Ubiquitylation and proteasomal degradation of the p21^Cip1, p27^Kip1 and p57^Kip2 CDK inhibitors

Abstract

The expression levels of the p21^Cip1 family CDK inhibitors (CKIs), p21^Cip1, p27^Kip1 and p57^Kip2, play a pivotal role in the precise regulation of cyclin-dependent kinase (CDK) activity, which is instrumental to proper cell cycle progression. The stabilities of p21^Cip1, p27^Kip1 and p57^Kip2 are all tightly and differentially regulated by ubiquitylation and proteasome-mediated degradation during various stages of the cell cycle, either in steady state or in response to extracellular stimuli, which often elicit site-specific phosphorylation of CKIs triggering their degradation.

Introduction

Ubiquitin (Ub)/proteasome-regulated proteolysis is a key mechanism for regulating many cellular and organismal processes. Conjugation of Ub to substrate proteins requires three enzymes: the Ub-activating enzyme (E1), a Ub-conjugating enzyme (E2), and a Ub ligase (E3). E1 activates Ub through the formation of a thiol-ester bond between the C-terminus of Ub and the active site cysteine (Cys). The activated Ub is then trans-esterified to a conserved Cys of an E2. The E3 ligase interacts with both E2 and the substrate and facilitates polyubiquitylation of the substrate.¹ There are two distinct types of E3 ligases: the enzymatic HECT (homologous to E6-AP C-terminus) domain E3s and the adaptor E3s. HECT domain E3s form a thioester with Ub, which can then be transferred directly to the substrate. The adaptor E3s, containing a RING, SP-RING finger, variant RING/PHD (plant homeodomain) (also named LAP [leukemia-associated protein] domain), or a U box do not form a thioester with Ub but function as adaptors to facilitate the transfer of ubiquitin between a charged E2 and a substrates.^2–5 E3s have a key role in defining the substrate specificity of ubiquitylation.

Cell cycle progression is governed by cyclin-dependent kinases (CDKs), which are activated by cyclin binding and inhibited by CDK inhibitors (CKIs). Although CDK protein levels do not change significantly, the dynamic activities of CDKs during the cell cycle are regulated through expression, ubiquitylation, and degradation of cyclins and CKIs, both temporally and spatially, in addition to phosphorylation and dephosphorylation.¹ CKIs can be divided into two families. The Inhibitor of CDKs (INK4) family includes p15^INK4B, p16^INK4A, p18^INK4C and p19^INK4D; these CKIs specifically bind CDK4 and CDK6 and inhibit cyclin D association. The other family, the kinase inhibitor protein (KIP) or CDK-interacting protein (CIP) family, includes p21^Cip1, p27^Kip1 and p57^Kip2; these CKIs bind and inhibit all cyclin-bound CDKs.^6,7 The ubiquitylation and proteasome-mediated degradation of p16^INK4A,⁸ and p19^INK4D (which depends on binding to CDK4)⁹ has been reported. In this review, we discuss the ubiqutylation and degradation of mammalian p21^Cip1, p27^Kip1, including its analogues Sic1 and Far1 [in Saccharomyces (S.) cerevisiae] and Rum1 (in fission yeast) and p57^Kip2.

The Skp1/cullin/F-box (SCF) complexes, functioning as E3 ligases, play a pivotal role in ubiquitylation of p21^Cip1, p27^Kip1 and p57^Kip2. The SCF complexes consist of three invariant subunits—Skp1, CUL1 [cell division cycle (Cdc) 53p in yeast], and the RING finger protein RBX1 (RING box protein-1) (also known as regulator of cullins-1, Roc1)—as well as a variable component known as an F-box protein. F-box proteins bind to Skp1 through their F box and recognize substrates through other domains in the F-box protein.¹⁰ The human genome encodes 69 F-box proteins.¹¹ Three classes of F-box proteins have been defined on the basis of their substrate recognition motifs: Fbw, F-box proteins containing WD40-repeat domains; Fbl, containing leucine-rich repeats (LRR) domains; and Fbx, containing other domains. The Fbw and Fbl classes of F-box proteins can recognize target phosphodegrons, specific sequences of amino acids in proteins that direct their degradation in a phosphorylation-dependent manner, through their WD40 or LRR domains.¹ The CUL1 scaffold binds RBX1, a RING finger protein that recruits the charged E2, and Skp1. The human genome encodes eight Cullin proteins [CUL1-4A, 4B, 5, 7 and 9 (formerly PARC)] that form similar Cullin-RING Ligase (CRL) complexes.¹¹ Neddylation of CUL1 is necessary for SCF activity as a Ub ligase, because unneddylated CUL1 holds RBX1 in a closed form bound by an inhibitory protein CAND1, thus preventing Skp1-F-box protein binding. Neddylation of CUL1 loosens the RBX1 structure and dissociates the CAND1 inhibitor, allowing binding of the Skp1-F-box protein complex. The reoriented RBX1 bridges the gap between the associated E2 and substrate bound to the F-box protein, facilitating both initial ubiquitylation and subsequent polyubiquitylation.¹¹ SCF complexes, working in concert with the UBC3/ Cdc34 E2 (and other E2s, such as Ubc4), which binds to Rbx1, control G₁-S progression and target CKIs (p21^Cip1, p27^Kip1 and p57^Kip2 in mammals and Sic1 and Far1 in budding yeast) for degradation.¹

p21^Cip1

Expression of the p21^Cip1 mammalian CKI is tightly regulated by signals that control cell division, and also in response to DNA damage, which activates cell cycle checkpoints. Shortly after its discovery, p21^Cip1 was shown to be a short-lived protein that is degraded by the proteasome.¹² Subsequent work has shown that p21^Cip1 can be degraded through both ubiqiutin-dependent and independent mechanisms.

Ubiquitylation-dependent proteasomal degradation of p21^Cip1

Proteolysis of p21^Cip1 has been shown to require a functional Ub-activating enzyme (E1) and conjugation of the Ub-like protein, Nedd8, whose linkage to CULs is necessary for their Ub ligase activity.^13,14 Consistent with a role for SCF complexes, p21^Cip1 was shown to be ubiquitylated at four distinct Lys residues located in the C-terminal region by SCF^Skp2.15 p21^Cip1 interacts with and is phosphorylated on serine (S)130 by cyclin E/ CDK2, which promotes Skp2-dependent p21^Cip1 degradation in S phase^16,17 (Fig. 1). Mutation of all six Lys into Arg, disruption of the two nuclear export signal sequences in p21^Cip1, or treatment with the nuclear-export inhibitor leptomycin B blocks H₂O₂-induced p21^Cip1 degradation, implying that cytoplasmic localization is required for p21^Cip1 degradation.¹⁸ Cytoplasmic p21^Cip1 degradation is also induced by activation of extracellular signal-regulated kinase 2 (ERK2). ERK2 directly interacts with and phosphorylates p21^Cip1 at threonine (T)57 and S130, promoting translocation of p21^Cip1 into the cytoplasm and Ub-dependent degradation by unidentified E3 ligases, thereby resulting in cell cycle progression.¹⁹ In contrast, in transforming growth factor-β1 (TGFβ1) treated cells, p38 and JNK1 are activated and phosphorylate p21^Cip1 at S130 leading to its stabilization. Whether p21^Cip1 phosphorylation at S130 by different kinases somehow results in p21^Cip1 being able to associate with distinct regulators that affect p21^Cip1 stability differentially warrants further investigation. Similarly, T57 within the CDK binding domain of p21^Cip1 is also a substrate of glycogen synthase kinase (GSK)3,²⁰ or JNK,²¹ which result in degradation or stabilization of p21^Cip1, respectively.

Figure 1.

Phosphorylated p21^Cip1 is degraded by distinct E3 ligases. (A) E3 ligases involved in p21^Cip1 degradation. p21^Cip1 is ubiquitylated and degraded in late G₁ and S phases by SCF^Skp2 and CRL4^Cdt2, and in G₂ phase by APC/C^Cdc20 in the nucleus. A portion p21^Cip1 is phosphorylated and translocated into the cytosol where it is ubiquitylated and degraded by proteasomes. (B) Schematic structure of p21^Cip1 showing the regulatory phosphorylation sites and the cognate protein kinases. CDI: CDK inhibitor domain.

HER-2/Neu-activated AKT associates with p21^Cip1 and phosphorylates it at T145, resulting in cytoplasmic localization of p21^Cip1 and promotion of cell growth, whereas a nuclear p21^Cip1 T145A mutant preferentially suppresses growth of transformed cells.²² AKT-dependent phosphorylation of p21^Cip1 at T145 prevents complex formation of p21^Cip1 with proliferating cell nuclear antigen (PCNA) and decreases the binding of CDK2 and CDK4 to p21^Cip1, thereby attenuating the inhibitory activity of p21^Cip1 on CDK and DNA replication.^23,24 In addition to T145, S146 of p21^Cip1 is also phosphorylated by AKT, and this phosphorylation is suggested to stabilize p21^Cip1.24 How phosphorylation of two adjacent residues by AKT leads to opposite effects on p21^Cip1 stability is not yet clear.

Ubiquitylation of p21^Cip1 by the SCF^Skp2 complex requires functional interaction between p21^Cip1 and the cyclin E/CDK2 complex. Mutation of both the cyclin E recruitment motif (RXL) and the CDK2-binding motif (FNF) at the N-terminus of p21^Cip1 abolishes its ubiquitylation by Skp2 (Fig. 1).¹⁵ p21^Cip1 accumulates in Skp2^−/− mouse embryonic fibroblasts during S phase, suggesting that SCF^Skp2 plays a role in p21^Cip1 degradation during the G₁/S transition.²⁵ In addition to Skp2, overexpression of p53-inducible RING-finger protein (p53RFP) causes degradation of p21^Cip1,²⁶ suggesting that multiple E3 ligases are involved in p21^Cip1 ubiquitylation. Additional evidence shows that, in spite of regulation of p21^Cip1 by SCF^Skp2 during the G₁/S transition, p21^Cip1 reaccumulates during G₂ and is degraded again in prometaphase through the interaction of the anaphase-promoting complex/cyclosome (APC/C)^Cdc20 E3 ligase with the destruction box (D box) of p21^Cip1 (Fig. 1). A D-box mutant of p21^Cip1 is resistant to degradation by APC/C^Cdc20 E3 ligase, but not to degradation by SCF^Skp2. Silencing of APC/C^Cdc20 induced accumulation and binding of p21^Cip1 to CDK1, thereby inhibiting CDK1 activity during prometaphase, implying that p21^Cip1 degradation by APC/C^Cdc20 is required for the full activation of CDK1 necessary for mitotic events.²⁵ In contrast to the normal turnover of p21^Cip1 mediated by the SCF^Skp2 and APC/C^Cdc20 complexes at different phases of the cell cycle, ionizing radiation-induced Ub-dependent degradation of p21^Cip1 requires the CRL4^Cdt2 E3 ligase (composed of the Cul4A/B, DDB1, and the DCAF subunit Cdt2) and p21^Cip1 binding to PCNA,^27–30 suggesting that p21^Cip1 stability is regulated by distinct mechanisms in response to different extracellular stimuli.

Since p21^Cip1 accumulates in cells harboring a temperature-sensitive mutation in Ub E1 at the restrictive temperature,^13,14 degradation of p21^Cip1 requires active ubiquitylation, but this does not have to be due to ubiquitylation of p21^Cip1 itself. Although p21^Cip1 ubiquitylation can be detected, there has been a lively debate about which residues in p21^Cip1, if any, need to be ubiquitylated for its degradation, and this issue is complicated by the fact that p21^Cip1 can be degraded by the proteasome in a ubiquitin-independent fashion (see below). Exogenously expressed p21^Cip1 K6R, harboring mutations of all six internal Lys to Arg to prevent ubiquitylation, is still ubiquitylated.¹³ Moreover, p21^Cip1 K6R is unstable, and its abundance also increases in response to proteasome inhibition indistinguishably from WT p21^Cip1.31 These findings suggested that Lys-independent ubiquitylation and proteasomal degradation is involved in regulation of p21^Cip1 stability, and this idea is supported by the evidence that N-terminally-tagged p21^Cip1 can be ubiquitylated on its free α-NH₂ group.^13,14 However, this occurs because the N-terminal Tyr residue of the tag is not acetylated. In contrast, the majority of endogenous p21^Cip1 is acetylated at its N-terminal Ser residue,³² meaning that N-terminal ubiquitylation of p21^Cip1 is unlikely to be physiologically important. Nevertheless, with the recent discovery that N-terminally-acetylated residues, including N-acetyl Ser, can act as a degradation signal (degron) and specify a short half life for yeast proteins, the role of the acetylated N-terminus in p21^Cip1 degradation needs further investigation.³³ Such an ^AcN-degron might dictate the basal half life of p21^Cip1, and involve ubiquitylation of internal Lys residues in p21^Cip1 leading to its degradation by the proteasome.

p21^Cip1 degradation can be inhibited by nucleophosmin (NPM)/B23, a multifunctional protein that binds p21^Cip1. Actinomycin D stimulation increases the interaction and nucleoplasmic translocation of NPM and p21^Cip1 and stabilizes p21^Cip1.34 However, whether NPM-induced p21^CIP1 stabilization is due to retention of p21^Cip1 in nucleus thereby preventing p21^Cip1 degradation in the cytosol remains unclear. p21^Cip1 can also be stabilized by interaction with WISp39, a tetratricopeptide repeat (TPR) protein. The C-terminal TPR domain of WISp39 recruits Hsp90 and forms a trimeric complex to prevent proteasomal degradation of p21^Cip1. Point mutations within the C-terminal TPR domain of WISp39, which abolish the binding of WISp39 to Hsp90 but not to p21^Cip1, fail to stabilize p21, suggesting an essential role of Hsp90 in stabilization of p21^Cip1.35

Ubiquitylation-independent proteasomal degradation of p21^Cip1

Although the degradation of p21^Cip1 is generally believed to depend on proteasomes, other observations suggest that proteasomal degradation of p21^Cip1 does not obligatorily require its ubiquitylation. For instance, MDM2 promotes p21^Cip1 degradation independently of ubiquitylation.³⁶ Ubiquitylation- and Skp2-independent proteasomal degradation of p21^Cip1 also occurs in response to ultraviolet irradiation, which is mediated by phosphorylation of p21^Cip1 at S114 via GSK3β, a downstream effector of the ATR DNA damage signaling pathway.^37,38 Further, the C8 α-subunit of the 20S proteasome has been shown to interact directly with the non-ubiquitylated C-terminus of p21^Cip1 and thereby mediate the degradation of p21^Cip1.39 The role of the C8 α-subunit of the 20S proteasome in p21^Cip1 is also evidenced by Ras signaling-induced p21^Cip1 stabilization. Ras activation promotes the formation of p21^Cip1-cyclin D1 complexes and prevents p21^Cip1 from associating with the C8 α-subunit of the 20S proteasome.⁴⁰ This observation is further supported by the finding that degradation of unbound p21^Cip1 by the 20S proteasome occurs in an ATP- and Ub-independent manner. This process is directly mediated by the proteasome activator REGγ, which binds and activates the 20S proteasome.⁴¹ Depletion of REGγ results in increased p21^Cip1 protein levels and changes in cell cycling and proliferation.

Obviously, further detailed studies on the mechanisms by which p21^Cip1 is regulated in quiescent and cycling cells will be required for a full understanding of the relative importance of the Ub-dependent and Ub-independent processes for degradation of p21^Cip1.

p27^Kip1, Sic1, Rum1 and Far1 p27^Kip1

Regulation of p27^Kip1 stability by phosphorylation of p27^Kip1 at T187 by cyclin/CDK complexes

The most well-studied mammalian CKI is p27^Kip1, which is abundant in quiescent and G₁ cells and is downregulated in proliferating cells and in S- and G₂-phase cells (Fig. 2). p27^Kip1 acts in G₀ and early G₁ to inhibit G₁ cyclin/CDK2 complexes, with the primary target being E-type cyclin/ CDK2.^42–44 Before the onset of S-phase, cyclin E/CDK2 binds to and phosphorylates p27^Kip1 at T187,^45,46 with p27^Kip1 complexed to cyclin E/CDK2 being a target for phosphorylation by a second, active cyclin E/CDK2 complex. Phosphorylation of T187 creates a phosphodegron that recruits Skp2 through LRR domain binding, resulting in p27^Kip1 ubiquitylation by SCF^Skp2 E3 ligase containing either the Rbx1/Roc1 RING finger protein^47–49 or the Ro52 RING finger protein.⁵⁰ In addition to using the CUL1-containing SCF complex, Skp2 interacts with CUL4A-DDB1-associated COP9 signalosomes to induce proteolysis of p27^Kip1.51 Depletion of COP9 signalosome subunits 4 and 5 decreases the levels of Skp2 protein, stabilizes p27^Kip1, and impairs cell proliferation, and this can be partially reversed by suppression of p27^Kip1.52

Figure 2.

Phosphorylated 27^Kip1 is degraded by distinct E3 ligases. (A) E3 ligases involved in p27^Kip1 degradation. p27^Kip1 is ubiquitylated and degraded in late G₁, S and G₂ phases by SCF^Skp2 in the nucleus. p27^Kip1 phosphorylated at S10 is ubiquitylated by the KPC complex when exported to the cytoplasm. (B) Schematic structure of p27^Kip1 showing the regulatory phosphorylation sites and the cognate protein kinases. CDI, CDK inhibitor domain.

The SCF^Skp2-mediated ubiquitylation of p27^Kip1 requires an additional factor, CDK subunit 1 (Cks1), which is a member of the highly conserved Suc1/Cks family of proteins that bind to some cyclin/CDK complexes and phosphorylated proteins. Cks1 binds to the LRR domain and C-terminal tail of Skp2,⁵³ in which a negatively charged residue Asp331 is essential for the interaction.⁵⁴ The formation of complexes between Cks1 and Skp2 causes conformational changes in both proteins and significantly stabilizes the interaction between Skp2 and Skp1 and enhances the binding of Skp2 to p27^Kip1 phosphorylated at T187.^53,55 p27^Kip1 binds to both Cks1 and Skp2 by inserting the side chain of an invariant Glu185 into the interface between Skp2 and Cks1 and interacts with a Cks1 phosphate-binding site via the phosphorylated T187 side chain of p27^Kip1.53 Phosphorylated p27^Kip1, complexed to cyclin E/CDK2, binds to the SCF^Skp2/Cks1 complex in a cooperative manner. Cyclin E/CDK2 contributes to p27^Kip1 binding to the SCF^Skp2/Cks1 by T187 phosphorylation as well as by a potential direct interaction between cyclin E/CDK2 and the SCF^Skp2/Cks1 complex.⁵⁶ In contrast, the interaction between phospho-T187 p27^Kip1 and Cks1 is dramatically reduced by the cis-trans peptidylprolyl isomerase Pin1. Pin1 binds to the phosphorylated Thr187-Pro motif in p27^Kip1 and can cause cis-trans isomerization of this bond altering the conformation of p27^Kip1. p27^Kip1 exhibits increased polyubiquitylation and has a shorter half life in Pin1^−/− mouse embryonic fibroblasts than in wild-type cells.⁵⁷

In addition to cyclin E, cyclin D and cyclin A are also involved in p27^Kip1 degradation by distinct mechanisms. Depletion of cyclin D1 causes p27^Kip1 accumulation, with a simultaneous decrease in CUL1 neddylation and increased binding to CAND1, which blocks the accessibility of CUL1 to Skp1 and Skp2, thus preventing the formation of the SCF^Skp2 complex.⁵⁸ How cyclin D1 is able to promote SCF^Skp2 complex formation, however, remains unclear. Cyclin A/CDK2, but not cyclin B/CDK1, forms a stable complex with T187-phosphorylated p27^Kip1 to stimulate p27^Kip1 ubiquitylation.^46,59 Distinct interaction regions in cyclin A separately bind to Skp2 and p27^Kip1, and Skp2-cyclin A interaction directly protects cyclin A-CDK2 from inhibition by p27^Kip1 through competitive binding. Disruption of cyclin A-Skp2 binding, which does not affect p27^Kip1-cyclin A interaction, compromises Skp2’s proliferation stimulatory activity without affecting its ability to degrade p27^Kip1 and p21^Cip1.60

Although SCF^Skp2 mediates the nuclear ubiquitylation of CDK2-bound p27^Kip1 during S/G₂ and M phases, when p27^Kip1 is translocated into the cytoplasm its stability is controlled in an SCF^Skp2-independent manner (see below).^61–63 In addition, other E3 ligases, such as a HECT domain family ubiquitin ligase E6-AP,⁶⁴ p53-inducible protein with RING-H2 domain (Pirh2),⁶⁵ and CUL4A/CUL4B-based E3 ligases in response to active Wnt signaling^66–68 are also reportedly involved in the ubiquitylation and proteasomal degradation of p27^Kip1,⁶⁴ in an extracellular stimuli- or cell type-specific manner.

Regulation of p27^Kip1 stability by phosphorylation of p27^Kip1 at S10

Stimulation by hepatocyte growth factor (HGF) increases p27^Kip1 phosphorylation at S10 and induces nuclear export of p27^Kip1,⁶⁹ (Fig. 2). In keeping with this, ERK activation induces the cytoplasmic localization of p27^Kip1,⁷⁰ and p27^Kip1 S10A is refractory to Ras-induced cytoplasmic translocation.⁷¹ p27^Kip1 phosphorylated on S10 at the G₀-G₁ transition binds to the nuclear export protein CRM1, leading to its subsequent cytoplasmic translocation.^62,63 The stability of p27^Kip1 exported into the cytoplasm can be regulated by interaction with and ubiquitylation by the E3 complex KPC (Kip1 ubiquitylation-promoting complex), which consists of KPC1 and KPC2. KPC2 contains a Ub-like domain and two ubiquitin-associated (UBA) domains. KPC1 possesses a C-terminal RING-finger domain, and its N-terminal region is involved in the interaction with free p27^Kip1. Thus, KPC may control the degradation of p27^Kip1 exported from the nucleus in G₁ phase.^72–74 However, it remains unknown whether a direct modification of p27^Kip1, such as phosphorylation at S10, is required for KPC-mediated p27^Kip1 degradation. KPC1 itself can be regulated by the USP19 deubiquitinating enzyme, which interacts with and stabilizes KPC1, thereby modulating p27^Kip1 levels and cell proliferation.⁷⁵

p27^Kip1 S10 phosphorylation can depend on both tissue and cell type as well as cell cycle stage. CDK5, an unconventional neuronal CDK that is activated in postmitotic neurons but not in proliferative cells, directly phosphorylates p27^Kip1 at S10, which contributes to neuronal migration in the developing cerebral cortex.⁷⁶ CDK5 is also able to phosphorylate T187 in vitro, but the physiological significance remains unclear.⁷⁶ Other protein kinases can phosphorylate p27^Kip1 at S10 at different stages of the cell cycle: Mirk/DYRK1B, which is maximally active in G₀, phosphorylates and stabilizes p27^Kip1 in quiescent cells;⁷⁷ human kinase-interacting stathmin (KIS), which is activated by mitogens, binds the C-terminal domain of p27^Kip1, phosphorylates S10 at the G₀-G₁ transition, and promotes its nuclear export to the cytoplasm.⁷⁸ In addition, corneal endothelial cells treated with fibroblast growth factor (FGF)-2 possess distinct polyubiquitylation pathways for phospho-T187 and phospho-S10 p27^Kip1; phospho-T187 p27^Kip1 is ubiquitylated through nuclear SCF^Skp2 during late G₁ phase, whereas phospho-S10 p27^Kip1 is ubiquitylated by cytosolic KPC during early G₁ phase.⁷⁹ These results further support the conclusion that p27^Kip1 stability is regulated by phosphorylation at different residues at various stages of the cell cycle. While S10 phosphorylation stabilizes p27^Kip1 in resting cells (G₀ phase), this phosphorylation promotes p27^Kip1 nuclear export and cytoplasmic degradation during early and late G₁ phases.⁸⁰ Consistently, the S10A p27^Kip1 mutant has reduced stability compared with WT p27^Kip1 in G₀ phase whereas mutation of S10 into Asp or Glu to mimic phosphorylation stabilizes p27^Kip1.80 A higher proportion of S10A p27^Kip1 is found in association with cyclin/CDK complexes than WT p27^Kip1, thus promoting p27^Kip1 S10A assembly into cyclin-CDK complexes, which is, in turn, necessary for p27^Kip1 turnover.^71,81 p27^Kip1 S10A knock-in mice have normal body size but exhibit organ-specific reductions in p27^Kip1 expression in brain, thymus, spleen and testis.⁸¹ The reason for this organ specificity in the downregulation of p27^Kip1 S10A remains an enigma.

Regulation of p27^Kip1 stability by phosphorylation of p27^Kip1 at tyrosines 74 and 88 and threonines 157 and 198

Distribution of p27^Kip1 into cyclin/CDK complexes seems to be affected by phosphorylation of p27^Kip1 at T198, an event that can be mediated by AKT or p90 ribosomal protein S6 kinases (RSK)^82–84 (Fig. 2). Phosphorylation at this residue stabilizes free p27^Kip1, whereas loss of this phosphorylation site promotes its binding to CDK2-containing complexes.⁸⁴

Tyrosine (Y) 74 and Y88 of p27^Kip1 can be phosphorylated by the Src-family kinases c-Src and Yes, whereas Lyn and the oncoprotein BCR-ABL appear to phosphorylate predominantly Y88. p27^Kip1 phosphorylation at Y74/88 reduces its steady-state binding to cyclin E-CDK2, thus restoring partial CDK activity. The activated CDK2 phosphorylates p27^Kip1 on T187, which in turn promotes SCF^Skp2-dependent degradation of p27^Kip1.85,86 Consistently, reduced p27^Kip1 expression levels are observed in Src-activated breast cancer lines, correlating with Src activation in primary human breast cancers.⁸⁵ In contrast, p27^Kip1 phosphorylation at Y88/89 was also proposed to regulate CDK4 activity via a distinct mechanism. p27^Kip1 associates with cyclin D-CDK4 constitutively. However, Y88 and Y89 phosphorylated preferentially in proliferating cells converts p27^Kip1 to a non-inhibitor of cyclin D-CDK4 by dislodging p27^Kip1 from the catalytic cleft of CDK4 to allow ATP binding.^87,88

p27^Kip1 T157 can be a substrate of multiple protein kinases. The Pim kinase family members (Pim1, Pim2 and Pim3) bind to and phosphorylate p27^Kip1 at both T157 and T198. Pim-mediated phosphorylation induces p27^Kip1 binding to 14-3-3 proteins, resulting in its nuclear export and proteasome-dependent degradation.⁸⁹ Phosphorylation of p27^Kip1 at T157 is also observed upon AKT and SGK1 activation, which interrupts association of p27^Kip1 with importin-α, thus preventing re-entry of p27^Kip1 into the nucleus.^90,91 p27^Kip1 phosphorylation at S83 by the protein kinase CK2,⁹² and at S178 downstream of the phosphoinositide 3-kinase pathway⁹³ are also reported, although the biological significance of these phosphorylations remains unclear.

Significance of the regulation of p27^Kip1 stability

The degradation of p27^Kip1 is necessary for entry into S phase, as overexpression of a nondegradable p27^Kip1 mutant (T187A), microjection of Skp2 antibody, or antisense oligonucleotides targeting Skp2 arrests cells in G₁ phase.^94,95 Cells derived from p27^Kip1 T187A knock-in mice exhibit rising levels of p27^Kip1 T187A in late G₁/S/G₂ cells and a defect in cell cycle progression and cell proliferation.⁹⁶ Mice expressing this protein develop normally, and for unknown reasons, grow to be larger than WT mice. RBX1, which encodes an SCF subunit, is an essential gene for mouse embryogenesis, and disruption of RBX1 causes embryonic lethality due to reduced proliferation as a result of p27^Kip1 accumulation. Simultaneous loss of p27^Kip1 extends the life span of RBX1-deficient embryos from E6.5 to E9.5,⁹⁷ indicating that a failure to downregulate p27^Kip1 as well as other RBX1 substrates, is detrimental to embryonic development. Both Cks1- and Skp2-deficient mice are smaller than their WT littermates, and their cells tend to proliferate more slowly, which correlates with elevated levels of p27^{Kip1.98–100} Most of the cellular abnormalities apparent in Skp2^−/− mice are not evident in Skp2^−/−-p27^Kip1−/− double-mutant mice, suggesting that p27^Kip1 is a major physiological target of SCF^Skp2.101 This is further supported by the fact that eliminating p27^Kip1 phosphorylation on T187 in p27^Kip1−/− T187A knock-in mice reproduces the effects of Skp2 knockout in preventing spontaneous tumorigenesis in Rb1^+/− mice.¹⁰² In addition, the p27^Kip1 T187A mutation inhibits progression of intestinal adenomas to carcinomas in a mouse model,¹⁰³ and the absence of Skp2, which correlates with increased expression of p27^Kip1, decreases the leukemogenicity of BCR-ABL in a murine model of chronic myelogenous leukemia (CML).¹⁰⁴ Furthermore, Skp2 deficiency restricts tumorigenesis induced by inactivation of Pten or Arf, which is mediated by induction of cellular senescence resulting from concomitant upregulation of p27^Kip1, p21^Cip1 and Atf4.¹⁰⁵ These results provide additional in vivo evidence that p27^Kip1 regulation by Skp2 contributes to the development of some tumor types. However, in certain tumor types p27^Kip1 T187 phosphorylation-induced protein degradation might not be the primary mechanism for regulation of p27^Kip1 stability. For instance, expression of p27^Kip1 T187A is not higher than that of WT p27^Kip1, and the expression of both WT and the T187A mutant of p27^Kip1 is downregulated at the transcriptional level during development of activated K-Ras-induced non-small cell lung cancers in mice.¹⁰³ Moreover, regulation of the p27^Kip1 transcript levels in human breast cancers has been reported.¹⁰³

In agreement with the requirement for phosphorylation by and association with cyclins/CDKs for Skp2-mediated p27^Kip1 degradation, the expression levels of p27^CK− in p27^CK− knock-in cells, which carry point mutations in protein interaction domains that abolish p27^Kip1 binding to cyclins and CDKs and consequently cannot inhibit cyclins/CDKs, are increased in various tissues and mouse embryonic fibroblasts.⁷¹ As expected, an increased growth rate and body size and general organomegaly are observed in p27^CK− knock-in mice as well as in p27^Kip1−/− mice. p27^Kip1−/− mice also exhibit sterility (in females) and disrupted retinal architecture and develop multiorgan hyperplasia and pituitary tumors.^106–108 Furthermore, a decline or loss of p27^Kip1 protein, with an associated increase in Skp2 protein levels, has been detected in many human cancers and is of prognostic significance.⁷ In addition, indirect regulation of p27^Kip1 may also contribute to human cancer development. For instance, inactivating mutations in the Fbw7 F-box protein (also known as hCdc4 and archipelago) in a number of different human cancers, which result in loss of SCF^Fbw7-mediated cyclin E ubiquitylation and accumulation of cyclin E,^109,110 may enhance p27^Kip1 degradation indirectly by contributing to p27^Kip1 binding to SCF^Skp2.56

The Sic1p budding yeast CKI

In S. cerevisiae, the CDK inhibitor Sic1p, which is functionally and structurally analogous to the mammalian p27^Kip1,¹¹¹ must be degraded for the onset of B-type cyclin/CDK (Clb/Cdc28) activity and consequent DNA replication.¹¹¹ Sic1p is phosphorylated in late G₁ phase by G₁ cyclin/CDK (Cln/Cdc28) activity, and the phosphorylated Sic1p binds to the WD40 repeat domain of the Cdc4 F-box protein.^112–114 Cdc4 dimerization via its D domain, which facilitates Ub conjugation but not substrate recognition, is required for SCF^Cdc4 function and Sic1p ubiquitylation.^115,116 Six N-terminal lysines of Sic1p serve as the major ubiquitylation sites.¹¹⁷ Sic1p has nine suboptimal Cdc4 phosphodegrons (CPDs);^118,119 Sic1p mutants that lack multiple CDK phosphorylation sites are stabilized, and hence arrest cells in G₁ phase.¹¹² The nine Sic1p phosphorylation sites form three separate Cdc4-recognition phosphodegrons (T2/5/9, T33/45/48 and S69/76/80), and each degron contains two essential phosphorylated residues and binds to Cdc4 with similar affinities. Double phosphorylation of a single degron is necessary and sufficient for binding to Skp1-Cdc4.¹¹⁶ The nine phosphorylation sites can be replaced by a single high affinity CPD, leading to the proposal that multiple suboptimal CPDs in Sic1p serve as a mechanism for setting a threshold of cyclin/CDK phosphorylation before Sic1p degradation can be triggered.¹¹⁹ Kinetic analysis reveals that Sic1p ubiquitylation has an initiation step, which is a rate-limiting attachment of the first Ub followed by the acidic tail loop of Cdc34 (an E2)-dependent rapid synthesis of K48-linked Ub chains.¹²⁰ Subsequently, polyubiquitin chains are built on SCF substrates by sequential transfers of single ubiquitins.¹²¹ Protein kinase CK2-mediated phosphorylation of Cdc34 on the acidic tail domain (S207/216 in yeast and S203/222/231 in humans) stimulates Cdc34-SCF^Cdc4 ubiquitylation activity toward Sic1p and cell cycle progression.¹²² The multiubiquitin chain binding proteins (MCBPs) Rad23 and Rpn10 contribute to recruitment of ubiquitylated Sic1p to the 26S proteasome, where Rpn11 metalloprotease is essential for deubiquitylation and degradation of Sic1p.^123,124 Intriguingly, analysis using an in vitro biochemical reconstitution system shows that Sic1p degradation is essential for triggering the ATP hydrolysis-dependent dissociation and disassembly of the 19S regulatory particles from the 26S proteasome, implying that this is a general process for degradation of other proteins.¹²⁵

During sporulation, Sic1p degradation is independent of Cdc28 but requires Ime2 protein kinase activity.^126,127 The meiosis-specific kinase Ime2 phosphorylates only a subset of the Sic1p sites corresponding to CDK sites, which, by itself, is insufficient to promote Sic1p binding to Cdc4 and Sic1 degradation¹²⁸ The identity of the other kinase(s) that phosphorylate Sic1 in meiosis to promote Sic1p degradation in combination with Ime2 remains obscure. Intriguingly, Ime2p kinase is degraded by SCF^Grr1p upon glucose stimulation, which increases Sic1p levels, resulting in blockage of meiotic DNA replication.¹²⁹

The Rum1 fission yeast CKI

Rum1, the sole CKI in fission yeast, is essential for proper regulation of the G₁/S transition, inhibiting Cdc13/Cdc2 complex activity. Rum1, the fission yeast analogue of p27^Kip1 and Sic1p,¹³⁰ is a substrate of SCF complexes containing the fission yeast orthologues of Cdc4, Pop1 and Pop2.^131–134 Rum1 degradation requires phosphorylation of Rum1 at T58 and T62 by Cig1/Cdc2 (cyclin/CDK) and covalent attachment of NEDD8 to the Cul-family proteins Pcu1 and Pcu4 (Cul-1 and Cul-4A/Cul-4B orthologues, respectively).^135,136 Pop1 interacts with the N-terminal domain of Pop2 and forms heterooligomeric complexes, which bind and direct polyubiquitylation of Rum1.^137,138

The Far1 budding yeast CKI

In addition to Sic1p, a second CKI is present in budding yeast, Far1. While Sic1 specifically inhibits Clb/Cdc28 kinases,¹¹¹ Far1 inhibits Cln1,2/Cdc28 and Cln3/Cdc28 complexes during the pheromone response.^139,140 Despite the functional similarities between Far1 and p27^Kip1, these proteins share only a very small amount of amino acid identity.^43,69,141 Far1 is required to arrest the cell cycle of S. cerevisiae in response to mating factor. Far1 is phosphorylated at S87 by Cln2/Cdc28 and degraded by SCF^Cdc4 in the nucleus, through recognition of the pS87 phosphodegron by the Cdc4 WD40 repeat domain.¹⁴² In response to mating pheromone, a fraction of Far1 is stabilized after it is exported into the cytoplasm by Ste21/ Msn5, whereas blockage of nuclear export destabilizes Far1.¹⁴³

p57^Kip2

p57^Kip2, the third member of the p21^Cip1 family of CKIs, is most closely related to p27^Kip.1 p57^Kip2 is primarily expressed in terminally differentiated cells and associated with G₁ CDKs, and this can cause cell cycle arrest in the G₁ phase.¹⁴⁴ p57^Kip2, which accumulates following serum starvation, causing cell cycle arrest of osteoblastic cells, is rapidly degraded upon transforming growth factor (TGF)β1 stimulation.¹⁴⁵ TGFβ1-stimulated ubiquitylation and proteasomal degradation of p57^K1p2 does not influence the levels of p21^Cip1 and p27^Kip1 proteins, indicating that p57^Kip2 degradation in response to TGFβ1 is mediated by a distinct mechanism. One specific mechanism of p57^Kip2 degradation is mediated through TGFβ1-activated, Smad-dependent transcription of the gene for the F-box protein FBL12,^146,147 (Fig. 3). FBL12 forms an SCF^FBL12 complex that binds to and ubiquitylates mouse p57^K1p2 phosphorylated at T329 (equivalent to human p57^Kip2 T310), which is conserved between the COOH-terminal QT domains of p57^Kip2 and p27^Kip1. Inhibition of FBL12 suppresses TGFβ-induced degradation of p57^Kip2, increases the steady-state level of p57^Kip2, and promotes the differentiation of primary osteoblasts.¹⁴⁷

Figure 3.

Phosphorylated p57^Kip2 is degraded by distinct E3 ligases. (A) E3 ligases involved in p57^Kip2 degradation. p57^KIP2 phosphorylated at T329 is ubiquitylated and degraded in late G₁ and S phases by SCF^FBL12 and SCF^Skp2. (B) Schematic structure of p57^Kip2 showing the single regulatory phosphorylation site. CDI, CDK inhibitor domain.

SCF^Skp2 is another E3 ligase responsible for regulating the cellular level of p57^Kip2 by targeting it for ubiquitylation and proteolysis.¹⁴⁸ Overexpression of WT Skp2 promotes degradation of p57^Kip2, whereas expression of a dominant-negative mutant of Skp2 prolongs the half-life of p57^Kip2. p57^Kip2 interacts with Skp2, and mutation of T310 in human p57^Kip2 abrogates Skp2-induced p57^Kip2 degradation, suggesting that phosphorylation at this site is required for SCF^Skp2-mediated ubiquitylation. Similar to the role of cyclin/CDK in p27^KIP1 ubiquitylation, purified recombinant SCF^Skp2 complex ubiquitinates p57^Kip2 and this is dependent on the presence of the cyclin E/CDK2 complex. Skp2^−/− cells have abnormal accumulation of p57^Kip2,¹⁴⁸ suggesting that SCF^FBL12 cannot compensate for the deficiency of Skp2 in the ubiquitylation and degradation of p57^Kip2.

Whereas the lack of p21^Cip1 or p27^Kip1 does not show gross defects in embryonic development,¹⁴⁹ most p57^Kip2-null mice die after birth and display severe developmental defects resulting from increased apoptosis and delayed differentiation.^144,150 Most of the developmental defects apparent in tissues of the p57^Kip2 knockout mouse are corrected by replacing the p57^Kip2 gene with the p27^Kip1 gene, although the fact that a few developmental defects remain suggests that p57^Kip2 also has specific functions.¹⁵¹

Conclusion

The precise regulation of CDK activity is instrumental to cell cycle progression. Unlike the activity of many other protein kinases, which are often themselves regulated by direct ubiquitylation and degradation of the protein kinase itself,¹ CDK activity is controlled by regulation of cyclins and CKIs. The stability of p21^Cip1, p27^Kip1 and p57^Kip2 are tightly and differentially regulated by the Ub/proteasome system, in a manner that depends on many factors including the nature of extracellular stimuli, cell cycle stage, differences in subcellular context in different tissues and cells, interaction of CKIs with other regulatory proteins, such as Cks1 for p27^Kip1 and NPM for p21^Cip1, involvement of distinct E3 ligases, phosphorylation by distinct protein kinases, and a distinct subcellular compartment for degradation.

p21^Cip1, p27^Kip1 and p57^Kip2 are all targeted by more than one E3 ligase for ubiquitylation. The function of E3 ligases can be overlapping, as illustrated by p21^Cip1 degradation, where the Cul4-DDB1 and the SCF^Skp2 E3 ligases are redundant with each other in promoting the degradation of p21^Cip1 during an unperturbed S phase of the cell cycle.²⁹ However, these CKIs exhibit specifically regulated ubiquitylation by distinct E3 ligases in response to different extracellular stimuli, such as mitotic stimulation, ionizing radiation, or stress signaling. Under many conditions, phosphorylation of p21^Cip1, p27^Kip1 and p57^Kip2 is important for their degradation. Many of the regulatory phosphorylation sites lie outside the CDI region in the unique domains of these related proteins, allowing specific regulation of stability and subcellular localization by phosphorylation. Phosphorylation of p27^Kip1 at S10 or T187 will determine the localization and stages of the cell cycle where degradation of p27^Kip1 is mediated by distinct E3 ligases, whereas phosphorylation of p21^Cip1 at T57, S130 or S114 by CDK2, ERK2 or GSK3β will determine its degradation by a Ub-dependent or Ub-independent proteasomal system in cells with or without various stimuli. Given that CKI inactivation by accelerated degradation and mislocalization occurs in cancers,¹⁵² interrupting degradation of p21^Cip1, p27^Kip1 and p57^Kip2 by targeting the different regulatory steps could open new avenues for cancer therapy.

Abbreviations

CDKs

cyclin-dependent kinases

CKIs

CDK inhibitors

KIP

kinase inhibitor protein

CIP

CDK interacting protein

Ub

ubiquitin

E1

the Ub-activating enzyme

E2

Ub-conjugating enzyme

E3

Ub ligase

HECT

homologous to E6-AP C-terminus

PHD

plant homeodomain

LAP

leukemia-associated protein

SCF

Skp1/cullin/F-box

LRR

leucine-rich repeats

CRL

cullin-RING ligase

Cks1

CDK subunit 1

Pirh2

protein with RING-H2 domain

HGF

hepatocyte growth factor

UBA domain

ubiquitin-associated domain

FGF

fibroblast growth factor

MCBP

multiubiquitin chain binding protein

ERK

extracellular signal-regulated kinase

PCNA

proliferating cell nuclear antigen

APC/C

anaphase-promoting complex/cyclosome

GSK

glycogen synthase kinase

TGF

transforming growth factor

D box

destruction box

KPC

Kip1 ubiquitylation-promoting complex

RSK

p90 ribosomal protein S6 kinase