Mechanosensitive p27Kip1 Regulation and Cell Cycle Entry in Vascular Smooth Muscle Cells
Background— Cyclic stretch plays an important role in the homeostasis of vessel structure. Increased forces might, however, contribute to remodeling processes, resulting in vascular proliferative diseases. The initial molecular events necessary for mechanosensitive cell cycle entry of quiescent smooth muscle cells are poorly understood.
Methods and Results— In this study, we demonstrate that mechanical strain resulted in a rapid, integrin-dependent but mitogen-independent activation of phosphoinositide 3-kinase (PI3-K)/protein kinase B (Akt) in quiescent vascular smooth muscle cells. Subsequently, downstream ALL 1 fused gene from chromosome X (AFX)-like forkhead transcription factors were inactivated, leading to transcriptional downregulation of p27Kip1. This contrasted with the posttranscriptional protein reduction of p27Kip1 in cells stimulated with serum mitogens. Stretch-mediated p27Kip1 downregulation was accompanied by activation of cyclin-dependent kinase 2, hyperphosphorylation of retinoblastoma protein, and proliferation. Forkhead transcription factor inactivation and p27Kip1 downregulation were prevented by the PI3-K inhibitors wortmannin and LY294002. Pharmacological blockade of other kinases, such as p42/44, p38, and protein kinase A or C, did not influence the mechanosensitive gene regulation. p27Kip1 downregulation and cell cycle entry were, however, prevented by overexpression of a constitutively inactive form of Akt or constitutively active forms of forkhead transcription factors.
Conclusions— Our data demonstrate that the earliest cell cycle events can occur in a solely mechanosensitive fashion. Vascular smooth muscle cells are, furthermore, able to use transcriptional or posttranscriptional mechanisms to regulate p27Kip1, depending on the stimulus to which they are exposed. This observation has novel implications for understanding of vascular proliferative diseases.
Received November 13, 2002; revision received March 25, 2003; accepted March 27, 2003.
In the vasculature, cells are constantly exposed to alternating mechanical forces, and the cell cycle plays an enormous role in maintaining vessel structure and allowing its adaptation to acute and chronic changes. Through these remodeling processes, altered mechanical forces can contribute to pathological changes of the vessel by inducing vascular smooth muscle cell (VSMC) migration, proliferation, and hypertrophy, which have been considered key events in the development of atherosclerosis, postangioplasty restenosis, and venous bypass graft failure.1–3 Although signaling pathways responsive to mechanical force have come to light,1,4 the initial molecular events necessary for mechanosensitive cell cycle entry of quiescent VSMCs are poorly understood.
The cyclin-dependent kinase (Cdk) inhibitor p27Kip1 plays a critical role in the entry of quiescent, G0 phase cells into the cell cycle.5 Being present at maximal levels in quiescent cells, the amount of p27Kip1 decreases as cells are stimulated to enter the cell cycle, thereby allowing the activation of G1 phase Cdk/cyclin complexes with subsequent cell cycle progression.
Although the growth factor–induced decline in the amount of p27Kip1 had first been demonstrated to occur solely through posttranscriptional processes, such as increased degradation or decreased protein synthesis,6,7 it has recently become evident that downregulation of p27Kip1 can also occur at the transcriptional level. AFX-like forkhead transcription factors were found to be involved in p27Kip1 gene transactivation.8 Furthermore, ALL 1 fused gene from chromosome X (AFX)-like forkhead transcription factor phosphorylation by phosphoinositide 3-kinase (PI3-K)–activated protein kinase B (Akt) resulted in their inactivation.9 It is, however, not clear whether cells use transcriptional or posttranscriptional mechanisms of p27Kip1 downregulation differentially, depending on the stimulus that leads to cell cycle entry.
With the aim of elucidating the earliest mechanosensitive cell cycle events, we were able to demonstrate that mechanical forces rapidly activate Akt in quiescent, serum-deprived, VSMCs, leading to AFX-like forkhead transcription factor inactivation. This event resulted in the transcriptional downregulation of p27Kip1 and cell cycle entry, without a change in the rate of its protein degradation. Activation of this signaling cascade was dependent on extracellular matrix/integrin interactions and independent of acutely released or newly synthesized growth factors. Our finding contrasts with the mitogen-induced cell cycle entry of VSMCs, which predominantly depends on increased p27Kip1 protein degradation without a change in its mRNA level. Our data not only demonstrate the early mechanosensitive cell cycle regulation but also indicate that vascular cells are able to translate mechanical forces differentially into a signal toward cell cycle entry.
The following antibodies were used: rabbit polyclonal anti-pAkt (Ser 473), anti-pFKHR (Thr 247), anti-pFKHRL1 (Thr 32), anti-pan–Akt1, and anti-phistone H3 (Ser 10) (all from New England Biolabs, Frankfurt, Germany). Mouse monoclonal anti-p27Kip1, anti-retinoblastoma (RB), anti-Cdk2 (Santa Cruz Biotechnology, Santa Cruz, Calif), anti-basic fibroblast growth factor (bFGF), anti-insulinlike growth factor (IGF), and anti–platelet-derived growth factor (PDGF) (R&D Systems, GMBH, Weisbaden, Germany) were also used. Secondary antibodies included the following: goat anti-rabbit immunoglobulin G and goat anti-mouse immunoglobulin G, both of which were linked to horseradish peroxidase (Santa Cruz Biotechnology Inc), Arg-Gly-Asp (GRGDSP), and Arg-Gly-Asp control peptide (GRGESP) (both from Bachem Biochemica, Heidelberg, Germany). Wortmannin was obtained from Sigma (Diesenhofen, Germany); LY294002, from Biolmol (Plymouth Meeting, Pa); and PD98059, SB203580, KT5720, staurosporine, and the tyrphostines AG1478, AG1296, and AG1024, from Calbiochem (San Diego, Calif). 5-Bromo-2′-deoxyuridine (BrdU) and the anti-BrdU staining kit were obtained from Zymed (San Francisco, Calif).
Cell Culture, Stretch Apparatus, and Experimental Conditions
Primary cultures of VSMCs were initiated as previously descibed.10 The cells were seeded (≈10 000 cells/mL) onto 6-well, fibronectin-coated plates (FlexI plates, Flexercell). Studies were conducted on VSMCs (passage 7 to 12) after they had achieved confluence in 10% fetal bovine serum/Dulbecco’s modified Eagle’s medium/Ham’s F12 medium, followed by serum withdrawal for 2 days to achieve quiescence. On the day of the experiment, fresh, serum-free medium was substituted, and cyclic stretch was applied with a commercially available apparatus (125% resting length, 0.5 Hz; Flexercell) in a tissue-culture incubator.
Preparation of Cellular Lysates, Immunoblot Analysis, and Histone H1 Kinase Assay
Specific protein content in the cell lysates was analyzed by immunoblot, as previously described.10 In brief, the supernatant was run on a polyacrylamide gel and blotted onto nitrocellulose (Hybond-ECL, Amersham) by wet electroblotting. After being blocked, the blots were incubated with primary antibody (1:1000 dilution for anti-pAkt; 1:500 for anti-pFKHR; 1:500 for anti-pFKHRL1; 1:2000 for anti–pan-Akt1; 1:200 for anti-p27Kip1; and 1:100 for anti-RB) for 1 hour at room temperature. Specific proteins were then detected by enhanced chemiluminescence (ECL+, Amersham) after being labeled with horseradish peroxidase–labeled secondary antibody (1:2000 for 1 hour) according to the manufacturer’s instructions.
For the histone H1 kinase assay, lysates were labeled with an anti-Cdk2 antibody (1 μg/250 μg protein), and immune complexes bound to protein A/G–agarose beads (Oncogene Sciences) were assayed by addition of kinase buffer, histone H1, and [32P]ATP (3000 Ci/mmol, DuPont–New England Nuclear), as described previously.10 The samples were boiled for 5 minutes, electrophoresed through a 12% sodium dodecyl sulfate–polyacrylamide gel, dried, and exposed to x-ray film.
Transfection Procedure and Plasmids
For overexpression studies, a commercially available lipid formulation (Fugene, Roche) was used. For a 35-mm dish, 6 μL liposomes were added to 100 μL Opti-MEM (GIBCO BRL) and mixed with 2 μg DNA of either plasmid before being added to the mixture with the cells, resulting in a 20% to 30% transfection efficiency. Successfully cotransfected cells were subsequently selected by magnetic-activated cell sorting (MACS; see following section). npAkt was kindly donated by K. Walsh, Tufts University School of Medicine, Boston, Mass; pECE.FKHR-L1 and pECE.FKHR-L1.A3 by M. Greenberg, Harvard, Boston, Mass; and pBabe, AFX, and pMT2HA-AFX.A3 by B.M.T. Burgering, University Medical Center, Utrecht, the Netherlands.
Analysis of Proliferation
For flow cytometric analysis, cells were harvested by trypsinization, fixed overnight with 75% methanol, washed, and incubated with 100 μg/mL RNase (Oncogene) and 10 μg/mL propidium iodide in phosphate-buffered saline for 1 hour at 37°C. Samples were analyzed for DNA content with a high-speed cell sorter (EPICs Altra, Beckman Coulter), and data were analyzed by computer with commercially available software (Multicycle, Phoenix Flow Systems).
For additional determination of DNA synthesis, 10 mmol/L BrdU was added to the medium, and cells were stretched for 24 hour. After fixation, BrdU-positive nuclei were quantified after visualization with a monoclonal anti-BrdU antibody and subsequent diaminobenzidine staining, according to the manufacturer’s instructions (Zymed). VSMCs in mitosis were quantified after immunocytochemical staining for phosphorylated histone H3 and subsequent fluorescence microscopy (Cell Signaling).
Relative mRNA quantification was performed with the Sequence Detection System 7700 (PE Applied Biosystems) and real-time reverse transcription–polymerase chain reaction (RT-PCR). By applying comparative quantification, the target gene was normalized to an internal-standard gene, as described before.11,12 For internal calibration, mRNA transcribed from the gene encoding glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used. Amplification efficiency of GAPDH and p27KIP1 primer/probe sets were approximately equal and amounted to 1.0 (=100%).
For cDNA synthesis and real-time RT-PCR, reagents, primers, and probes were applied as described before.12 The sequences, amplicon sizes, and exon localization of primers and probes used were as follows: GAPDH (amplicon size, 121 bp): GAPDH forward, 5′-GTGATGGGTGTGAACCACGAG-3′ (exon e5), and GAPDH reverse, 5′-CCACGATGCCAAAGTTGTCA-3′ (exon e6); GAPDH-hybridization probe: 5′-CTCAAGATTGTCAGCAATGCA-TCCTGCAC-3′ (exon e5–e6); p27Kip1 (amplicon size, 75 bp): p27Kip1 forward, 5′-GCAGTGTCCAGGGATGAGGA-3′(exon e1); p27Kip1 reverse, 5′-TCTGTTCTGTTGGCCCTTTTGT-3′ (exon e2); and p27Kip1 hybridization probe, 5′-ACCTGCGGCAGAAGATT-CTTCTTCGC-3′ (exon e1–e2).
Magnetic-Activated Cell Sorting
Cells were cotransfected by using equimolar amounts of pMACS.Kk-II and an expression plasmid as indicated. After 24 hours, transduced cells were trypsinized and magnetically labeled with MACSelect Kk-II MicroBeads (Miltenyi Biotec). Transfected (Kk-II–positive) cells were then separated on an MS+/RS+ separation columns (Miltenyi Biotec) and subsequently lysed in the appropriate buffer. Positively selected cells were >85% positive for Kk-II expression.
Data are given as mean±SEM. Statistical analysis was performed by ANOVA. Post hoc analysis was performed by the method of Bonferroni. All experiments, including the immunoblots, were independently repeated at least 3 times.
Cyclic Stretch Results in Cell Cycle Entry of VSMCs
After subjecting quiescent VSMCs to cyclic stretch (24 hours, 125% resting length, 0.5 Hz), we observed a rapid downregulation of p27Kip1 within 10 to 12 hours, which reached a nadir after 24 hours (data not shown). p27Kip1 downregulation was accompanied by enzymatic activation of Cdk2 and hyperphosphorylation of the RB protein, indicating cell cycle entry (Figure 1a). Indeed, 24 hours of cyclic stretch resulted in a decline of VSMCs in the G0/G1 phase, from 93.1±1.8% to 83.4±1.4% (P<0.01), similar to the 1.9-fold increase in the number of proliferating cells, as quantified by fluorescence-activated cell sorting of propidium iodide–stained VSMCs, which indicated entry into the cell cycle (Figure 3b). In additional, BrdU incorporation had significantly increased (Figure 1b). Furthermore, cell cycle progression through mitosis was quantified by fluorescence staining for phosphorylated histone H3. The number of cells positive for this marker increased significantly within 24 hour of cyclic stretch (control, 2.6±0.8%; stretched, 8.4±1.4%; n=3, P<0.05).
Inhibition of PI3-K/Akt Stabilizes p27Kip1 and Prevents Stretch-Induced Cell Cycle Entry
Strain-induced downregulation of p27Kip1 protein levels was prevented when PI3-K/Akt signaling was inhibited with the specific PI3-K inhibitors wortmannin or LY294002. Accordingly, activation of Cdk2 was prevented, RB remained hypophosphorylated (Figure 1a), and cell cycle entry was blocked (Figures 1b and 3⇑b). Also, the stretch-induced increase in the number of cells positive for phosphorylated histone H3 was prevented (control, 2.6±0.8%; stretched, 8.4±1.4%; stretched+wortmannin, 3.2±1.0%; and stretched+LY294002, 2.8±0.8%; n=3, P<0.05). Similar data were obtained when a nonphosphorylatable, inactive form of Akt (npAkt) was transiently overexpressed (Figures 2a and 3⇑b). Conversely, we were unable to stabilize p27Kip1 through inhibition of other central signaling cascades, such as p42/44 and p38 mitogen-activated protein kinase, protein kinase A, or protein kinase C, by using their specific inhibitors PD98059, SB203580, KT5720, or staurosporine, respectively (Figure 2b).
Cyclic Strain–Activated Akt Mediates AFX-Like Forkhead Transcription Factor Phosphorylation, Resulting in p27Kip1 Transcriptional Repression
Cyclic strain of quiescent VSMCs resulted in rapid and marked phosphorylation of Akt, (reaching maximal levels at 10 to 15 minutes and decreasing thereafter), as it did during serum stimulation, although serum seemed more potent (data not shown). Forkhead transcription factors, which represent a direct downstream target of Akt,9 were found to be phosphorylated and thereby inactivated (Figure 2c). Either state was prevented by inhibition of PI3-K (wortmannin and LY294002) but not by inhibition of other signaling cascades (only the results of p42/44 and p38 inhibition by PD98059 and SB203580, respectively, are shown).
We subsequently used overexpression studies to further characterize the involvement of AFX-like forkhead transcription factors in strain-induced p27Kip1 downregulation. Overexpression of a nonphosphorylatable, constitutively active form of FKHRL1 (FKHRL1.A3) or AFX (AFX.A3), but not a control vector (GFP), prevented the stretch-induced reduction of p27Kip1 protein levels (Figure 3a), resembling those seen in quiescent VSMCs. Furthermore, overexpression of FKHRL1.A3 and AFX.A3 prevented cell cycle entry of stretched cells (Figure 3b). When we used wild-type AFX or FKHRL1 for overexpression studies, p27Kip1 levels were only restored partially, pointing toward its intact regulation during stretch.
Cyclic Strain Alters p27Kip1 Gene Transcription but Not Its Protein Degradation
Because growth factor–induced downregulation of p27Kip1 had been demonstrated to principally occur through posttranscriptional processes,6,7 we analyzed its degradation kinetics during cyclic stretch and compared it with the kinetics of quiescent cells and cells subjected to serum stimulation (Figure 4). We determined a half-life of 40±6 minutes for p27Kip1 in quiescent VSMCs, which was reduced to 17±8 minutes (P<0.05, n=3) when the cells were exposed to serum, demonstrating increased degradation. However, cyclic stretch did not change the degradation kinetics of p27Kip1 at all (half life, 38±8 minutes) compared with quiescent cells (Figure 4). Conversely, cyclic stretch resulted in a rapid downregulation of p27Kip1 mRNA levels, as quantified by real-time RT-PCR and compared with quiescent, nonstretched cells (Figure 5; P<0.001, n=3). These data demonstrate that both growth factors and mechanical strain result in p27Kip1 downregulation, although through different mechanisms.
Soluble Factors Are Not Involved in Stretch-Induced Akt/Forkhead Transcription Factor Signaling
We further sought to identify the upstream mechanisms underlying the rapid strain-induced activation of the PI3-K/Akt signaling pathway. Mechanical stress results in synthesis and release of growth factors, such as IGF, PDGF, and bFGF, from VSMCs, which might activate several signaling cascades in an autocrine or paracrine manner.13–15 We tested the involvement of these factors in mediating stretch-induced Akt activation and p27Kip1 transcriptional downregulation by adding neutralizing antibodies against IGF, PDGF, or bFGF to the media before the cell stretching experiments. The effectiveness of these neutralizing antibodies in blocking the target molecules had been determined in preceding experiments (data not shown). None of these antibodies prevented Akt activation or p27Kip1 downregulation during stretch (Figure 6a). Even applying all antibodies simultaneously did not affect this mechanosensitive signal transduction (data not shown). Furthermore, we examined the effect of conditioned medium collected from VSMCs after 15 minutes of cyclic stretch. Treatment of quiescent VSMCs with conditioned medium did not affect Akt phosphorylation (data not shown).
To exclude a possible growth factor–independent but growth factor receptor–mediated activation of the PI3-K/Akt pathway, we added specific inhibitors of growth factor receptors, the tyrphostines AG1478 (EGF), AG1296 (PDGF), and AG1024 (IGF-1) to quiescent VSMCs before stretch. Neither stretch-induced Akt phosphorylation nor p27Kip1 downregulation was prevented by these tyrphostines (Figure 6b). Our data indicate that early stretch-induced Akt phosphorylation and later p27Kip1 downregulation are not caused by a release of growth factors, growth factor receptor activation, or other soluble factors.
RGD Proteins Perturb the Ability of VSMCs to Sense Mechanical Strain
To test the involvement of extracellular matrix/integrin interactions, we examined whether the integrin-binding peptide RGD would interfere with the response of quiescent VSMCs to mechanical strain. Addition of RGD peptide to the medium before cyclic stretch dose-dependently blocked phosphorylation of Akt and prevented downregulation of p27Kip1 (Figure 7), suggesting that an intact matrix/integrin interaction is critical for mechanosensitive p27Kip1 downregulation by way of PI3-K/Akt/AFX-like forkhead transcription factor signaling. Consequently, addition of RGD peptide, but not control peptide, to quiescent VSMCs was able to prevent stretch-induced proliferation (Figure 3b).
Mechanotransduction plays a critical role in vascular homeostasis. However, in contrast to growth factor–induced cell cycle entry and progression, not much is known about its direct influence on cell cycle regulation. Although moderate cyclic stretch, as occurs under physiological conditions, seems essential for maintaining the vessel wall structure and for inhibiting growth factor–stimulated proliferation of VSMCs,16 increased stretch, more like that resembling pathological conditions as occurs in severe hypertension, in venous bypass grafts, or during balloon angioplasty, was reported to induce proliferation of VSMCs.4 Whether mechanical force directly exerts influence on the initial events of cell cycle entry and whether this might be able to trigger the development of vascular proliferative diseases were, however, unclear. Our study provides evidence that VSMCs, when exposed to an enhanced mechanical force, use extracellular matrix/integrin interactions to activate Akt, which,, in turn,, phosphorylates and inactivates AFX-like forkhead transcription factors, leading to transcriptional repression of the G0 phase gatekeeper, p27Kip1. This chain of events is growth factor independent and initiates cell cycle entry and progression, as indicated by Cdk2 activation, RB hyperphosphorylation, and subsequent increased DNA synthesis and mitosis.
p27Kip1 has primarily been demonstrated to be regulated by way of posttranscriptional processes (inhibition of synthesis or ubiquination and subsequent proteasome-dependent degradation) when growth factors are used to initiate cell cycle entry.6,7 Indeed, serum-induced downregulation of p27Kip1 in quiescent VSMCs resulted mainly from its protein degradation and furthermore, was independent of PI3-K/Akt signaling, because the inhibitors wortmannin or LY294002 did not result in stabilization of p27Kip1 protein levels, a finding previously reported by us.17 Conversely, extracellular matrix/integrin interactions, initiated by mechanical strain in the absence of growth factors, nearly exclusively resulted in the transcriptional repression p27Kip1 due to inhibition of AFX-like forkhead transcription factor activity. Forkhead transcription factors have already been functionally implicated in the regulation of metabolism, cell proliferation, and apoptosis.9 Recently, it has been demonstrated that they are directly phosphorylated and inactivated by Akt,8 identifying them as additional targets of PI3-K/Akt signaling.
In our study, extracellular matrix/integrin interaction resulted in forkhead transcription factor–dependent transcriptional downregulation of p27Kip1. The kinetic of the mRNA decrease, as well as the experiments with conditioned medium and neutralizing antibodies, implies independence from newly synthesized growth factors. Using tyrphostines, we were also able to rule out an involvement of growth factor receptors, which could have been activated independently of growth factors themselves.
Although we were able to dissect p27Kip1 downregulation in vitro, it seems likely that vascular cells in vivo use both translational and posttranslational mechanisms of p27Kip1 regulation during arterial remodeling. Simultaneously, a highly mitogenic milieu is locally created by pathological trauma to the vessel wall, as occurs in vascular proliferative diseases.13,15 Synergy between mitogenic stimulation and mechanical force has been suggested before.15,18,19 Our data, however, imply that the early steps of stretch-induced cell cycle entry might be triggered by mechanotransduction rather than mitogenic activation.
Protein kinase C and MAPKs have been shown to be activated in force-induced VSMC proliferation (for a review, please see Li and Xu4). Although we also determined rapid activation of extracellular signal–regulated kinase 1/2 when VSMCs were stretched in vitro (data not shown), its inhibition as well as that of p38 MAPK, protein kinase A, or protein kinase C did not result in stabilization of p27Kip1, thus excluding involvement of these signaling pathways in its force-induced early downregulation.
Further in vivo studies will examine the role of mechanosensitive p27Kip1 downregulation in vascular remodeling processes and the development of vascular proliferative diseases. Recent in vivo studies have demonstrated that apolipoprotein E/p27Kip1 double-knockout mice develop enhanced proliferation of VSMCs and accelerated atherosclerosis. Interestingly, a very moderate decrease of p27Kip1 expression levels (p27−/+ mice) was sufficient to predispose to atherosclerosis.20 Also, gene transfer of p27Kip1 was able to inhibit neointima formation in the rat carotid artery.21 Controversially, however, p27Kip1-deficient mice did not show an altered arterial wall proliferative response to femoral artery transluminal injury.22
Mechanosensitive regulation of p27Kip1 through PI3-K/Akt and AFX-like forkhead transcription factors might play an important role in physiological vascular remodeling processes and the pathophysiology of vascular proliferative diseases. Our observations have novel implications for the understanding of vascular proliferative disease processes.
Ruediger C. Braun-Dullaeus is supported by the Deutsche Forschungsgemeinschaft (Sonderforschungsbereich 547/A7 and 547/C1). Daniel Sedding is a scholar of the Deutsche Forschungsgemeinschaft (Graduiertenkolleg 534).
Sherr CJ, Roberts JM. CDK inhibitors: positive and negative regulators of G1-phase progression. Genes Dev. 1999; 13: 1501–1512.
Hengst L, Reed SI. Translational control of p27Kip1 accumulation during the cell cycle. Science. 1996; 271: 1861–1864.
Pagano M, Tam SW, Theodoras AM, et al. Role of the ubiquitin-proteasome pathway in regulating abundance of the cyclin-dependent kinase inhibitor p27. Science. 1995; 269: 682–685.
Cheng GC, Libby P, Grodzinsky AJ, et al. Induction of DNA synthesis by a single transient mechanical stimulus of human vascular smooth muscle cells: role of fibroblast growth factor-2. Circulation. 1996; 93: 99–105.
Chapman GB, Durante W, Hellums JD, et al. Physiological cyclic stretch causes cell cycle arrest in cultured vascular smooth muscle cells. Am J Physiol Heart Circ Physiol. 2000; 278: H748–H754.
Braun-Dullaeus RC, Mann MJ, Seay U, et al. Cell cycle protein expression in vascular smooth muscle cells in vitro and in vivo is regulated through phosphatidylinositol 3-kinase and mammalian target of rapamycin. Arterioscler Thromb Vasc Biol. 2001; 21: 1152–1158.
Fang F, Orend G, Watanabe N, et al. Dependence of cyclin E-CDK2 kinase activity on cell anchorage. Science. 1996; 271: 499–502.
Diez-Juan A, Andres V. The growth suppressor p27(Kip1) protects against diet-induced atherosclerosis. FASEB J. 2001; 15: 1989–1995.