Oncogene (1997) 15, 749 ± 758  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Ultraviolet light-induced G2 phase cell cycle checkpoint blocks cdc25-dependent progression into mitosis

BG Gabrielli, JM Clark, AK McCormack and KAO Ellem

Queensland Cancer Fund Research Laboratory, Joint Oncology Program, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Herston, Queensland 4029, Australia

In response to low doses of ultraviolet (U.V.) radiation, kinase), and is essential for the catalytic activity of cells undergo a G2 delay. In this study we have shown cdc2, whereas phosphorylation of Thr14 and Tyr15 is that the G2 delay results in the accumulation of inactive inhibitory, and catalysed by a member of the wee1/ forms of cyclin B1/cdc2 and both the G2 and mitotic mik1 family of kinases (Morgan, 1995). The latently complexes of cyclin A/cdk. This appears to be through a active form of cyclin B/cdc2 is maintained in the block in the cdc25-dependent activation of these cytoplasm until prophase, when the majority is complexes. The expression and localisation of cyclin A translocated into the nucleus (Pines and Hunter, and cyclin B1/cdk complexes are similar in U.V.-induced 1991; Bailly et al., 1992). At the same time, the G2 delay and normal early G2 phase cells. Cdc25B and inhibitory Thr14 and Tyr15 phosphates are removed cdc25C also accumulate to normal G2 levels in U.V. by a member of the cdc25 family of dual speci®city irradiated cells, but the mitotic phosphorylation asso- phosphatases, and fully active cyclin B/cdc2 is able to ciated with increased activity of both cdc25B and cdc25C perform its many mitotic functions. is absent. The cdc25B accumulates in the nucleus of U.V. In yeast only a single form of cdc25 exists, but in irradiated cells and in normal G2 phase cells. Thus the mammalian systems three isoforms have been identi®ed block in cyclin B/cdc2 activation is in part due to the (Gabrielli, 1994). Cdc25A has been demonstrated to physical separation of cyclin B/cdc2, localised in the regulate progression into S phase, while both cdc25B cytoplasm, from the cdc25B and cdc25C phosphatases and cdc25C have roles in the G2/M transition (Millar localised in the nucleus. The data positions the U.V.- et al., 1991; Jinno et al., 1994; Gabrielli et al., 1996). induced G2 checkpoint at either the S/G2 transition or The activity of cdc25 is regulated at both a early G2 phase, prior to the activation of cyclin A/cdk2. translational and post-translational level. Cdc25C is expressed at a constant level throughout the cell cycle, Keywords: cdc25; cell cycle checkpoint; U.V. while the expression of cdc25A and cdc25B has been demonstrated to be cell cycle dependent in some cell lines (Jinno et al., 1994; Gabrielli et al., 1996). The activity of both cdc25A and cdc25C is further Introduction modulated by phosphorylation (Ho€man et al., 1993; Ho€mann et al., 1994). Progression through the cell cycle is regulated by the A further level of regulation of the cyclin/cdks has ordered activation of cyclin dependent kinases (cdks). been identi®ed, through the binding of low molecular In yeast, a single cdk, cdc2/CDC28, associates with weight protein inhibitors of these complexes. Two multiple cyclin partners to regulate both the START families of these proteins have been identi®ed, the and G2/M checkpoints, whereas in higher eukaryotes, p21Cip1 family including p21Cip1,p27Kip1 and p57Kip2, and a large family of cdks have so far been identi®ed, each the p16CDKN2A family including p16CDKN2A,p15CDKN2B, of which appears to regulate discrete cell cycle p18CDKN2C and p19CDKN2D (Sherr and Roberts, 1995). functions (Meyerson et al., 1992; van den Heuvel and The members of the p21 family bind to a wide range of Harlow, 1993). The best characterised in terms of cyclin/cdks, whereas the p16 family proteins have a function and regulation is cyclin B/cdc2, the prototypic much more restricted speci®city, binding only to cyclin cyclin/cdk. This complex was ®rst puri®ed from D/cdk4 and cdk6 (Sherr and Roberts, 1995). These Xenopus eggs as maturation promoting factor (MPF), proteins appear to respond to extracellular in¯uences which was the only factor required to initiate mitosis to produce cell cycle delays, for example, increased when injected into prophase arrested frog oocytes expression of p21 is induced by exposure to ionising (Lohka et al., 1988; Gautier et al., 1988). The histone radiation and is involved in a G1 delay (Dulic et al., kinase activity of cyclin B/cdc2 is regulated through a 1994), while p16 induced by irradiation with ultraviolet complex series of protein associations, phosphoryla- (U.V.) light is correlated with a G2 delay (Wang et al., tions and dephosphorylations. Cyclin B is synthesised 1996). during S and G2 phases, and immediately binds to the Exposure of cells to low doses of U.V. light catalytically inactive cdc2 monomer in the cytoplasm. produces delays in G1, S and G2/M phases of the Upon association with cyclin B, the cdc2 subunit cell cycle, depending on the dose and cell type used (Lu undergoes a series of phosphorylations. The phosphor- and Lane, 1993; Wang and Ellem, 1994; Barker et al., ylation of Thr 161 is catalysed by CAK (cdk activating 1995; Petrocelli et al., 1996; Poon et al., 1996). A role for p53-induced expression of p21, and the redistribu- Correspondence: B G Gabrielli tion of p27 among G1 and S phase cyclin/cdks in U.V.- Received 21 November 1996; revised 6 May 1997; accepted 6 May induced G1 and possibly S phase delays has been 1997 demonstrated (Lu and Lane, 1993; Barker et al., 1995; U.V. induced G2 delay BG Gabrielli et al 750 Poon et al., 1995; Petrocelli et al., 1996). The inhibitory was complexed with cyclin B1 to the same extent as in tyrosine phosphorylation of cdk4 has also been normal S/G2 cells, and that the cdc2 was predomi- implicated in the U.V.-induced G1 delay (Terada et nantly the slower migrating, Thr14 Tyr15 phosphory- al., 1995). The U.V.-induced G2 delay is independent lated form, which was converted to the faster of p53 function (Herzinger et al., 1995; Wang et al., migrating, dephosphorylated, active form at mitosis 1996), and involves a block in the cdc25-dependent (Figure 3a). This was con®rmed by phosphotyrosine activation of cyclin B/cdc2 (Herzinger et al., 1995; immunoblotting of the same cyclin B1 immunopreci- Poon et al., 1996). In this report we show that the pitates, which showed high levels of phosphotyrosine U.V.-induced G2 delay is mediated through a block in on cdc2 in both U.V. irradiated G2 delayed and the cdc25-dependent activation of cyclin B1/cdc2, normal S/G2 cyclin B1/cdc2 complexes (Figure 3b). cyclin A/cdk2 and cyclin A/cdc2, and that this block This suggested that the G2 delay was due to a block in occurs at the S/G2 transition or early G2. The block in the cdc25-dependent activation of the cyclin B1/cdc2 cdc25 function is not due to a reduction in the levels of either cdc25B or cdc25C proteins, but rather to the nuclear accumulation of cdc25B, and the inhibition of mitotic phosphorylation and activation of both cdc25B and cdc25C.

Results

Activation and nuclear translocation of the cyclin B1/ cdc2 complex is blocked by U.V. irradiation In response to low doses of U.V.C., HeLa cells are delayed in S and G2/M phases of the cell cycle (Wang and Ellem, 1994). Irradiation of an asynchronously growing population of the melanoma cell line A2058 also resulted in an accumulation of cells in G2/M (Figure 1a). No signi®cant accumulation or delay of cells in S phase was observed with this cell line, which was in contrast to HeLa cells irradiated under identical conditions. Following a delay of up to 20 h in G2/M, the cells recommenced progression through the cell cycle at 36 h after irradiation, which can be seen by a sharp increase in G1 and concomitant decrease in G2/ M phase cells at this time point (Figure 1b). We have investigated further the nature of the G2/M delay in HeLa and A2058 cell lines. The expression and activity of cyclin B1/cdc2, a key regulator of entry into mitosis, was analysed by immunoblotting and assaying the immunoprecipitated kinase from lysates of asynchronously growing HeLa cells (AS) which are 60% G1 phase cells, U.V. irradiated G2 delayed cells (U.V.), and late S/G2 (G2) and G2/M (M) cells from a thymidine synchrony (see Materials and methods). Immunoblotting with an anti-cyclin B1 antibody showed that cyclin B1 accumulated to normal G2 levels in the U.V. irradiated G2 delayed cells (Figure 2a), but only the G2/M phase cyclin B1/cdc2 was catalytically active (Figure 2b). Analysis of samples of similarly treated populations of A2058 cells again showed normal G2 phase accumulation of cyclin B1 and block in activation of the cyclin B1/cdc2 complex in the U.V. irradiated G2 delayed cells as was seen in HeLa (Figure 2c and d). Indirect immuno¯uorescence showed that the U.V. irradiated G2 delayed HeLa cells accumulated high levels of cyclin B1 in the cytoplasm, as in normal G2 phase cells (Pines and Hunter, 1991; Bailly et al., 1992), although due to the delay in transit through G2 in the irradiated sample up to 60% of the population showed bright cytoplasmic staining for cyclin Figure 1 U.V. irradiation of A2058 cells produces a G2 delay. B1, compared to 520% of a normal asynchronous (a) Flow cytometry histograms of DNA content of asynchronous cultures of A2058 cells irradiated with 10 Jm72 U.V.C. and culture (Figure 2e and f). harvested at the indicated times (hours) after irradiation. (b) Analysis of the cyclin B1/cdc2 complexes in U.V. Quantitation of the percentage of the population in each phase of irradiated G2 delayed HeLa cells, showed that cdc2 the cell cycle from (a) U.V. induced G2 delay BG Gabrielli et al 751 a c HeLa A2058 Activation of cyclin A/cdk complexes is blocked by U.V. irradiation Analysis of cyclin A expression showed that the level AS UV G2 M AS UV G2 M of cyclin A in the U.V. irradiated G2 delayed cells was lower than in the S/G2 cells in both HeLa and A2058 b d cultures (Figure 4a and c), but this simply re¯ected the lower degree of synchrony obtained with U.V. irradiation of an asynchronous culture compared to AS UV G2 M AS UV G2 M cultures synchronised by a thymidine block. The reduced level of cyclin A in the G2/M samples is in ef accordance with the demonstrated destruction of cyclin Asynchronous UV irradiated A during mitosis prior to the destruction of cyclin B in anaphase (Minshull et al., 1990; Pines and Hunter, 1990). Assay of the histone kinase activity of cyclin A immunoprecipitates from these samples showed that the level in the U.V. irradiated G2 delayed samples was equivalent to the S/G2 synchronised cells, and was always less than that of G2/M cells (Figure 4b and d). The di€erence in the levels of cyclin A associated histone kinase activity in the S/G2 samples of HeLa and A2058 cells was due to the relative levels of cyclin A/cdk2 activity in S phase in the two cell lines. HeLa Figure 2 U.V.-induced G2 delay results in a block of cyclin B1/ cdc2 activation and translocation to the nucleus. (a and c) cells have been shown to have high levels of S phase Immunoblot for cyclin B1 from whole cell lysates of asynchro- cyclin A/cdk2 activity, and only a relatively small nous (AS) UV irradiated G2 delayed (UV) cultures, and late S/G2 increase in the cyclin A/cdk2 histone kinase activity in (G2) and G2/M (M) cells from thymidine block cultures. (b and G2 and mitosis (Pines and Hunter, 1990; Rosenblatt et d) Histone H1 kinase assays of cyclin B1 immunoprecipitates from the indicated cell samples. (e and f) Indirect immunofluor- al., 1992). In A2058 cells, a comparatively low level of escent staining of control and U.V. irradiated, G2 delayed HeLa cyclin A associated histone kinase activity was seen cells with anti-cyclin B1 antibody. Note that the U.V. irradiated throughout S phase, with an abrupt increase in cyclin culture showed predominantly cytoplasmic staining, whereas the A associated kinase activity detected as the cells asynchronous control contained cells with cyclin B1 expression progressed from S into G2 phase, preceding the ranging from no expression (e, centre ®eld) through to prophase cells with high levels of cyclin B1 in both the cytoplasm and activation of cyclin B1/cdc2 (Figure 4e). The major nucleus (bottom ®eld) component of the increased cyclin A associated histone kinase activity detected in G2/M samples was due to the activation of cyclin A/cdk2, as anti-cyclin A and ab cdc2 phosphotyrosine anti-cdk2 immunoprecipitate histone kinase assays from both HeLa (Figure 4f) and A2058 (data not shown) were essentially identical, and an increase in anti-cdc2 immunoprecipitate histone kinase activity was only seen in G2/M phase cells (Figure 4g). This AS UV G2 M AS UV G2 M indicated that the activation of both cyclin A/cdk2 in G2, and cyclin A/cdc2 at G2/M, was blocked in U.V. c irradiated G2 delayed cells. Work from other laboratories has shown that the — H1 cdk inhibitor proteins p21 and p27 have a role in the U.V.-induced G1 cell cycle delay, through the binding cdc25B –+ + and inhibition of cyclin D/cdk4 and cyclin A/cdk2 vanadate –– + complexes (Poon et al., 1995, 1996; Petrocelli et al., 1996). Cdk2 immunoprecipitates from asynchronous Figure 3 Cyclin B1/cdc2 is maintained in the tyrosine phosphorylated form in G2 delayed cells. Cyclin B1 immunopre- control and U.V. irradiated G2 delayed HeLa and cipitates from lysates of the indicated HeLa cell samples (as in A2058 cells revealed that HeLa cdk2 was not Figure 2) were either immunoblotted with an anti-cdc2 associated with p21 and that the level of p21 monoclonal antibody (a) or anti-phosphotyrosine monoclonal associated with cdk2 in A2058 cells decreased antibody (b). (c) Cyclin B1 immunoprecipitates from U.V. following U.V. exposure (Figure 5a). This was in irradiated G2 delayed HeLa cells were incubated with or without puri®ed bacterially expressed GST-cdc25B and sodium vanadate, contrast to neonatal foreskin ®broblasts (NFF), where then assayed for histone H1 kinase activity the level of p21 associated with cdk2 increased dramatically following U.V. exposure. The level of p21 associated with cdk2 re¯ected the changes in total complex. The recovery of the histone kinase activity of p21 levels (data not shown). P27 was also found to be cyclin B1 immunoprecipitates from G2 delayed HeLa associated with cdk2 in all three cell lines, although the cells using bacterially expressed cdc25B further levels decreased in the U.V. irradiated G2 delayed supported this model (Figure 3c). The inhibition of HeLa and A2058 samples compared with the the cyclin B1/cdc2 activation by sodium vanadate asynchronous controls (Figure 5a). Decreased levels proved that the increased histone kinase activity was of p27 associated with cdk2 were also detected in S/G2 due to the action of the recombinant cdc25B. and G2/M HeLa cell samples compared to the U.V. induced G2 delay BG Gabrielli et al 752

a c a HeLa A2058 NFF HeLa A2058 cdk2

AS UV G2 M AS UV G2 M p21 b d p27

AS UV AS UV AS UV AS UV G2 M AS UV G2 M

e b UV S/G2 G2/M

L cdk2

H1K

vanadate +–+–+–

Figure 5 Cdc25B activates the U.V delayed and normal S/G2 cyclin A/cdk2. (a) Cdk2 immunoprecitates from either asynchro- nous (AS) or U.V. irradiated cultures (UV) of the neonatal f g foreskin ®broblasts (NFF), HeLa and A2058 cells collected 20 Cdk2 Cdc2 (A2058) or 24 h (NFF and HeLa) after irradiation with 10 Jm72 U.V.C. were immunoblotted for cdk2, p21 and p27. (b) Cyclin A immunoprecipitates from either U.V. irradiated G2 delayed, S/G2 and G2/M HeLa cell samples were incubated with recombinant AS UV G2 M AS UV G2 M GST-cdc25B either in the presence or absence of sodium Figure 4 U.V.-induced G2 delay results in the block of cyclin A/ vanadate, and either assayed for histone H1 kinase or cdk activation. (a and c) Lysates from the indicated cell samples immunoblotted for cdk2. The level of H1 phosphorylation was (as in Figure 2) were immunoblotted for cyclin A. (b and d) quantitated by phosphorimaging, and the relative levels are Cyclin A immunoprecipitates from the indicated cell samples were shown assayed for histone H1 kinase activity. (e) Flow cytometry, cyclin A and cyclin B1 immunoprecipitate histone H1 kinase activity from A2058 samples at times after release from a thymidine HeLa cells with recombinant cdc25B resulted in the synchrony block. (f) Cdk2 immunoprecipitates from the indicated activation of the cyclin A complexes to similar HeLa samples assayed for histone H1 kinase activity. (g) Cdc2 immunoprecipitates from identical samples to those used in (f) maximal levels, and the activation was inhibited by were assayed for histone H1 kinase activity addition of sodium vanadate (Figure 5b). Recombinant cdc25B was also found to activate cyclin A immuno- precipitates from similar A2058 cell samples, and predominantly G1 phase asynchronous samples (data similar results were obtained using cdk2 immunopre- not shown), suggesting that the decreased association cipitates from both cell lines (data not shown). in the U.V. irradiated cells simply re¯ected a decreased Immunoblotting of the cyclin A immunoprecipitates association normally occurring in G2 phase, rather for cdk2 revealed similar levels of cdk2 in each sample, than a U.V.-induced e€ect. Cyclin A immunoprecipi- and cyclin A associated exclusively with the faster tates from similar samples did not contain detectible migrating form of cdk2, corresponding to the Thr160 levels of either p21 or p27, suggesting that the p21 and phosphorylated form of cdk2 (Gu et al., 1992). This p27 detected was associated with other cyclin/cdk2 again pointed to a U.V.-induced block in the cdc25- complexes. No p21 or p27 was found in cyclin B1 dependent progression through G2 into mitosis. immunoprecipitates from similar HeLa and A2058 samples. Expression of the mitotic cdc25s is una€ected in U.V.- The lack of p21 or p27 involvement in the U.V.- induced G2 delay induced block in cyclin A/cdk2 activation in G2 indicated that this complex was inhibited in some We have demonstrated that the cdc25-dependent other manner. The inactive cyclin A/cdk2 complex in activation of both cyclin A/cdk2 and cyclin B1/cdc2 late S/G2 phase cells has been demonstrated to be was blocked in the U.V. irradiated G2 delayed cells. phosphorylated on Tyr15, and that this phosphoryla- Therefore we assessed the expression of the two cdc25 tion is lost on activation of the complex in G2 (Gu et isoforms which are known to have a role in regulating al., 1992). Incubation of cyclin A immunoprecipitates mitosis, cdc25B and cdc25C. The level of cdc25C was from U.V. irradiated G2 delayed, S/G2 and G2/M found to be unchanged throughout the normal cell cycle U.V. induced G2 delay BG Gabrielli et al 753 in both HeLa and A2058 cells, and this was not a€ected the lysis bu€er (unpublished observations). The in G2 delayed cells (Figure 6a). A low level of the slower appearance of the electrophorectically retarded form migrating, hyperphosphorylated form of cdc25C was of cdc25B was previously shown to coincide with the detected in the G2/M samples from both cell lines, but cdc2-dependent increase in microtube nucleation at the was not detected in either U.V. delayed, or synchronised centrosome and changes in microtubule dynamics S/G2 samples (Figure 6a, ®lled arrowhead). (Gabrielli et al., 1996), suggesting that this was the The levels of both cdc25B mRNA and protein have mitotically active form of cdc25B, similar to the been demonstrated to vary through the cell cycle, being hyperphosphorylated form of cdc25C (Ho€mann et maximal at G2/M (Nagata et al., 1991; Gabrielli et al., al., 1993). The activity of both cdc25B and cdc25C was 1996). The level of cdc25B mRNA and protein assessed by the ability of immunoprecipitated cdc25B accumulated to nearly S/G2 levels in the U.V. and cdc25C to activate a Thr14 and Tyr15 phosphory- irradiated G2 delayed HeLa and A2058 cells (Figure lated cdc2/cyclin B1 complex (PY15) produced in vitro. 6b, c and d). Again, the di€erences in the levels of The cdc25 antibodies have previously been shown to be cdc25B transcript and protein between the U.V. speci®c for their respective isoform (Gabrielli et al., delayed and synchronised S/G2 cells was due to the 1996). The relative cdc25 activity of the immunopreci- relatively higher percentage of G2 phase cells in the pitated cdc25 was quantitated by the increased histone synchronised S/G2 compared to U.V. delayed samples. H1 kinase activity of the activated PY15 substrate. The cdc25B protein in both the S/G2 and U.V. Immunoprecipitated cdc25B from U.V. irradiated G2 irradiated G2 delayed cells ran as a single band, delay, S/G2 and G2/M HeLa cells increased PY15 H1 whereas a slower migrating form of cdc25B was kinase activity over the preimmune levels (the PY15 detected in the G2/M phase samples from both HeLa substrate possesses a low level of H1 kinase activity), and A2058 cells (Figure 6c and d, ®lled arrowheads). with the G2/M sample showing the largest increase (Figure 7a). The di€erence between the U.V. irradiated G2 and S/G2 samples was due to the lower level of Activation and cytoplasmic accumulation of cdc25B is cdc25B in the former (Figure 7a, immunoblot). The blocked in U.V.-induced G2 delay activity of cdc25B in the G2/M samples was less than The slowest migrating form of cdc25B was suspected of twice that from S/G2 samples, but this small increase being a phosphorylated form of the protein, as it was in activity re¯ected the relatively low abundance of the dicult to detect in cdc25B immunoprecipitates from electrophoretically retarded form of cdc25B in these G2/M phase cells without the inclusion of microcystin, samples. The increased H1 kinase activity was shown a potent inhibitor of phosphatases PP1 and PP2A, in to be due to the action of the immunoprecipitated cdc25B, as it was blocked by the inclusion of sodium vanadate in the incubation (Figure 7a). Similar immunoprecipitate activity assays were a performed for cdc25C. The cdc25C from an asynchro- HeLa A2058 nous culture (predominantly G1 phase cells) had little L L cdc25C activity towards PY15, whereas both the U.V. irradiated and S/G2 samples contained a low, though AS UV G2 M AS UV G2 M reproducible level of activity, and G2/M cdc25C was b greater than fourfold more active than cdc25C from either G2 sample (Figure 7b). The level of immuno- precipitated cdc25C protein did not vary between samples, as seen in the immunoblots of whole cell AS S G2 M UV lysates (Figure 6a), and again only a very low level of c the hyperphosphorylated cdc25C was detected in the G2/M samples, as an indistinct, lower mobility smear (Figure 7b, immunoblot). The speci®city of the L activation assay was demonstrated by the ability of sodium vanadate to inhibit the increase in PY15 H1 AS S G2 M UV kinase activity. The absence of the lower mobility forms of both cdc25B and cdc25C in the U.V. d irradiated G2 delayed cells indicated that the mitotic phosphorylation and activation of both cdc25 isoforms L was blocked, although both cdc25B and cdc25C from AS UV G2 M both G2 samples were capable of activating the Figure 6 U.V. does not a€ect accumulation of cdc25B and exogenous PY15 substrate, albeit to a reduced extent. cdc25C, but blocks their mitotic phosphorylation. (a) Lysates We had previously demonstrated that the mitotic from the indicated cell samples (as in Figure 2) were activity of cdc25B is tightly correlated with its immunoblotted for cdc25C. The hyperphosphorylated form is accumulation in the perinuclear region and more indicated (®lled arrowhead). (b) Northern analysis of the indicated HeLa cell samples including a thymidine arrested S broadly in the cytoplasm (Gabrielli et al., 1996). phase sample (S), for the expression of cdc25B transcript. (c) Indirect immuno¯uorescent staining of U.V. irradiated Immunoblotting for cdc25B from identical samples to those used G2 delayed HeLa cells showed a striking accumulation in (b). The hyperphosphorylated form of cdc25B (®lled arrow- of cdc25B staining in the nucleus and negative staining head) and IgG heavy chain (open arrowhead) are indicated. (d) Lysates of A2058 cells similar to those in (a) were immunoblotted of the nucleoli (Figure 8c). This had not previously been for cdc25B. Again, the lower mobility, phosphorylated form is noted in immunostaining experiments of asynchronous indicated (®lled arrowhead) cultures, although closer inspection of these did show a U.V. induced G2 delay BG Gabrielli et al 754

a a cdc25B microtubules

L cdc25B

H1K

b

b

L cdc25C

H1K

c

Figure 7 U.V.-induced G2 delay blocks mitotic activation of cdc25B and cdc25C. (a) Cdc25B immunoprecipitates from the indicated HeLa cell samples were either immunoblotted for cdc25B or assayed for cdc25 activity by quantitating the activation of PY15 in terms of the increased histone H1 kinase activity either in the presence or absence of sodium vanadate. The level of endogenous H1 phosphorylation by unmodi®ed PY15 was Figure 8 U.V.-induced G2 delay results in the nuclear assessed by incubation with a preimmune IgG immunoprecipitate accumulation of cdc25B. Indirect immuno¯uorescent staining of from a G2/M sample. The level of H1 phosphorylation was synchronised late S/G2 phase cells (a), S/G2 cells stained with quantitated by phosphorimaging. The mean of duplicate cdc25B antibody preincubated with bacterially expressed GST- determinations are shown. (b) Cdc25C immunoprecipitates from cdc25B (b), or U.V. irradiated G2 delayed HeLa cells (c), with the indicated HeLa cell samples were immunoblotted or assayed either anti-cdc25B or anti b-tubulin (microtubules) antibody for cdc25C activity as in (a)

Discussion very low percentage of cells with increased nuclear staining suggesting that this may represent a transient Activation of a G2 checkpoint is a common mechanism nuclear accumulation of cdc25B in early G2 phase. This by which cells respond to genotoxic stress. However, the was con®rmed by indirect immuno¯uorescent staining response mechanisms leading to the G2 delay di€er of synchronised late S/G2 phase HeLa cells which depending on the agent causing the stress, although they revealed a high level of nuclear cdc25B staining in a appear to converge at cdc2, resulting in an accumula- majority of cells (Figure 8a), similar to the U.V. tion of the Tyr15 phosphorylated form of cdc2. irradiated G2 delayed cells, although a small propor- Exposure to ionising radiation in S phase inhibits tion of the synchronised population cells also showed cyclin B mRNA accumulation, while exposure in G2 the cytoplasmic localisation in prophase and mitotic inhibits cyclin B translation, both of which suppress cells, as previously described (Gabrielli et al., 1996). The cyclin B/cdc2 activation at mitosis (Muschel et al., increased nuclear staining in both samples correlated 1991), whereas cyclin B levels are una€ected following with the elevated levels of cdc25B detected by exposure to nitrogen mustard, but cyclin A mRNA and immunoblotting (Figure 6c and d), and the speci®city protein levels exceed control levels in the G2 delay of the staining demonstrated by using antibody induced by both ionising radiation and nitrogen solutions preincubated with GST-cdc25B, which re- mustard (O'Connor et al., 1993; Muschel et al., 1993). duced the staining to near background (Figure 8b). In this study, cyclin B1 accumulated to normal G2 Counterstaining with a b-tubulin monoclonal antibody levels, and formed complexes with cdc2 which are showed the microtubule network to be the normal maintained as the inactive Tyr15 phosphorylated form interphase, basket-like array, with no mitotic increase of in the U.V. induced G2 checkpoint. A U.V.-induced microtubule nucleation on the centrosomes or changes accumulation of Tyr15 phosphorylated cdc2/cyclin B in microtubule dynamics were observed. has also been reported in keratinocytes and ®broblasts U.V. induced G2 delay BG Gabrielli et al 755 (Herzinger et al., 1995; Poon et al., 1996). Our data Expression of cdc25B and cdc25C was found to be con®rms these earlier ®ndings, but also shows that the una€ected in the U.V. irradiated G2 delayed cells, but inactive cyclin B12/cdc2 complex is maintained in the the slower migrating forms of both cdc25B and cdc25C cytoplasm as seen in normal G2 phase cells, indicating were absent in the G2 delayed as well as the normal S/ that not only is the mitotic activation of this complex G2 cells. The mitotic hyperphosphorylation of cdc25C inhibited, but its translocation from the cytoplasm to has been associated with an increase in its activity the nucleus is also blocked. The translocation of cyclin (Ho€mann et al., 1993), which we have con®rmed in B1/cdc2 to the nucleus appears to be controlled by this study, although the levels of hyperphosphorylated phosphorylation of the cytoplasmic retention signal cdc25C detected in the G2/M HeLa and A2058 (Pines and Hunter, 1994; Li et al., 1997), and in Xenopus samples were very low. The loss of the hyperpho- the nuclear translocation has been demonstrated to be sphorylated form of cdc25C has also been observed in essential for GVBD, a marker of entry into metaphase G2 delays in response to both ionising radiation and of meiosis I (Li et al., 1997). Thus blocking the nitrogen mustard, with the corresponding loss of translocation of cyclin B1/cdc2 to the nucleus can mitotic activation (O'Connor et al., 1994; Barth et e€ectively inhibit entry into mitosis. This has been al., 1996). This is presumably simply a consequence of demonstrated in cells overexpressing a non-phosphor- the block of cyclin B/cdc2 activation, which is ylatable Thr14 and Tyr15 mutant of cdc2. These cells proposed to be responsible for the mitotic hyperpho- have a reduced G2 delay following exposure to ionising sphorylation and activation of cdc25C, via an radiation (Jin et al., 1996), but during the G2 delayed, autoampli®cation mechanism (Ho€mann et al., 1993). cells had high levels of cytoplasmic cyclin B/cdc2 The slower migrating form of cdc25B observed in activity, indicating that the correct localisation of mitotic samples also appeared to be the result of a cyclin B/cdc2 activity is essential for mitosis. mitotic phosphorylation, and was correlated with an We have also found that cyclin A accumulates to increase in its activity. We have found that in vitro levels approaching normal G2 levels in the U.V. phosphorylation of recombinant cdc25B, using mitotic irradiated G2 delayed cells, and that the cyclin A/cdk HeLa cell extracts, produced a 4 ± 5-fold increase in complexes are maintained in an inactive form which cdc25B activity (unpublished data). It is probable that can be activated to mitotic levels by cdc25 in vitro. this mitotic activation is also due to an auto- The cyclin A/cdk includes both the large peak of ampli®cation mechanism similar to that for cdc25C. cyclin A/cdk2 normally activated in G2, and the However, cdc25B from both G2 delayed and normal relatively minor peak of cyclin A/cdc2 which is S/G2 phase samples had a signi®cant level of activity activated as the cells enter mitosis (Rosenblatt et al., using a recombinant PY15 substrate in vitro, whereas 1992; Pagano et al., 1993). The G2 phase activation of only a very low level of cdc25C activity was found in cyclin A/cdk2 is clearly blocked, and our data suggests these samples. We have also found that the bacterially that the activation of the small pool of cyclin A/cdc2 expressed, unphosphorylated cdc25B is 4fourfold is also blocked, as the low level of cdc2 H1 kinase more active than cdc25C, but the hyperphosphory- activity in the U.V. irradiated G2 delayed cells was lated forms of the two recombinant proteins have equivalent to that in the S/G2 sample. This is in similar levels of activity in vitro (manuscript in contrast to cells exposed to nitrogen mustard, where preparation), mimicking the in vivo activities of these cyclin A/cdk2 is active during the G2 delay induced two enzymes. Thus the block in mitotic activation of by this agent (O'Connor et al., 1993). The function of cdc25B in the U.V. irradiated G2 delayed and normal cyclin A/cdk2 activity in G2 and mitosis is still early G2 cells, cannot explain the observed block in unclear, but there is good experimental evidence cyclin A/cdk2 and cyclin B1/cdc2 activation. suggesting a key role for cyclin A/cdk activity in A partial explanation may be found in the cellular regulating the timing of cyclin B/cdc2 activation and localisation of the cdc25s and cyclin/cdks. The nuclear progression into mitosis (Walker and Maller, 1991; accumulation of cdc25B in both the U.V. irradiated Pagano et al., 1992, 1993). and normal S/G2 phase cells was unexpected, although There is evidence for a role for p53 in a G2/M cell ectopically expressed cdc25B also accumulated in the cycle delay (Agarwal et al., 1995), but the G2 delay nucleus of the majority of transfected cells, indicating investigated in this study is independent of p53 that cdc25B possesses functional nuclear localisation function, as HeLa cells lack a functional p53 protein sequences. The nuclear accumulation of cdc25B in the (Wang et al., 1996), and the lack of an increase in the G2 delayed cells would also contribute to the block in level of p21 in the A2058 cells suggests that these cells cyclin B1/cdc2 activation, as both cdc25B and cdc25C also lack a U.V.-induced p53 response. By comparison, are now localised in the nucleus, whereas the inactive neonatal ®broblasts displayed a dramatic increase in cyclin B1/cdc2 complex is localised in the cytoplasm of p53-dependent p21 expression and binding to cdk2, the U.V. irradiated G2 delayed and normal G2 cells, associated with a U.V.-induced G1/S phase delay as and is thus physically separated from the potentially described by others (Petrocelli et al., 1996; Poon et al., activating phosphatases. In this respect, the G2 delayed 1996). This indicates the presence of a second DNA cells appear to be identical to normal early G2 phase damage induced checkpoint mechanism, independent cells, but whereas this is only a transitory phase in cells of p53, which is activated either by DNA that escapes normally progressing through the cell cycle, following the p53-dependent mechanism, or which occurs after U.V. irradiation the cells are held at this point in the the G1 checkpoint through which p53 operates. cell cycle until a signal for resumption of normal Evidence presented here suggests that this is via a progression through into mitosis is generated. This block in the cdc25-dependent activation of both cyclin would presumably include a signal for relocation of A/cdk complexes and cyclin B1/cdc2 in G2 are cdc25B into the cytoplasm as was previously observed regulated by a cdc25 phosphatase. (Gabrielli et al., 1996). U.V. induced G2 delay BG Gabrielli et al 756 Separate compartmentation, however, cannot provide separate compartmentation of the cdc25B and cdc2/ an explanation for the lack of activation of the G2 and cyclin B1. It is unclear as to the mechanism by which mitotic cyclin A/cdk complexes in the U.V. irradiated active cdc25B is blocked from activating cyclin A/cdk2 G2 delayed and normal G2 cells, as these complexes are and cyclin A/cdc2, both of which are localised to the almost exclusively nuclear (Pines and Hunter, 1991; nucleus with cdc25B in G2. Bailly et al., 1992; Pagano et al., 1992). Recombinant cdc25B can activate the immunoprecipitated cyclin A/ cdk2 and presumably cyclin A/cdc2, from U.V. irradiated G2 delayed and normal S/G2 cells in vitro, Materials and methods suggesting that some other factor is blocking the cdc25- dependent activation of the cyclin A/cdk, and that factor Cell culture, synchrony and U.V. treatment is lost during the immunoprecipitation of the cyclin A/ HeLa (cervical adenocarcinoma derived) and A2058 cdk2 complex. One possibility is that p95WEE1Hu, (human melanoma) cells were grown in RPMI 1640 which is also localised to the nucleus (McGowan and medium supplemented with 5 and 10% Serum Supreme Russell, 1995), could maintain the cyclin A/cdk2 in its (Biowhittaker), respectively, and 0.1 mg/ml streptomycin Tyr15 phosphorylated inactive state. This kinase has and 100 U/ml penicillin. Synchronised cells were obtained by seeding 10 cm plates with 1 ± 1.56106 cells, then the been demonstrated to be speci®c for the phosphoryla- following day adding 2 or 2.5 mM thymidine to A2058 and tion of Tyr15, and is active throughout the cell cycle HeLa cultures, respectively, for 16 h. The cells were except mitosis when its activity is severely reduced released from the thymidine block by washing three times (Parker et al., 1995; McGowan and Russell, 1995). The with prewarmed serum-free RPMI 1640, and replacing this signal for the resumption of cell cycle progression may with fresh medium supplemented with 24 mM deoxycytidine be the inactivation of this kinase, which may be sucient and thymidine. S/G2 phase cells were collected 8 h after to shift the balance to favour the cdc25-dependent release, and G2/M cells were collected when 425% cells activation of cyclin A/cdk2. A triggering role for cyclin had a rounded mitotic morphology. The cell cycle stage A/cdk in the subsequent activation of cyclin B/cdc2 and was determined using ¯ow cytometry as previously entry into mitosis has been proposed (Minshull et al, described (Gabrielli et al., 1996). This protocol routinely gave 50 ± 60% synchrony, assessed by ¯ow cytometry and 1990; Lehner and O'Farrell, 1990). cyclin B1 histone kinase assay, respectively (see Figure 4e). Alternatively, the delay may be attributed solely to U.V. irradiated G2 delayed cells were obtained by the regulation of cdc25 activity. The identi®cation of a irradiating asynchronously growing cultures with 10 Jm72 cdc25-dependent pathway in the G2 delay has UVC as described previously (Wang and Ellem, 1994), with precedents in both ®ssion yeast and Xenopus systems, the cells collected 18 ± 24 h later, at which time 40 ± 60% of where incomplete DNA replication can block the this cell population were in G2/M (see Figure 1b). cdc25-dependent activation of mitotic cyclin/cdks (Enoch and Nurse, 1990; Kumagai and Dunphy, Immunoblotting and immunoprecipitate kinase assays 1991; Gabrielli et al., 1992). In both systems, mitotic levels of cdc25 are present, but the cdc25 is maintained Equal cell numbers were lysed in NETN (20 mM Tris in an inactive form until replication is completed pH 8.0, 1 mM EDTA, 100 mM NaCl, 0.5% NP-40) (Izumi et al., 1992; Kovelman and Russell, 1996). This supplemented with 300 mM NaCl and 5 mg/ml leupeptin, parallels our ®ndings for cdc25C, where the level of pepstatin A and aprotinin, and 0.5 mM PMSF, 10 mM NaF and 0.1 mM sodium vanadate. The cleared supernatant was cdc25C protein was unchanged, but only displayed a then used for either immunoblotting or immunoprecipitate signi®cant level of activity in G2/M samples. In the kinase assay. Immunoblotting for cyclin B1 and cdc25B U.V. response, DNA replication is completed prior to was performed as described previously (Gabrielli et al., the G2 delay, demonstrated by an accumulation of 1996), and cyclin A and cdc25C immunoblots were cells with 4N DNA content. Therefore the cdc25- performed in a similar manner. Brie¯y, 20 mg protein was dependent G2 delay is presumably invoked by the resolved on 10% SDS ± PAGE gels, electrotransferred to presence of damaged DNA, and this delay allows for nitrocellulose ®lters and probed with either an anity repair of the U.V.-induced DNA damage. puri®ed anti-cdc25C polyclonal antibody (Gabrielli et al., In summary, this study has shown that the G2 phase 1996) or anti-cyclin A monoclonal antibody (Pharmingen), checkpoint invoked following low dose U.V.C. and detected using ECL (DuPont). For kinase assays, immunoprecipitations were performed from 1 mg of cell exposure results in the accumulation of the inactive lysate. Cyclin B1 immunoprecipitate kinase assays were as Tyr15 phosphorylated forms of cdk2/cyclin A, cdc2/ described previously (Gabrielli et al., 1996), and cyclin A cyclin B1 and presumably cdc2/cyclin A, which immunoprecipitate kinase assays were performed in a e€ectively blocks progression into mitosis. This block similar manner, using 1 mg anti-cyclin A antibody. of both cyclin A/cdk and cyclin B1/cdc2 activation Immunoblotting for cyclin B1 associated cdc2, cyclin A puts the U.V.-induced G2 delay at the S/G2 transition associated cdk2, and cdk2 associated p21 and p27, was or early G2 before the activation of cyclin A/cdk in performed by resolving the immunoprecipitates on 12% G2, rather than the G2/M transition. The mechanism SDS ± PAGE gels, and probing the nitrocellulose transfers of the G2 checkpoint is independent of p53 and p21, with either anti-cdc2 (Santa Cruz), anti-phosphotyrosine and appears to be due to inhibition of the cdc25- monoclonal antibodies, anti-cdk2 polyclonal antibody (Santa Cruz), anti-p21 (Calbiochem) or anti-p27 (Trans- dependent dephosphorylation and activation of the duction Laboratories) monoclonal antibodies, using ECL mitotic cdk/cyclins. This is in part due to a block in the detection. mitotic hyperphosphorylation and activation of cdc25B and cdc25C, which in the case of cdc25C appears to be essential for its activity. However, normal G2 phase Northern blotting cdc25B retains a signi®cant level of activity, but is Total RNA was prepared from 107 HeLa cells from the blocked from activating cdc2/cyclin B1 due to the indicated cell cycle stage, prepared as described above, U.V. induced G2 delay BG Gabrielli et al 757 using RNA extraction columns (Qiagen). Twenty mgof speci®c antibodies (Gabrielli et al., 1996) bound to 30 mlof RNA was run out on a 1% agarose/formaldehyde gel and 50% suspension of protein A-Sepharose, for 3 h at 48C. The blotted onto Hybond-N membrane (Amersham). Equiva- precipitates were washed twice with NETN with 5 mM DTT, lent lane loading were monitored by the ethidium bromide then twice with NETN containing 5 mM DTT and 1 M NaCl, staining of the ribosomal RNA bands. Blots were and once with phosphatase bu€er. The activity of the hybridised with a 2.2 kb fragment of cdc25B cDNA. immunoprecipitated cdc25 was assayed by quantitating the increase in activity of a Thr14 and Tyr15 phosphorylated cdc2/cyclin B1 (PY15) produced in vitro. The production of Cdc25 assays PY15 will be detailed elsewhere (manuscript in preparation). Bacterially expressed GST-cdc25B wild type was prepared Brie¯y, baculovirus expressed human cyclin B1 GST fusion as described previously (Gabrielli et al., 1996). The was bound to glutathione-agarose then incubated in a HeLa bacterial lysates were diluted in NETN supplemented with cell lysate with the addition of 2 mM ATP, 15 mM MgCl2 and protease inhibitors (as described above) and 2 mM DTT, the cdk inhibitor olomucine (0.1 mM). The agarose bound then incubated with 300 ml of 50% suspension of GST-cyclin B1/cdc2 was washed and cleaved from the GST glutathione-agarose for 2 h at 48C. The precipitates were moiety with thrombin. The released Thr14, Tyr15 and washed twice in NETN supplemented with 2 mM DTT, Thr161 phosphorylated cdc2/cyclin B1 was then stored at twice with NETN supplemented with 1 M NaCl and 2 mM 7708C until use. Five ml of PY15 substrate was diluted with DTT and once with phosphatase bu€er (20 mM Tris 15 ml of phosphatase bu€er and incubated with the pH 8.2, 2 mM EDTA, 2 mM DTT, 150 mM NaCl, 0.1% immunoprecipitated cdc25 for 30 min at 308C. The reaction TX-100), and the GST-cdc25B was eluted with 200 mlof was stopped by addition of 5 mlof5mMsodium vanadate, phosphatase bu€er containing 10 mM glutathione, for 1 h and the increase in H1 kinase activity measured as described at 48C. Thirty ml aliquots of the eluted GST-cdc25B were above. added to cyclin A or cyclin B1 immunoprecipitates with or without addition of 5 mlof5mM sodium vanadate, and Immuno¯uorescence studies incubated for 30 min at 308C, when the reaction was stopped with the addition of 5 mlof5mM sodium Indirect immuno¯uorescent staining for cdc25B, cyclin B1 vanadate. The resultant histone kinase activity of the and b-tubulin were performed as described previously cyclin immunoprecipitates was then measured as previously (Gabrielli et al., 1996). described. For cdc25 immunoprecipitate assays, each cdc25 was immunoprecipitated from 1.5 mg of HeLa cell lysate, prepared as described above and supplemented with 2 mM Acknowledgements microcystin and 5 mM DTT. Lysates were precleared by The authors are grateful to Dr Colin De Souza for the incubating with 30 ml of a 50% suspension of protein A- critical reading of this manuscript. This work was Sepharose for 30 min at 48C. The supernatant was then supported by the Queensland Cancer Fund. BG is a incubated with either anity puri®ed, cdc25B or cdc25C Fiuczek Fellow.

References

AgarwalML,AgarwalA,TaylorWRandStarkGR.(1995). Izumi T, Walker DH and Maller JL. (1992). Mol. Biol. Cell, Proc. Natl. Acad. Sci. USA, 92, 8493 ± 8497. 3, 927 ± 939. Bailly E, Pines J, Hunter T and Bornens M. (1992). J. Cell Jin P, Gu Y and Morgan DO. (1996). J. Cell Biol., 134, 963 ± Sci., 101, 529 ± 545. 970. Barker D, Dixon K, Medrano EE, Smalara D, Im S, Mitchell Jinno S, Suto K, Nagata A, Igarashi M, Kanaota Y, Nojima D, Babcock G and Abdel-Malek ZA. (1995). Cancer Res., H and Okayama H. (1994). EMBO J., 13, 1549 ± 1556. 55, 4041 ± 4046. Kovelman R and Russell P. (1996). Mol. Cell. Biol., 16, 86 ± Barth H, Ho€mann I and Kinzel V. (1996). Cancer Res., 56, 93. 2268 ± 2272. Kumagai A and Dunphy WG. (1991). Cell, 64, 903 ± 914. Dulic V, Kaufamann WK, Wilson SJ, Tlsty TD, Lees E, Lehner CF and O'Farrell PH. (1990). Cell, 61, 535 ± 547. Harper JW, Elledge SJ and Reed SI. (1994). Cell, 76, Li J, Meyer AN and Donoghue DJ. (1997). Proc. Natl. Acad. 1013 ± 1023. Sci. USA, 94, 502 ± 507. Enoch T and Nurse P. (1990). Cell, 60, 665 ± 673. Lohke MJ, Hayes MK and Maller JL. (1988). Proc. Natl. Gabrielli B. (1994). Adv. Prot. Phosphatases, 8, 209 ± 226. Acad. Sci. USA, 85, 3009 ± 3013. Gabrielli BG, Lee MS, Walker DH, Piwnica-Worms H and Lu X and Lane DP. (1993). Cell, 75, 765 ± 778. Maller JL. (1992). J. Biol. Chem., 267, 18040 ± 18046. McGowan CH and Russell P. (1995). EMBO J., 14, 2166 ± Gabrielli BG, De Souza CPC, Tonks ID, Clark JM, 2175. Hayward NK and Ellem KAO. (1996). J. Cell. Sci., 109, Meyerson M, Enders GH, Wu C-L, Su L-K, Gorka C, 1081 ± 1093. Nelson C, Harlow E and Tsai L-H. (1992). EMBO J., 11, Gautier J, Norbury C, Lohka M, Nurse P and Maller J. 2909 ± 2917. (1988). Cell, 54, 433 ± 439. Millar JB, Blevitt J, Gerace L, Sadhu K, Featherstone C and Gu Y, Rosenblatt J and Morgan DO. (1992). EMBO J., 11, Russell P. (1991). Proc. Natl. Acad. Sci. USA, 88, 10500 ± 3995 ± 4005. 10504. Herzinger T, Funk JO, Hillmer K, Eick D, Wolf DA and Minshull J, Golsteyn R, Hill CS and Hunt T. (1990). EMBO Kind P. (1995). Oncogene, 11, 2151 ± 2156. J., 9, 2865 ± 2875. van den Heuval S and Harlow E. (1993). Science, 262, 2050 ± Morgan DO. (1995). Nature, 374, 131 ± 134. 2054. Muschel RJ, Zhang HB, Iliakis G and McKenna WG. Ho€mann I, Clarke PR, Marcote MJ, Karsenti E and (1991). Cancer Res., 51, 5113 ± 5117. Draetta G. (1993). EMBO J., 12, 53 ± 63. Muschel RJ, Zhang HB and McKenna WG. (1993). Cancer Ho€mann I, Draetta G and Karsenti E. (1994). EMBO J., Res., 53, 1128 ± 1135. 13, 4302 ± 4310. U.V. induced G2 delay BG Gabrielli et al 758 Nagata A, Makoto I, Jinno S, Suto K and Okayama H. Pines J and Hunter T. (1990). Nature, 346, 760 ± 763. (1991). New Biol., 3, 959 ± 968. Pines J and Hunter T. (1991). J. Cell Biol., 115, 1±17. O'Connor PM, Ferris DK, Pagano M, Draetta G, Pines J, Pines J and Hunter T. (1994). EMBO J., 16, 3772 ± 3781. Hunter T, Longo DL and Kohn KW. (1993). J. Biol. Poon RYC, Toyoshima H and Hunter T. (1995). Mol. Biol. Chem., 268, 8298 ± 8308. Cell, 6, 1197 ± 1213. O'Connor PM, Ferris DK, Ho€mann I, Jackman J, Draetta Poon RYC, Jiang W, Toyoshima H and Hunter T. (1996). J. G and Kohn KW. (1994). Proc.Natl.Acad.Sci.USA,91, Biol. Chem., 271, 13283 ± 13291. 9480 ± 9484. Rosenblatt J, Gu Y and Morgan DO. (1992). Proc. Natl. Pagano M, Pepperkok R, Verde F, Ansorge W and Draetta Acad. Sci. USA, 89, 2824 ± 2828. G. (1992). EMBO J., 11, 961 ± 971. Sherr CJ and Roberts JM. (1995). Genes Dev., 9, 1149 ± 1163. Pagano M, Pepperkok R, Lukas J, Baldin V, Ansorge W, Terada Y, Tatsuka M, Jinno S and Okayama H. (1995). Bartek J and Draetta G. (1993). J. Cell. Biol., 121, 101 ± Nature, 376, 358 ± 362. 111. Walker DH and Maller JL. (1991). Nature, 354, 314 ± 317. Parker LL, Sylvestre PJ, Byrnes MJ, Liu F and Piwnica- Wang X-Q and Ellem KAO. (1994). Exp. Cell Res., 212, Worms H. (1995). Proc. Natl. Acad. Sci. USA, 92, 9638 ± 176 ± 189. 9642. Wang XQ, Gabrielli BG, Milligan A, Dickinson JL, Antalis Petrocelli T, Poon R, Drucker DJ, Slingerland JM and TM and Ellem KAO. (1996). Cancer Res., 56, 2510 ± 2514. Rosen CF. (1996). Oncogene, 12, 1387 ± 1396.