The Cyclin-dependent Kinase Cdk2 Regulates Thymocyte Apoptosis

p53 gene–deficient, Bcl-2 transgenic (Tg), DO.11.10 OVA TCR-α/β Tg, and P14 TCR-α/β Tg mice have been described previously (30–33). BALB/c mice were purchased from Taconic Farms. All mice were kept at the Animal Facility of the Ontario Cancer Institute in accordance with institutional guidelines.

Freshly isolated thymocytes from BALB/c mice were cultured in RPMI 1640 medium (10% FCS, 10⁻⁵ M β-mercaptoethanol) in the absence or presence of dexamethasone (Sigma), heat shock, γ-irradiation, anti-CD95 Ab (clone Jo91; PharMingen), anti-CD3ε (clone 145-2C11; PharMingen), PMA (12.5 ng/ml), or etoposide (2.5 μg/ml) for different time periods and at different concentrations as indicated in the figure legends (34). Optimal concentrations and activation regimes for the induction of apoptosis were determined in pilot studies. The specific Cdk2 blockers olomoucine (Calbiochem) and roscovitine (gift of Dr. Meijer, CNRS, Roscoff, France) were dissolved in DMSO (Sigma). Titration experiments determined that 100 μM olomoucine and 50 μM roscovitine were the most effective concentrations for inhibiting Cdk2 activity and apoptosis in thymocyte cultures. DMSO had no effect on Cdk2 activity or the induction or prevention of apoptosis in response to all stimuli tested. The cell cycle blockers TGF-β (1 nM; R&D Systems) and rapamycin (2 ng/ml; Calbiochem) were used at concentrations that optimally blocked cell cycle progression in T cell lymphoma cells.

Total cell numbers of viable and apoptotic cells were determined by trypan blue exclusion. Relative percentages of viable and apoptotic CD4⁺CD8⁺ thymocytes were determined by triple staining with anti-CD4–PE, anti-CD8–FITC, and the vital chromogenic dye, 7AAD (34). The results were expressed as the percentage of viable thymocytes remaining after 22 h, calculated as follows: (number of viable CD4⁺CD8⁺ thymocytes after stimulation)/(number of viable CD4⁺CD8⁺ thymocytes cultured under the same conditions in the absence of stimulation) × 100. For the detection of cycling and apoptotic cells, thymocytes were stained with propidium iodide (PI). After different periods of stimulation, thymocytes were harvested, washed once in PBS (0.5% glucose), and fixed in cold 70% ethanol overnight. Fixed cells were pelleted to remove ethanol and stained with PI (final concentration 50 μmol/ml) for 30 min at room temperature. Apoptosis-mediated membrane changes were determined via staining with Annexin V (R&D Systems). PI and Annexin V staining of thymocytes was determined by cytofluorometry using a FACSCalibur™ (Becton Dickinson).

After different periods of stimulation, thymocytes were harvested and lysed, and proteins were immunoprecipitated using Abs against Cdk2 (amino acids [aa] 283–298), Cdk4 (aa 282–303), and Cdc2 (aa 278–297) (all from Santa Cruz Biotechnology). Cdk2 and Cdc2 kinase activities in immunoprecipitates were assayed using [γ-³²P]ATP (3,000 cpm/pmol) and histone H1 (2 μg/ml; Boehringer Mannheim) or p53 (2 μg/ml; PharMingen) as substrates. Cdk4 activity was determined using glutathione S-transferase (GST)-Rb as a substrate (2 μg/ml; PharMingen). H1, p53, or GST-Rb phosphorylation was assayed by autoradiography after SDS-PAGE separation. The levels of immunoprecipitated Cdk2, Cdc2, and Cdk4 were determined by Coomassie blue staining and Western blotting.

Thymocytes were lysed in 1% NP-40 lysis buffer. Proteins were separated by SDS-PAGE, transferred onto nitrocellulose membranes, and incubated with Abs reactive to Cdc2, Cdk2, Cdk4, Cdk7, Pctaire-2, Cdc25A, cyclins A, E, D1, D2, B, and D3, E2F-1, p27^Kip1, caspase 2, and p53 (clone 240) (all from Santa Cruz Biotechnology), p21 (Calbiochem), Bcl-XL (Transduction Laboratories), Bcl-2 (PharMingen), caspase 3 (gift of Dr. R. Sekaly, McGill University, Montreal, Quebec, Canada), and Rb (clone G3-245 reactive to an aa 300–380 epitope of Rb, PharMingen; and clone C-15 reactive against aa 914–928, Santa Cruz Biotechnology). The anti-caspase 8–specific Abs were developed in our Institute and were a kind gift of Dr. R. Hakem (Amgen Institute). Immunoprecipitations were performed using protein A–Sepharose. Optimal Ab concentrations and conditions for immunoprecipitations were determined in pilot studies.

The mitochondrial transmembrane potential (ΔΨ_m) results from the asymmetric distribution of protons across the inner mitochondrial membrane, giving rise to a chemical (pH) and electric gradient (35, 36). The inner side of the inner mitochondrial membrane is negatively charged. As a consequence, the cationic lipophilic fluorochrome 3,3′-dihexyloxacarbocyanine iodide (DiOC₆(3)) is distributed on the mitochondrial matrix as a function of the Nernst equation, correlating with ΔΨ_m. DiOC₆(3) can be used to measure variations in the ΔΨ_m on a per-cell basis. Cells induced to undergo apoptosis manifest an early reduction in the incorporation of ΔΨ_m-sensitive dyes, indicating a disruption of ΔΨ_m. For DiOC₆(3) staining, 10⁶ thymocytes were incubated with DiOC₆(3) (final concentration 20 nM in PBS) for 20 min at 37°C. DiOC₆(3) staining was analyzed immediately using a FACSCalibur™.

Thymocytes were purified from P14 Tg mice, which express an α/β TCR (TCR Vα2Vβ8) specific for a peptide (p33) of the lympholytic choriomeningitis virus (LCMV). P14 Tg thymocytes (10⁶/well) were cultured on a monolayer of confluent and adherent MC57/L fibroblasts (H-2^b/b) in RPMI medium (5% FCS, 10⁻⁵ M β-mercaptoethanol). MC57/L APCs were pulsed with different concentrations of the deleting LCMV-p33 peptide for 2 h before coculture with thymocytes. Thymocytes were harvested after 22 h incubation and stained with anti-CD4–PE, anti-CD8–FITC, and the dye, 7AAD. Percent survival was calculated as follows: (total number of viable CD4⁺CD8⁺ thymocytes cultured at different LCMV-peptide concentrations)/(total number of viable CD4⁺ CD8⁺ thymocytes cultured with MC57/L cells at 37°C in the absence of peptide) × 100. Total numbers of viable and dead cells were determined by trypan blue exclusion. For detection of Cdk2 activity in peptide-activated thymocytes, P14 Tg thymocytes (10⁷/well) were cultured on a monolayer of confluent (noncycling) and adherent MC57/L fibroblasts pulsed with 10⁻⁵ M of the deleting p33 peptide (37, 38). After 5 h incubation, thymocytes were separated from adherent fibroblasts and subjected to Cdk2 kinase assays as above.

DO.11.10 males were mated with estrous BALB/c females. On day 16 of gestation, pregnant females were killed and embryonic thymi were harvested and placed in culture. Thymi were microdissected and placed on the surface of 0.8-mm filters (Nucleopore) resting on Gelfoam gelatin sponges (Upjohn) in RPMI 1640 medium supplemented with 10% FCS. Each sponge was placed into a 3.5-cm plastic dish in 2 ml of medium. Cultures were incubated at 37°C. Chicken (c)OVA protein was added at 1 mg/ml on day 1 of culture, and thymi were analyzed 20 and 40 h later. Cell numbers were determined by counting in the presence of trypan blue.

To determine whether the cell cycle machinery has a role in apoptosis of noncycling G1 CD4⁺CD8⁺ thymocytes, we analyzed the activity of several Cdks in these cells. Freshly isolated thymocytes were treated with different death stimuli such as dexamethasone, heat shock, γ-irradiation, and CD95 for different time points, and Cdk activities of stimulated as well as control thymocytes were assessed in in vitro kinase assays. Surprisingly, increased activity of the cyclin-dependent kinase Cdk2 was detected within 30 min of dexamethasone activation and peaked at 5 h (Fig. 1 A). Cdk2 activity was rapidly increased in response to all apoptotic stimuli tested, including dexamethasone, anti-Fas (CD95) cross-linking, heat shock, or γ-irradiation (Fig. 1 B). No changes in the kinase activities of Cdk4 or Cdc2 were observed after induction of apoptosis (Fig. 1, A and C). Cdc2 activation was also not observed using an anti-Cdc2 phosphorylation epitope-specific Ab indicative of Cdc2 activation. Cdc2 and Cdk4 activities were readily detectable in cycling T lymphoma cells (not shown). The expression levels of molecules involved in the cell cycle, such as Cdk2, Cdk4, Cdc2, Cdk7, Pctaire-2, Cdc25A, cyclins A, B, D1, D2, D3, and E, p21, p27^Kip1, and E2F-1, did not change 5 h after stimuli (not shown). Cdk2 was found to bind to cyclin A and E thymocytes after treatment with dexamethasone and γ-irradiation. Immunoprecipitations of both cyclin A and cyclin E showed that after γ-irradiation or dexamethasone, histone H1 was phosphorylated, suggesting that both cyclin A and cyclin E have a role in Cdk2 activation (Fig. 1 E). These results show that the induction of thymocyte apoptosis by dexamethasone, γ-irradiation, heat shock, or CD95 cross-linking leads to the activation of the cell cycle regulator Cdk2.

To test whether Cdk2 activity was required for the induction of thymocyte apoptosis, the effects of two specific inhibitors of Cdk2, olomoucine and roscovitine (39, 40), were examined. These inhibitors are purine analogues that selectively inhibit the activity of Cdk2 and Cdc2 by specific binding to the ATP-binding pocket. Both molecules completely inhibited dexamethasone-induced Cdk2 activation in thymocytes (Fig. 1 D), and blocked thymocyte apoptosis after stimulation with dexamethasone, heat shock, γ-irradiation, PMA, or the DNA damaging agent, etoposide (Fig. 2 A). PI staining confirmed that thymocytes were in the G1 phase of the cell cycle and that induction of apoptosis did not correlate with cell cycle progression (Fig. 2 D). A “point of no return” was reached between 15 and 30 min after treatment with the apoptotic stimulus such that a Cdk2 blocker added after this time was unable to prevent apoptosis (Fig. 2 B). Addition of other cell cycle blockers such as TGF-β1 or rapamycin, used at the optimal concentrations, had no effect on the kinetics or extent of thymocyte death (Fig. 2 C). Interestingly, although CD95 cross-linking led to strong Cdk2 activation (Fig. 1 B), Cdk2 blockers did not inhibit CD95-mediated thymocyte death (Fig. 2 A). In fact, CD95-mediated apoptosis was consistently enhanced in the presence of the Cdk2 blockers. These results imply that CD95 uses a pathway other than the one used by the other inducers, i.e., a receptor/ caspase 8 pathway instead of a nucleus/mitochondrial/ caspase 9 pathway (41). It should be noted that in our screen, CD95 activation is the only thymocyte death stimulus so far that cannot be blocked by Cdk2 inhibition.

The process of clonal deletion and selection-triggered thymocyte death is a fundamental mechanism required for the maintenance of lymphocyte homeostasis and immunotolerance. CD4⁺CD8⁺ thymocytes expressing TCRs which recognize self-antigens with high affinity/ avidity are clonally deleted via apoptosis, leading to the removal of T cells that express TCRs with potentially harmful self-reactivity (thymic tolerance [29]). To test whether Cdk2 is a physiological regulator of thymocyte apoptosis, we induced apoptosis of CD4⁺CD8⁺ immature thymocytes by anti-CD3 cross-linking (42). Fig. 2 A shows that Cdk2 inhibitors were able to block anti-CD3–mediated apoptosis of immature thymocytes.

To further investigate the role of Cdk2 in clonal deletion, we used an in vitro negative selection system using thymocytes from P14 Tg mice. P14 Tg mice express a rearranged TCR α/β chain reactive to the p33 peptide of LCMV. P14 Tg CD4⁺CD8⁺ thymocytes undergo apoptosis after culture with APCs pulsed with different concentrations of the deleting p33 peptide. Thymocytes from P14 Tg mice underwent apoptosis in a p33 peptide dose- dependent fashion which was inhibited by the addition of Cdk2 blockers (Fig. 3 A). Importantly, induction of peptide-specific apoptosis of P14 Tg thymocytes triggered Cdk2 kinase activity (Fig. 3 B). Inhibition of Cdk2 did not interfere with TCR-mediated proximal signaling events or with TCR internalization, which is a functional measure of antigen receptor–mediated activation (not shown). Moreover, inhibition of Cdk2 by olomoucine blocked OVA-mediated negative selection of CD4⁺CD8⁺ OVA-specific TCR Tg thymocytes in fetal thymic organ cultures (FTOCs; Fig. 3 C). Cdk2 blockers did not interfere with positive thymocyte selection in reaggregation culture assays (not shown), indicating that Cdk2 has a specific role in peptide-specific thymocyte apoptosis.

Where does Cdk2 function in the hierarchy of apoptosis? Disruption of ΔΨ_m due to the opening of mitochondrial pores has been invariably associated with apoptosis and is an early common denominator of cell death (43). Alterations in mitochondria lead to the release through the outer mitochondrial membrane of molecules such as cytochrome c and the apoptosis-inducing factor (AIF), and the activation of the caspase cascade (7, 35, 36, 44–48).

To assess whether Cdk2 acts upstream or downstream of mitochondrial events, we examined changes in ΔΨ_m using cytometry and the fluorochromic dye, DiOC₆(3). Thymocytes were stimulated either with dexamethasone or anti-CD95, and the mitochondria changes of ΔΨ_m were assessed at different time points. The first changes in thymocyte ΔΨ_m were observed 2 h after dexamethasone treatment, and ΔΨ_m was significantly disrupted after 5 h (Fig. 4 A). Addition of Cdk2 inhibitors blocked dexamethasone-induced losses of ΔΨ_m (Fig. 4 A). CD95-mediated ΔΨ_m disruption and apoptosis still occurred in the presence of Cdk2 inhibitors (Figs. 2 A and 4 A), implying that Cdk2 inhibition per se does not interfere with opening of mitochondrial pores. Since ΔΨ_m is regulated by Bcl-2 family members (43, 49), we also tested Cdk2 activation in Bcl-2 Tg thymocytes (31, 50). Although overexpression of Bcl-2 protected thymocytes from dexamethasone- and irradiation-induced cell death and disruption of ΔΨ_m (31, 50), Cdk2 was still activated in Bcl-2 Tg thymocytes in response to these apoptotic stimuli (not shown).

Caspase activation is a crucial event in apoptosis, and caspases can function upstream or downstream of mitochondrial ΔΨ_m disruption (8, 9). To determine where Cdk2 acts during apoptosis with regard to the caspase activation cascade, processing of different caspases was assessed in thymocytes after treatment with different apoptotic stimuli in the presence or absence of Cdk2 inhibitors. Within 2 h after dexamethasone and γ-irradiation, caspase 3 (Cpp32) and caspase 8 activation was observed in thymocytes whereas caspase 2 (nedd2) processing was first observed 3 h after death induction. Caspase activation peaked at 5 h after induction of cell death (Fig. 4 B, and data not shown). However, activation of caspase 3 (Fig. 4 B), caspase 2 (Fig. 4 C), or caspase 8 (not shown) did not occur after blocking of Cdk2 kinase activity. These results show that Cdk2 acts upstream of Bcl-2, ΔΨ_m, and caspases.

During apoptosis, various cell cycle regulatory molecules such as p21 and Rb are proteolytically cleaved by caspases. In particular, proteolytic processing of the G1 to S cell cycle gatekeeper Rb (ΔRb) has been previously reported in TNF- and CD95-treated tumor cell lines (51, 52). Rb and Cdk2 were found to coimmunoprecipitate in developing thymocytes (not shown). Induction of thymocyte apoptosis in response to dexamethasone, irradiation, heat shock, or anti-CD95 correlated with the appearance of a second smaller Rb protein (ΔRb; Fig. 4 D). Although it has been shown that Rb is cleaved by caspase 3 (53) and in thymocytes ΔRb was found to be a proteolytic cleavage product of Rb mediated by caspases, ΔRb was still observed in caspase 3 gene–deficient mice (not shown). The earliest detectable Rb cleavage (ΔRb) occurred 5 h after dexamethasone treatment (Fig. 4 D). Cdk2 inhibitors or transgenic overexpression of Bcl-2 in thymocytes prevented cleavage of Rb in response to dexamethasone (Fig. 4 D, and data not shown). These results demonstrate that Cdk2 acts upstream of mitochondrial pore opening, Bcl-2, caspase activation, and proteolytic cleavage of the cell cycle regulator Rb. The functional consequences of Rb cleavage are not known. Since ΔRb can only be observed downstream of the caspase effector phase and still binds to Cdk2 (not shown), the generation of ΔRb might function as a regulatory feedback loop that could influence Cdk2 and/or E2F-1 activity.

How is Cdk2 activity mechanistically linked to apoptotic mitochondrial events? Various members of the Bcl-2 family of mitochondrial gatekeepers are phosphorylated on serine/threonine residues (9, 49). Although Bcl-2 and Bcl-XL contain consensus sites for Cdk2 activity, we could not detect Cdk2-mediated phosphorylation of either Bcl-2 or Bcl-XL in in vitro kinase assays (not shown). The tumor suppressor p53 is a substrate for Cdk2 in the DNA repair response (54), and thymocytes mutated in p53 are resistant to γ-irradiation–induced apoptosis but still susceptible to dexamethasone and antigen receptor–mediated cell death (22, 23). The effect of the p53 mutation has been mapped upstream of apoptotic mitochondrial events (55).

Therefore, we tested whether p53 is a target for Cdk2 activity during thymocyte apoptosis after γ-irradiation. In vitro kinase assays using immunoprecipitated Cdk2 from γ-irradiated and dexamethasone-treated thymocytes showed that Cdk2 can phosphorylate p53 (Fig. 5 A). Moreover, p53 was found to associate with Cdk2 in thymocytes (Fig. 5 B). To test the effect of Cdk2 activity on p53 expression, we analyzed the levels of p53 protein in γ-irradiated thymocytes in the presence or absence of Cdk2 inhibitors. Although p53 protein accumulated to significant levels after treatment of cells with γ-irradiation alone, little p53 accumulation was observed when cells were treated with γ-irradiation in the presence of Cdk2 blockers (Fig. 5 C). Irradiation-induced p53 protein accumulation was caused by enhanced p53 protein stability but not by p53 gene transactivation (not shown).

To further corroborate the regulation of p53 by Cdk2, we examined the expression of the p53-inducible death promoter Bax (56–58) by Northern blotting. Induction of thymocyte apoptosis by γ-irradiation led to an increase in Bax transcripts, and Bax transactivation was found to depend on Cdk2 activity (Fig. 5 D). In dying thymocytes we found only induction of the p53-regulated death promoter Bax but not transactivation of the p53-regulated gene p21 (not shown). Cdk2 kinase activity was normally induced in γ-irradiated p53^−/− thymocytes (not shown), indicating that p53 is downstream of Cdk2 in the thymocyte death signaling cascade. Since thymocytes from p53^−/− mice are not resistant to dexamethasone or antigen receptor–mediated apoptosis, other molecules must exist that link Cdk2 activation to cell death.

The identification of Cdk2 as a master regulator of cell death provides the first evidence for a shared signaling pathway that integrates multiple death signaling pathways into a common death effector cascade in developing thymocytes. The hierarchy of Cdk2 action suggests that Cdk2 is the earliest known common signaling element required for thymocyte apoptosis in response to environmental and developmental cues such as negative selection. This hypothesis is based on the following findings: (a) all nonspecific (γ-irradiation, heat shock, dexamethasone) and specific (peptide-mediated thymocyte cell death) apoptotic stimuli tested induce rapid activity of the cyclin-dependent kinase Cdk2 in noncycling thymocytes; (b) Cdk2 acts upstream of the opening of mitochondrial pores, Bcl-2 family proteins, caspase activation, p53, and proteolytic processing of Rb; (c) inhibition of Cdk2 completely protects thymocytes from γ-irradiation, heat shock, dexamethasone, PMA, anti-CD3, and peptide-mediated cell death; (d) Cdk2 and the tumor suppressor, p53, constitutively associate in thymocytes and activated Cdk2 isolated from apoptotic thymocytes can phosphorylate p53; (e) Cdk2 regulates p53 protein accumulation and transactivation of the p53-inducible death promoter, Bax, after γ-irradiation. These data provide the first link between the cell cycle machinery and apoptosis in normal development and differentiation and indicate that Cdk2 is a crucial kinase that mediates cell death in thymocyte maturation and thymocyte selection.

Our results indicate that after γ-irradiation, Cdk2 phosphorylates and stabilizes p53, which then transactivates the death promoter Bax. Interestingly, in dying thymocytes we found only induction of the p53-regulated death promoter Bax but not transactivation of the p53-regulated gene p21 (59), suggesting that only certain gene loci are accessible for p53 transactivation or that other cofactors act in concert with p53 to modulate gene expression in a tissue- and lineage-specific manner. Although p53 protein levels were increased in thymocytes after dexamethasone and γ-irradiation, it has been shown in p53 gene–deficient mice that p53 protects thymocytes only from DNA damage, and not from dexamethasone or antigen receptor–mediated cell death (22, 23), implying that other downstream molecules exist that link Cdk2 to apoptosis. Preliminary evidence from our laboratory implies that the glucocorticoid receptor which is required for dexamethasone-mediated cell death can be phosphorylated by activated Cdk2 and coimmunoprecipitates with Cdk2 in dying thymocytes. Besides association with the ligand, phosphorylation of the glucocorticoid receptor is required for its translocation from the cytoplasm into the nucleus (60–62). In addition to the glucocorticoid receptor, other orphan steroid receptors such as Nur77 might be molecular targets for Cdk2 kinase activity.

It has been shown that mitochondria are early checkpoints that integrate multiple death signaling pathways into a common Ced4/caspase-regulated effector mechanism. Opening of mitochondrial pores, mitochondrial swelling, disruption of ΔΨ_m, and release of proapoptotic molecules, including cytochrome c and apoptosis-inducing factor (AIF), from the mitochondrial intermembrane spaces into the cytoplasm have all been implicated as fundamental mechanisms that initiate and propel the effector phase of apoptosis. Posttranslational modification and the balance between death suppressors, such as Bcl-2 and Bcl-XL, and death promoters, including Bax and Bad, are crucial mechanisms of mitochondrial integrity and the apoptotic effector phase (9, 49). Our results show that Cdk2 acts upstream of ΔΨ_m disruption, Bax and Bcl-2, and caspase activation in developing thymocytes. Moreover, whereas loss of the mitochondrial transmembrane potential can only be observed 2 h after addition of death stimuli, Cdk2 kinase activity is induced very rapidly and a “point of no return” was reached between 15 and 30 min after treatment with the apoptotic stimulus, such that a Cdk2 blocker added after this time was unable to prevent apoptosis. Thus, our results indicate that Cdk2 is the earliest known common denominator that can integrate many independent apoptotic signals into one common effector pathway.

Although CD95 (Fas) stimulation induced Cdk2 activity in thymocytes, inhibition of Cdk2 did not block CD95-mediated apoptosis. In fact, Cdk2 inhibition enhanced the susceptibility to CD95 killing. So far, CD95-mediated apoptosis is the only death signal in thymocytes that does not rely on Cdk2 activation. Although apoptosis in response to γ-irradiation, heat shock, dexamethasone, or peptide-specific negative thymocyte selection requires active transcription of death genes, apoptosis after CD95 killing can occur in the presence of RNA or protein synthesis inhibitors (9). Moreover, enucleated cells can undergo apoptosis after CD95 activation, suggesting that all components necessary for CD95-mediated apoptosis are present in cells and that CD95 activation can directly trigger the apoptotic machinery. It should be noted that both Cdk2 inhibitors roscovitine and olomoucine do not prevent transcription or translation in noncycling neurons (63) and that inhibition of Cdk2 did not regulate transactivation of the FasL in thymocytes (not shown).

The specific Cdk2 blockers olomoucine and roscovitine are purine analogues that inhibit Cdk2 kinase activity by specifically binding to the Cdk2 ATP-binding site (39, 40). At concentrations at which olomoucine and roscovitine blocked apoptosis in in vivo thymocyte cultures, these inhibitors had no effects on in vitro kinase activity of PKCα-ζ, Cdk4, Cdk6, Abl, cAMP- or GMP-dependent protein kinases (PKA, PKG), mitogen-activated protein kinase (MAPK; extracellular signal regulatory kinase [ERK]1, ERK2), Src family kinases, glycogen synthase kinase 3 (GSK3), casein kinase, receptor tyrosine kinases, myosin light chain kinase, p38/HOG, or stress-activated protein kinase (SAPK)/c-Jun NH₂-terminal kinases (JNKs) (39; and data not shown). Through their unique selectivity for Cdk2, roscovitine and olomoucine provide a unique opportunity to study the role of Cdk2 in thymocyte apoptosis. However, we cannot exclude that roscovitine and olomoucine inhibit a yet unidentified kinase, and our results need to be confirmed using genetic model systems. Thus, genetic systems for inducible and thymocyte-specific inactivation/activation of Cdk2 need to be developed in the future. However, our results showing that all death stimuli lead to Cdk2 kinase activity in thymocytes and that two different Cdk2 kinase inhibitors, but not other inhibitors that block G1 to S progression, inhibit thymocyte apoptosis strongly suggest that Cdk2 is a key kinase involved in thymocyte apoptosis.

Cdk2 is crucial for the progression from the G1 to the S phase of the cell cycle. Inhibition of Cdk2 activity in vitro has been shown to protect cultured sympathetic neurons and heart muscle cells from apoptosis (26, 27). Our results in noncycling CD4⁺CD8⁺ thymocytes provide the first evidence that Cdk2 has a crucial role in the induction of cell death during normal development. However, it has been shown that Cdk2 inhibition can also lead to cell death in tumor cell lines, and Cdks are frequently deregulated in tumors (28). Similarly, we found that in contrast to developing, noncycling thymocytes, inhibition of Cdk2 in four different thymic lymphoma cell lines led to rapid apoptosis and sensitized T cell tumors to anti-CD3, dexamethasone, or γ-irradiation–mediated cell death (not shown). These results suggest that Cdk2 has functions in the apoptotic processes that regulate normal development which are distinct from those in tumorigenesis and transformation. Thus, specific inhibition of Cdk2 could be exploited to sensitize tumor cells to apoptosis by anticancer drugs, whereas molecular inhibition of Cdk2 might protect normal, noncycling cells from the adverse effects of the same drugs.

We thank L. Meijer and R. Sekaly for reagents; R. Hakem, K. Bachmaier, Y.Y. Kong, H. Nishina, A. Oliveira-dos-Santos, L. Zhang, V. Stambulic, L. Harrington, J. Krassopolous, D. Bentley, J. Woodgett, P. Ohashi, T.W. Mak, and G. Kroemer for helpful discussions; and M. Saunders for scientific editing.

aa

amino acid(s)

Cdk

cyclin-dependent kinase

ΔΨ_m

mitochondrial transmembrane potential

FTOC

fetal thymic organ culture

GST

glutathione S-transferase

LCMV

lympholytic choriomeningitis virus

PI

propidium iodide

PK

protein kinase

Rb

retinoblastoma protein

Tg

transgenic