Cyclin A and Cyclin B Dissociate from ~ 3 4 " ~ " ~ with Half-times of 4 and 15 h, Respectively, Regardless of the Phase of the Cell Cycle*

The exchange rate of cyclin A and cyclin B between endogenous ~ 3 4 " ~ " and bacterially expressed GST-cdc2 was measured in concentrated Xenopus egg extracts. The half-lives of the cyclin A.~34'~' and the cyclin Be~34'~' complexes are estimated as 4 and 15 h, respec-tively. There is no significant difference in these affini- ties when they are measured in mitosis, interphase, or during transition between these two states. The tight association between cyclin and ~ 3 4 ' ~ ' in the cell cycle may be significant for mechanisms of cyclin destruction. Cell cycle transitions depend on the activity of members of the cyclin-dependent kinase (cdk) family (1-5). These enzymes are composed of a protein kinase subunit encoded by the cdc2 gene or one of its close relatives (e.g. cdk2-cdk6 (6)) in combination with a cyclin subunit. The protein kinase activity of these heterodimeric complexes seems to require that the cdk subunit should be phosphorylated on threonine 161 (in ~ 3 4 ' ~ ' ) or threonine 160 (in p3Fdk2), and not be phosphorylated on tyrosine 15 or threonine 14 (for reviews, Norbury and Nurse, (7) and Solomon, (8)). The cyclin (K33R) of GST-cdc2 and GSTedkZ were constructed from GST- cdc2 and GST-cdk2 cDNAs by the polymerase chain reaction-based method essentially as described by Horton and Pease (31) using muta- genic oligonucleotides and their complements, for GST-cdk2. Expression and Purification GST-cdc2, GST-cdc2K33R, and GSTcdk2K33R"TO prepare proteins in Escherichia BL21(DE3) containing the appropriate plasmid was grown to an A,,, of about 1.0 at 37 "C. Synthesis of the recombinant proteins was induced by the addition of 100 p~ isopropyl-l-thio-fh-galactopyrano-side, followed by incubation for 16 h at 23 "C. The cells were harvested by centrifugation and suspended in 50 mM Tris-HC1, pH 7.5, mM EDTA, 1 m~ Dm, 2 pghl aprotinin, 15 pg/ml benzamidine, 1 pg/ml leupeptin, 0.25 mM phenylmethylsulfonyl fluoride, and 2 mg/ml ly- sozyme and left at 4 "C for 15 min. All subsequent steps were carried out at 4 "C. The cells were lysed by sonication and the lysate was centri- fuged at 18,000 x g for 30 min. The supernatant was then filtered through a 0.45-p filter (Millipore) and applied onto a 0.5-ml of GSH-Sepharose (Pharmacia Biotech Inc.), equilibrated with phosphate-buffered (170 m~ NaCI, 3 mM KCl, 10 m~ Na,HPO,, 2 m~ KH,PO,) supplemented with 0.25 M KCl, 0.1% Tween 20, 1 mM Dl", 0.25 mM phenylmethylsulfonyl

type cyclins to the nucleus, B-types t o the cytoplasm (10)(11)(12)(13) and can also control the timing of activation of the complex (14,15). Coordination of the processes of the cell cycle by the transient association of cdks with different cyclin partners is particularly well illustrated by the case of Saccharomyces cereuisiae, which contains only one known cdk, CDC28. This kinase subunit is found in combination with Clnl, Cln2, or Cln3 in late G, to S-phase, with Clb3, -4, -5, and -6 from the start of S-phase onwards, and with Clbl and-2 during late G, and "phase (5). It is clearly important for cell cycle progression that the correct form of CDC28 kinase be present and active at the correct time.
For example, cells lacking the CLN gene products are blocked at Start, and cannot be rescued by mitotic cyclins (4) and cells deficient in Clb5 and -6 show very slow passage through Sphase (16). The molecular basis for these observations is not yet clear, and while it is highly probable that different forms of the kinase phosphorylate different substrates, very few of these targets have yet been identified (see Nigg (17)). If passage through the cell cycle is controlled by the activation of cdks by different cyclins at distinct points, then it must be possible to replace one set of cyclins by another at the appropriate time in the cell cycle. In general, this seems to be achieved by the synthesis or selective destruction of cyclins at the appropriate time, rather than by competition between cyclins for kinase subunits. The ~3 4 "~" ' protein is normally very stable throughout the cell cycle (18), but cyclins undergo changes in concentration (19)(20)(21). In clam embryos undergoing rapid cleavage, for example, both Aand B-type cyclins are stable except for a brief interval that starts a minute or two before the metaphase-anaphase transition and ends abruptly 5 min later (22).
As different cyclin proteins which can bind to the same cdk subunit are present in the cell at the same time (141, and the cyclin partner in the complex probably plays an important role in determining which substrates the kinase phosphorylates and the subcellular localization of the complex, the stability of t h e ~y c l i n~p 3 4 '~"~ complex under physiological conditions is clearly of interest. This problem has not previously been investigated. In the course of experiments designed to test whether cyclin A bound to ~3 4 '~" ' was degraded at the same time and at the same rate as cyclin A bound t o a kinase-deficient mutant of p34cdc2K33R,' it became clear that it was relevant to measure the rate of exchange of cyclin A off ~34"~"'. At the same time we established a procedure t o measure the half-life of ~yclin~p34'~~' complexes in Xenopus egg extracts. The use of such extracts allows the measurements to take place under near physiological conditions while being relatively easy to perform. In addition, Xenopus egg extracts, which are arrested in metaphase, can be triggered to exit from "phase and enter interphase by the addition of Ca2+ (23-271, allowing the half-life of cy~lin.p34"~"~ complexes to be measured at different stages of the cell cycle using identical methodology.
In this paper, we present evidence that once combined, cyclin A and cyclin B remain bound to p34'dc2 for several hours; the half-lives of these complexes are long compared to the length of the cell cycle. Cyclin B seems to be somewhat more tightly bound than cyclin A. We also find that there is no significant difference in these affinities when they are measured in mitosis, interphase, or the transition between these two stages.

MATERIALS AND METHODS
Constructs-The Xenopus cyclin A1 and cyclin B2 cDNA clones in pGEM (Promega) vectors were previously described by Minshull et al. (28). The destruction box mutant of cyclin A in which the sequence mutants (K33R) of GST-cdc2 and GSTedkZ were constructed from GST-cdc2 and GST-cdk2 cDNAs by the polymerase chain reaction-based method essentially as described by Horton and Pease (31) using mutagenic oligonucleotides and their complements, 5"GTTGCAATGLG for GST-cdk2.
Expression and Purification of GST-cdc2, GST-cdc2K33R, GST-cdk2, and GSTcdk2K33R"TO prepare proteins in bacteria, Escherichia coli strain BL21(DE3) containing the appropriate plasmid was grown to an A,,, of about 1.0 at 37 "C. Synthesis of the recombinant proteins was induced by the addition of 100 p~ isopropyl-l-thio-fh-galactopyranoside, followed by incubation for 16 h a t 23 "C. The cells were harvested by centrifugation and suspended in 50 m M Tris-HC1, pH 7.5, 2 m M EDTA, 1 m~ D m , 2 p g h l aprotinin, 15 pg/ml benzamidine, 1 pg/ml leupeptin, 0.25 m M phenylmethylsulfonyl fluoride, and 2 mg/ml lysozyme and left a t 4 "C for 15 min. All subsequent steps were carried out at 4 "C. The cells were lysed by sonication and the lysate was centrifuged at 18,000 x g for 30 min. The supernatant was then filtered through a 0 . 4 5 -p filter (Millipore) and applied onto a 0.5-ml column of GSH-Sepharose (Pharmacia Biotech Inc.), equilibrated with phosphate- Immunoblotting-Blotting of polyacrylamide gels was carried out using a Hoefer semi-dry blotting apparatus according to Harlow and Lane (32). Cyclin A was detected using a monoclonal anti-Xenopus cy-clinA antibody XLAl-3, isolated in this laboratory by Dolores Harrison. This antibody recognizes an epitope lying between residues 88 and 106 of Xenopus cyclin AI. Monoclonal antibody A17, a giR from Dr. Julian Gannon, was used to detect ~3 4 '~. Both primary antibodies were detected with rabbit anti-mouse horseradish peroxidase-conjugated second antibody (DAKO) and the Amersham enhanced chemiluminescent system, ECL (Amersham, United Kingdom). The absorbances of the bands on the films were quantified by scanning densitometry. For quantitative studies, a Sau3A fragment of a cyclin AI cDNA clone, encoding residues 56-419, was subcloned into the BamHI site of pET3a and expressed in BL21(DE3). The insoluble protein was harvested and its concentration determined by comparison with bovine serum albumin standards. This cyclin A standard, which has a molecular weight of 42,136, was used to construct the standard curve shown in Fig. 1. A similar approach, usingxenopus ~3 4 '~'~ expressed in bacteria, was used to quantitate ~3 4 '~'~.
Gel Filtration-Xenopus egg extracts were incubated for 2 h at 23 "C with or without added cyclin A mRNA, in the presence of 30 pCi of [36S]methionine to monitor cyclin synthesis. A parallel reaction was set up to which no exogenous mRNA was added. The samples were clarified by centrifugation in a microcentrifuge at 4 "C for 30 min. Tkanslation in Xenopus Egg Extracts-Xenopus egg extracts, known as cytostatic factor arrested extracts, were prepared as described by Murray (33), frozen in liquid nitrogen, and used for the M-phase assays. Interphase extracts were prepared by incubating the stored egg extracts with 0.4 m M CaC1, for 30 min to inactivate maturation promoting factor (23)(24)(25)(26)(27). The assay for the metaphase-anaphase transition was carried out in cytostatic factor-arrested extract in which cyclin destruction was triggered by the addition of 0.4 m M CaCl,. These Xenopus egg extracts were supplemented with 0.1 volume of rabbit reticulocyte -GAAAATTCGA-3' for GST-cdc2 and GTGGCGCWABAAAATCCGC -70 "C. lysate in order to boost the translation of added mRNA. Capped cyclin mRNA was prepared using T7 RNA polymerase as described previously (29). The cyclin mRNA (100 ng/pl) was translated in the Xenopus egg extracts in the presence of [35Slmethionine at 23 "C for 80 min and cycloheximide (100 J~M) was added to stop the translation.
In parallel reactions, GST-cdc2 protein (800 IIM; 50 ng/pl) was added either at the start of the translation, or after 80 min of translation, along with the cycloheximide (see Fig. 4). After translation was arrested by the cycloheximide, both extracts were incubated for a further 80 min at 23 "C. In most cases, no exogenous cyclin B mRNA was added, but labeled cyclin B was made from the endogenous cyclin mRNA pool.
GSH-Sepharose Affinity Chromatography-To assess the amount of labeled cyclin associated with GST-cdc2 protein, 10-pl samples of the reactions were taken at various times, diluted with 400 pl of buffer B (50 n" Tris-C1, pH 7.4, 5 n" NaF, 250 m M NaC1, 5 m~ EDTA, 5 m M EGTA, 0.1% (v/v) Nonidet P-40, 1 pg/ml leupeptin, 2 pg/ml aprotinin, 10 pg/ml soybean trypsin inhibitor, 15 pg/ml benzamidine), mixed with 15-20 pl of glutathione-Sepharose (Pharmacia), and rotated for 1 h at 4 "C. The beads were harvested by centrifugation and washed three times with buffer B. The bead-bound proteins were eluted with SDS sample buffer and analyzed by SDS-PAGE followed by autoradiography. The radioactive cyclin bands were quantified by scanning densitometry of an autoradiogram made on Hyperfilm (Amersham) or with a PhosphorImager (Molecular Dynamics).

Concentrations of Cyclins and ~3 4~' '
in Xenopus Egg Extracts-In order to determine the degree of saturation of endogenous ~3 4~~' with endogenous cyclins, the concentrations of cyclins and ~3 4~' in cell-free extracts made from Xenopus eggs were measured. It was also necessary to establish how much additional cyclin synthesis occurred when exogenous cyclin mRNA was added to the extracts. A bacterially-synthesized protein that had been used to immunize mice was used as a standard for quantitative immunoblotting. This protein lacks the N-terminal 56 residues of cyclin Al, and has a molecular mass of 42.1 kDa. The monoclonal antibody XLA1-3 reacts with an epitope lying between residues 88 and 106 of Xenopus ~y c l i n A 1 .~ Fig. 1, lane 10, shows that this antibody could detect as little as 0.05 ng of cyclin Al. Endogenous cyclin A was not detectable at this exposure in 0.4 p1 ofXenopus egg extract (Fig.   1, lane 11 1, although it could be recognized in very long exposures of the ECL immunoblot (not shown). Using a different antibody and detection system we previously estimated the concentration of cyclin A1 in whole Xenopus eggs (as opposed to cell-free extracts) as being 0.5 nM (34). After translation of added cyclin A1 mRNA in the egg extract for 2 h, however, cyclin A1 was easily detectable, at about 2 n~ (Fig. 1, lanes 12 and 13). Addition of rabbit reticulocyte lysate to 10% (lanes 14 and 25) or 50% (lanes 16 and 17) by volume considerably increased the translation of added cyclin A1 mRNA, giving 13 n~ and 58 n~ cyclin Al, respectively, in the experiment shown in Fig. 1, and attaining a maximum of 75 n~ in some experiments where equal mixtures of egg and reticulocyte extracts were used. There was considerable batch-to-batch variation between extracts prepared in parallel from the eggs of different frogs; Table I gives the range of values determined in the same way for four such extracts, with extract 1 being shown in Fig. 1.
The concentration of ~34'~'' in Xenopus egg extracts was measured in essentially the same way, and was found to range from 470 to 600 nM in three different egg extracts (Table 111, a slightly lower value than our previous estimates (34). To confirm that the egg extracts contained free, uncomplexed ~34'~"', as was originally shown by Draetta and Beach (35) in human cells and as would be expected in the egg extracts if the total cyclin concentration (including both A and B types) is no more than 5-10 nM (about 2 n~ for cyclin B1 and 3 nM for cyclin B2 according to Kobayashi Table ILA, lunes 1-10, bacterial cyclin A standards; lune 11, pure egg extract, no added mRNA lunes 12-17, 100 ng/pl added cyclin A1 mRNA, lunes 12 and 13, no added reticulocyte lysate; lunes 14 and 15, 10% by volume added reticulocyte lysate; lanes 16 and 17, 50% added reticulocyte lysate. In lunes 11-14 and 16,0.4 pl of extract were loaded, lunes 15 and 17 were loaded with 0.2 pl of extract. Note that the bacterial standard was missing 56 residues from the N terminus and migrates faster than "real" cyclin Al. Immunoreactive bands were detected by using the enhanced chemiluminescence (ECL) system from Amersham. B , the absorbances of the bands on the film from A were determined by scanning densitometry and plotted to give a calibration curve, which was used to convert the values of the absorbances of the bands in lunes [12][13][14][15][16][17] to the concentrations of cyclin A listed in Table I  (extract 1). extract in which a cyclin A mRNA encoding a cyclin that was unable to bind to ~34'~'' had been translated (see "Materials and Methods") was analyzed by gel filtration and the position of the ~3 4 '~'~ determined by immunoblotting with a n a n t i -~3 4~~'~ monoclonal antibody (A17). As shown in Fig. 2, the majority of ~3 4 '~'~ (at least 80%) showed the expected behavior for a protein of 34 kDa, with only a small fraction migrating at higher apparent molecular masses. There was no very obvious difference between the behavior of ~3 4~' ' in the extract in which the mRNA encoding a cyclin A protein that could not bind to ~3 4 '~'~ had been translated, and extracts in which wild-type cyclin A mRNA or no exogenous mRNA had been translated (data not shown). This is a reflection of the small fraction of the large pool of endogenous monomeric ~3 4 '~'~ inXenopus egg extracts which is normally occupied by newly-made cyclins. This pool has an almost 10-fold overcapacity for binding newly synthesized mitotic cyclins.
GSTcdc2 Binds Cyclins A and B, but GSTcdk2 Binds only Cyclin A-We next tested the ability of bacterially-synthesized GST-cdc2 and GST-cdk2 proteins to bind [35S]methioninelabeled cyclins A and B produced by cell-free translation in Xenopus egg extracts. Exogenous cyclin A mRNA was added in order to increase the amount of labeled cyclin A. Either GST-cdc2 or GST-cdk2 protein were added at a final concentration of 800 n~ to parallel reactions at zero time, and further translation was inhibited by the addition of cycloheximide after 80 min at 23 "C. The GST-cdc2 or GST-cdk2 was recovered by affinity chromatography on GSH-Sepharose (see "Materials and Meth- is monomeric in Xenopus egg extracts. Fifty microliters of a translation reaction were analyzed on a Superdex 200 FPLC column (see "Materials and Methods") and fractions analyzed by SDS-PAGE and immunoblotting with monoclonal antibody A17 to detect ~3 4~" .
Immunoreactive bands were detected using the ECL system. The elution position of size markers run in parallel are indicated.
ods"), and the bound proteins analyzed by SDS-PAGE and autoradiography. If no GST-cdc2 or GST-cdk2 were added, or if blank Sepharose beads were used, essentially no radioactively labeled proteins were recovered by this procedure (data not shown; but see Fig. 5, lane 11). The added protein did not significantly affect the translation of endogenous or added mRNA (compare lanes 1 and 3 of Fig. 5A). Fig. 3 Fig. 3, lane 3, shows, both cyclins A and B were recovered on GSH-Sepharose beads when GST-cdc2 was added to the reactions. Approximately 25% of the input labeled cyclin A was recovered on the GSH-Sepharose when 800 n M GST-cdc2 was added to the extract. In contrast, although GST-cdk2 bound cyclin A as efficiently as GST-cdc2, the recovery of cyclin B on GST-cdk2 was extremely low (Fig. 3, lane 4).
We also tested the ability of inactive mutant versions of ~3 4 '~~' and ~33"'~', in which the conserved lysine 33 residue was mutated to arginine (K33R), to bind cyclins. Under the conditions used here, the inactive mutants, GST-cdc2K33R and GST-cdk2K33R, were equally efficient as "traps" for cyclins when compared to their wild-type counterparts (data not shown).
Experimental Design to Measure Rates of Dissociation of Cyclin~dc2 Complexes-The foregoing results indicated that endogenous ~3 4~~~' or added GST-cdc2 can both bind cyclins efficiently. It should therefore be possible to measure the rate of dissociation of cyclins from ~34'~'' under physiological conditions by measuring the rate at which radioactively labeled cyclins that have first been allowed to form complexes with endogenous ~3 4~' during a translation reaction can subsequently by "trapped" by GST-cdc2 that is added after translation is complete. Fig. 4 shows the design of the experiment. Xenopus egg extracts were incubated with [35Slmethionine for 80 min to allow the synthesis of radioactive cyclins, which rapidly combined with the large pool of endogenous ~34'~''. Further translation was then stopped by the addition of cycloheximide. In the "exchange" experiment, bacterially synthesized GST-cdc2 protein (800 I~M final concentration) was added to the extract at this point. At intervals thereafter, samples were removed for affinity chromatography on GSH-Sepharose. Under these conditions, labeled cyclins that dissociate from pre-existing cyclin.cdc2 complexes have a chance of binding to the added GST-cdc2, in which case they can be recovered by affinity chromatography on GSH-Sepharose. The amount of labeled cyclin harvested on the GSH-Sepharose was determined by SDS-PAGE and autoradiography. To control for the amount of correctly-folded GST-cdc2 added to the extracts, and to ensure that labeled cyclins could bind efficiently to the added GST-cdc2 in competition with the endogenous p34cdc2, a "control" incubation was carried out in which the same concentration (800 nM) of GST-cdc2 was added at the beginning of the translation, rather than at the end. The amount of labeled cyclin that was recovered on GSH-Sepharose when the GST-cdc2 was present during the period of labeled cyclin synthesis was measured at the end of the 80-min translation, and again after a further 80 min in the presence of cycloheximide. The level of radioactive cyclin bound to the GSH beads under these (A) Control -GST-cdc2 added at start  conditions represented the maximum amount that could possibly bind in the exchange reaction when GST-cdc2 was added after the cycloheximide, if the cyclin were to completely equilibrate between the endogenous ~34'~'' and the added GST-cdc2. The exchange of cyclin between endogenous ~34'~'' and the added GST-cdc2 was expressed as the ratio of the radioactivity in cyclin recovered on GSH-Sepharose a t time t to the amount recovered when the GST-cdc2 was added at the start of translation. This method of expressing the data corrects for variations in both the amount of GST-cdc2 added, and for possible batch to batch variability in the proportion of correctly folded protein.
Cyclin B.~34'~'' and Cyclin A*p34cd"2 Form Stable Complexes-In the control incubation, in which GST-cdc2 was present from the start of the incubation, the amount of labeled cyclins A and B recovered by GST-cdc2 was the same a t the end of the incubation as at the time of cycloheximide addition (Fig.  5A, lanes 9 and 10). In contrast, in the exchange experiment, when GST-cdc2 was added at the end of the translation, essentially no labeled cyclins were retained on GSH-Sepharose when a sample was taken immediately after the addition of the protein to the translation reaction (Fig. 5, lane 11 ). All the components, labeled and unlabeled, were present at the same concentrations in lanes 9 and 10 and 11-16; only the time of addition of the GST-cdc2 was different. The amount of cyclin recovered by the GSH-Sepharose gradually increased with time in the case where the GST-cdc2 was added at the end of the translation reaction, so that by 80 min, just over 10% of the recoverable cyclin A was bound by the beads and just under 5% of the cyclin B. The control incubation shows that the GST-cdc2 was able to compete effectively with the endogenous ~34'~'' for cyclin binding. The added GST-cdc2 was able to bind cyclins without delay after addition to the extract in which translation  Fig. 4; radiolabeled cyclins were analyzed by SDS-PAGE followed by autoradiography. A, lunes 1-8 show samples from the two translation reactions; lunes 1 and 2, control; lunes 3-9, exchange reaction. Lunes 9-16 show GSH-Sepharose-bound material. Lanes 9 and 10, control reaction; lunes 11-16, exchange reaction. The position of cyclin A1 is indicated by an arrow, and the B-type cyclins by an asterisk. Times of incubation refer to time after addition of cycloheximide. B , the halflives of the cyclin.cdc2 complexes. The amount of GSH-Sepharosebound labeled cyclin, shown in the autoradiogram in A was quantified using a PhosphorImager and plotted as described in the text. C, graph B is shown with expanded scale. Error burs show the range and mean of the values obtained in three independent experiments using different A (0) is compared with the mean values for the B-type cyclins (0). egg extracts. The mean values from these three experiments for cyclin was proceeding (i.e. it did not require significant time for correct refolding in the Xenopus extract; data not shown). There are at least two possible explanations for the finding that the recovery of cyclins on GST-cdc2 is a slow, time-dependent reaction. We favor the explanation that the rate of cyclin bind-  ing to added GST-cdc2 is a measure of the rate of dissociation of pre-formed cy~lin.p34'~'~ complexes. An alternative possibility is that these extracts contain a certain amount of cyclin that is not bound to ~3 4 '~'~ (it might have folded incorrectly, for example) and that the gradually increasing recovery represents the rate of release from this pool. In this case, however, one would have to conclude that the rate of dissociation of pre-formed ~yclin.p34'~'~ complexes is even lower than that proposed in the following section.  Fig. 5, B and C , shows that up to 80 min, the data fit a reasonable straight line as expected, and that cyclin B .~3 4 '~'~ complexes appear to be significantly more stable than cyclin A.p34cdc2 complexes. The half-lives of both types of complex were long: cyclin complexes having an apparent halflife of about 4 h and that of the cyclin B.~34'*~ complex was estimated to be about 15 h. A similar experiment using the kinase-deficient mutant GST-cdc2K33R, instead of GST-cdc2, was carried out. There was no significant difference in the measured rate of dissociation of the ~yclinlp34'~'~ complexes using this reagent as the trap (data not shown).

The Half-life of the Cy~lin-p34'~'~ Complex-To
As a check of our methodology, we varied the amount of added GST-cdc2; in one experiment GST-cdc2 was added at a final concentration of either 160 or 800 nM. The fraction of recovered cyclin in the exchange reaction compared to the control incubation were very similar in both cases, as shown in Table 111 Ca2+ (0.4 mM) was added toXenopus egg extract at zero time, and I"S1methionine and cyclin B2 mRNA were added a t 30 min. Cycloheximide was added a t 90 min, and samples were taken at the indicated times for histone H I kinase assay. Lane 1, before Ca2+ addition; lune 2, before addition of ["'Slmethionine and mRNA (30 rnin); lune 3,90 min; lune 4, 100 min; lune 5, 120 min; lune 6, 170 min; lune 7, control extract to which no additions were made was sampled at 170 min. In lunes 1, 2, and 7 only R'PO, radioactivity was present, whereas the bands seen in lunes 3-6 are due to both 35S and "PO4, except that very little histone H1 kinase activity was present in these lanes. dition of 0.4 mM CaCl, to Xenopus egg extracts to trigger cyclin destruction and the loss of kinase activity associated with p3edC2, as shown in Fig. 6B, comparing lanes 1 and 2. In order to avoid the problem that translation of added cyclin mRNAcan "drive" extracts back into mitosis (37), cyclin B2 mRNA was translated in these experiments rather than cyclin A mRNA. Cyclin B-dependent activation of ~3 4 " "~ kinase has a long lag phase, in contrast to cyclin A (36).'. Cyclin B2 mRNA was well translated in these extracts (Fig. 6 A , lanes 3 4 , and histone H1 kinase activity did not reappear during the incubation. We have found it unusual for Ca2+-treated egg extracts to re-enter mitosis under these conditions. Fig. 7, A and B , shows that the dissociation rate of cyclin B 2 .~3 4 '~'~ i n interphase is the same as that in "phase (tH = 15 h). Thus, within the limits of accuracy of these experiments, there is no change in the rate of dissociation of the cyclin B2.~34'~'' complex in interphase compared to metaphase, and the complexes are long-lived compared with the length of the cell cycle in the early cleavage stages of Xenopus embryos.
The dissociation rate of cyclin.cdc2 complexes during the metaphase to anaphase transition was also investigated. Cyclins A and B are rapidly proteolyzed at this point in the cell cycle, with half-lives of about 5-10 min. This destruction of cyclins made it impossible to measure exchange by the method we used for the other types of extracts. We therefore employed an "indestructible" mutant of cyclin A, in which two residues (Arg4' and Leu44) in the destruction box were altered to alanine; RTVL + ATVA. This mutant cyclin is not degraded when the destruction machinery is activated (29). The mRNA for ATVAcyclin A was translated in egg extract for 80 min as before, with or without added GST-cdc2. Cycloheximide (100 PM) and 0.4 mM CaCl, were then added to the extracts, together with GST-cdc2 in the case of the reaction previously lacking it. Fig. 8A shows that the endogenous B-type cyclins disappeared between 20 and 40 min after addition of Ca2+, but that the mutant cyclin A survived (lanes 1-81. The rate of binding of the mutant cyclin A to the GSH-Sepharose in the exchange reaction occurred at a very similar rate to the previous experiment. There was no evidence for any increased rate of dissociation of the cyclin A-p34'd'2 complex during the period of rapid cyclin destruction. DISCUSSION In this paper, we show that GST-cdc2 acts as an efficient trap for both cyclins A and B when it is present during their synthesis in extracts ofXenopus eggs. In contrast, GST-cdk2 binds only cyclin A to a significant extent in these extracts. When GST-cdc2 is added to an extract containing radioactively labeled cyclins which are already bound to endogenous ~3 4 "~'~, however, the rate of binding of the cyclins to added tagged cdc2 protein is very slow, approaching equilibrium with a time constant of 4 h for cyclin A and 15 h for cyclin B. We interpret the slow binding that occurs under these conditions as being a measure of the rate of dissociation of complexes between cyclin and endogenous ~34'~'', as it was likely that all of the newly translated cyclins were bound to endogenous ~3 4 '~" i n these experiments, given that there is a 40-fold molar excess of ~34'~'' over cyclins in Xenopus egg extracts. It is conceivable, however, that uncomplexed cyclin was present in these extracts. If this were the case, then the recovery of labeled cyclin on GSH-Sepharose could be due to the gradual binding of this uncomplexed cyclin to the added GST-cdc2. This would mean, however, that the rate of dissociation of pre-formed ~yclin.p34'~'~ complexes is even lower than estimated here. Added GST-cdc2 can rapidly bind to cyclins A and B in egg extract, and these complexes have the expected histone H1 kinase activity (not shown). Nevertheless, we have not demonstrated that bacterially expressed GST-cdc2 can bind to cyclins A and B in an manner that is completely equivalent to the endogenous ~34'~'' in Xenopus egg extracts, which is a difficult point to establish definitively. In the experiments carried out in this paper, however, the exchanges of cyclins onto GST-cdc2 from endogenous ~3 4~~~ were always expressed as a proportion of the amount of the particular cyclin that could bind to GST-cdc2 when i t was competing directly with the endogenous ~34'~'' in the control reaction. Thus, even if the binding of cyclins to GST-cdc2 is not quite as tight or as rapid as the binding of cyclins to the endogenous ~34'~'', we believe that this does not affect the results of the experiments described in this paper.
The tight association between cyclins and ~34'~'' means that each ~yclin.p34"~'~ complex is effectively insulated from all others, making i t possible to regulate the activity and the timing of activation of each kind of ~3 4 '~'~ kinase independently (38). An alternative possibility, that cyclins would activate ~3 4 '~~' and then the active ~34'~'' was able to change its cyclin partner, is not supported by the evidence presented here. The high stability of ~yclin.p34'~'~ complexes also implies that the interaction between the two components is highly energetically favorable, probably involving numerous noncovalent bonds. Unfortunately, little is known about the interacting surfaces of the proteins and the residues involved in the interaction, a state of affairs that will probably persist until the structure of the ~yclin.p34'~~' holoenzyme has been established (see Marcote et al. (39)).
The tight association between cyclin and ~3 4 " "~ also has important consequences for the mechanism of cyclin destruction. At the metaphase-anaphase transition, maturation promoting factor is rapidly inactivated by cyclin destruction, and if it is not, cell cycle progression cannot occur properly (24,40). The concentration of ~34'~'' protein remains roughly constant in normal, rapidly growing cells (18), whereas the levels of cyclins A and B undergo characteristic oscillations (41). Since we show here that cyclin and ~3 4~~" are bound tightly together even in cell cycle extracts that are undergoing the metaphase to anaphase transition, i t would appear that the only way to dissociate cyclin.cdc2 complexes is by proteolysis of the cyclin subunit. As the binding of cyclin to ~3 4 " "~ is required for cyclin proteolysis at the metaphase-anaphase transition (27), it is likely that the cyclin.cdc2 complex is the minimum substrate recognized by the cyclin-specific protease at the metaphaseanaphase transition.
Members of the cyclin family share a region of homology known as the cyclin box, which probably represents the core of the domain responsible for interaction with ~3 4 '~'~ and related kinases (29). This region alone is not sufficient for binding, however; in the case of mitotic cyclins, residues at the extreme C terminus are also required for stable binding to ~34'~''. In contrast, the N terminus of the protein is not required for binding to p34""' (13,29,42). If the only method of dissociation of cyclins from ~34'~'' during the metaphase-anaphase transition is that of proteolytic cleavage of the cyclin subunit, then the organization of the cyclin protein suggests that the best way to achieve this dissociation would be cleavage first at the C terminus of the cyclin protein. The cyclin would then dissociate from the complex as it could no longer bind to ~34'~''. First cleavage at the N terminus of the cyclin protein, on the other hand, could be disastrous for the cell, since this produces forms of cyclin that are resistant to further degradation (24,25,29,43), but still able to bind to and activate ~3 4 '~'~ (13,29,42). This raises an interesting paradox, since the destruction box, the region of mitotic cyclins required for their proteolysis, is located in the N terminus of the protein. It is of course possible that destruction occurs by a processive mechanism, which would avoid this problem.
If the experiments described in this paper do indeed measure the dissociation of cyclin from endogenous ~3 4 '~'~, they clearly indicate that cyclin B is more tightly bound to ~3 4 '~~' than is cyclin A; the dissociation of cyclin A from ~3 4 '~'~ occurs approximately three times faster than that of cyclin B1 and B2 from ~3 4 " "~. As we show here, in these concentrated egg extracts the B-type cyclins bind to a significant extent only to ~3 4 '~'~, whereas under the same conditions, cyclin A can bind equally well to both ~34'~'' and to ~3 3 " '~~. As cyclin A can bind to two alternative partners, the binding of cyclin A to ~3 4 '~'~ may involve certain compromises with respect to maximizing its cdk binding energy that the B-type cyclins do not have to make. This could lead to the slightly less tight binding of cyclin A to ~34'~'' than B-type cyclins. It would be interesting to measure the stability of cyclin E-p3Fdk2 and cyclin D-cdk complexes, as cyclin E appears to be a dedicated partner for ~33"'~' in the same way that the B-type cyclins are specialized for binding to ~34'~'' (44)(45)(46), whereas D-type cyclins are reported to be able to bind to a number of different cdk subunits (47,481.