Autoregulatory Mechanisms of Phosphorylation of Checkpoint Kinase 1

In response to replication perturbation or DNA damage, cells activate elegant genome surveillance pathways, called cell-cycle checkpoints, to counter these assaults. Central to these surveillance pathways are 2 protein kinases, the upstream kinase, ATR (ataxia telangiectasia mutated and Rad3 related), and its downstream target kinase, checkpoint kinase 1 (Chk1). Complete loss of CHK1 or ATR leads to embryonic lethality in mice (1–3). On the other hand, partial loss of these 2 genes, for instance loss of one copy of CHK1 or a hypomorphic mutation in ATR, increased genome instability and caused spontaneous cell death even in the absence of extrinsic stress (4–6). These findings suggest that these 2 proteins play key roles in monitoring the DNA replication and in maintaining the genome integrity (7). Therefore, targeting ATR and Chk1 has the potential to selectively enhance the antitumor effect for tumors that undergo increased replicative stress (8, 9).

Chk1 primarily responds to replication fork interference in the S phase and DNA damage at the G₂ phase (10–12). Chk1 is composed of a highly conserved kinase domain at the N-terminal half and a regulatory region at the C-terminal half. The C-terminus contains a Ser/Gln (SQ) motif and 2 highly conserved motifs (CM1 and CM2, Fig. 1A). Recent studies from this laboratory and others showed a model of Chk1 activation that requires protein conformation change of Chk1. Under normal growth conditions, Chk1 seems to adopt a “closed” conformation through an intramolecular interaction between the N-terminus and the C-terminus (13–15). This closed conformation not only suppresses the kinase activity of Chk1 but also stabilizes the protein (14, 16). Upon DNA damage, Chk1 undergoes ATR-dependent phosphorylation on chromatin (17, 18). This phosphorylation seems to disrupt the intramolecular interaction, leading to an “open” conformation of Chk1 followed by checkpoint activation (16).

Phosphorylation of Chk1 at 2 conserved ATR sites, Ser-317 and Ser-345, has long been viewed as the gold standard for the activation of replication checkpoints. However, precise molecular mechanisms controlling Chk1 phosphorylation are less well understood. In this study, we uncover a number of novel mechanisms buried within the Chk1 polypeptide that control Chk1 protein phosphorylation, checkpoint activation, and the maintenance of cell viability.

HEK293T, HeLa, U2-OS, and A549 cells were cultured in Dulbecco's Modified Eagle's Medium with 10% FBS. Transfection was carried out with either calcium phosphate or Lipofectamine 2000 (Invitrogen). For the cell proliferation assay, HEK293T cells were transfected with GFP, GFP-Chk1 wild type (WT), or the L449R mutant for 48 hours. The cells were reseeded at a density of 1 × 10⁴ cells per well in 6-well plates and cultured for 8 days. On each day, the number of GFP-positive cells within a colony was counted under fluorescence microscopy.

Myc- or GFP-tagged vectors expressing Chk1 WT or mutants were generated using standard PCRs. Point mutations were carried out using the Quick Change Mutagenesis Kit (Stratagene). Primer information will be provided upon request.

Cell-cycle analyses and immunoblotting were carried out as previously described (17, 19). Anti-Chk1 (DCS-1310 and G4) and anti-ATR (N-19) antibodies were from Santa Cruz. Anti–phospho-S317-Chk1, anti–phospho-S345-Chk1, anti–phospho-S1981-ATM, and anti–phospho-S216-Cdc25C were from Cell Signaling. Anti-MCM7 and anti–Cyclin B were from BD Pharmingen. Anti-Cdc25A was from NeoMarkers.

HeLa Tet/Off cells grown on glass cover slips were transfected with tetracycline-controlled GFP, GFP-Chk1 WT, or the L449R mutant for 24 hours, synchronized at the G₂–M phase with 100 ng/mL nocodazole for 20 hours in the presence of doxcycline, washed twice with 1 × PBS, and released into fresh medium without doxcycline. After a 16-hour release from nocodazole, cells were pulse labeled with 10 μmol/L 5-ethnyl-2′-deoxyuridine (EdU), a nucleotide analog, for 20 minutes at 0-, 4-, 8-, and 10-hour time period. Cells were then washed with ice-cold PBS and fixed with 3.7% formaldehyde at room temperature for 10 minutes and followed with the Click-iT kit to measure EdU incorporation according to the manufacturer's instruction (Invitrogen).

The kinase CM1 and CM2 domains of Chk1 (Fig. 1A) are highly conserved among different species. However, whether these domains are required for initiating Chk1 phosphorylation at ATR sites is unknown. To address this question, we generated Myc-tagged mutants in which the kinase domain (the C mutant), the CM2 domain (the 1-421 mutant) or the CM1 plus CM2 domain (the 1-368 mutant) of human Chk1 were deleted (Fig. 1A). All these mutants contain the 2 ATR phosphorylation sites, Ser-317 and Ser-345. HEK293T cells expressing these mutants were treated with a DNA-damaging agent, the topoisomerase 1 inhibitor, camptothecin (CPT), and immunoblotted with anti–phospho-Chk1 antibodies. The results showed that deletion of the kinase CM2 or the CM1 plus CM2 domain did not abolish Chk1 phosphorylation (Fig. 1B, lanes 10–12). These data suggested that none of these 3 conserved domains were essential for Chk1 phosphorylation at ATR sites.

Previous studies reported that mutating the Ser-317 to Ala abolished phosphorylation at the Ser-345 site of Chk1 (15, 20). Here we asked whether the phospho-mimic S317E mutation would induce Chk1 phosphorylation at Ser-345. Consistent with previous publications, the S317A mutant failed to undergo phosphorylation at either the Ser-317 or the Ser-345 site (Fig. 1B, lane 6). However, the S317E mutant also failed to be phosphorylated at the Ser-345 site (Fig. 1B, lane 7), indicating that S317E is not a true phosphomimic mutation or the Ser-317 residue is critical for phosphorylation at Ser-345. In contrast, mutating the Ser-345 site to Ala or Glu only moderately reduced Chk1 phosphorylation at the Ser-317 site compared with the Chk1 WT (Fig. 1B, lanes 2, 8–9).

To further test this idea, we examined protein phosphorylation of more refined mutations of the Chk1 C-terminus expressing one or both phosphorylation sites (C1 to C5 in Fig. 1C) upon CPT treatment. The small fragment C1, which only contains the Ser-317 site, was phosphorylated at the Ser-317 site (Fig. 1D, lanes 3–4). However, the C2, C4, or C5 fragment, which only contains the Ser-345 site, was not phosphorylated at Ser-345 (Fig. 1D, lanes 5–6 and 9–12). Only when the fragment contains both Ser-317 and Ser-345 sites (i.e., the C3 fragment), can phosphorylation at Ser-345 be detected (Fig. 1D, lanes 7–8). These data suggested that either the Ser-317 residue or its phosphorylation is required for phosphorylation at Ser-345, but not the other way around (15, 20, 21).

We recently reported that 3 highly conserved Arg residues (R372/376/379, Supplementary Fig. S1B) in the CM1 region of Chk1 play an important role in maintaining Chk1 protein conformation (14). Thus, we asked whether mutating these residues could affect Chk1 phosphorylation upon DNA damage. Our results showed that Chk1 phosphorylation at both Ser-317 and Ser-345 in the 3RE mutant was significantly reduced compared with the Chk1 WT (Fig. 1B, lanes 2 and 4). This seemed to be because of the significantly increased cytoplasmic localization of this 3RE mutant (14). On the other hand, the 3RA mutant is located mainly in the nucleus like the WT (data not shown). Interestingly, although the 3RA mutant was highly phosphorylated at the Ser-317 site, phosphorylation at the Ser-345 site was significantly reduced compared with the Chk1 WT (Fig. 1B, lanes 2–3). We also noticed that the Chk1 (1-421) mutant exhibited more profound reduction in phosphorylation at the Ser-345 site than the Ser-317 site compared with the Chk1 WT (Fig. 1B, lanes 2 and 10). The Chk1 kinase dead (D148A) mutant exhibited reduced phosphorylation at both sites (Fig. 1B, lane 5), probably because this mutant is less stable than the Chk1 WT (Supplementary Fig. S1A). These results showed that high level phosphorylation at Ser-317 does not necessarily correlate with high-level phosphorylation at Ser-345. Thus, even though the CM1 and CM2 domains are not essential for initiating Chk1 phosphorylation, they seem to contribute to maximal phosphorylation at the Ser-345 site by DNA damage. These data are consistent with yeast Chk1 whose C-terminus contributed to the full activation of Chk1 (22).

Together, these results indicated that phosphorylation at Ser-317 is necessary, but not sufficient, for high level of Chk1 phosphorylation at the Ser-345 site. This is in line with the idea that phosphorylation at the Ser-345 site is the final determinant of full activation of Chk1 (21). Therefore, we focused mainly on Chk1 phosphorylation at the Ser-345 site for the following studies.

During our analysis, we unexpectedly discovered that the Chk1 C-terminus devoid of the kinase domain was constitutively phosphorylated at both Ser-317 and Ser-345 sites (Fig. 1D and 2A, lane 1). CPT treatment did not further increase protein phosphorylation compared with the basal state (Fig. 1D and 2A, lanes 1–2). We also noticed a weak constitutive phosphorylation of the C3 fragment in the absence of DNA damage (Fig. 3A lane 5 and Fig. 3B lane 6), similar to the entire Chk1 C-terminus (Fig. 3B, lane 3). These results suggested that the N-terminal kinase domain suppresses phosphorylation of residues located at the C-terminus of Chk1 in the absence of DNA damage. To understand the biologic implications of this constitutive phosphorylation of the Chk1 C-terminus, we asked whether this Chk1 C-terminus displayed properties similar to those that govern phosphorylation regulation as the full-length (FL) Chk1.

First, we observed that the constitutive phosphorylation of this fragment was completely abolished when Ser-317 was mutated to Ala or Glu (Fig. 2A, lanes 3–4). Mutating the Ser-345 to Ala or Glu significantly reduced the constitutive phosphorylation of the Chk1 C-terminus (Fig. 2A, lanes 5–6); however, treating these cells with CPT increased Ser-317 phosphorylation (data not shown). These results suggested that the Chk1 C-terminus undergoes phosphorylation in a way similar to the Chk1 FL protein (Fig. 1A). Second, we asked whether ATR is also responsible for this constitutive phosphorylation of the Chk1 C-terminus. Our results showed that inhibiting ATR, and to a lesser extent, ATM, but not DNA-PK, reduced the levels of phosphorylated proteins of this Chk1 C-terminal fragment (Fig. 2B). In parallel, similar effects were observed on phosphorylation of the Chk1 FL induced by CPT (Fig. 2C). Furthermore, we showed that depletion of ATR, but not ATM, clearly reduced the level of constitutive phosphorylation of the Chk1 C-terminal fragment under normal growth condition (Fig. 2D, lane 2). Owing to the fact that ATM-dependent activation of ATR only occurs in the presence of DNA damage (7), these results suggested that ATR is the predominant kinase that causes the constitutive phosphorylation of the Chk1 C-terminal fragment in nontreated cells.

Previously, the C-terminus of Chk1 was proposed to form an intramolecular interaction with the N-terminal kinase domain (13–15), so that the catalytic activity of the open kinase domain of Chk1 is suppressed under normal conditions (ref. 16; see model in Supplementary Fig. S5). Here we showed that deletion of the N-terminal kinase domain led to constitutive phosphorylation of the C-terminus of Chk1, suggesting that another purpose of this intramolecular interaction might be to suppress Chk1 phosphorylation in the absence of DNA damage.

If this hypothesis is correct, then we would expect that interrupting the interaction between the N-terminal kinase domain and the C-terminus of Chk1 should lead to constitutive phosphorylation of endogenous Chk1 under normal conditions. To address this issue, we overexpressed those small fragments of Chk1 (Fig. 1C) into HEK293T cells and examined phosphorylation of endogenous Chk1. This experimental design was based on the assumption that one or more than one of those exogenous small Chk1 fragments will compete with the C-terminus of endogenous Chk1 for the interaction with the N-terminal kinase domain of the same Chk1 molecule and disrupt the closed conformation of Chk1. As a result, the C-terminus of endogenous Chk1 is now exposed to undergo constitutive phosphorylation at ATR sites (see model in Supplementary Fig. S5). Our results showed that expression of the C5 fragment, and to a lesser extent, the C6 fragment, but not other small fragments, induced phosphorylation of endogenous Chk1 under normal growth conditions (Fig. 3A, lanes 9 and 11). Treatment with a replicative stress, hydroxyurea, only moderately further increased the phosphorylation signal of endogenous Chk1 proteins in the presence of the Chk1 C5 fragment (Fig. 3A, lanes 9–10).

We further showed that overexpression of the C5 fragment, and to a lesser extent, the entire C-terminus, induced strong phosphorylation of endogenous Chk1 proteins in another cell line, HeLa (Fig. 3B, lanes 3 and 8), indicating that this is not a cell line–specific effect. On the other hand, the Chk1 FL, the N-terminal kinase domain, or other fragments failed to do so (Supplementary Fig. S2A). The reason why the C5 fragment, which contains the CM1 and CM2 domains, caused the strongest phosphorylation of endogenous Chk1 is probably because this fragment interacts most strongly with the N-terminal kinase domain (Supplementary Fig. S2C), thereby providing the maximal interference of the closed conformation of endogenous Chk1 (see the model in Supplementary Fig. S5).

If the CM1 and CM2 domains were critical for preventing phosphorylation of Chk1 under normal conditions, then we would expect to identify key residues within these domains, whose mutation should disrupt the intramolecular interaction and lead to constitutive phosphorylation of Chk1 in the absence of DNA damage. To address this issue, we generated GFP-tagged FL Chk1 vectors, in which essentially every residue within the CM1 and CM2 domains was mutated and examined protein phosphorylation with or without CPT treatment. The majority of these point mutations did not show constitutive phosphorylation of Chk1 (data not shown). However, mutating one of 2 residues in the CM2 domain (G448 or L449) led to constitutive phosphorylation of GFP-Chk1 in the absence of DNA damage (Fig. 4, lanes 9 and 11). Phosphorylation of endogenous proteins, including Chk1 and ATM, was not observed in cells expressing these 2 mutants (Fig. 4, endogenous Chk1), indicating the lack of a pan-cellular DNA damage response. CPT treatment moderately increased the phosphorylation signal of these 2 mutants (Fig. 4, lanes 9–12). These data suggested that these 2 residues (G448 and L449) play important roles in suppressing constitutive phosphorylation of Chk1 under normal conditions.

To further understand the physiologic relevance of the constitutive phosphorylation of these Chk1 mutants, we first asked whether it is a cell line–specific effect or not. We consistently detected high levels of constitutive phosphorylation of Chk1 mutants in HeLa, U2-OS, A549, or HCT116 cell lines (Fig. 5A in HeLa cells, lanes 2–3; data not shown). In contrast, phosphorylation of the Chk1 WT was only detected by CPT treatment (Fig. 5A, lanes 1 and 4). Thus, constitutive phosphorylation of these 2 Chk1 mutants is not restricted to one system or cell line.

Second, we asked whether this constitutive phosphorylation is also ATR dependent. Inhibiting ATR, but not other kinases, reduced the level of constitutive phosphorylation of the Chk1 L449R mutant (Fig. 5B, lane 2). Considering that cells were only treated with caffeine for hours although they had expressed the L449R mutant for days, the reduction in Chk1 phosphorylation is significant. Depletion of ATR significantly reduced phosphorylation of the Chk1 L449R mutant, in a way similar to endogenous Chk1 (Fig. 5C). Together, these data strongly indicated that constitutive phosphorylation of the Chk1 L449R mutant is ATR dependent.

Third, we asked whether the L449R mutant follows a similar cell-cycle–dependent expression pattern as endogenous Chk1 whose expression peaks in the S to G₂ phase (23). The results showed that indeed the level of the GFP-Chk1 L449R mutant was the highest from S to G₂ phase, similar to endogenous Chk1 (Fig. 5D, compare the anti-GFP and the anti-Chk1 blots). No phosphorylation was detected for endogenous Chk1 in the absence of DNA damage; however, phosphorylation of the L449R mutant was detected throughout the cell cycle, with the highest in the S phase (Fig. 5D, lane 3 in the anti-pS345 blot). Together, these data suggested that the Chk1 L449R mutant undergoes the same regulation as the endogenous Chk1 protein.

To understand the biologic significance of the constitutive phosphorylation of Chk1, we first asked whether it induced an artificial S-phase checkpoint. To this end, we transfected HeLa Tet/Off cells with tetracycline-regulated expression vectors for GFP, GFP-Chk1 WT, or the L449R mutant and examined expression of Cdc25A. Cdc25A is a key Chk1 downstream target that plays a crucial role in regulating both the entry and the progression of S phase (24). Activation of Chk1 leads to the proteasome-dependent degradation of Cdc25A followed by S-phase progression inhibition (24). The results showed that the Chk1 WT only slightly reduced the level of Cdc25A compared with GFP alone (Fig. 6A, lanes 3–4), in agreement with our previous report (14). In contrast, expression of the L449R mutant almost completely blocked Cdc25A expression (Fig. 6A, lane 5), as did the DNA damage agent hydroxyurea (Fig. 6A, lane 2). Consistent with the reduction of Cdc25A, only hydroxyurea treatment and the L449R mutant showed Chk1 phosphorylation at Ser-345 (Fig. 6A, lanes 2 and 5). In vitro kinase assay showed that the L449R mutant exhibited much stronger autophosphorylation than the Chk1 WT (Supplementary Fig. S2D). These results showed that the L449R mutant is functionally more active than the Chk1 WT, both in vitro and in vivo.

Subsequently, we tested the effect of the Chk1 L449R mutant on the S-phase progression. HeLa Tet/Off cells blocked at the G₂–M phase by nocodazole were released into the cell cycle with concomitant expression of GFP, GFP-Chk1 WT, or the L449R mutant. Our preliminary data showed that a 16-hour release after nocodazole treatment would allow normal HeLa cells to start entering the S phase (data not shown). Thus, we monitored DNA synthesis over a 10-hour period beginning at 16 hours of nocodazole release by measuring the incorporation of EdU, a nucleotide analog (see Supplementary Fig. S3A for experimental design). Fluorescence microscopy revealed that the GFP-Chk1 WT and the L449R mutant were nearly equally expressed at the end of the 16 hours of release (Supplementary Fig. S3B). Importantly, we found that less cells expressing the GFP-Chk1 L449R were incorporating EdU compared with the GFP-Chk1 WT or the GFP alone (Fig. 6B, 0 hour), indicating a delayed S-phase entry. During the subsequent 10-hour chase period, the number of cells that incorporated EdU in the GFP-Chk1 WT or GFP control group dropped much more significantly than in the L449R group (Fig. 6B, 4–10 hours). This indicated that at a time point when control cells are exiting S phase, GFP-Chk1 L449R-expressing cells remain in the S phase. We also noticed a slightly less reduction in EdU-positive cells in the GFP-Chk1 WT group than the GFP alone (Fig. 6B). This is consistent with the Cdc25A expression profile (Fig. 6A). These data indicated a delayed S-phase entry and prolonged S-phase progression caused by the L449R mutant, and to a much lesser extent, the Chk1 WT, compared with the GFP control.

To confirm the prolonged S-phase progression, we analyzed the percentage of late S-phase cells during that 10-hour chase period. Whereas early S-phase cells had a pan-nuclear EdU staining pattern, late S-phase cells exhibited punctuate or more focal EdU staining pattern (Supplementary Fig. S3C). The results showed that GFP-Chk1 L449R-expressing cells had a significantly lower percentage of late S-phase cells than the GFP-Chk1 WT or the GFP control, especially at later time points (Fig. 6C). Again, the S-phase progression in the GFP-Chk1 WT was slower than the GFP control (Fig. 6C, more obvious at 0–4 hour). Cell-cycle analyses confirmed that the L449R mutant-expressing cells progressed through S phase much more slowly than GFP control or the Chk1 WT (Supplementary Fig. S4). Together, these data showed that constitutive phosphorylation of Chk1 leads to prolonged S-phase progression.

The cell-cycle analyses showed significantly increased dead cell population for cells expressing the Chk1 L449R mutant (the sub-G₁ population in Supplementary Fig. S4), indicating that expression of constitutively active Chk1 is counterproductive to cell viability. To further test this idea, we transfected HEK293T cells with vectors expressing the GFP control, GFP-Chk1 WT or GFP-Chk1 L449R mutant, and counted GFP-positive cell numbers in each clone over 8 days. The results showed that although cells expressing GFP control expanded exponentially, cells expressing GFP-Chk1 WT had a significant delay in expansion; however, no clone expansion was observed for cells expressing the GFP-Chk1 L449R mutant (Fig. 7). Similar results were observed for HeLa, U2-OS, and HCT116 cell lines. These data suggested that constitutive activation of Chk1 suppresses tumor cell growth.

In this study, we provide evidence that the Chk1 polypeptide contains a number of critical regulatory mechanisms mediating its own phosphorylation, checkpoint activation, and cell viability. Checkpoint activation requires phosphorylation of Chk1 at both Ser-317 and Ser-345 residues. However, whereas Ser-345 is essential for cell viability, Ser-317 is not (21). This led to the idea that Ser-345 phosphorylation is the final determinant of checkpoint activation. Our data support this idea by showing that DNA damage–induced Ser-345 phosphorylation of Chk1 is tightly regulated. First, either the phosphate group or the serine residue at the position of Ser-317 is critical for inducing phosphorylation of Chk1 at Ser-345 (15, 20, 21). Second, phosphorylation of Ser-317 is not sufficient for maximal phosphorylation at Ser-345. It seems that the relief from the N-terminal kinase domain and the involvement of the CM1 and CM2 domains are also required for high-level phosphorylation of Chk1 at Ser-345 (see the model in Supplementary Fig. S5). The C-terminal CM1 and CM2 motifs seem to have dual roles in regulating Chk1 phosphorylation. In the absence of DNA damage, the CM1 and CM2 domains contribute to the inhibitory effect on Chk1 phosphorylation by interacting with the N-terminal kinase domain. On the other hand, the CM1 and CM2 domain contributes to high-level phosphorylation of Chk1 in response to DNA damage, which might be through providing a proper conformation (22) for maximal Ser-345 phosphorylation by ATR.

Previously, Chk1 has been proposed to adopt a closed conformation through an intramolecular interaction between the N-terminal kinase domain and the C-terminal regulatory domain (13–15). This conformation not only suppresses the catalytic activity of Chk1, but also stabilizes the protein (14, 16). Data presented in this study added further insights into the fold-back structure of Chk1 and its roles in checkpoint regulation. These new data indicated that under normal circumstances, while the C-terminus of Chk1 masks the kinase domain, the N-terminal kinase domain of Chk1 simultaneously suppresses Chk1 phosphorylation by ATR. These mutual inhibitory effects may provide much safer mechanisms to ensure that no accidental activation of checkpoints, through either the exposure of the catalytic domain or the phosphorylation of the SQ sites of Chk1, will be achieved under normal growth conditions.

It has long been known that inadequate Chk1 phosphorylation leads to checkpoint defects, and consequently loss of cell viability (25). These new results indicate that cells may have evolved mechanisms to prevent accidental Chk1 phosphorylation when not needed. In agreement with this idea, activation of ATR by expressing the ATR-activating domain of TopBP1 or tethering TopBP1 or claspin to DNA led to artificial checkpoint activation in the absence of DNA damage. As a result, cells undergo permanent cell-cycle arrest or senescence (26, 27). However, one potential caveat of these methods is that they activated the entire ATR pathway. In this study, we used the constitutively active Chk1 mutant (L449R) as a model to show that activation of Chk1 only, but not the entire ATR pathway, is detrimental to cell viability without exogenous DNA damage. The existence of constant S phase checkpoints posed by this constitutively active Chk1 mutant is likely to eventually stop cell division and growth. Similarly, mutating the corresponding Leu residue in budding yeast Chk1 (L506) activated checkpoints in the absence of DNA damage (28, 29), indicating that mechanisms preventing Chk1 from being accidentally phosphorylated might be highly conserved.

An interesting question is how mutation of G448 or L449 leads to constitutive Chk1 phosphorylation at ATR sites? A possible explanation is that these 2 residues sit at the key interface between the N-terminal kinase domain and the C-terminal domain of Chk1. Therefore, mutating one of them fully exposes the Ser-345 site to ATR, leading to its phosphorylation in the absence of DNA damage (Supplementary Fig. S5). Clearly, structural studies are needed to provide answers to this question and other questions that are related to the conformational change of Chk1.

No potential conflicts of interest were disclosed.

Conception and design: Y. Zhang

Development of methodology: J. Wang, X. Han

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Wang, X. Han

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Wang, Y. Zhang

Writing, review, and/or revision of the manuscript: Y. Zhang

Study supervision: Y. Zhang

The authors thank Tony Hunter for discussion and suggestions, Zhenghe Wang and Paul MacDonald for critical reading of the manuscript, and also thank James Jacobberger and the core facility at Case Comprehensive Cancer Center for the help in cell-cycle analyses.

Y. Zhang is funded by the NCI Howard Temin Career Development Award (R00CA126173), NCI R01CA163214, and a pilot grant from the American Cancer Society (IRG-91-022-15).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.