controlled by periodic transcription of Cln1 and Cln2 and regulated protein degradation, respectively. The determinant for G₁ cyclin degradation lies in the C-terminal 170 aa since truncations of this domain cause stabilization of Cln2. The C terminus contains a PEST domain as well as six of the seven Cdc28 consensus phosphorylation sites. The PEST domain, found in all three yeast G₁ cyclins, was originally identified as a potential determinant of protein instability on the basis of its frequent occurrence in unstable proteins (33) but has yet to be functionally defined. Deletions in the C terminus of Cln2 and Cln3 that include the PEST motifs result in phenotypes consistent with hyperactivation of G₁ cyclins and partially stabilize the mutant proteins (36, 49). However, it is not clear whether the PEST sequences per se or other aspects of the PEST-containing region are the relevant determinants. We have shown that one of those determinants is phosphorylation. Thus, a yeast G₁ cyclin mutated in the seven potential Cdc28 phosphorylation sites is highly stabilized and renders cells insensitive to nutrient- and growth-inhibitory signals (24). Our present model suggests that the instability of the G₁ cyclins derives from Cdc28-dependent autophosphorylation of the cyclin subunit, which targets the cyclin for degradation (24, 37, 45, 49) by the SCF^Grr1. A similar model was proposed for mammalian cells based on findings reported for the G₁ cyclins E and D1 (6, 8, 48). These observations strongly suggest that precise expression and destruction of yeast G₁ cyclins are crucial for proper regulation of the cell cycle and that this mechanism is highly conserved from yeast to humans.

A key issue, however, remains unresolved: what is the molecular nature of the signal that promotes recognition by the F-box subunit of the SCF complex? Each of the known F-box proteins recognizes two or more substrates: Grr1 targets Cln1, Cln2, and Gic2 for degradation; Cdc4 recognizes Sic1, Far1, Gcn4, and Cdc6. In mammalian cells, the F-box protein Skp2 targets p27 and the transcription factor E2F1 (27) for destruction, whereas the F-box protein βTrCP targets IκBα (50) and β-catenin (14, 25, 26, 47). Yet there is little if any sequence homology between substrates recognized by the same F-box protein. Our recent observation that SCF^Grr1 interacts via the cationic surface of the Grr1 leucine-rich repeat (LRR) domain with phosphorylated substrates, including Cln2, begins to address this important question (17). Here, we extend these findings by demonstrating that a small domain within the carboxy terminus of Cln2, containing a PEST element as well as four crucial phosphorylation sites, converts a stable reporter protein into a substrate for the SCF^Grr1-mediated ubiquitin-proteasome pathway. Interestingly, while fusions of Cln2 domains to a glutathione S-transferase (GST) reporter protein render these chimerical proteins unstable, their degradation is phosphorylation and Grr1 independent. In contrast, fusions to stabilized Sic1 result in phosphorylation- and Grr1-dependent ubiquitination and degradation.

MATERIALS AND METHODS

Yeast strains and culture.

Strains were isogenic with 15Daub (MATa ade1 his2 leu2-3,112 trp1-1 ura3Δns bar1Δ [34]) or with W303 (MATa ade2-1 his3-1,15 ura3-1 leu2-3,112 trp1-1 can1-100). Culture conditions and medium were as indicated and were prepared by standard methods. Yeast strains were grown in selective medium or YAPD (1% yeast extract, 2% peptone, adenine [40 mg/liter], 2% glucose) at 23, 30, or 37°C as indicated. For expression of the Sic1 fusions, strains were grown overnight in glucose-containing selective medium, diluted to an optical density at 600 nm (OD₆₀₀) of 0.2 in YMRaff (yeast medium-2% raffinose) and grown for two generations. Then, galactose was added to a final concentration of 2% for 1 to 2 h. Expression of GST fusion proteins was induced with 500 μM CuSO₄ for 1 h.

Plasmids.

Primer sequences and details about plasmid constructions are available upon request. Briefly, single and multiple point mutations in CLN2 were generated by sequential PCR. PCR products were subcloned and sequenced to verify fidelity. To generate GST fusions, the appropriate Cln2 domains were amplified by PCR; for subcloning purposes, primers were designed to attach 5′ BamHI and 3′ XhoI sites to the amplified fragments. The various Cln2 domains were ligated into the pYEX-4T1 vector (Clontech Labs), generating GST fusions under control of the CUP1 promoter. To generate the ΔN-Sic1 fusions, the same CLN2 fragments were subcloned into plasmid pSL276, resulting in fusions of CLN2 domains with SIC1 with N-terminal deletions followed by a hemagglutinin (HA) and a hexahistidine (6His) tag. Plasmid pSL276, derived from RDB597 (a generous gift from R. Verma) containing SIC1-HA-6His under control of the GAL1 promoter, introduces an XhoI site at position 313 in the SIC1 coding sequence. Cloning of the CLN2 domains into BamHI-XhoI-cut SIC1 replaces the first 312 nucleotides of SIC1, which encode 104 aa, with CLN2 sequences.

To generate chromosomally 6His-tagged GRR1, the 3′ end of GRR1 was amplified by PCR and cloned into the pLEU2-6His vector (pSL212 [derived from pKAN-6His]; S. B. Haase, M. Wolff, and S. Reed, unpublished data), resulting in in-frame fusion of GRR1 to 6His. pSL212 was cut with HpaI for targeted integration into the GRR1 locus.

To create N-terminally 6His-tagged GRR1 for expression in SF9 insect cells, wild-type full-length GRR1 was cloned into pFASTBacHTc (Gibco BRL-Life Technologies), and recombinant baculovirus was generated according to the manufacturer's conditions.

PEST scores were determined with the program PESTFIND developed by Martin C. Rechsteiner and Scott W. Rogers (35) (http://www.at.embnet.org/embnet/tools/bio/PESTfind/).

Baculovirus expression and purification of Grr1.

For the expression of 6His-Grr1 in insect cells, SF9 cells were infected with recombinant baculovirus encoding the full-length Grr1 for 48 h. Cells were lysed by resuspension in SF9 lysis buffer (phosphate-buffered saline containing 0.5% NP-40, 5 mM β-mercaptoethanol, protease inhibitors {0.2 mM AEBSF [4-(z-aminoethyl)benzenesulfonyl fluoride], 1 mM EDTA, 20 μM leupeptin, 1 μM pepstatin}, and phosphatase inhibitors [10 mM NaF, 10 mM glycerol phosphate]) followed by incubation on ice for 20 min. Extracts were clarified by centrifugation for 10 min at 14,000 rpm (20,000 × g) in an Eppendorf 5810R microcentrifuge, aliquoted, and snap-frozen in liquid nitrogen. 6His-Grr1 was batch purified by incubation of SF9 extracts adjusted to 500 mM NaCl with Ni-nitrilotriacetic acid (Ni-NTA) beads for 2 h at 4°C, followed by washes containing 20, 50, 80, 120, and 200 mM imidazole. The 80 mM imidazole fraction was used for binding studies.

Immunoprecipitation and immunoblotting.

Yeast protein extracts were made with glass beads in lysis buffer (50 mM Tris-HCl [pH 7.5], 0.1% NP-40, 250 mM NaCl) containing protease inhibitors and phosphatase inhibitors (5 mM EDTA, 5 mM EGTA, 0.1 mM orthovanadate). Cells were lysed with glass beads by 5 to 10 cycles of vortexing for 1 min followed by a 1-min incubation on ice. Extracts were collected after centrifugation for 15 min at 14,000 rpm (20,000 × g) in an Eppendorf 5810R microcentrifuge. The protein concentrations of the lysate were determined by the Bio-Rad protein assay.

Extracts were fractionated on sodium dodecyl sulfate (SDS) gels and transferred onto polyvinylidene difluoride membrane (PolyScreen; NEN) by semidry blotting. Membranes were blocked with 5% nonfat dry milk in Tris-buffered saline-0.25% Tween and incubated with primary antibodies (1:1,000 in Tris-buffered saline-1% milk-0.25% Tween unless otherwise indicated) for 2 h to overnight followed by secondary antibodies (horseradish peroxidase-conjugated) (1:10,000; Bio-Rad) for 1 h. Antibodies used were anti-HA (mouse monoclonal; Babco), anti-Myc (mouse monoclonal 9E10) (1:250; Santa Cruz), anti-Cdc28 (goat, yC-20) (1:250; Santa Cruz), and anti-GST (rabbit polyclonal; Santa Cruz). Signals were detected on films by enhanced chemiluminescence (SuperSignalPico; Pierce). Anti-HA or anti-GST beads were made by covalently coupling the antibody (rabbit polyclonal anti-GST antibody [Santa Cruz] or mouse anti-HA ascites [Babco]) to protein A-Sepharose with dimethylpimelimidate.

For coimmunoprecipitations, extracts containing 1 to 2 mg of total proteins were incubated with the appropriate resin (anti-GST beads or anti-HA beads) in lysis buffer containing protease and phosphatase inhibitors for 2 h at 4°C. The beads were washed three times with lysis buffer and boiled with 2× SDS sample buffer, and bound proteins, along with extracts and flowthroughs, were analyzed by immunoblotting. For analysis of the interaction of Cln2^4T3S, Cln2^M46, or Cln2-Sic1 fusions with Grr1, the various Cln2 constructs were transformed into a yeast strain carrying chromosomally integrated GRR1-6His (or into a wild-type yeast strain as negative control). Extracts were made and coimmunoprecipitations performed using low-salt lysis buffer (100 mM NaCl). Total proteins (10 mg) were immunoprecipitated, and 9 of 10 of the bound fractions were analyzed for Grr1 and 1 of 10 was analyzed for Cln2-Sic1 fusions.

Analysis of phosphorylation.

The phosphorylation status of the recombinant fusion proteins was analyzed by treatment with calf intestine phosphatase (CIP) (10 U/μl; New England Biolabs) in the presence or absence of phosphatase inhibitors. Extracts (300 μg) were bound on anti-HA beads in lysis buffer with or without phosphatase inhibitors for 2 h at 4°C. After being washed three times with buffer with or without phosphatase inhibitors, the beads were incubated with 50 μl of alkaline buffer (50 mM Tris-HCl [pH 8.5], 150 mM NaCl) containing 20 U of CIP in the presence or absence of phosphatase inhibitors at 30°C for 60 min on a rotating support. Beads were washed twice with buffer and boiled in 2× SDS sample buffer. Proteins were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) and immunoblotting with anti-HA antibody.

Detection of ubiquitination.

Yeast strains containing the Cln2-Sic1 constructs were transformed with a plasmid carrying a tagged mutant ubiquitin (CUP1-Ubi-HIS-MYC-RA). This polyhistidine- and Myc-tagged K48R, G76A mutant ubiquitin competes with endogenous ubiquitin and is incorporated into ubiquitin chains, but the K48R substitution prevents further polymerization of ubiquitin and the G76A substitution inhibits hydrolysis by ubiquitin isopeptidases (24). Transformants were grown in selective glucose-containing medium and then diluted to an OD₆₀₀ of 0.2 in raffinose-containing medium and grown to an OD₆₀₀ of 0.5. Protein expression was induced with 250 μM CuSO₄ for 4 h and with 2% galactose for two additional hours. Glass bead extracts (7.5 mg) made in buffer G (6 M guanidine hydrochloride, 0.1 M NaH₂PO₄, 20 mM Tris-HCl, pH 8.0) were incubated with Ni²⁺-NTA Agarose (Qiagen) for 1 h at room temperature. Beads were washed three times with 2.5 ml of buffer G and four times with 5 ml of buffer C (50 mM Tris-HCl [pH 8.0], 500 mM NaCl). Bound proteins were eluted with 25 μl of buffer E (100 mM Tris-HCl [pH 6.8], 1% SDS, 100 mM dithiothreitol, 100 mM EDTA) for 10 min at room temperature on a rotating support, and the eluates were analyzed by SDS-PAGE and immunoblotting with anti-HA antibodies.

Determination of protein degradation and half-life.

Strains containing the various Cln2-Sic1 fusions under control of the GAL1 promoter were grown in glucose-containing medium, diluted to an OD₆₀₀ of 0.25 in raffinose-containing medium, and grown for two generations. Expression of the fusion proteins was induced by addition of galactose to a final concentration of 2% for 30 min to 1 h. At time point 0, cells were filtered and resuspended in glucose-containing medium to repress transcription. Samples were removed at the time points indicated by filtering 20 ml of culture on ice, washing the pellets briefly with water, and freezing the cell pellets in liquid nitrogen. Extracts were made with glass beads, and protein content was normalized using the Bio-Rad assay and analyzed by immunoblotting with monoclonal anti-HA antibodies or polyclonal anti-Cln2 antibodies (data not shown). Band intensities were quantified using PhosphorImager technology and ImageQuant software (Molecular Dynamics) or LabWorks software (UVP). For accurate calculations, we routinely included internal controls with serial dilutions of proteins and normalized the samples to the Cdc28 loading control. Protein half-lives were estimated by best-fit analyses of degradation curves.

Pulse-chase analysis.

Logarithmically growing cells expressing GST-Cln2 domain fusions from the CUP1 promoter were induced with CuSO₄ for 2 h, washed, resuspended in methionine-free medium, and labeled by incubation with 1 mCi of ³⁵S-protein labeling mix (Easy Tag Express-[³⁵S]-protein labeling mix; NEN) for 5 min at 30°C. Excess cold methionine and cysteine were added, and samples were collected at the time points indicated. Extracts were made with glass beads and subjected to denaturing immunoprecipitation with anti-GST beads. Beads were boiled with 2× SDS sample buffer and proteins bound were fractionated on SDS gels and visualized by PhosphorImager analysis as described above.

RESULTS

We have shown previously that Cdc28-mediated phosphorylation of Cln2 promotes its rapid degradation by the ubiquitin-proteasome pathway (24, 45). A replacement of the seven Cdc28 consensus phosphorylation sites in Cln2 by alanine (termed Cln2^4T3S) stabilized Cln2 eightfold. However, the Cln2 domain(s) responsible for recognition by the SCF degradation machinery is not defined. Therefore, we set out to delineate the domain(s) in Cln2 important for instability.

A cluster of four phosphorylation sites destabilizes Cln2.

First we asked whether all seven Cdc28 phosphorylation sites in Cln2 were important for degradation. We had previously observed that (i) alanine substitutions at sites 3 and 5 (mutant D35; Cdc28 phosphorylation sites are labeled 1 through 7 [Fig. 1A ]) increased the half-life of Cln2 to 35 min, (ii) substitution at site 7 had no effect on Cln2 stability, and (iii) deletion of residues 376 to 514 (eliminating sites 2 through 6) resulted in a highly stabilized Cln2 (24) (Fig. 1 and Table 1). These results suggested that phosphorylation sites 3 and 5 in combination with any or all of sites 2, 4, and 6 are important for Cln2 destabilization. We, therefore, created all possible permutations of alanine substitutions of sites 2, 4, and 6; combined those with the D35 mutant (Fig. 1A); and analyzed the half-life of the mutant proteins (Fig. 1B). Mutation of site 2 slightly but consistently destabilized Cln2. One particular mutant, termed Cln2^M46, that had sites 3, 4, 5 and 6 mutated, was stabilized to a degree comparable to that for the seven-site mutant Cln2^4T3S. Interestingly, these sites cluster in a 35-aa domain (the D domain) that overlaps marginally (9 aa) with the PEST domain, pointing to the D domain as an important determinant of Cln2 instability. When introduced in single copy into the yeast genome and expressed under its own promoter, the mutant Cln2^M46 resulted in higher protein steady-state levels and in elimination of some of the low-mobility phosphorylation forms of Cln2 (Fig. 2 and data not shown). The mean cell size of strain SLY349, containing a single copy of Cln2^M46 as its sole source of G₁ cyclin activity, was reduced by 40% compared to that of the isogenic Cln2^wt strain SLY156, suggesting a hyperactive Cln2/Cdc28 kinase due to high steady-state levels of Cln2^M46 (data not shown). Cln2^M46 was slightly but consistently more hyperactive in mating pheromone and cell size assays than was the double mutant Cln2^D35 (data not shown). Consistent with hyperactive G₁ cyclin kinase, Cln2^M46 cells continued to proliferate in the presence of the growth inhibitor alpha factor (data not shown). Interestingly, overexpression of the stabilized mutants Cln2^M46 and Cln2^4T3S, but not wild-type Cln2, from the GAL1 promoter is lethal, arresting cells in late M phase (C. Wittenberg and S. Lanker, unpublished). In conclusion, we have identified a cluster of four phosphorylation sites in Cln2 that are essential for its rapid degradation.

FIG. 1.

A cluster of four phosphorylation sites destabilizes Cln2. (A) Cln2 protein coding region. The seven Cdc28 phosphorylation sites are numbered (1, T311; 2, T381; 3, S396; 4, T405; 5, S427; 6, T430; 7, S518). The PEST domain is indicated as hatched box. Abbreviations: X, amino acid substitution to alanine; D, the region in mutant M46 harboring the four relevant substitutions that render Cln2 stable. (B) Stability of wild-type and mutant Cln2. Wild-type Cln21 and mutant Cln2 tagged with an HA epitope were pregrown under noninducing conditions (2% raffinose). GAL1-CLN2 or its derivatives were expressed for 45 min by addition of 2% galactose and then repressed by addition of 2% glucose. Extracts were prepared from samples taken at the indicated times and analyzed by immunoblotting using anti-HA antibodies. Equal loading of samples was verified by incubating the blots with anti-Cdc28 antibodies. (C) Graph representing degradation rates from at least two independent experiments. Values were normalized to Cdc28 as internal loading control and to a dilution series (see Materials and Methods).

TABLE 1.

Genotypes and phenotypes of relevant CLN2 mutants and CLN2 domain fusions

Genotype	Name	No. of sites mutated	Half-life (min)a	Binding
				Grr1	Cdc28
Cln2 mutants
CLN2	Wild type	0	8	++	+
CLN2(S518A)		1	6	NDc	+
CLN2(S396,427A)	D35	2	35	ND	+
CLN2(Δ376-514)	ΔHN	5b	55	ND	+
CLN2(Δ376-545)	ΔC	6b	60	ND	+
CLN2(T311,381,405,430A S396,427,518A)	4T3S	7	60	−	+
CLN2(S396,427A T405,430A)	M46	4	55	−	+
CLN2(S396,427D T405,430D)	M4D	4	55	ND	ND
cln2Δxs	Cln2Δxs	0	30	ND	−
cln2Δxs(T311,381,405,430A S396,427,518A)	ClnΔxs(4T3S)	0	30	ND	−
GST-Cln2 fusions
GST	GST	na	166	−	−
GST-CLN2C(376-545)	GST-2C(wt)	0	41	−	−
GST-CLN2(376-545; T405,430A S396,427A)	GST-2C(M46)	6	38	−	−
GST-CLN2(376-431)	GST-PD	0	93	−	−
GST-CLN2(396-431)	GST-D	0	134	−	−
Cln2-ΔN-Sic1 fusions
CLN2C(376-545)-ΔN105-SIC1	2C(wt)-ΔN-Sic1	0	18	++	+
CLN2C(376-545; T405,430A S396,427A)-ΔN105-SIC1	2C(M46)-ΔN-Sic1	4	∼180	−	+
CLN2C(396-431)-ΔN105-SIC1	D-Sic1	0	150	+	+
CLN2C(376-403)-ΔN105-SIC1	P-Sic1	0	90	−	+
CLN2C(396-431)-ΔN105-SIC1	PD-Sic1	0	41	++	+

^aCln2 half-life was determined by densitometric analysis.

^bNumber of sites deleted.

^cND, not determined.

FIG. 2.

Grr1 interacts with phosphorylated Cln2 but not with hypophosphorylated Cln2^4T3S or Cln2^M46. Lysates from strains expressing either untagged GRR1 (lanes 1 and 4) or chromosomally 6His-tagged GRR1 (lanes 2, 3, 5, 6, and 7 to 10) from its endogenous promoter and expressing wild-type (wt) CLN2HA (lanes 1, 2, 4, 5, 7, and 9), CLN^24T3S-HA (lanes 3 and 6), or CLN2^M46-HA (lanes 8 and 10) from the GAL1 promoter were immunoprecipitated (IP) with a monoclonal anti-HA antibody (lanes 4 to 6, 9, and 10). Immunoprecipitates were immunoblotted with anti-His antibody to detect Grr1-6His, with 12CA5 antibody to detect Cln2HA, or with anti-Cdc28 antibody. IgG, immunoglobulin G.

Grr1 interacts with phosphorylated Cln2 in vivo, but not with hypophosphorylated Cln2^4T3S or Cln2^M46.

Cln2 is ubiquitinated by SCF^Grr1 in complex with the E2 Cdc34. The F-box protein Grr1 has been shown to interact with phosphorylated G₁ cyclins Cln1 and Cln2 in vitro (39, 40); in vivo interaction has been demonstrated using overexpression of both Grr1 and Cln2 (17, 22). We evaluated whether the stabilization of the Cln2^4T3S and Cln2^M46 mutants was due to loss of interaction with Grr1. The abundance of Grr1 is very low (12); to preclude potential artifacts associated with overexpression of proteins, we constructed by targeted integration strains that carry C-terminally tagged Grr1 in single copy under control of the native Grr1 promoter. Here we demonstrate, using coimmunoprecipitation experiments, an in vivo interaction between endogenous Grr1 and phosphorylated Cln2 (Fig. 2, upper panel, lanes 5 and 9). Importantly, this interaction was strongly reduced when the mutants Cln2^4T3S and Cln2^M46 were expressed (lanes 6 and 10). However, wild-type as well as both mutant forms of Cln2 bound equally well to Cdc28 (Fig. 2, lower panel, lanes 4, 5, 6, 9, and 10), demonstrating that Cln2^4T3S and Cln2^M46 are folded correctly and maintain the ability to interact with and activate Cdc28. We conclude that Grr1 interacts in vivo with phosphorylated Cln2 and that mutations in phosphorylation sites that stabilize Cln2, Cln2^4T3S, and Cln2^M46 significantly reduce this interaction.

Destabilization of GST-Cln2 domain fusions does not depend on phosphorylation.

Previous studies have suggested that the PEST domain is necessary but not sufficient for Cln2 instability (36). Since we have implicated the phosphorylated D domain as an important protein instability determinant, we wanted to establish the relationship between the PEST and the D domains and ultimately identify the domain(s) necessary and sufficient to promote Cln2 degradation.

We fused the following Cln2 domains to the stable GST polypeptide (Fig. 3A): the Cln2 C terminus (2C; aa 376 to 545), the C terminus harboring the substitutions in each of the four essential Cdc28 phosphorylation sites (2C^M46), the D domain encompassing the four phosphoacceptor sites (aa 396 to 431), and the PD domain (aa 376 to 431). We investigated the stability of the GST fusions by pulse-chase analysis (data not shown). Quantification of at least two independent experiments indicated a half-life of about 170 min for GST and the control GST-SKP1, of about 40 min for GST-2C^wt and GST-2C^M46, of 90 min for GST-PD, and of 135 min for GST-D (Table 1). These data show that the C-terminal domain of Cln2 and, to a lesser degree, the PD domain, but not the D domain alone, are able to confer significant instability to GST. However, the observation that GST-2C^wt and GST-2C^M46 have similar half-lives indicates that, unlike the case for wild-type Cln2, degradation of these fusion constructs is not controlled by phosphorylation. We hypothesized that an SCF^Grr1-independent pathway, which is independent of substrate phosphorylation, degrades Cln2 domains fused to GST. In agreement with this, GST-Cln2 fusions did not interact with Cdc28 in vivo, were not phosphorylated in vivo or in an in vitro phosphorylation assay using insect cell-expressed active Cdc28/Cln2 kinase, and did not interact with Grr1 in vivo (data not shown).

FIG. 3.

The Cln2-2C and -PD domains confer phosphorylation-dependent instability upon a stable ΔN-Sic1 reporter protein. (A) Domains of Cln2 fused to reporter proteins. The top bar shows the CLN2 C terminus containing the PEST domain (open box) and six Cdc28 phosphorylation sites. Below are the Cln2 domains that were fused to the reporter proteins; mutations of serine or threonine residues to alanines are marked (X). (B) Wild-type Sic-HA, stabilized Sic^mut-HA harboring mutations in four phosphorylation sites, or ΔN-Sic1-HA fused to the indicated Cln2 domains was subjected to GAL1 promoter shutoff experiments as described in Fig. 1. Extracts from samples taken at the indicated time points after repression of the GAL1 promoter were analyzed by immunoblotting using anti-HA antibodies. As a loading control, anti-Cdc28 antibodies were used. (C) Graph representing the degradation rates of the various fusions.

Construction of Cln2 domains fused to a stable ΔN-Sic1 reporter.

Sic1 binds to and inhibits B-type cyclin/Cdc28 (Clb/Cdc28) kinase complexes via its C terminus (44). Importantly, Sic1 is a well-characterized substrate of SCF^Cdc4; the N-terminal 105 aa are necessary and sufficient for SCF^Cdc4-mediated ubiquitination and degradation (11, 39, 44). Based on our GST fusion experiments we hypothesized that fusions of Cln2 domains to stabilized Sic1 lacking the N terminus would target the Cln2-Sic1 fusions for SCF^Grr1-dependent degradation. Accordingly, we decided to fuse the following Cln2 domains—2C^wt, 2C^M46, PD, and D—as well as the PEST domain (P; aa 376 to 403) to an N-terminally truncated Sic1 (ΔN-Sic1; aa 106 to C terminus) containing an HA-6His double tag. The fusion genes were put under control of the inducible GAL1 promoter to allow for controlled expression and repression by galactose and glucose, respectively.

The PD-ΔN-Sic1 fusion is rapidly degraded.

First, we tested the stability of Cln2-Sic1 fusions by promoter shutoff experiments. Cells were grown in raffinose-containing medium; expression of the fusion proteins was induced with galactose for 1 h and then repressed with glucose. Samples were collected at the time points indicated after glucose addition. As shown previously (43), wild-type Sic1 was very unstable, with a half-life of less than 5 min, whereas a Sic1 phosphorylation site mutant is extremely stable (Fig. 3B, top two panels). Grafting the 2C domain onto the ΔN-Sic1 protein destabilized it dramatically, with a half-life of approximately 18 min (Fig. 3B, third panel). Importantly, however, the mutant 2C^M46 was highly stable (Fig. 3B, fourth panel; half-life more than 180 min). The PD fusion was rapidly degraded (Fig. 3B, last panel; half-life about 41 min), suggesting that we have identified a domain within Cln2 sufficient to confer significant instability to ΔN-Sic1. In contrast, the D and the P fusion proteins were stable (Fig. 3B, fifth and sixth panels) with half-lives of about 150 and 90 min, respectively. Since 2C^wt and PD carrying four intact phosphorylation sites were rapidly degraded whereas nonphosphorylatable 2C^M46 was highly stabilized, we suggest that degradation of these fusion proteins is phosphorylation dependent.

2C^wt, P, and PD fusions are phosphorylated.

We have demonstrated previously that Cln2 is phosphorylated in vivo and in vitro in a Cdc28-dependent manner (24). Cln2 phosphorylation was manifested on SDS-polyacrylamide gels as a series of low-mobility bands that collapsed into a single band upon treatment with calf intestine phosphatase. We also showed that phosphorylation of the stable M46 mutant Cln2 is reduced more than 90% compared to that of the wild type and runs as a single band on SDS-polyacrylamide gels. Therefore, we analyzed the electrophoretic behavior of the fusion proteins. The low-mobility bands of fusion proteins 2C^wt and PD were eliminated upon treatment with CIP in the absence of phosphatase inhibitors (Fig. 4A, lanes 4 and 8) but not in the presence of inhibitors (Fig. 4A, lanes 3 and 7). The P domain produced a double band in the presence of phosphatase inhibitors (Fig. 4A, lane 9). Upon treatment with CIP, most of the lower-mobility forms were converted to higher-mobility forms; the double bands persisted, suggesting that the P fusion contains an additional modification (Fig. 4A, lane 10). Both 2C^M46 and D did not exhibit any low-mobility forms (Fig. 4A, lanes 1 and 5) and were unchanged upon CIP treatment (Fig. 4A, lanes 2 and 6). This experiment demonstrates that the fusion proteins 2C^wt and PD and, to a lesser degree, P are modified by phosphorylation. As expected, the 2C^M46 did not display any modified forms. In this assay, the D fusion appears to be unphosphorylated; however, we cannot exclude the possibility that the D fusion is modified by phosphorylation that does not result in a shift on SDS-PAGE.

FIG. 4.

The Cln2-2C and -PD domain fusions are phosphorylated and ubiquitinated in vivo. (A) Cln2-ΔN-Sic1-HA fusion proteins were captured on anti-HA beads and incubated for 30 min with CIP in the presence (+) or absence (−) of phosphatase inhibitors, followed by immunoblotting using anti-HA antibodies. The encircled “P” designates phosphorylated species. (B) Extracts prepared from cells expressing the indicated Cln2-ΔN-Sic1-6His-HA fusions and 6His-Myc-tagged K48R, G76A mutant ubiquitin (6His-MYC-UBI-RA) were chromatographed over Ni-NTA beads in a buffer containing 6 M guanidine hydrochloride. Bound proteins and ubiquitin conjugates were analyzed by immunoblotting using anti-HA antibodies.

2C- and PD-ΔN-Sic1 fusions are ubiquitinated.

Our previous result showed that phosphorylated Cln2 was ubiquitinated in vivo in a Cdc53- and Cdc34-dependent manner (45), consistent with Cln2 degradation via the SCF^Grr1 pathway. Here we have examined whether the Cln2-ΔN-Sic1 fusions were ubiquitinated in vivo. To be able to detect in vivo polyubiquitinated proteins, we coexpressed a polyhistidine- and Myc-tagged K48R, G76A mutant ubiquitin gene (Ubi-6His-MYC-RA) together with the Cln2-Sic1-HA-6His fusions. Ubi-6His-MYC-RA competes with endogenous ubiquitin and is incorporated into ubiquitin chains, but the K48R substitution prevents multiubiquitin chain formation; proteins that contain short chains of ubiquitin (up to three ubiquitin moieties) are poor substrates for the proteasome. The G76A substitution inhibits hydrolysis by ubiquitin isopeptidases (24). Therefore, the double-mutant tagged ubiquitin was expected to enrich for ubiquitinated species in vivo. Denatured extracts from strains expressing the fusion constructs and Ubi-6His-MYC-RA were incubated with Ni²⁺-NTA beads, and after extensive washing, bound proteins were eluted and analyzed by Western blotting with anti-HA antibody. Note that under these conditions, ubiquitinated Sic1 fusion species will be detected as a lower-mobility smear above the nonubiquitinated fusion proteins. Consistent with their instability, the Sic1 fusions 2C^wt and PD displayed extensive ubiquitination (Fig. 4B, lanes 1 and 4), while 2C^M46, D, and P did not show significant ubiquitination (Fig. 4B, lanes 2, 3, and 5). We conclude that 2C^wt and, in particular, the PD domain contain signals sufficient to promote ubiquitination most likely via SCF^Grr1 (see below). Neither the PEST domain P, although phosphorylated to some degree, nor the D domain alone is sufficient to attract the ubiquitination machinery.

Grr1, but not Cdc4, binds to 2C-, PD, and D-ΔN-Sic1 fusions.

SCF specificity relies on the nature of the F-box protein: Sic1 is targeted for degradation via binding of Cdc4 to its phosphorylated N terminus, while Cln2 is targeted by Grr1. To investigate whether Grr1 would recognize the various Cln2 domains fused to ΔN-Sic1, the fusion constructs were transformed into yeast strains that carried C-terminally 6His-tagged Grr1 in single copy under control of the native GRR1 promoter. Expression of the HA-tagged fusion proteins was induced briefly with galactose, and coimmunoprecipitation with anti-HA beads was performed followed by SDS-PAGE and Western blotting. Grr1-6His bound to 2C^wt but not to 2C^M46 (Fig. 5, lanes 9 and 10). Grr1-6His coprecipitated with the PD domain (Fig. 5, lane 12) and, unexpectedly, the D domain (Fig. 5, lane 11) but not with the P domain (Fig. 5, lane 13). These results show that, as with full-length Cln2, phosphorylation of the 2C-Sic1 fusion is required for its interaction with Grr1, since the mutation of the phosphorylation sites in the fusion protein 2C^M46 abolished the interaction. Interestingly, the D domain was sufficient for binding to Grr1, albeit with reduced efficiency compared to PD (Fig. 5, compare lanes 11 and 12). In contrast, we were not able to detect an interaction between Cdc4 and 2C-ΔN-Sic1 fusions (Fig. 5B). We conclude that the 2C and the PD fusions to ΔN-Sic1 target these chimeras for recognition by Grr1, thus switching the F-box protein specificity for Sic1 from Cdc4 to Grr1.

FIG. 5.

Grr1 interacts with Cln2-2C, D, and PD fused to ΔN-Sic1-HA-6His. (A) Lysates from strains expressing either untagged GRR1 (lanes 1 and 8) or chromosomally 6His-tagged GRR1 (lanes 2, 3, 5, 6, and 7 to 10) from its endogenous promoter and those expressing wild-type CLN2HA (lanes 7 and 14), or the indicated Cln2 domain fusions to ΔN-Sic1-HA-6His from the GAL1 promoter, were immunoprecipitated (IP) with a monoclonal anti-HA antibody (α-HA) (lanes 8 to 14). Immunoprecipitates were immunoblotted with anti-His antibody (α-His) to detect Grr1-6His or with anti-HA antibody to detect Cln2HA. (B) Cdc4 fails to interact with 2C-ΔN-Sic1-HA-6His. Lysates from strains expressing chromosomally HA-tagged CDC4 from its endogenous promoter and expressing 2C-ΔN-Sic1-HA-6His (lanes 1 to 3) or Sic1-HA-6His (lanes 4 to 6) from the GAL1 promoter were incubated with Ni²⁺-NTA beads. Extracts (L), flowthrough (F), and bound proteins (B) were analyzed by immunoblotting with anti-HA antibody.

Degradation of Cln2-ΔN-Sic1 fusions does not depend on Clb/Cdc28 kinase.

What kinase(s) is involved in degradation of the Cln2-ΔN-Sic1 fusions? Our previous experiments demonstrated that functional Cdc28 as well as G₁ cyclins is essential for degradation of Cln2 (24, 37); however, another report suggested that Clb/Cdc28 kinase can phosphorylate G₁ cyclins, thus targeting them for degradation (4). To address the role that Clb/Cdc28 kinase might play in degradation of the Cln2-ΔN-Sic1 fusions, we first verified that the Cln2-ΔN-Sic1 fusions interacted with Cdc28, most likely via the Sic1 C terminus. Cln2-ΔN-Sic1 fusions were precipitated with anti-HA beads, and proteins bound were analyzed by Western blotting using anti-HA and anti-Cdc28 antibodies. All Cln2-ΔN-Sic1 fusion proteins were able to bind Cdc28 (Fig. 6A, lanes 6, 9, 12, 15, and 18), while the negative control Cln2Δxs (Cln2 with a deletion in the cyclin box [7]) was unable to interact with Cdc28 (lane 3).

FIG. 6.

(A) Cdc28 binds to Cln2-ΔN-Sic1 fusions. Strains expressing the indicated fusions (lanes 4 to 18) or a mutant Cln2Δxs lacking its cyclin box were subjected to coimmunoprecipitation and immunoblotting as described in the legend to Fig. 5, except that anti-Cdc28 antibody (α-Cdc28) was used to detect bound Cdc28. Abbreviations: E, extract; F, flowthrough; B, bound; α-HA, anti-HA antibody. (B) Deletion of the Sic1 C terminus results in loss of Clb-Cdc28 binding. Strains expressing 2C-ΔN-Sic1 (lanes 4 and 8) or Cln2 domains fused to ΔN-Sic1 lacking its C terminus (lanes 1 to 3 and 5 to 7) were subjected to coimmunoprecipitation (IP) and immunoblotting as described in the legend to Fig. 5, except that anti-Cdc28 antibody was used to detect bound Cdc28. (C) Deletion of the Sic1 C terminus does not affect the stability of 2C-ΔN-Sic1. 2C-ΔN-Sic1-HA and 2C-ΔN-Sic1-HA with a C-terminal deletion of the Clb/Cdc28 binding site were subjected to GAL1 promoter shutoff experiments as described in the legend to Fig. 1. Extracts from samples taken at the indicated time points after repression of the GAL1 promoter were analyzed by immunoblotting using anti-HA antibodies. As a loading control, anti-Cdc28 antibodies were used.

To address whether Clb/Cdc8 binding to the Sic1 C terminus was necessary for Cln2-ΔN-Sic1 degradation (43), we deleted the Clb/Cdc28 binding domain in the C terminus of the various Sic1 fusions (Sic1-ΔC) (44) (Fig. 6B) and analyzed the stability of the resulting deletion mutants. Degradation of 2C-ΔN-Sic1-ΔC fusions was unaltered (Fig. 6C), while the PD-ΔN-Sic1-ΔC fusions were slightly stabilized (data not shown), suggesting that C-terminal binding to Clb/Cdc28 is not essential for rapid degradation of the Cln2-ΔN-Sic1 fusions.

Is Clb/Cdc28 kinase essential for phosphorylation and degradation of Cln2-ΔN-Sic1 fusions in trans? If so, then inhibition of Clb/Cdc28 kinase by expression of undegradable Sic1 (Sic1^4A) (43) should alter the degradation rate of the Cln2-ΔN-Sic1 fusions. However, we observed at most a slight increase in stability of the Cln2-ΔN-Sic1 fusions when coexpressed with Sic1^4A (data not shown), suggesting that, as is the case for wild-type Cln2, the G₁ cyclin/Cdc28 kinase is the major kinase that phosphorylates the Cln2-ΔN-Sic1 fusions, initiating their ubiquitination and degradation.

Cdc28 is not required for interaction of phosphorylated Cln2 with Grr1 in vitro.

We envision two plausible explanations accounting for the observed dependence of Cln2 degradation on Cdc28: (i) Cdc28 might be necessary solely for the phosphorylation step, but not for subsequent steps including Grr1 binding; (ii) alternatively, interaction with Cdc28 might be important for recognition by Grr1 and, possibly, later steps including ubiquitination and degradation. To distinguish between these models, we tested whether the presence of Cdc28 in a complex with the fusion protein is required for the interaction between Grr1 and Cln2 in vitro. Cln2HA was expressed from a Gal-inducible promoter and precipitated with anti-HA beads (Fig. 7, lanes 1 and 3, lower panel). The beads were washed either with lysis buffer or with buffer containing 1% SDS to wash off the bound Cdc28 (Fig. 7, lanes 2 and 4) and incubated with extracts from insect cells expressing Grr1-6His. Grr1 was bound to Cln2 independently of whether Cdc28 was present or absent (Fig. 7, lanes 1 and 2, top panel). The negative controls with mock insect cell extract (Fig. 7, lanes 3 and 4) or mock yeast extract (Fig. 7, lanes 5 to 8) did not retain significant amounts of Grr1-6His. SCF-mediated degradation is a highly conserved process, and it is known that insect cells can complement missing yeast components of the core SCF complex (20, 40). To avoid the possibility that an insect cell-derived Cdc28 homolog might bind to Cln2, we purified Grr1-6His over Ni-NTA beads and repeated the experiment (Fig. 7, lanes 9 to 12). Again, Grr1 interacted with column-bound Cln2 irrespective of the presence or absence (Fig. 7, lanes 9 and 10, respectively) of Cdc28. These results show that Grr1 can interact with phosphorylated Cln2 in vitro in the absence of Cdc28.

FIG. 7.

Release of Cdc28 from phosphorylated Cln2 does not affect binding to Grr1. Extracts from strains expressing Cln2HA (lanes 1 to 4 and 9 to 12) or not expressing Cln2HA (lanes 5 to 8) were incubated with anti-HA beads for 1 h and washed two times under low-stringency (lysis buffer; lanes 1, 3, 5, 7, 9, and 11) or high-stringency (lysis buffer + 1% SDS) conditions, followed by incubation in extracts from insects cells expressing 6His-Grr1 (lanes 1, 2, 5, and 6), mock extracts (lanes 3, 4, 7, and 8), Ni-NTA-purified 6His-Grr1 (lanes 9, 10), or mock-purified extracts (lanes 11 and 12) for 1 h at 4°C. Beads were washed and boiled in 2× SDS sample buffer, and bound proteins were analyzed by immunoblotting using anti-His antibody to detect 6His-Grr1, anti-HA antibody to detect Cln2HA, and anti-Cdc28 antibody. IP, immunoprecipitation.

Cln2-ΔN-Sic1 fusions depend on functional SCF^Grr1 but not on functional SCF^Cdc4.

Undegradable Sic1 is lethal due to inhibition of S-phase initiation (43). We showed that fusion of Cln2 domains to undegradable ΔN-Sic1 destabilized ΔN-Sic1. Therefore, we hypothesized that fusions of Cln2 domains to ΔN-Sic1 should alleviate the lethality of ΔN-Sic1. Indeed, cells expressing 2C-ΔN-Sic1, PD-ΔN-Sic1, or wild-type Sic1 from the GAL1 promoter were able to grow on galactose-containing plates (Fig. 8A), whereas cells expressing stabilized Sic1 or 2C^M46-ΔN-Sic1 were not. In addition, since Cln2-ΔN-Sic1 fusion degradation is dependent on SCF^Grr1, Grr1 should become essential in cells expressing Cln2-ΔN-Sic1 fusions, whereas Grr1 is dispensable in wild-type cells. Indeed, we observed that grr1Δ strains expressing 2C-ΔN-Sic1 or PD-ΔN-Sic1 from the GAL1 promoter were not able to grow on galactose plates (Fig. 8B).

FIG. 8.

Cln2-ΔN-Sic1 fusions are degraded by SCF^Grr1. (A) Cln2 fusions alleviate the lethality of stabilized Sic1. Strains expressing Sic1; stabilized Sic1^4A (Sic1^m); empty vector (vec); Cln2; or fusions of stabilized Sic1 (ΔN-Sic1) with 2C (2C), mutant 2C^M46 (2C^m), and PD (PD)—all under control of the GAL1 promoter—were replica plated onto glucose- and galactose-containing plates and incubated at 30°C for 3 days. (B) Grr1 is essential in strains expressing Cln2-ΔN-Sic1 fusions. Wild type (wt) strains or strains with Grr1 deleted (grr1Δ) and expressing 2C-ΔN-Sic1 (2C), or PD-ΔN-Sic1 (PD) under control of the GAL1 promoter, were replica plated onto glucose- and galactose-containing plates and incubated at 30°C for 3 days. (C) Degradation of 2C and PD fusions to ΔN-Sic1 depends on functional SCF^Grr1. 2C-ΔN-Sic1-HA or Sic1-HA was expressed in a wt strain, in a grr1Δ strain, or in temperature-sensitive strains (cdc53ts and cdc4ts) grown at 37°C and subjected to GAL1 promoter shutoff experiments as described in the legend to Fig. 1. Extracts from samples taken at the indicated time points after repression of the GAL1 promoter were analyzed by immunoblotting using anti-HA antibodies. (D) Graph representing the degradation rate of 2C-ΔN-Sic1 in the indicated strains.

The finding that the Cln2-ΔN-Sic1 fusions bind to Grr1 (Fig. 5) suggested that their degradation was no longer dependent on SCF^Cdc4 but rather was dependent on functional SCF^Grr1. To test this, we expressed the fusion constructs and wild-type Sic1 in strains bearing inactivated alleles of grr1, cdc53, and cdc4. The 2C- and PD-ΔN-Sic1 fusions were stable in strains with grr1 deleted (grr1Δ) and in strains with a temperature-sensitive cdc53 [cdc53(Ts)] allele grown at the restrictive temperature (37°C) (Fig. 8C and D). In contrast, stability of the Cln2-ΔN-Sic1 fusions was unaltered in cdc4(Ts) strains at 37°C. As expected (13), Sic1 was completely stable in cdc4(Ts) strains at 37°C (Fig. 8C). We conclude that fusion of 2C or PD to ΔN-Sic1 switches degradation of Sic1 from an SCF^Cdc4-dependent to an SCF^Grr1-dependent pathway.

DISCUSSION

Cln2 PD domain: transferable signal for SCF^Grr1-mediated degradation.

One of the key issues in ubiquitin-mediated degradation is substrate recognition. In this report, we have identified a domain in the Cln2 C terminus, the PD domain, which is sufficient to convert a stable reporter polypeptide, ΔN-Sic1, into an efficient substrate for SCF^Grr1-mediated degradation via the ubiquitin-proteasome pathway. The PD domain encompasses the PEST sequence (P) and four Cdc28 phosphorylation sites crucial for Cln2 instability (D) (Fig. 1 and 3) (24). Substrate phosphorylation has emerged as a hallmark for recognition by the SCF-dependent proteolysis (11, 24, 39, 44). We show that the PD-ΔN-Sic1 fusion and the 2C^wt-ΔN-Sic1 fusion are unstable proteins that are phosphorylated and ubiquitinated in vivo and are recognized by Grr1. Thus, our data strongly suggest that, similarly to wild-type Cln2, ΔN-Sic1 fusions are degraded via the SCF^Grr1 pathway. In contrast, neither the phosphorylation mutant 2C^M46 nor the P or D domain was sufficient to target the ΔN-Sic1 fusion for rapid degradation. Interestingly, the D domain is able to interact with Grr1 in a coprecipitation experiment, albeit to a lower degree than the PD or the 2C domain. However, the D fusion does not display a ubiquitin ladder in vivo, nor were we able to detect phosphorylation, although we do not rule out the possibility that the D fusion is phosphorylated in vivo without causing the characteristic shift in mobility on SDS-polyacrylamide gels. The P fusion is modified in vivo, most likely by phosphorylation, since phosphatase treatment reduces the size of the double band. Neither D nor P fusions are ubiquitinated in vivo, consistent with their inability to target ΔN-Sic1 for rapid degradation.

PEST domain: protein instability motif?

The role of the PEST motif in protein degradation has been under investigation since its first description (33, 35). No clear function has yet been assigned to PEST domains. Indeed, one example, ubiquitin-mediated degradation of the yeast Matα2 repressor, clearly does not depend on the PEST motif (19). An earlier report investigating Cln2 protein turnover implicated the PEST domain as necessary but not sufficient for rapid degradation (36). The authors based their conclusion on the finding that elimination of a 37-residue segment encompassing the PEST domain stabilized Cln2 but that the isolated PEST motif was unable to confer instability when fused to a stable reporter protein. In agreement with their conclusion, we demonstrate here that the PEST motif alone is not sufficient to significantly destabilize ΔN-Sic1, although we do observe a slight reduction in the half-life of the P-ΔN-Sic1 fusion compared to that of stabilized Sic1 or D-ΔN-Sic1 (Fig. 5). In addition, the P domain fusion is not detectably ubiquitinated and does not bind to Grr1; thus, we conclude that the PEST motif is important but not sufficient for Cln2 degradation: it requires the phosphorylation-dependent signaling of the adjacent D domain to perform its function within the ubiquitin-dependent degradation pathway.

The PEST motif seems to be necessary for the degradation of some but not all SCF substrates. Since the PEST motif has never been demonstrated to be sufficient for signal-induced degradation, we hypothesize that the PEST domain represents a constitutive signal for degradation that can be modulated, often by phosphorylation, to become a signal-dependent motif for the SCF degradation machinery. Our data with the GST fusions demonstrate that the isolated PEST motif can destabilize reporter proteins, but the lack of proper phosphorylation results in SCF^Grr1-independent, intermediate instability. A similar conclusion was reached for PEST domain- and phosphorylation-dependent degradation of Cln3 (49). These authors concluded that PEST elements might function to confer phosphorylation dependence on degradation. We note, however, that the Cln3 PEST domain appears to be phosphorylated when fused to beta-galactosidase, a protein that does not bind Cdc28; in contrast, our Cln2 PEST domain fusion to GST is not phosphorylated, a difference that awaits further experimentation to be understood.

Phosphorylation-dependent recognition of Cln2 by Grr1.

While several models can be imagined to account for the results shown above, we favor the multisite phospho-epitope model for degradation of Cln2 (Fig. 9). According to this model, phosphorylated Cln2 residues participate directly in binding to Grr1. Initially, Grr1 (via the C terminus [see below]) might bind weakly to the hypophosphorylated D domain of Cln2/Cdc28 complexes, consistent with the observed residual interaction between the Cln2 phosphorylation mutants and Grr1 (Fig. 2). Subsequently, phosphorylation at multiple sites generates additional, high-affinity binding sites for Grr1, thus stabilizing the interaction. High-affinity binding allows the SCF^Grr1/Cdc34 complex to efficiently ubiquitinate Cln2. Multiubiquitinated Cln2 is finally recognized by the proteasome and rapidly degraded, freeing the bound Cdc28 subunit as well as the SCF^Grr1 complex.

FIG. 9.

Model of phosphorylation-induced binding of SCF^Grr1 to Cln2. For an explanation, see text. Abbreviations: CB, cyclin box; P, Cln2 PEST domain; D, Cln2 D domain; F, Grr1 F box; LRR, Grr1 LRR domain; 34, Cdc34; 53, Cdc53.

We propose that multistep phosphorylation of Cln2 by Cln2/Cdc28 occurs in trans, setting a threshold level for G₁ cyclin kinase activity. This establishes a minimal G₁ phase duration, ensuring that essential G₁ phase events dependent on G₁ cyclin kinase activity have been initiated prior to the degradation of G₁ cyclins. A recent report analyzing recognition of phosphorylated Sic1 by Cdc4 proposed a very similar mechanism (29). These authors demonstrate that degradation of Sic1 requires multisite phosphorylation of Sic1, thereby establishing a minimal G₁ phase period required for proper DNA replication. They propose a model whereby multisite phosphorylation of Sic1 converts the gradual accumulation of Cln-Cdc28 kinase into a switch-like response for Sic1 degradation and onset of S phase. Such a switch-like loss of Cln2 would only be effective if Cln2 is phosphorylated in trans. In support of this idea, a gradual increase in Cln2 phosphorylation during G₁ has been documented (28). This model suggests that although Cln2 must bind Cdc28 for Cln2 to become phosphorylated, Cln2 in that complex is not a substrate for its bound Cdc28 activity. Yet we have shown that Cln2 lacking its cyclin box is not a good substrate for phosphorylation in vivo (24). Furthermore, we show here that Cln2 fusions to GST are not phosphorylated. We also show that, although fusions to ΔN-Sic1 or ΔN-Sic1 lacking the C terminus are phosphorylated in vivo, phosphorylation is not dependent upon binding of Clb/Cdc28 to Sic1. There are a number of possible explanations for these observations. It is possible that recognition of Cln2 as a substrate requires sequences outside the PD domain. We suppose that, in the context of intact Cln2, Cdc28 binding is required for that recognition, whereas Sic1 provides the appropriate context in the absence of bound CDK. The fact that Sic1 is a Cln2/Cdc28 substrate suggests that interaction with Cln2 can occur but does not lead to stable binding. We show here that Cdc28 binding to the inhibitory site on Sic1 is not required for degradation of Cln2/ΔN-Sic1 chimeras. However, that does not preclude a contribution of a distinct binding site on Sic1 for Cln2/Cdc28 kinase. It is unclear whether sequences outside the Sic1 N terminus might contribute to such an interaction. Clearly, further experimentation is needed to fully understand this issue. Cln2 C-terminal phosphorylation has been shown to regulate the cellular localization of Cln2 (10, 28). Interestingly, the Cln2^4T3S phosphorylation mutant was shown to accumulate in the nucleus, raising the possibility that the stability of different Cln2 phosphorylation mutants is related to localization rather than to binding to Grr1. While this is an interesting mechanism, published data demonstrating that Grr1 is localized both in the nucleus and in the cytoplasm would argue against it (3). It will be of interest to correlate the localization of differently mutated Cln2 mutants and of the Cln2-ΔN-Sic1 fusions to their stability.

We have recently developed a structural model for the Grr1 LRR (17). We predicted a high density of positively charged residues on the concave surface of the LRR and hypothesized that certain basic residues on the concave surface are essential for interaction with phosphorylated Cln2. Indeed, when these basic residues were mutated to neutral or acidic residues, interaction with Cln2 was lost. As a result, Cln2 and Gic2, another SCF^Grr1 substrate (18), were strongly stabilized, demonstrating that Grr1 interacts with phosphorylated targets via basic residues on the concave surface of the Grr1 (LRR) domain (17). Thus, we suggest that electrostatic interactions between basic residues on Grr1 LRR and phosphorylated residues on Cln2 contribute a portion of the energy required for that interaction. In an attempt to create a constitutively unstable Cln2, we generated a mutant Cln2 that carried glutamic acid substitutions in the four phosphoacceptor sites within the D domain (Cln2^4D). However, Cln2^4D remained functional, failed to bind to Grr1 in vivo, and was highly stable (data not shown), suggesting that a negative charge at those four Cln2 residues is not sufficient for interaction with Grr1.

It is noteworthy that additional domains besides the Grr1 LRR domain appear to contribute to the interaction between Grr1 and Cln2. We have shown that the Grr1 C terminus is important for stable interaction with Cln2 (17), a finding that might be explained by the recent structural model of human Skp2, a Grr1 homolog (38). This model shows that the Skp2 C terminus wraps back along the entire LRR fold, thus covering the putative substrate interaction domain. Thus, by analogy to the Grr1 C terminus, the Skp2 C terminus might play an important role in substrate recognition.

As demonstrated in this work, four phosphorylated residues within 33 aa in the Cln2 C terminus serve as an essential recognition signal; similarly, multiple phosphorylations within the N terminus target Sic1 for degradation (29, 44). Therefore, binding of F-box proteins to their substrates often appears to depend on a cluster of several phosphorylated residues. In contrast, other phosphorylation-dependent protein-protein interactions often involve single phospho residues or short continuous phosphopeptide sequences, such as phosphoserine motifs recognized by 14-3-3 proteins or phosphotyrosine by SH2-containing proteins.

SCF specificity: F-box hypothesis.

Bai et al. (1) proposed that F-box proteins serve to link different substrates to Skp1 and the SCF machinery. This hypothesis was confirmed in vitro by demonstrating that SCF^Cdc4 binds tightly to and ubiquitinates Sic1, but not Cln1 (11, 40), whereas SCF^Grr1 specifically binds to and ubiquitinates phosphorylated Cln1 and Cln2 but does not recognize Sic1. Here we have demonstrated that the F-box specificity for Sic1 can be switched in vivo from Cdc4 to Grr1 by exchanging the Sic1 signaling domain with the Cln2 PD domain. These results confirm the modularity concept of the SCF machinery and represent, to our knowledge, the first demonstration of a chimeric substrate for the SCF pathway. In a reciprocal approach, Zhou and coworkers (51) recently demonstrated that a mammalian F-box protein, βTrCP, fused to a pocket protein binding peptide (βTrCP^E7N), was able to bind and ubiquitinated stable pocket proteins, including pRB and p107, causing their rapid destruction by the proteasome. These results not only strengthen the F-box hypothesis but also demonstrate that substrate binding—often catalyzed by substrate phosphorylation—is likely the key requirement for degradation by the SCF machinery.

Most SCF-substrate interactions in yeast and mammalian cells have been shown to depend on phosphorylation. Our description of the determinants of Cln2-Grr1 interaction and Cln2 instability may serve as a general model for the study of substrate recognition by SCF complexes.