In budding and fission yeast, the highly regulated action of a single kinase subunit (cdc28 or cdc2, respectively) drives the cell cycle forward.47 In higher eukaryotes, cell-cycle control is more complex, and several proteins homologous to cdc2 (termed the cyclin-dependent kinases or cdks) have been identified.48 Cyclin-dependent kinases are protein kinases that vary in size between 30 and 40 kb and share greater than 40 percent sequence identity. In addition to amino acid homology, cdks share many functional and regulatory features with yeast cdc2/28.49,50 Almost all cdks require association with protein subunits called cyclins to become active kinases. Cdks also contain conserved amino acid residues that modulate kinase activity when phosphorylated or dephosphorylated. Additionally, specific regulatory molecules that bind and inhibit cdk subunits inhibit cdk activity. Each of these regulatory mechanisms is discussed in detail below. Remarkably, while differences among organisms do exist, this multi-tiered regulatory system has been conserved from yeast to humans.
Each phase of the cell cycle is characterized by a unique pattern of cdk activity (Fig. 23-6).51–53 In mammalian cells, eight cdks have been identified, and most are active (and required) only in specific phases of the cell cycle. Progression through G1 phase depends upon the activities of cdk2, cdk3, cdk4, and cdk6. The recently described cdk8 protein also may function primarily in G1, and may be involved in transcriptional regulation. Cdk2 and cdc2 are active in S-phase, and cdc2 kinase activity also governs mitotic entry and exit. In distinction to its kindred, cdk5 does not appear to be intimately involved in cell-cycle progression, but may, instead, play a role in the developing nervous system, where it associates with the non-cyclin activator p35.54,55
Patterns of cyclin-cdk activity during the mammalian cell cycle. The expression patterns of the key mammalian cyclins is superimposed upon the cell cycle, along with their respective cdk partners. The approximate position of the R POINT is shown (adapted from Sherr51).
Cyclins: Activating Subunits of CDK Enzymes
Monomeric cdk subunits are essentially devoid of enzymatic activity, and kinase activation requires the association of cdks with cyclins.49 An active cdk is thus a heterodimeric enzyme consisting of regulatory (cyclin) and enzymatic (cdk) subunits. The cyclins are a group of related proteins that contain a conserved region of homology (the cyclin box) and are usually expressed in a cell cycle specific fashion. Cyclin expression is rate-limiting for cdk activation, and control of cyclin expression is a fundamental mechanism underlying cdk periodicity. In general, cyclin levels are determined by both transcriptional control and regulated proteolysis by the ubiquitin-proteosome system.
The recently solved crystal structure of cyclin A bound to cdk2 reveals that cyclins activate cdks in at least two ways. Cyclin binding induces conformational changes in the cdk that first reorients the configuration of the ATP phosphate groups to facilitate phosphotransfer to protein substrates, and second moves the T-loop of the cdk out of a position that would otherwise block entry of protein substrates into the active site (Fig. 23-7).56,57
Crystal structure of cdk2, cyclin A-cdk2, and cyclin A-cdk2-p27 complexes. A, The structure of monomeric cdk2. The T-loop is indicated in yellow, and the PSTAIRE motif in red. An ATP molecule in indicated within the active site. B, The structure of cdk2 bound to an amino-terminal truncated version of cyclin A. Cyclin binding re-orients the PSTAIRE helix and moves the T-loop, resulting in the re-positioning of ATP-phosphate groups within the complex and allowing substrate accessibility to the active site. C, The structure of cdk-activating kinase (CAK)-phosphorylated cyclin A bound to cdk2. The yellow ball indicates the position of thr160 within the T-loop. D, The structure of a ternary complex of the amino terminus of p27 bound to cyclin A -cdk2. Separate domains of P27 interact with both cyclin A and cdk2. The structure of this complex reveals that p27 inhibits cdk activity by distorting the structure of the active site, and by binding within the catalytic cleft and preventing ATP binding.
The specificity of cdk action at different times in the cell cycle is in large part determined by its particular cyclin subunit. This functional diversity of cyclins was first established in the yeast S. cerevisiae where specific cyclins have been identified that are required for G1, S phase, and mitosis. Budding yeast express three functionally redundant G1 cyclins (cln1, cln2, and cln3), which are required for passage through START.58–62 While differences between the cln genes have been described, mutant yeast cells with mutated cln alleles can still enter S phase as long as one of these three genes remains functional. Transcription of cln1 and cln2 is controlled by the Swi4/Swi6 transcription factor, 63–65 and cln activity positively reinforces further cln expression.66–68 Thus, cln1 and cln2 mRNAs rise during G1, reaching peak levels around START.
Once START has been traversed, cln activity is no longer required for subsequent cell-cycle progression. Instead, other cyclins associate with and activate cdc28 during other phases of the cell cycle. Complexes containing cdc28 and the cyclins clb5 and clb6 are required for S phase, 69,70 and the four B-type cyclins, clb1 to clb4 are required for mitosis.71
START marks a point of transition in the yeast cell cycle where G1 cyclin expression ends and mitotic cyclin expression begins. This transition comes about because these two classes of cyclin modulate each other's expression. Cln/cdc28 kinase activity directly increases the expression of the clb genes, and conversely clb-cdc28 kinase activity represses cln expression.72 Not only do G1 cyclins promote the expression of the genes for mitotic cyclins but, as described below, they also increase the stability and functional activity of clb proteins. Furthermore, the cln proteins themselves are rapidly degraded after START. This is discussed below in the section on cell-cycle regulated proteolysis. Together, these controls insure ordered progression through the cell cycle by establishing alternating periods of the cell cycle where either G1 or mitotic cyclins are expressed and functionally active.
Mammalian cyclins C, D1, and E were first identified in a screen for mammalian genes that could complement yeast cyclin mutations.73–75 At the same time Cyclin D1 was identified by two other approaches—as a mitogen-responsive gene in a macrophage cell line76 and as a gene located at a chromosome inversion breakpoint in a parathyroid tumor (and in this guise was originally named PRAD1).77 A dozen mammalian cyclin genes have been identified that are both structurally and functionally homologous to yeast cyclins.51,52 Like the yeast cyclins, many of these molecules exhibit cell-cycle-dependent periodicity in their expression and activity (Fig. 23-6).
The primary mammalian G1 cyclins are the D-type cyclins and cyclin E. These cyclins associate with the cdk-4/6, and cdk2 subunits, respectively. There are three D-type cyclins (D1, D2, and D3), which are expressed in a cell-type specific fashion.51,52,78 The G1 role of the D-type cyclins is revealed by their pattern of expression and by their functional properties. Cyclin D expression begins in early G1 when quiescent cells are stimulated to reenter the cell cycle, and cyclin D expression remains at high levels as long as mitogens are present. In other words, the expression of these labile proteins (t½<20 min) is not intrinsically periodic, but instead depends upon the presence of cell-type specific mitogens. Inhibition of cyclin D1 function blocks the cell cycle in G1, demonstrating the necessity of cyclin D for the cell cycle.79,80 Also, enforced overexpression of cyclin D1 shortens the G1 phase of the cell cycle, and partially diminishes the mitogen requirement for cell proliferation, demonstrating that cyclin D1 levels are limiting for G1 progression.80,81
Cyclin E activity is also required in G1, although probably somewhat after cyclin D activity.81–83 Cyclin E protein expression peaks at the G1-S boundary, and then decays as S-phase progresses.82,84,85 Determinants of cyclin E periodicity include both transcriptional control by E2F and regulated proteolysis (see below). Overexpression of cyclin E results in G1 contraction and decreased mitogen requirements, 81,86 and cyclin E kinase activity is required for S-phase entry.82,83 Activation of cyclin D and cyclin E-associated kinases may biochemically constitute the restriction point, the mammalian equivalent of START control in yeast.
Later cell-cycle transitions are governed by the cyclin A and cyclin B proteins. Cyclin A associates with both the cdk2 and cdc2 subunits, and cyclin A kinase activity is required at the start of S phase and at the G2-M transition.87–89 Cyclin B associates with cdc2, and, like the yeast clb proteins, cyclin B-cdc2 kinase activity regulates both mitotic entry and exit.90
The cyclin H protein associates with cdk7, and this heterodimer constitutes the cdk-activating kinase (CAK).91 CAK is also a component of the human transcription factor TFIIH, and is capable of phosphorylating the carboxy-terminal domain (CTD) of RNA polymerase II. Cyclin C is classed as a G1 cyclin, although its role is not yet defined.73 Cyclin C has recently been shown to associate with cdk8, and cyclin C-cdk8 complexes also have RNA polymerase II CTD kinase activity, although they do not co-purify with TFIIH.92 Phosphorylation of the CTD by cyclin-cdk complexes may couple cell-cycle events to the cellular transcriptional machinery.
Comparatively little is known about the remaining cyclins that have been identified to date, including cyclins F, G and I. Cyclin F is the largest cyclin, with a molecular weight of 87, and is most closely related to cyclins A and B. Cyclin F mRNA peaks in G2 and cyclin F protein accumulates in interphase and is destroyed during mitosis.93 Cyclin G mRNA does not fluctuate in a cell cycle-dependent fashion, but is induced by both the p53 protein and growth stimulation of quiescent cells.94 The cyclin I protein is expressed most highly in post-mitotic tissues, including muscle and neurons, and may have a unique regulatory role.95
CDK Regulation by Phosphorylation and Dephosphorylation
In addition to cyclin binding, phosphorylation and dephosphorylation of conserved cdk residues provides another important level of control over kinase activity.49 Cdks can be either activated or inactivated by phosphorylation (Fig. 23-8). The site of activating phosphorylation is a conserved threonine residue in the so-called “T-loop” (e.g., threonine 161 in cdc2, threonine 160 in cdk2).56 The binding of cyclin to the cdk, and phosphorylation of this residues together move the T-loop away from the catalytic cleft of the enzyme, thereby providing access to protein substrates.96 Thr160 is phosphorylated by CAK (cyclin H-cdk7), and this phosphorylation is required for cdk activation.97–100 CAK activity, however, is neither cell-cycle regulated nor limiting, and the major determinant of Thr160 phosphorylation is probably cyclin binding.91
CDK regulation by cyclin binding and cdk phosphorylation. As described in the text, activation of cdks requires cyclin binding and cdk phosphorylation at thr160 by the cdk-activating kinase (CAK). Subsequent phosphorylation of Tyr15 by the wee1 and mik1 kinases and dephosphorylation by cdc25 phosphatases further regulates kinase activity.
Cdks can also be phosphorylated on a specific amino-terminal tyrosine residue (e.g., tyrosine 15 in cdc2 and cdk2). Tyrosine phosphorylated cdk2 is catalytically inactive, even if it is phosphorylated on the activating threonine within the T-loop.98,100–103 The kinases that phosphorylate tyr15 are evolutionarily conserved and are known as the wee1 and mik1 kinases.104–107 Conversely, dephosphorylation of tyrosine by the cdc25 phosphatase activates the cdk.108–112 Regulation of wee1 and cdc25 is complex, 113,114 but the bottom-line is that the relative activities of these enzymes set a threshold for cdk activation and determine mitotic entry. Three mammalian cdc25 homologues have been identified (cdc25a, cdc25b, and cdc25c), and each may have a unique cell-cycle role.115,116 Cdc25a is active in G1 and may be induced by raf-dependent pathways.117
CKIs: Inhibitory Subunits of CDK Enzymes
All organisms express proteins that directly bind to and inhibit cdk activity.118,119 These cdk-inhibitors (CKIs) provide another important strategy by which cdk activity is regulated in response to diverse stimuli. In budding yeast, two kinds of inhibitors have been described. One type is inducible and links the cell cycle to extracellular signals; the other type is an intrinsic component of the mitotic cycle. The best example of the first type of CKI is the FAR1 protein. Mating pheromones induce FAR1, a protein that binds to and inhibits the cln-cdc28 kinase and thereby causes the yeast cell cycle to arrest at START. Another important CDK inhibitor in budding yeast is Sic1, but it is a constitutive element in the mitotic clock and is not known to be induced by extrinsic proliferative signals.120,121 At the conclusion of each mitosis, Sic1 protein levels rise, inhibiting the clb-cdc28 kinases and facilitating the transition from anaphase to the next G1. Sic1 protein remains at high levels during G1 until activation of the cln-cdc28 kinases at START induce its degradation. This is one mechanism that links activation of the S-phase clb-cdc28 kinases to passage through START.
Mammalian cells express two classes of CKIs that are distinguished by their cdk targets: the Cip/Kip family of CKI's are universal cdk inhibitors, whereas the INK4 proteins are specific cdk4/6 inhibitors (Fig. 23-9).118 The Cip/Kip family consists of three members: p21, p27, and p57. Overexpression of these molecules causes a G1 arrest in cultured cells, and they are able to inhibit most cyclin-cdk complexes in vitro. These molecules bind to assembled cyclin-cdk complexes much more avidly than to monomeric cdk or cyclin subunits. p21 was first identified as a component of cyclin-CDK complexes in proliferating cells123 and as a protein induced as cells in vitro became senescent.124 Shortly thereafter, it was cloned by a number of independent approaches.125–128 The p21 protein contains two functional domains, an amino terminal cdk interaction region that is sufficient for cdk inhibition, and a carboxy-terminal region that binds PCNA, a processivity factor associated with DNA polymerase delta.129–131
Mammalian CDK-inhibitors are classed by their cyclin-cdk targets. The Cip/Kip proteins (p21, p27, p57) are universal cdk inhibitors that inactivate all cyclin-cdk complexes (with the possible exception of cyclin B-cdc2). In contrast, the INK4 proteins (p15, p16, p18, p19) specifically bind and inhibit only cdk4/6, and cyclin D-cdk4/6 complexes.
Two biological roles have been suggested for p21.118,122 The first is in contributing to the cell-cycle arrest that occurs in cells with damaged DNA.127 This is discussed in greater detail below. Additionally, p21 has been suggested as a facilitator of withdrawal from the cell cycle in cells undergoing terminal differentiation.132,133
The CKI p27Kip1 is structurally related to p21Cip1.134,135 p21 and p27 share significant amino terminal homology within the cdk inhibitory domain, but p27 does not contain the PCNA interaction region. p27 is not a p53 response gene, but p27 levels respond instead to a variety of extracellular mitogenic and antimitogenic signals.118 In general, p27 levels are high in nondividing cells and low in proliferating cells. The regulation of p27 is complex, with transcriptional, translational and post-translational mechanisms all implicated in different biological contexts.136
The mechanism of cdk inhibition by p27 has been clarified by the crystal structure of p27 bound to cyclin A-cdk2 (Fig. 23-7).96 In the ternary p27-cyclin A-cdk2 complex, separate domains in p27 interacts with the cyclin and the cdk. Although p27 does not significantly alter the structure of the cyclin, it may bind to a site on the cyclin that the cyclin would ordinarily use for interactions with protein substrates. In this way, p27 might inhibit phosphorylation of physiologically important substrates without inhibiting the catalytic activity of the cdk enzyme. In addition, p27 does have dramatic effects on cdk structure. p27 disrupts the structure of the N-terminal lobe of the cdk, widening and distorting the ATP-binding site. In fact, p27 itself inserts into the catalytic cleft and directly interacts with the amino acids that would bind ATP. This would completely prevent cdk-binding of ATP and completely inhibit catalytic activity.
Less is known about p57, the most recently isolated family member that was cloned by virtue of its homology with p27.137,138 Both the amino and carboxy-terminal domains of p57 are related to p27. Compared with p27, however, p57 expression is relatively restricted to terminally differentiated tissues.
The INK4 family of CKIs includes four structurally related proteins (p15, p16, p18, p19), each of which contains four ankyrin repeats.118 The first member of this family to be identified, p16, was found to be associated with cdk4 in transformed cells, 138 and subsequently fingered as a candidate tumor suppressor in familial melanoma.139,140 INK4 proteins bind to monomeric cdk4/6 subunits, preventing their association with D-type cyclins, and INK4 proteins can also inhibit the activity of cyclin D-cdk4/6 complexes. The other INK4 proteins are expressed ubiquitously in mouse tissues and cultured cells, and the expression of p19 does oscillate with the cell cycle.141,142 While p15 is involved in the anti-proliferative response to TGF-B, the physiological roles of the INK4 protein remain unknown. The frequent deletions of p15 and p16 in primary tumors and the high spontaneous tumor rate in p16-deficient mice indicate that these proteins play a critical role in maintaining normal growth control.143