Cdk2-dependent phosphorylation of p27 facilitates its Myc-induced release from cyclin E/cdk2 complexes

Cdk2-dependent phosphorylation of p27 facilitates its Myc-induced releasefrom cyclin E/cdk2 complexes

Daniel MuÈ ller1, Caroline Bouchard1, Bettina Rudolph1,4, Philipp Steiner1,5, Ingo Stuckmann2,Rainer Sa�rich3, Wilhelm Ansorge3, Wieland Huttner2 and Martin Eilers1,6

1Zentrum fuÈr Molekulare Biologie Heidelberg (ZMBH), Im Neuenheimer Feld 282, 69120 Heidelberg, 2Department ofNeurobiology, University of Heidelberg, Im Neuenheimer Feld 364, 69120 Heidelberg, Germany, and 3Biochemical InstrumentationProgramme, EMBL, Meyerhofstr.1, 69117 Heidelberg, Germany

Activation of Myc triggers a rapid induction of cyclin E/cdk2 kinase activity and degradation of p27. Overtdegradation of p27 is preceded by a speci®c dissociationof p27 from cyclin E/cdk2, but not from cyclin D/cdk4complexes. We now show that cyclin E/cdk2 phosphor-ylates p27 at a carboxy-terminal threonine residue(T187) in vitro; mutation of this residue to valinestabilises cyclin E/cdk2 complexes. This reaction is notsigni®cantly inhibited by high concentrations of p27,suggesting that cdk2 bound to p27 is catalytically active.In vivo, p27 bound to cyclins E and A, but not to D-typecyclins is phosphorylated. Myc-induced release of p27from cdk2 requires cdk2 kinase activity and is delayed ina T187V mutant of p27. After induction of Myc, p27phosphorylated at threonine 187 transiently accumulatesin a non cdk2 bound form. Our data suggest amechanism in which p27 is released from cyclin E/cdk2upon phosphorylation; in Myc-transformed cells, releaseis e�cient as phosphorylated p27 is transiently bound ina non-cdk2 containing complex and subsequentlydegraded.

Keywords: Myc; p27; cyclin E; cdk2

Introduction

The proto-oncogene c-myc encodes a helix ± loop ±helix/leucine zipper protein (Myc) that activates genesas part of a heterodimeric complex with a partnerprotein termed Max (for recent reviews, see Bernards,1995; Henriksson and LuÈ scher, 1996). Myc/Maxcomplexes bind to speci®c DNA sequences with acentral CAC(G/A)TG motif (Blackwell et al., 1993;Prendergast and Zi�, 1991). They activate transcriptiondue to potent transactivation domains localized in theamino-terminus of Myc (Desbarats et al., 1996; Kato etal., 1990). Several genes that are regulated by Myc/Max complexes in vivo have been isolated (Bello-Fernandez et al., 1993; Eilers et al., 1991; Galaktionovet al., 1996; Grandori et al., 1996; Miltenberger et al.,1995). In quiescent cells, in which c-myc is usually not

expressed, Max homodimers accumulate, which bind tothe same sequence yet fail to activate transcription(Kao et al., 1992). Other proteins, termed Mad, Mxiand Mnt, have been identi®ed that bind to Max, butnot Myc (Ayer et al., 1993; Hurlin et al., 1995, 1997;Zervos et al., 1993). They often accumulate duringcellular di�erentiation (Ayer and Eisenman, 1993).Mad/Max and Mxi/Max recognize the same DNAelement as Myc/Max complexes, but act as transcrip-tional repressors via recruitment of a co-repressorprotein termed sin3 (e.g. Ayer et al., 1995; Laherty etal., 1997; Rao et al., 1996; Schreiber Agus et al., 1995;Sommer et al., 1997). Ectopic expression of c-myc alsoinduces loss of expression of a number of genes andthis may not be mediated via complex formation withMax (Peukert et al., 1997; Philipp et al., 1994; Roy etal., 1993; Shrivastava et al., 1993).c-myc is a central regulator of both proliferation and

apoptosis in mammalian and avian cells. For example,constitutive expression of Myc or activation ofconditional alleles (MycER) in culture disrupts thegrowth factor dependence of cell cycle progressionobserved in normal cells (Eilers et al., 1991; Evan et al.,1992; Keath et al., 1984). Transgenic animals, in whichc-myc is expressed under the control of an IgH-enhancer, display a distinct preneoplastic phenotypewhich is characterized by an enhanced rate ofproliferation and an increase in the number of preBcells (Langdon et al., 1986). Ectopic expression ofMad-1 protein blocks CSF-1 induced mitogenesis,demonstrating an essential role for Myc proteins inG1 progression (Roussel et al., 1996); this is supportedby experiments using antisense oligonucleotides toinhibit expression of Myc (Heikkila et al., 1987;Prochownik et al., 1988). Also, knock-out animals inwhich the c-myc gene has been disrupted, die early inembryogenesis (Davis et al., 1993).A number of observations show that Myc exerts its

e�ect on cellular proliferation at least in part throughactivation of cyclin E/cdk2 kinase activity. First,induction of conditional alleles of Myc in RAT1-MycER cells leads to a rapid induction of cyclin E/cdk2 kinase activity, which precedes any overt changein cellular proliferation (Steiner et al., 1995). Mycupregulates cyclin E/cdk2 kinase activity not onlyduring the G0/G1 transition, but also disrupts the cellcycle dependent regulation of cyclin E/cdk2 kinaseactivity normally observed in exponentially proliferat-ing cells (Pusch et al., 1997). Thus, activation of Mycleads to strongly elevated levels of cyclin E/cdk2activity very early in the G1 phase of the cell cycle.Second, dominant negative alleles of Myc inhibit cyclin

Correspondence: M EilersPresent addresses: 4 ICRF, P.O. Box 123, 44 Lincoln's Inn Fields,London WC2A 3PX, UK; 5Whitehead Institute for BiomedicalResearch, 9 Cambridge Centre, Cambridge, MA 02142, USA;6 Institute for Molecular Biology and Tumor Research (IMT);University of Marburg, 35033 Marburg, GermanyReceived 26 May 1997; revised 18 July 1997; accepted 21 July 1997

Oncogene (1997) 15, 2561 ± 2576 1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

E/cdk2 kinase activity in exponentially growing cells(Berns et al., 1997). Third, activation of cyclin E/cdk2kinase is required for mitogenic responses to activatedMyc and is at least in part su�cient to account for

them: for example, activation of cyclin A geneexpression by Myc is mediated by cyclin E/cdk2kinase (Rudolph et al., 1996). Fourth, expressing ofcdc25A, a phosphatase thought to be involved in

– hu

man

cE

– hu

man

cE

Control Ab

α-human cyclin E

Myc: – – + +

— histone H1

con

tro

l Ab

α-h

um

an c

E

con

tro

l Ab

α-h

um

an c

E

Myc: – +

human cyclin E

α-human cyclin E

α-human cyclin E

a

b

human cyclin E human

cyclin E

IgG

inp

ut

con

tro

l (1)

con

tro

l (2)

α-p

27 (

po

lycl

.)

α-h

um

an c

yclin

E (

mo

no

cl.)

con

tro

l (m

on

ocl

.)

cycl

in E

p27

(1)

p27

(2)

pre

imm

un

e

cycl

in E

p27

(1)

p27

(2)

pre

imm

un

e

– 4-OHT + 4-OHT

c

human cyclin E

histone H1

Figure 1 Induction of Myc activates the kinase activity associated with constitutively expressed human cyclin E. (a) Left: Westernblot documenting expression of human cyclin E in RAT1-MycER (hcE) cells. Right: Human cyclin E protein levels and associatedkinase activity in con¯uent RAT1-MycER (hcE) cells before (7) and 12 h after (+) activation of Myc. (b) Release from p27 isrequired for activation of human cyclin E/cdk2 by Myc. Upper panel: Anti-human cyclin E Western blot documenting the speci®cassociation of human cyclin E with p27 in resting RAT1-MycER(hcE) cells). Lower panel: Immune complex kinase assays before(`74-OHT') or after activation of Myc (`+4-OHT') with either anti-human cyclin E antibodies or two di�erent anti-p27 antibodiesas indicated. (c) Human cyclin E/cdk2 complexes isolated from Myc-induced cells can be activated by treatment with cdc25A invitro. Anti-human cyclin E immunepricipates from either uninduced RAT1-MycER(hcE) cells (7) or from cells incubated for 12 hwith 200 nM 4-OHT (+) were treated with recombinant cdc25A in vitro as described previously (Steiner et al., 1995). Complexeswere reisolated and histone H1 kinase activity determined. The panel shows the speci®c histone H1 activity in untreated samples(`basal') and the additional kinase activity released by treatment with cdc25A (`latent'). Cells labelled 4-OHT/PSI were incubated for12 h with 4-OHT in the presence of the proteasome inhibitor, PSI (see Figure 2)

Phosphorylation of p27D MuÈller et al

2562

activation of cyclin E/cdk2, is upregulated by Myc viaan E-box element in the second intron of the gene

(Galaktionov et al., 1996) (but see: Perez-Roger et al.,1997).

a b

c

d

con

tro

l

4-O

HT

4-O

HT

/PS

I

4-O

HT

/ICE

-in

h

p27

cyclin E

cdk2

con

tro

l

4-O

HT

4-O

HT

/PS

I

p27

human cyclin E

cdk2

histone H1

con

tro

l

4-O

HT

/PS

I

4-O

HT

con

tro

l

4-O

HT

/PS

I

4-O

HT

IP: α hcE WB: α hcE

IP: α cdk2 WB: α hcE

4-OHT – + – +

cyclin A protein

IP: α p27 WB: α cdk2

IP: α p27 WB: α hcE

p27 protein

p27 associated with cyclin A

PSI – – + +

Figure 2 Inhibition of proteasome-mediated degradation of p27 does not abolish Myc-induced activation of cyclin E/cdk2complexes. (a) Western blot documenting the total amount of p27, cdk2, and cyclin E present in lysates of RAT1-MycER cellsbefore and 12 h after addition of 4-OHT, either by itself or together with PSI or ICE inhibitor I. (b) Histone H1 kinase assays ofcyclin E immune-precipitates from the same cells. The panel shows the quantitation of triplicate assays. (c) PSI does not blockactivation of human cyclin E/cdk2 kinase by Myc. Left: Western blot and human cyclin E-associated histone H1 kinase assays fromRAT1-MycER(hcE) cells under the same conditions. Right: Western blots documenting partial dissociation of human cyclin E/cdk2complexes from p27 in the presence of proteasome inhibitors. Note that the uppermost band seen in anti-human cyclin E immune-precipitates is the heavy chain of the precipitating antibody. (d) Western blots of total cyclin A, total p27 and p27 contained incyclin A immune-precipitates documenting loss of p27 from cyclin A/cdk2 complexes in RAT1-MycER cells 12 h after activation ofMyc, either in the presence or the absence of PSI


2563

In resting RAT1-MycER ®broblasts, cyclin E/cdk2kinase is inactive despite the presence of considerableamounts of cyclin E, cdk2 and cdc25A proteins and ofassembled cyclin E/cdk2 complexes (Perez-Roger et al.,1997; Steiner et al., 1995). In a pulse-chase experiment,activation of Myc induced a shift of pre-formed high

molecular cyclin E/cdk2 complexes to a lowermolecular weight, suggesting that induction of Mycmight lead to the activation of pre-existing cdk2complexes. Recent experiments demonstrated thatMyc antagonises growth suppression by low amountsof p27, suggesting that it may direct the synthesis of

IP: co E co E co E

time (hrs)

a bIP: co p27 co p27 co p27

IP: co p27 co p27 co p27

IP: co ck2 co ck2 co ck2

D3

D1

p27

time (hrs) 0 4 9

0 4 9

input

0 4 9

cdk2 —

cyclin A —

cyclin E —

p27 —

cdk2/p27 —

cyclin A/p27 —

cyclin E/p27 —

time after addition of 4-OHT (hours)c

0 4 6 8 10 12 24

Figure 3 Loss of p27 from cyclin E/cdk2 complexes precedes degradation of total p27. (a) Histone-H1 kinase assays documentingactivation of cyclin E/cdk2 kinase activity in RAT1-MycER cells after activation of Myc. As control, antibodies directed againsthuman Mad proteins were used. (b) Western blots documenting the amount of cyclins D1, D3, of cdk2 and of p27/cyclin D1, p27/cyclin D3 and p27/cdk2 complexes isolated from RAT1-MycER cells before and after the indicated times of addition of 4-OHT. (c)Western blots documenting the amount of cyclins E and A, cdk2, p27, and of cyclin A/p27, cyclin E/p27 and cdk2/p27 complexes atdi�erent time points after addition of 4-OHT


2564

proteins that can bind to and sequester p27 (Vlach etal., 1996). We now show that activation of cyclin E/cdk2 kinase by Myc in resting RAT1 ®broblasts occursby a similar mechanism. Further, we show that p27 is®rst selectively lost from cdk2 complexes, but not fromcyclin D/cdk4 complexes after activation of Myc andshow that this speci®city is due to phosphorylation ofp27 by cdk2 at a carboxy-terminal threonine residue.Finally, we show that binding of p27 does not abolishthe catalytic activity of cdk2 towards p27; takentogether the data suggest a simple model as to howthe speci®c activation of cyclin E/cdk2 kinase activityupon activation by Myc may be achieved.

Results

Previous work had shown that activation of Myctriggers a rapid induction of cyclin E/cdk2 kinaseactivity in resting RAT1-MycER ®broblasts; thisactivation occurred despite the presence of consider-able amounts of cyclin E protein and of assembledcyclin E/cdk2 complexes in arrested RAT1-MycER®broblasts (Steiner et al., 1995). To facilitate theanalysis of this process, we generated a cell line thatexpressed human cyclin E constitutively from aretroviral promoter (RAT1-MycER(hcE); Figure 1a).Western blotting with monoclonal antibodies directedagainst human cyclin E detected the protein speci®callyin transfected cells, but not in lysates from control cellsand no crossreaction with rodent cyclin E was observed(Figure 1a). To test whether induction of Myc a�ectedthe kinase activity associated with human cyclin E,cells were grown to con¯uence and arrested for severaldays. Subsequently, MycER chimeras were activatedby addition of 200 nM hydroxy-tamoxifen (4-OHT)and lysates prepared twelve hours later. Addition of 4-OHT did not enhance expression levels of humancyclin E protein (Figure 1a). Indeed, a moderatedecrease in the amount of human cyclin E proteinrelative to control cells was observed, which was due todegradation of cyclin E (see Figure 2) (Clurman et al.,1996; Won and Reed, 1996). No kinase activityassociated with human cyclin E could be detected inresting cells, whereas stimulated cells showed a highcyclin E speci®c kinase activity (Figure 1a). Thus, theconstitutively expressed human cyclin E closely mimicsthe behaviour of the endogenous cyclin E gene underthese experimental conditions.Immune-precipitations revealed that human cyclin E

was associated with p27 in resting cells (Figure 1b). ByWestern blotting, neither p21 nor p57 could bedetected in extracts from RAT1 cells; this is supportedby depletion experiments showing that p27 accountsfor most of the inhibitory activity present in heat-treated lysates from RAT1-MycER cells (data notshown). To test whether release from p27 was requiredfor induction of human cyclin E/cdk2, immune-precipitations were carried out with either antibodiesdirected against human cyclin E or two di�erentantisera directed against p27. As reported above,Myc strongly upregulated human cyclin E-associatedkinase activity. At no point did we detect any histoneH1 kinase activity associated with p27 (Figure 1b),despite the fact that anti-p27 immune-precipitatescontained signi®cant amounts of both cdk2 and cyclin

E (see Figure 2). We concluded from this experimentthat only the fraction of cyclin E/cdk2 that had lostp27 became catalytically active upon induction of Myc.Thus release from p27 is required for activation ofhuman cyclin E/cdk2 kinase by Myc in RAT1-MycERcells.Recently, cdc25A, a phosphatase involved in the

regulation of cdk2 kinase activity, has been shown tobe a target for transcriptional induction by Myc(Galaktionov et al., 1996). To determine whethercomplexes formed with human cyclin E were depho-sphorylated at threonine-14 or tyrosine-15, we treatedhuman cyclin E/cdk2 complexes from induced cells andnon-induced cells with recombinant cdc25A in vitro.From these data, we calculated that more than 70% ofhuman cyclin E/cdk2 complexes present after inductionof Myc were still inhibited by phosphorylation of cdk2at either threonine-14, tyrosine-15 or both, suggestingthat cdc25A remains limiting for full activation ofhuman cyclin E/cdk2 complexes (Figure 1c); this issimilar to ®ndings for the endogenous cyclin E gene(Steiner et al., 1995). The amount of cdc25A protein isnot a�ected by activation of MycER in RAT1®broblasts under conditions where e�cient inductionof cyclin E/cdk2 kinase occurs (Perez-Roger et al.,1997; Pusch et al., 1997; Vlach et al., 1996); andcdc25A protein is detectable in resting RAT1-MycERcells (Perez-Roger et al., 1997). Thus, no evidencecould be obtained that changes in the amount ofcdc25A protein are involved in activation of cyclin E/cdk2 kinase by MycER in this particular cell line.

total

cdk2

cyclin D2

cyclin E

cyclin A

a b1 2 3 4

1 2 3 4

1 2 3 4

p27 —

p27

– phosphatase

+ phosphatase

—

—

—

— —

—

—

—

—

—

—

—

Figure 4 p27 bound to cdk2 is hyperphosphorylated. (a) Anti-p27 Western blot of an isoelectric-focusing-2D gel of total cellularlysates from RAT1-MycER cells either with or without treatmentwith calf intestine phosphatase. (b) Anti-p27 Western blot of anisoelectric-focusing-2D gel of anti-cdk2, anti-cyclin D2, anti-cyclinA, and anti-cyclin E immune-precipitates from mouse telence-phalic vesicles. Anti-cyclin E and cyclin A immune-precipitateswere analysed by ¯uoroimager after Western blotting forenhanced sensitivity, the remainder by exposure to ®lm


2565

Myc does not a�ect the expression of p27 mRNAin RAT1-MycER cells (Steiner et al., 1995). Duringthe G1/S-transition, levels of p27 protein have beenreported to be regulated by two distinct mechanisms:®rst, translation of p27 is regulated in HeLa cells(Hengst and Reed, 1996) and, second, p27 isdegraded in a proteasome-dependent manner (Paga-no et al., 1995). To test which of both, if anymodulates protein levels of p27 in RAT1-MycERcells after induction of Myc, 4-OHT was added to

RAT1-MycER cells either alone or together with aspeci®c inhibitor of proteasome function (PSI:Figueiredo Pereira et al., 1994; Traenckner et al.,1995) or of ICE proteases (ICE inhibitor I, Bachem:An and Knox, 1996; Jacobsen et al., 1996) as acontrol. In the presence of PSI, but not in thepresence of ICE inhibitor, p27 protein levels wereuna�ected by Myc (Figure 2a). We concluded thatinduction of Myc stimulates proteasome-mediateddegradation of p27.

IP: cyclin E – cE con –

histone H1: – – – – – – + + – + – – – – + + – – –

p27 (nM): 5 50 500 5 50 500 50 500 – – 5 50 500 – – 5 50 500

cyclin E

histone H1

p27

nM 100 500 100 500 100 500 100 500 100 500 2500

cyclin E

histone H1

p27

wt S10A T187V T187E p27∆ H1

a

b

c

wt

S10

A

T18

7V

T18

7V

p27

∆

T18

7E

Figure 5 In vitro phosphorylation of p27 by cyclin E/cdk2 kinase. (a) Cyclin E/cdk2 complexes were isolated by immune-precipitation with anti-human cyclin E antibodies (`cE') from Sf9 insect cells that had been co-infected with baculoviruses expressingboth cyclin E and cdk2. As control, either no antibody (`7') or an anti-c-abl antibody (`con') was used as indicated. Isolatedcomplexes were incubated either with histone H1 or with the indicated concentration of p27 or both. (b) Phosphorylation of p27occurs mainly at threonine 187. The indicated proteins were expressed as histidine-tagged proteins in E. coli and puri®ed by a�nitychromatography. They were incubated at the indicated concentrations with cyclin E/cdk2 complexes isolated from baculovirus-infected cells. (c) Western blot of 40 ng of each puri®ed p27 protein. p27D is more unstable in E. coli and runs aberrantly on a SDSgel


2566

To assess how inhibition of p27 degradation a�ectsinduction of cyclin E/cdk2 kinase by Myc, cyclin E/cdk2 kinase activity was determined from lysates ofRAT1-MycER cells. Addition of PSI did not blockinduction of cyclin E/cdk2 kinase by Myc (Figure 2b).PSI somewhat stabilized cyclin E protein: thus, weestimate that there is about a twofold decrease in thespeci®c activity of cyclin E/cdk2 kinase if Myc isactivated in the presence of PSI compared to control(induced) cells. Similarly, addition of PSI did not blockthe Myc-induced upregulation of human cyclin E/cdk2kinase activity in RAT1-MycER (hcE) cells (Figure2c). Immune-precipitations revealed that there wastwofold less human cyclin E and cdk2 associated withp27, but unaltered amounts of human cyclin E/cdk2complexes in lysates from 4-OHT/PSI treated relativeto uninduced cells (Figure 2c). Also, the increasedamount of cyclin A present in induced relative touninduced control cells was not associated with p27,even if degradation of p27 was blocked by addition ofPSI (Figure 2d). Thus, degradation of p27 is notabsolutely required for Myc-induced dissociation ofp27/cyclin E/cdk2 and of p27/cyclin A/cdk2 complexesand for activation of cyclin E/cdk2 kinase activity.Taken together, our data show that activation of cyclinE/cdk2 complexes by MycER in resting cells occurs bya mechanism that is similar to the sequestration of p27observed in Myc-transformed cells infected withrecombinant retroviruses that ectopically expresselevated levels of p27 (Vlach et al., 1996).Several time course experiments revealed that

activation of cyclin E/cdk2 often preceded the loss ofp27 from total cell extracts (data not shown). Wewondered, therefore, whether there might be apreferential loss of p27 from cdk2 complexes at earlytime points after induction of Myc and preparedlysates from con¯uent RAT1-MycER cells before andat several time points after induction of Myc. Fromthese lysates, we determined in parallel cyclin E/cdk2kinase activity (Figure 3a) and the amount of p27complexed either with cyclin D1, cyclin D3 and withcdk2 and cyclin A (Figure 3b and c). Similar toobservations reported before, activation of Myc rapidlyinduced cyclin E/cdk2 kinase activity. There was analmost complete disappearance of p27 from cdk2(Figure 3b and c) and from cyclin A (Figure 3c)complexes at a time point at which there was nodetectable decrease in the total amount of p27 and theamount of p27 complexed with either cyclin D1 orcyclin D3 (Figure 3b). There was a twofold increase inthe amount of endogenous cyclin E, but no increase inthe amount of cyclin E/p27 complexes after inductionof Myc, suggesting that a subset of cyclin E complexeshad lost p27 upon activation of Myc (Figure 3c). Weconcluded that at early time points, Myc-induceddissociation of p27 was speci®c for p27 bound to cdkcomplexes.This observation prompted us to perform two

experiments. First, we tested whether cyclin E/cdk2complexes isolated from Myc-induced cells wereresistant to inhibition by recombinant p27. We foundthat recombinant p27 inhibited cyclin E/cdk2 kinaseactivity from Myc-induced cells with an inhibitionconstant of 2.6+0.1 nM and of 4.1+0.25 nM forcomplexes isolated from non-induced cells (at 50 mMATP; data not shown). Thus, activation of Myc does

not render cyclin E/cdk2 complexes resistant toinhibition by p27.Next we wondered whether p27 bound to cdk2

di�ered physically from p27 associated with D-typecyclin/cdk complexes. p27 has previously been shownto resolve into a number of discrete spots with di�erentisoelectric points upon 2D gel electrophoresis (Celis etal., 1995). In lysates from RAT1-MycER cells, p27resolved into four major spots (labelled 1 to 4 in Figure4a). Treatment of cell lysates with phosphataserevealed that forms 2-4 were due to di�erentialphosphorylation of p27 (Figure 4a).From lysates of RAT1-MycER cells, we were able to

visualize total p27 and p27 bound to cyclin D1. Incontrast to the total cellular p27, anti-cyclin D1immune-precipitates contained only spot 1 and asmall amount of spot 2, but lacked the more acidicforms 3 and 4 (not shown). Control precipitationsshowed that cyclin D1 was associated with cdk4 underthese conditions, precluding the possibility that lack ofphosphorylation was due to the lack of a kinasesubunit in the complex. We were unable to identify p27in cyclin E, cyclin A or cdk2 complexes due to the

a 4 mM ATP

b

50 µM ATP

p27

histone H1

p27

– –

p27 p27– –

Figure 6 Lack of inhibition of p27 phosphorylation by p27.Cyclin E/cdk2 complexes were isolated by immune-precipitationwith anti-human cyclin E antibodies from baculovirus-infectedSf9 cells and incubated with histone H1 in the presence of theindicated concentrations of p27. The experiment was performed inthe presence of 4 mM ATP (a, left) and in the presence of 50 mMATP (a, right). a shows the autoradiogram of the kinase reaction.b shows a quantitation of the phosphorylation of p27 and histoneH1 at each concentration of p27 relative to the maximalphosphorylation that is observed for each protein


2567

lower sensitivity of Western blots after 2D gelelectrophoresis (not shown).To obtain a complete set of data, we exploited our

observation that embryonic (E13) mouse brainexpresses high levels of p27. This tissue expressedhigh levels of cyclin D2, but not cyclin D3 or cyclin D1as judged by Western blotting. Total cell lysatesshowed a pattern of phosphorylated forms of p27identical to RAT1-MycER cells upon 2D gel-electro-phoresis (Figure 4b). Lysates were prepared, immune-precipitated with speci®c antibodies directed againstcdk2, cyclin D2, cyclin A, and cyclin E and separatedby 2D gel electrophoresis; as a control, total cell lysatewas subjected to the same procedure (Figure 4b). Thetwo most basic forms of p27 (spots 1 and 2) werefound in immune-precipitates with anti-cdk2-, cyclinD2-, cyclin A- and cyclin E-antibodies. The two mostacidic phosphorylated forms of p27 (spots 3 and 4)were found only in association with cyclin E and A andwith cdk2, but not in anti-cyclin D2 immune-precipitates. The data show that p27 bound to cyclinE and cyclin A/cdk2 complexes di�ers from p27 boundto D-type cyclins as it is phosphorylated at at least oneadditional position. Thus, either cdk2 phosphorylatesp27 directly or p27 bound to cdk2 is subject tophosphorylation by another kinase.To determine whether cdk2 can phosphorylate p27

directly, we co-infected Sf9 cells with baculoviruses that

express human cyclin E and cdk2. After infection,lysates were prepared and immune-precipitated with ananti-cyclin E antibody. Immune-precipitates wereincubated with increasing concentrations of recombi-nant p27 (Figure 5a). Phosphorylation of p27 wasobserved at 50 and 500 nM ®nal concentration of p27.Phosphorylation of both histone H1 and of p27 wasobserved when both proteins were mixed. Assumingthat the protein-G sepharose is saturated with cyclin E/cdk2 complexes, we estimate that these reactionscontain an amount of cdk2 equivalent to 100 nM finalconcentration; this may provide an explanation as towhy little inhibition of histone H1 phosphorylation isobserved under the experimental conditions. To controlwhether either the immune-precipitates or the prepara-tion of recombinant p27 were contaminated withanother kinase, identical amounts of Sf9 lysates wereimmune-precipitated with an irrelevant antibody: nophosphorylation of either p27 or histone H1 wasobserved in these reactions (`con'). Finally, incubationof either puri®ed p27 or histone H1 without insect celllysate did not yield phosphorylated protein. Weconcluded that p27 is both an inhibitor and asubstrate for cyclin E/cdk2 kinase.To determine which amino acid of p27 was

phosphorylated by cdk2 in vitro, we expressed ashistidine-tagged proteins and puri®ed p27 proteins thatcarry mutations at amino acids 187 (T187V and

p27 T187E wt-p27

nM: 0 0.05 0.5 5 50 0 0.05 0.5 5 50 0

— Histone H1

p27 T187E p27S10A

nM: 0 0.05 0.5 5 50 0 0.05 0.5 5 50

a

bbound bound

input A B

wt p27 — wt p27 —

p27 T187V — p27 T187V —

short exposure long exposure

— Histone H1

input A B


2568

T187E) and at amino acid 10 (S10A) as well as aprotein in which the last 15 amino acids were deleted(p27D). Equivalent amounts of each puri®ed protein(see Figure 5e) were incubated with cyclin E/cdk2kinase immune-precipitated from infected cells and thedegree of phosphorylation determined. We observedthat the p27T187V and p27T187E protein incorporated75% less phosphate than either the wild-type p27 orp27S10A; no signi®cant phosphorylation of thetruncated protein was found (Figure 5b). We con-cluded that p27 is predominantly phosphorylated bycyclin E/cdk2 at threonine 187; the remainingphosphorylation occurs most likely at serine 183(which is lost from p27D but present in the othermutants).Phosphorylation of p27 by cyclin E/cdk2 was

observed even when phosphorylation of histone H1was partially inhibited (see Figure 5a). To con®rm thisobservation, increasing amounts of p27 were added toan in vitro kinase assay containing histone H1 assubstrate. In these reactions, p27 phosphorylation bycyclin E/cdk2 was not inhibited even at a concentrationof p27 that inhibited phosphorylation of histone H1 bymore than 95% (Figure 6a and b). Similar results wereobtained both in the presence of 4 mM and 50 mM ATP

®nal concentration, although the total amount of p27that was phosphorylated was lower at 50 mM ATP. Weconcluded that p27 does not signi®cantly inhibit itsown phosphorylation by cyclin E/cdk2 kinase. In thesereactions, we also observed phosphorylation of thecyclin E protein (Figure 5a). From a reaction thatcontains neither histone H1 nor p27, the amount ofphosphorylated cyclin E can be used as an internalstandard for the amount of active cyclin E/cdk2complexes present in the reaction. Assuming that bothp27 (this report) and cyclin E (Clurman et al., 1996;Won and Reed, 1996) are phosphorylated by cdk2 at asingle residue, we calculate from the radioactivityincorporated into cyclin E and p27 that up to threemolecules of p27 are phosphorylated in the presence of4 mM ATP per active cyclin E/cdk2 complex inmultiple independent reactions. We conclude thatturnover of p27 occurs in these reactions. Inhibitionof histone H1 phosphorylation by p27 thus appears toresult from competition for access to cyclin E/cdk2rather than from inhibition of the catalytic activity ofcdk2 itself.To determine whether phosphorylated p27 di�ers in

its interaction with cyclin E/cdk2 complexes from non-phosphorylated p27, increasing amounts of either wild

Figure 7 In vitro analysis of mutant p27 proteins (a) Inhibition of histone H1 kinase activity of cyclin E/cdk2 by recombinant p27proteins. The indicated alleles of p27 were expressed as histidine-tagged proteins in E. coli and puri®ed by Ni-agarosechromatography. They were incubated at the indicated ®nal concentrations with cyclin E/cdk2 complexes isolated by immune-precipitation from induced RAT1-MycER cells. The amounts of protein used are documented in Figure 5. (b) Cyclin E/cdk2complexes immune-precipitated from infected Sf9 cells were incubated with either recombinant p27wt or p27T187V proteins. Theinput lanes correspond to 10% of the loaded material. Cyclin E/cdk2 complexes were reisolated and subjected either to three roundsof washing in bu�er containing 0.1% NP40 (sample a). Alternatively, they were washed as above, followed by an incubation for 1 hat 378C and reisolation (sample b). Shown is a Western blot of the bound p27 proteins. (c) Stability of di�erent alleles of p27.Shown is the percentage of input p27 bound to cyclin E/cdk2 after washing as described above for three di�erent alleles of p27. (d)Phosphorylated p27 is preferentially released from cyclin E/cdk2 complexes. Recombinant p27 was incubated with cyclin E/cdk2complexes isolated from baculovirus-infected Sf9 cells as described above. Bound and free p27 were separated by centrifugation,bound p27 was eluted by boiling in SDS and both bound and eluted p27 fractionated by gel ®ltration chromatography. The panelsshow Western blots of the peak fractions for cyclin E, cdk2 and p27 and an autoradiography of phosphorylated p27

c

bound released

Fraction No.: 4 5 6 4 5 6

cyclin E

cdk2

p27

d

32P-p27


2569

type recombinant p27 or of p27S10A, p27T187V,p27T187E or p27D (not shown) mutant proteins wereincubated with cyclin E/cdk2 complexes isolated fromgrowing RAT1-MycER cells and the histone H1 kinaseactivity determined. We found that neither mutation atresidues 187 or 10 nor truncation of the last ®fteenamino acids a�ected the amount of p27 needed forhalf-maximal inhibition of cyclin E/cdk2 in vitro(Figure 7a), suggesting that alterations at amino acid187 do not a�ect the inhibition constant of p27 forcyclin E/cdk2 complexes. This is supported by ourobservation that p27 isolated by heat-treatment fromMycER cells induced with 4-OHT in the presence ofPSI is fully capable of inhibiting cyclin E/cdk2 kinase(data not shown). We next tested whether mutation atthreonine 187 a�ected the stability of a ternary cyclinE/cdk2/p27 complex. Therefore, we preincubatedsaturating amounts of either p27wt or p27T187V withcyclin E/cdk2 complexes immune-precipitated frombaculovirus infected cells; the complexes were theneither subjected to three rounds of washing in bu�ercontaining detergent to determine the amount of stablybound p27 (Figure 7b: sample designated `A').Alternatively, complexes were reisolated and washedand then incubated for another hour at 308C beforethey were reisolated again (Figure 7b: sampledesignated `B'). The amount of stably bound p27 wasdetermined by Western blotting. In these experiments,p27 wt and p27T187V di�ered signi®cantly from eachother as binding of p27T187V to cyclin E/cdk2 wassigni®cantly more stable than that of p27wt, suggestingthat mutation at that site lowers the o�-rate of p27from preassembled cyclin E/cdk2 complexes. Deletionof the last 15 amino acids further enhanced thestability of ternary cyclin E/cdk2/p27 complexes,suggesting that the carboxy-terminus of p27 functionsas a negative regulator of complex stability. Aquantitation of the results is shown in Figure 7c.To show directly that phosphorylation of p27 by

cyclin E/cdk2 enhances its dissociation from cdkcomplexes, a limiting amount of recombinant p27was bound to cyclin E/cdk2 in vitro and incubated inthe presence of g[32P]ATP as before. After theincubation, p27 released from cyclin E/cdk2 wasisolated by centrifugation and gel ®ltration on a G-50column. Bound p27 was dissociated by boiling in SDSand similarly separated over a G-50 column. Westernblots of the fractions recovered from the G-50 columnsrevealed that little cyclin E and no cdk2 leaked fromthe beads during the incubation (Figure 7d). Weobserved that only a small fraction of total p27 wasreleased from cyclin E/cdk2 during the incubation.Autoradiography of the same fractions showed that amuch higher fraction of phosphorylated p27 wasreleased from cyclin E/cdk2 complexes during theincubation. From a comparison of both blots, weestimate that there is an at least tenfold enrichment ofphosphorylated p27 in the eluted fractions. The dataprovide direct evidence that phosphorylation of p27 bycyclin E/cdk2 facilitates its release from cyclin E/cdk2complexes in vitro.We speculated, therefore, that phosphorylation of

p27 by cdk2 might stimulate its dissociation fromcyclin E/cdk2 complexes in vivo upon activation ofMyc. This hypothesis makes a number of testablepredictions: First, both dissociation and degradation

should depend on cdk2 kinase activity in vivo. Twoexperiments were set up to test this prediction. First,we tested whether a speci®c inhibitor of cdk2 kinaseactivity, roscovitine, interfered with Myc-induced lossof p27 from cyclin E/cdk2 complexes (Abraham et al.,1995; Vesely et al., 1994). RAT1-MycER cells wereinduced either in the absence or presence of roscov-itine. Lysates were prepared, immune-precipitated withantibodies directed against cdk2, the precipitatesseparated by SDS ±PAGE and probed with antibodiesdirected against p27. As observed before, loss of p27from cdk2 complexes preceded the Myc-induceddegradation of p27 by several hours in the absence ofroscovitine (Figure 8a). Loss of p27 from cdk2complexes was observed in the absence of roscovitine,but was signi®cantly delayed in the presence of thecdk2 inhibitor (Figure 8a). Control experimentsestablished that roscovitine did not interfere with theability of Myc to induce the expression of a directtarget gene, prothymosin-a (not shown). Roscovitinealso blocked the decrease in the total amount of p27present in cell lysates observed at later time points(Figure 8a), suggesting that cdk2 kinase activity isrequired both for Myc-induced dissociation of p27 anddegradation.To extend these results, expression plasmids encod-

ing either dominant negative alleles of cdk2, of cdk4 orencoding the cdk inhibitors p16 and p21 weremicroinjected into quiescent RAT1-MycER cells. Cellswere subsequently induced for twelve hours, ®xed andstained with antibodies directed against p27. Injectionof empty expression vector did not a�ect Myc-induceddegradation of p27 (data not shown). In contrast,expression of dominant negative alleles of cdk2 or ofthe cdk inhibitor, p21, abolished Myc-induced degra-dation of p27 (Figure 8b). Ectopic expression of p16 orof dominant negative alleles of cdk4 also inhibitedMyc-induced degradation of p27, most likely becausep16 (but not dominant negative alleles of cdk2) inhibitstransactivation by Myc (Haas et al., 1997). Takentogether, the data show that cdk2 kinase activity isrequired for both Myc-induced dissociation of p27from cyclin E/cdk2 complexes and its subsequentdegradation.Second, one would predict that there should be an

accumulation of phosphorylated p27 after induction ofMyc. To test this prediction, we infected RAT1-MycER cells at low density with recombinant retro-viruses that express either p27wt, p27T187V, p27T187Eor p27S10A. Under these experimental conditions,endogenous p27 is not detectable. 4-OHT was added,cells were harvested at several time points and lysatesprobed with an antibody directed against p27; tovisualize potential phosphorylated forms of the protein,SDS ±PAGE was run signi®cantly longer than usual.In these experiments, we observed a slower

migrating band that appeared seven hours afteractivation of Myc; treatment of boiled cell lysateswith calf intestine phosphatase revealed that this bandwas due to phosphorylation of p27 (Figure 9a and b).This band was observed in cells infected with virusesexpressing either wild-type p27 or p27S10A, but not incells infected with viruses expressing p27T187V andp27T187E (Figure 9a). We concluded that there is atransient accumulation of p27 phosphorylated atresidue 187 in vivo after induction of Myc.


2570

To determine whether phosphorylated p27 wasbound to cdk2 under these conditions, cell lysateswere precipitated with anti-cdk2 antibodies (Figure 9a).We observed a transient initial increase in the amount

of p27 bound to cdk2, re¯ecting an early increase inthe total amount of p27; this probably re¯ects the®nding that the retroviral LTR that drives expressionof p27 is stimulated by activation of Myc (ME,unpublished). As observed for endogenous p27,ectopically expressed p27 is rapidly dissociated fromcdk2 complexes after activation of Myc and phos-phorylated p27 did not accumulate bound to cdk2.Parallel immune-precipitations also did not reveal anaccumulation of p27 on cyclins D1 and D3 (notshown). Wild-type p27, p27T187E and p27S10A wererapidly lost from cdk2 complexes after induction ofMyc; in contrast, binding of p27T187V to cdk2 wassigni®cantly more extensive and it dissociated moreslowly from cdk2 complexes, although the total level ofexpression was identical to that of p27T187E andp27S10A.Western blots of total p27 from these cells revealed

that wtp27 and p27S10A disappeared from total cellextracts at later time points, suggesting that they werebeing degraded; disappearance of p27T187V was, incontrast, signi®cantly delayed, whereas the total levelof p27T187E were somewhat variable (Figure 9c).These data provide direct biochemical evidence for thetransient accumulation of phosphorylated p27 in anon-cdk2 bound form before its subsequent degrada-tion upon induction of Myc. Second, the observationthat p27T187V is retarded re¯ects its behaviour in vitroand suggests that phosphorylation at threonine 187facilitates Myc-induced dissociation of p27 from cyclinE/cdk2 complexes in vivo.To test whether these biochemical data correlated

with the rate of proliferation of these cells, RAT1-MycER cells were infected with either empty vector orretroviruses expressing di�erent mutant alleles of p27.Cells were seeded at low density and grown in non-oestrogen depleted medium. Under these conditions,the MycER protein is partially active due to the lowamounts of oestrogen present in serum. We chose thisprotocol as it circumvents the high rate of Myc-induced apoptosis observed when 4-OHT is added forlonger time periods. The cell number was determinedat several days after plating. Control RAT1 cells werealmost completely growth-arrested upon infection withvirus expressing wild-type p27 (Figure 9d); in contrast,RAT1-MycER cells grew in the presence of wild-typep27, albeit at a reduced rate compared to non infectedcells (Figure 9d). In several independent experimentswe found that the growth rate of cells infected with thep27T187V virus was even slower than that of cellsinfected with viruses encoding either wild-type p27,p27S10A or p27T187E (Figure 9d). Thus, mutation ofthreonine 187 to valine retards both Myc-induceddissociation from cyclin E/cdk2 complexes in vivo andretards Myc-dependent proliferation.

Discussion

Previous work had shown that induction of Mycresults in a rapid upregulation of cyclin E/cdk2 kinaseactivity and had suggested that Myc triggers theactivation of pre-existing cyclin E/cdk2 complexes(Steiner et al., 1995). We now show that thisactivation is due to a dissociation of p27 from cdk2complexes; this precedes the actual degradation of p27.

hrs 4-OHT

total p27 —

0 4 9 13 4 9 13

– roscovitine + roscovitinea

IP: co cdk2 co cdk2 co cdk2 co cdk2

0 4 9 13

– roscovitine

p27 associated with cdk2

—

co cdk2 co cdk2 co cdk2

4 9 13

IP:

p27 associated with cdk2

—

+ roscovitine

b

Figure 8 cdk2 kinase activity is required for Myc-inducedsequestration and degradation of p27 in vivo. (a) Preincubationof RAT1-MycER cells with the cdk2 inhibitor, roscovitine, delaysMyc-induced loss of p27 from cyclin E/cdk2 complexes anddegradation of p27. Shown are Western blots documenting thetotal amount of p27 and of p27 present in anti-cdk2 immune-precipitates either before or after the indicated times afteraddition of 4-OHT. Where indicated, roscovitine was addedtogether with 4-OHT. An antibody directed against Mad-1 wasused as control. (b) Myc induced degradation of p27 requiresboth cdk2 and cdk4 kinase activity. The graph documents thepercentage of p27-positive RAT1-MycER cells before and twelveafter induction of Myc as judged by indirect immune¯uorescencewith a monoclonal anti-p27 antibody. Plasmids encoding cdk2kn,cdk4 kn (van den Heuvel and Harlow, 1993), p16 and p21 wereinjected four hours before addition of 4-OHT


2571

Dissociation is speci®c, as cyclin D/cdk4 complexes arestable for an extended period of time after induction ofMyc; this correlates with previous observations thatthere is a slow increase in cyclin D1-dependent kinaseactivity after induction of Myc (Steiner et al., 1995).Myc-induced dissociation of p27 from cdk2 com-

plexes is facilitated by a cdk2-dependent phosphoryla-tion of p27 at threonine 187. This notion is supportedby four lines of evidence: First p27 was found to be asubstrate for cdk2 in vitro. Cdk2 phosphorylates p27predominantly at threonine 187. In vivo, p27 bound tocyclins E and cyclin A is hyperphosphorylated; incontrast, p27 bound to cyclins D1 and D2 is largelyhypophosphorylated. Second, cdk2 kinase activity isrequired for dissociation of p27 from cyclin E/cdk2complexes and degradation of p27 after induction ofMyc. Third, mutation of threonine 187 to valinestabilises ternary cyclin E/cdk2/p27 complexes in vitroand delays Myc-induced dissociation of p27 from cdk2in vivo. Fourth, p27 phosphorylated at threonine 187accumulates after induction of Myc in an non-cdk2bound form. We suggest that the requirement forcarboxy-terminal phosphorylation of p27 explains the

speci®c dissociation of p27 from cyclin E/cdk2complexes and the speci®c activation of cyclin E/cdk2kinase activity that is observed after activation of Myc(Steiner et al., 1995).Mutation of threonine 187 to either valine or

glutamate does not alter the inhibition constant bywhich p27 inhibits cyclin E/cdk2 complexes in vitro.Also, phosphorylated p27 binds to cyclin E/cdk2complexes in vitro and binding is speci®c as it iscompeted by cold p27 (DM, unpublished). This isconsistent with our observation that phosphorylatedp27 is associated with cdk2 complexes in vivo.However, phosphorylation at threonine 187 decreasesthe stability of ternary cyclin E/cdk2/p27 complexesupon repeated washing and reisolation, suggesting thatmutation at this site a�ects the on- and o�-rate ofcyclin E/cdk2/p27 complexes. The data show thatphosphorylation of p27 per se cannot be su�cient togenerate a pool of free cyclin E/cdk2 complexes in vivo.We suggest that Myc directs the synthesis of proteinsthat bind to and ultimately degrade p27 that istransiently released from cyclin E/cdk2 complexesupon phosphorylation (Figure 10). Indeed, direct

—

—

—

—

—

—

—

—

—

—

—

—

p27wt —

a

b c

p27 S10A —

p27T187V —

p 27T187E —

0 4 7 14

total p27

inh: – – +

CIP

wt p27

p27

p27 P

—

—

—

—

—

—

—wt p27

— p27 T187V

time after addition of 4-OHT (hours)

0 4 10 13 15 20

0 4 7 14

—

—

—

—

—

—

—

—

—

—

—

—

α-cdk2 IPhrs 4-OHT


2572

evidence was found for the transient accumulation ofphosphorylated p27 in a non-cdk2 bound form uponinduction of Myc. Proteins bound to p27 at this stageappear to be distinct from D-type cyclins as there is nonet increase of p27 bound to cyclins D1 and cyclin D3after induction of Myc and no phosphorylated p27 wasfound bound to D-type cyclins. A somewhat similarmodel has recently been proposed by Amati andcolleagues to explain their ®nding that Myc trans-formed cells are partially (but not completely, seeRudolph et al., 1996) resistant to growth inhibition byectopic expression of p27 (Vlach et al., 1996). Whatmight the identity of these proteins be? There areseveral obvious parralels between Myc-induced dis-sociation of cyclin E/cdk2/p27 complexes and protea-some-dependent degradation of p27. First, activationof Myc induces both dissociation of p27 from cyclin E/cdk2 complexes and its subsequent degradation.Second, both Myc-induced dissociation and degrada-tion of p27 depend on cdk2 kinase activity. Finally,mutation of threonine 187 to valine delays Myc-induced dissociation of p27 from cyclin E/cdk2complexes and inhibits degradation of p27 in prolifer-ating cells (Shea� et al., 1997). It seems possible,therefore, that dissociation of p27 upon induction ofMyc may re¯ect the ®rst step in degradation of p27. Ifdegradation of p27 has similarities to the degradationof G1 cyclins in S. cerevisiae, our observations may beequivalent to the recognition of phosphorylated Cln2by a Cdc53/Cdc34/Cdc4 complex which precedes itsdegradation (e.g. Bai et al., 1996; Lanker et al., 1996;Willems et al., 1996). In this sense, the Myc-dependent`sequestration' of p27 reported by Amati andcolleagues may refer to the formation of a similar

complex (Vlach et al., 1996). However, we have noevidence that p27 accumulating after induction of Mycis ubiquitinated. Thus, the precise relationship betweenthe proteasome-dependent degradation of p27 and thedissociation observed after induction of Myc remainsto be determined.If most or all cyclin E/cdk2 is bound to p27 in

resting cells, how can Myc induce dissociation of p27from these complexes? As one solution, it has beensuggested that the moderate activation of cyclin Eexpression by Myc (Jansen-DuÈ rr et al., 1993) mighttrigger the synthesis of a small amount of free andactive cyclin E/cdk2 complexes which then start apositive feedback-loop that phosphorylates andreleases p27 from pre-existing ternary complexes(Perez-Roger et al., 1997). This model assumes thateither p27 bound to one cyclin E/cdk2 complex canbe phosphorylated by a second cyclin E/cdk2complex or that the on-and o� rates for p27 tocyclin E/cdk2 are high enough to allow a few activecomplexes to phosphorylate multiple p27 molecules invivo.A tacit assumption in this model is that a cyclin E/

cdk2 complex with a bound molecule of p27 is inactiveas a kinase towards p27. Our data are not easilycompatible with this assumption: we ®nd that cyclin E/cdk2 complexes actively phosphorylate p27 even whentheir kinase activity towards histone H1 is inhibited bymore than 95%. Assuming that the protein G-sepharose is saturated by cyclin E/cdk2 complexes,the molar excess of p27 relative to cyclin E/cdk2 isapproximately eightfold at a concentration of 750 nMp27. Thus, if p27 phosphorylation by cyclin E/cdk2required the presence of one p27-free complex, a strong

d

Figure 9 In vivo analysis of mutant alleles of p27 in RAT1-MycER cells. (a) RAT1-MycER cells were infected with retrovirusesexpressing the indicated alleles of p27 and resistant cells selected in puromycin. They were then plated at low cell density. At thestart of the experiment, 200 nM 4-OHT was added and lysates prepared at the indicated times afterwards. An aliquot of the lysatewas probed directly by Western blotting with anti-p27 antibody (left panel); a second aliquot was immune-precipitated withantibodies directed against cdk2 and the precipitates probed with an anti-p27 antibody (right panel). In this experiment, p27 becameundetectable after 20 h. (b) p27wt is phosphorylated. p27wt extracted from cell lysates by boiling was treated with calf intestinalphosphatase either in the absence or presence of phosphatase inhibitors and analysed by Western blotting. (c) Western blotsdocumenting the amount of total p27wt and of p27T187V in infected RAT1 MycER at di�erent time points after addition of 4-OHT. (d) Left: Growth curves of RAT1 or RAT1-MycER cells that were either non infected or infected with retroviruses expressingp27wt. Right: Growth curves of RAT1-MycER cells infected with viruses expressing the indicated alleles of p27. Cells were seededat a density of 16105 cells. Non-infected RAT1 and RAT1-MycER cells grow essentially at the same rate (Pusch et al., 1997)


2573

inhibition should occur. This is not observed. Further,if phsophorylation of cyclin E is taken as a measure forthe amount of active cyclin E/cdk2 complexes, wecalculate that up to three molecules of p27 arephosphorylated per cyclin E/cdk2 complex present inthese reactions. Thus, multiple rounds of p27phosphorylation occur per cyclin E/cdk2 complexwithout apparent inhibition by p27; again, the lack ofinhibition under these conditions is hard to reconcilewith a model that requires the presence of p27-freecomplexes for phosphorylation of p27 and stronglyfavours a model in which p27 is phosphorylated by asingle cyclin E/cdk2 complex. Therefore, the potentialfunction of a positive feedback loop is somewhatunclear. The ®ndings also alleviate the need to invoke atransient complex in which p27 associates with cyclinE, but not cdk2 as suggested by Amati and colleagues(Vlach et al., 1997, in press).Alternatively, phosphorylation and degradation of

p27 might be increased in response to elevated levels ofcdc25A protein that is induced by Myc (Galaktionov etal., 1996). Indeed, our data show that cdk2 activity isrequired for Myc to act upon p27. The ®nding thatcdc25A activity remains rate-limiting for full activationof cyclin E/cdk2 complexes after activation of Myc(Steiner et al., 1995) does not argue against such aview, assuming that there is some turnover even ofnon-phosphorylated p27 that allows a small number ofactive cdk2 complexes to phosphorylate a large numberof p27 molecules in vivo.At present, there appear to be three ®ndings that are

not explained by the hypothesis that Myc actsupstream of cdc25A expression to induce cyclin E/cdk2 kinase activity: ®rst, induction of cyclin E/cdk2kinase activity in RAT1-MycER cells occurs withoutsigni®cant changes in the amount of cdc25A protein(Perez-Roger et al., 1997; Pusch et al., 1997; Vlach etal., 1996). Second, phosphorylation of p27 does notappear to a�ect its a�nity to cyclin E/cdk2 complexesin vitro, yet p27 dissociates from cyclin E/cdk2complexes and is degraded in vivo upon activation ofMyc. Third, cdc25A cannot substitute for Myc in theactivation of cyclin E/cdk2 kinase activity, whereasprotein that bind to and sequester p27 can overcome aG1 arrest imposed by dominant negative alleles of Myc(Berns et al., 1997).Resting RAT1-MycER cells contain signi®cant

amounts of cyclin E, cdk2 and of cdc25A proteins;taken together with data reported in this paper, thisstrongly suggests that phosphorylation of p27 by cyclin

E/cdk2 occurs constitutively in these cells and thatphosphorylation of exogenous substrates is blockeddue to steric hindrance by p27. By far the simplestmodel describing the regulation of cyclin E/cdk2 kinasein RAT1-MycER cells is one in which phosphorylationof p27 by cyclin E/cdk2 does not depend on Mycfunction; phosphorylation by cyclin E/cdk2 enhancesthe on- and o�-rate from complexes whereas p27bound to D-type cyclin complexes is non-phosphory-lated and more stable. Activation of Myc induces thesynthesis of proteins that bind to and ultimatelydegrade p27 that is released from cyclin E/cdk2complexes upon phosphorylation (Figure 10). Thismodel makes the testable prediction that the abilityof a cell to sequester and degrade p27 should requireMyc function independently of the requirement forcyclin E/cdk2 to phosphorylate p27.

Materials and methods

Tissue culture and microinjection experiments

All experiments were carried out in RAT1-MycER (Eilerset al., 1989) or RAT1-MycER2 (Littlewood et al., 1995)cells as indicated. Cells were cultured as describedpreviously (Steiner et al., 1995). Brie¯y, cells were grownto con¯uence and then refed every two days for a period of4 ± 6 days before addition of 4-hydroxy-tamoxifen (4-OHT)to a ®nal concentration of 200 nM. Microinjections werecarried out as described (Rudolph et al., 1996). To generateRAT1-MycER(hcE) cells, RAT1-MycER cells were trans-fected with a pBabe-Puro vector (Morgenstern and Land,1990) carrying a full length human cyclin E cDNA (kindgift of Rolf MuÈ ller). Resistant cells were selected inmedium containing 2.5 mg/ml puromycin and 200 mg/mlneomycin and pooled. Mutant alleles of p27 cloned inpBabe-Puro were a kind gift of J Vlach and B Amati.Retroviruses were generated by transient transfection ofBOSC23 cells (Pear et al., 1993). PSI (Cbz-IleGlu(o-tBu)Ala-Leucinal) and ICE Inhibitor I (Ac-Tyr-Val-Ala-Aspartic acid aldehyde; Bachem) were used at a ®nalconcentration of 50 mM each. Roscovitine was used asdescribed (Rudolph et al., 1996).

Preparation of telencephalic vesicles

E13 mouse embryos were prepared according to (Copp andCockroft, 1990). Telencephalic vesicles were dissected inice-cold PBS and immediately frozen in liquid nitrogen.Between one and ®fteen telencephalic vesicles were pooledand extracts prepared in NP-40 lysis bu�er containing bothprotease and phosphatase inhibitors (Steiner et al., 1995).

Figure 10 Model describing the ®ndings. For details, see Discussion


2574

IEF 2D-Gel electrophoresis

Immune-precipitations were carried out and washed in NP-40 lysis bu�er as described before (Steiner et al., 1995).Protein-G sepharose beads were frozen and thawed inO'Farrell lysis bu�er (O'Farrell, 1975). IEF-2D-Gelelectrophoresis was carried out according to (O'Farrell,1975) except that the following mixture of ampholytes(Pharmacia) was used: pH 3.5 ± 9.5: 3.5% (v/v); pH 5 ± 7:1.75% (v/v); pH 7 ± 9: 1.75% (v/v). 12%-PAGE was usedas the second dimension. Western blots were carried outusing an enhanced chemoluminescence protocol accordingto the manufacturer's instructions (Amersham); whereindicated, a ¯uoroimager was used for detection. Forphosphatase treatment, 100 mg cell lysate was boiled toremove the bulk of cellular protein and clari®ed bycentrifugation. Control experiments showed that boilingdid not a�ect the pattern of spots seen in the 2D gels (datanot shown). Clari®ed lysates were incubated with 100 unitsof calf intestine phosphatase (CIP) for 30 min at 8Caccording to the manufacturer's instructions (Boehringer).

Cyclin E/cdk2 kinase assays

Cyclin E/cdk2 kinase assays using histone H1 as substratewere carried out as described previously (Steiner et al.,1995); to analyse human cyclin E kinase activity, we used acommercial antibody (HE-111; Santa Cruz). As control, ananti-mad-1 antibody was used (Santa Cruz). To detectphosphorylation of p27, Sf9 insect cells were co-infectedwith baculoviruses expressing cyclin E and cdk2 (Gu et al.,1993). Lysates were prepared and precipitated with eitheranti-cyclin E or anti c-abl (Santa Cruz) antibodies; ascontrol 0.5 mg of antibody was used in each reaction withsaturating amounts of protein-G sepharose and cell lysate.Incubations were carried out in a ®nal volume of 20 mlcontaining the indicated ®nal concentration of p27 and ofATP and 2 ± 5 mCi [32P]ATP. 1.25 mg histone H1 was addedwhere indicated.

Recombinant p27 protein was puri®ed by denaturing Ni-a�nity chromatography according to the manufacturer's

instructions (Quiagen) and used to raise a polyclonalantiserum in rabbits. Depletion and immunecomplex kinaseexperiments were carried out using both this antiserum and acommercial antibody raised against the 19 carboxy-terminalamino acids (C-19; Santa Cruz). For depletion experiments,200 mg of total cell lysate was incubated for 1 min at 1008C,clari®ed by centrifugation and incubated with 5 ml C-19 or20 ml polyclonal antiserum prebound to 20 ml Protein-Gsepharose. Two rounds of depletion were performed, one forfour hours and the subsequent one overnight. Inhibitionconstants were determined from Lineweaver-Burk diagrams.

Antibodies

Anti-p27 Western blots and immune ¯uorescence experi-ments were performed with a monoclonal antibody (SignalTransduction Laboratories). Cyclin D2 was detected withmonoclonal antibody DCS-3, cyclin D3 with monoclonalantibody DCS-22; both were generous gifts from JiriBartek. Other antibodies were used as described (Ru-dolph et al., 1996).

AcknowledgementsWe thank Rainer Frank for the synthesis of oligonucleo-tides, David Morgan for the gift of recombinant baculo-viruses, Karl-Heinz Nasheuer for puri®ed cyclin E/cdk2kinase, Joan MassagueÂ for plasmids encoding p27, LaurentMeijer for roscovitine, Jaromir Vlach and Bruno Amati forthe gift of retroviruses expressing mutant alleles of p27 andRolf MuÈ ller for human cyclin E cDNA, Ingrid Ho�mannfor recombinant cdc25A and Ilona Hallwax and AnnetteRuppert for excellent technical support. In particular, wethank Sherwin Wilk for numerous gifts of PSI and J Bartekfor antibodies against D-type cyclins. This work wassupported by the `Zellzyklusschwerpunkt' of the DeutscheForschungsgemeinschaft. CB acknowledges receipt of aEMBO long-term fellowship.

References

Abraham RT, Acquarone M, Anderson A, Asensi A, Bell R,Berger , Bergounioux C, Brunn G, Buquet-Fagot C, FagotD, Glab N, Goudeau H, Goudeau M, Guerrier P,Houghton PJ, Hendriks H, Kloareg B, Lippai M, MarieD, Maro B, Meijer L, Mester J, Mulner-Lorillon O, PouletSA, Schierenberg E, Schutte B, Vaulot D and VerlhacMH. (1995). Biol. Cell, 83, 105 ± 120.

An S and Knox KA. (1996). FEBS Lett., 386, 115 ± 122.Ayer DE and Eisenman RN. (1993). Genes Dev., 7, 2110 ±2119.

Ayer DE, Kretzner L and Eisenman RN. (1993). Cell, 72,211 ± 222.

Ayer DE, Lawrence QA and Eisenman RN. (1995). Cell, 80,767 ± 776.

Bai C, Sen P, Hofmann K, Ma L, Goebl M, Harper JW andElledge SJ. (1996). Cell, 86, 263 ± 274.

Bello-Fernandez C, Packham G and Cleveland JL. (1993).Proc. Natl. Acad. Sci. USA, 90, 7804 ± 7808.

Bernards R. (1995). Curr. Biol., 5, 859 ± 861.Berns K, Hymans M and Bernards R. (1997). Oncogene, inpress.

Blackwell TK, Huang J, Ma A, Kretzner L, Alt FW,Eisenman RN and Weintraub H. (1993). Mol. Cell. Biol.,13, 5216 ± 5224.

Celis JE, Rasmussen HH, Gromur P, Olsen E, Madsen P,Le�ers H, Honore B, Dejgaard K, Vorum H andKristensen DB. (1995). Electrophoresis, 16, 2177 ± 2240.

Clurman BE, Shea� RJ, Thress K, Groudine M and RobertsJM. (1996). Genes Dev., 10, 1979 ± 1990.

Copp AJ and Cockroft DL. (1990). PostimplantationMammalian Embryos A practical approach. OxfordUniversity Press, Oxford.

Davis AC, Wims M, Spotts GD, Hann SR and Bradley A.(1993). Genes Dev., 7, 671 ± 682.

Desbarats L, Gaubatz S and Eilers M. (1996). Genes Dev., 10,447 ± 460.

Eilers M, Picard D, Yamamoto K and Bishop JM. (1989).Nature, 340, 66 ± 68.

Eilers M, Schirm S and Bishop JM. (1991). EMBO J., 10,133 ± 141.

Evan GI, Wyllie AH, Gilbert CS, Littlewood TD, Land H,Brooks M, Waters CM, Penn LZ and Hancock DC.(1992). Cell, 69, 119 ± 128.

Figueiredo Pereira ME, Berg KA and Wilk S. (1994). J.Neurochem., 63, 1578 ± 1581.

Galaktionov K, Chen X and Beach D. (1996). Nature, 382,511 ± 517.

Grandori C, Mac J, Siebelt F, Ayer DE and Eisenman RN.(1996). EMBO J., 15, 4344 ± 4357.

Gu Y, Turek CW and Morgan DO. (1993). Nature, 366,707 ± 710.

Haas K, Staller P, Geisen C, Bartek J, Eilers M and MoÈ roÈ yT. (1997). Oncogene, in press.


2575

Heikkila R, Schwab G, Wickstrom E, Loke SL, Pluznik DH,Watt R and Neckers LM. (1987). Nature, 328, 445 ± 449.

Hengst L and Reed S. (1996). Science, 271, 1861 ± 1864.Henriksson M and LuÈ scher B. (1996). Cancer Research, 68,109 ± 182.

Hurlin PJ, Queva C and Eisenman RN. (1997). Genes Dev.,11, 44 ± 58.

Hurlin PJ, Queva C, Koskinen PJ, Steingrimsson E, AyerDE, Copeland NG, Jenkins NA and Eisenman RN.(1995). EMBO J., 14, 5646 ± 5659.

Jacobsen MD, Weil M and Ra� MC. (1996). J. Cell Biol.,133, 1041 ± 1051.

Jansen-DuÈ rr P, Meichle A, Steiner P, Pagano M, Finke K,Botz J, Wessbecher J, Draetta G and Eilers M. (1993).Proc. Natl. Acad. Sci. USA, 90, 3685 ± 3689.

Kato G, Lee WMF, Chen L and Dang CV. (1992). GenesDev., 6, 81 ± 92.

Kato GJ, Barrett J, Villa-Garcia M and Dang CV. (1990).Mol. Cell. Biol., 10, 5914 ± 5920.

Keath EJ, Caimi PG and Cole MD. (1984). Cell, 39, 339 ±348.

Laherty CD, Yang W-M, Sun J-M, Davie JR, Seto E andEisenman RN. (1997). Cell, 89, 349 ± 356.

Langdon WY, Harris AW, Cory S and Adams JM. (1986).Cell, 47, 11 ± 18.

Lanker S, Valdivieso MH and Wittenberg C. (1996). Science,271, 1597 ± 1601.

Littlewood TD, Hancock DC, Danielian PS, Parker MG andEvan GI. (1995). Nuc. Acid. Res., 23, 1686 ± 1690.

Miltenberger RJ, Sukow KA and Farnham PJ. (1995). Mol.Cell. Biol., 15, 2527 ± 2535.

Morgenstern JP and Land H. (1990). Nucleic Acids Res., 18,1068.

O'Farrell PH. (1975). J. Biol. Chem., 250, 4007 ± 4021.Pagano M, Tam SW, Theodoras AM, Beer Romero P, DelSal G, Chau V, Yew PR, Draetta GF and Rolfe M. (1995).Science, 269, 682 ± 685.

Pear WS, Nolan GP, Scott ML and Baltimore D. (1993).Proc. Natl. Acad. Sci. USA, 90, 8392 ± 8396.

Perez-Roger I, Solomon D, Sewing A and Land H. (1997).Oncogene, in press.

Peukert K, Staller P, Schneider A, Carmichael G, HaÈ nel Fand Eilers M. (1997). EMBO J., in press.

Philipp A, Schneider A, VaÈ srik I, Finke K, Xiong Y, BeachD, Alitalo K and Eilers M. (1994). Mol. Cell. Biol., 14,4032 ± 4043.

Prendergast GC and Zi� EB. (1991). Science, 251, 186 ± 189.Prochownik EV, Kukowska J and Rodgers C. (1988). Mol.

Cell. Biol., 8, 3683 ± 3695.Pusch O, Bernaschek G, Eilers M and HengstschlaÈ ger M.(1997). Oncogene, in press.

Rao G, Alland L, Guida P, Schreiber Agus N, Chen K, ChinL, Rochelle JM, Seldin MF, Skoultchi AI and DePinhoRA. (1996). Oncogene, 12, 1165 ± 1172.

Roussel MF, Ashmun RA, Sherr CJ, Eisenman RN and AyerDE. (1996). Mol. Cell. Biol., 16, 2796 ± 2801.

Roy AL, Carruthers C, Gutjahr T and Roeder RG. (1993).Nature, 365, 359 ± 361.

Rudolph B, Zwicker J, Sa�rich R, Henglein B, MuÈ ller R,Ansorge W, and Eilers M. (1996). EMBO J., 15, 3065 ±3076.

Schreiber Agus N, Chin L, Chen K, Torres R, Rao G, GuidaP, Skoultchi AI and DePinho RA. (1995). Cell, 80, 777 ±786.

Shea� RJ, Groudine M, Gordon M, Roberts JM andClurman BE. (1997). Genes Dev., 11, 1464 ± 1478.

Shrivastava A, Saleque S, Kalpana GV, Artandi S, Go� SPand Calame K. (1993). Science, 262, 1889 ± 1891.

Sommer A, Hilfenhaus S, Menkel A, Kremmer E, Seiser C,Loidl P and LuÈ scher B. (1997). Curr. Biol., 7, 357 ± 365.

Steiner P, Philipp A, Lukas J, Godden-Kent D, Pagano M,Mittnacht S, Bartek J and Eilers M. (1995). EMBO J., 14,4814 ± 4826.

Traenckner EB, Pahl HL, Henkel T, Schmidt KN, Wilk Sand Baeuerle PA. (1995). EMBO J., 14, 2876 ± 2883.

van den Heuvel S and Harlow E. (1993). Science, 262, 2050 ±2053.

Vesely J, Havlicek L, Strnad M, Blow JJ, Donella-Deana A,Pinna L, Letham DS, Kato JY, Deautivaud L, Leclerc Sand Meijer L. (1994). Eur. J. Biochem., 224, 771 ± 786.

Vlach J, Hennecke S, Alevizopoulos K, Conti D and AmatiB. (1996). EMBO J., 15, 6595 ± 6604.

Willems AR, Lanker S, Patton EE, Craig KL, Nason TF,Mathias N, Kobayashi R, Wittenberg C and Tylers M.(1996). Cell, 86, 453 ± 463.

Won KA and Reed SI. (1996). EMBO J., 15, 4182 ± 4193.Zervos AS, Gyuris J and Brent R. (1993). Cell, 72, 223 ± 232.


2576

Documents

Cdk2-dependent phosphorylation of p27 facilitates its Myc-induced release from cyclin E/cdk2 complexes