Expression levels of p18^INK4C modify the cellular efficacy of cyclin-dependent kinase inhibitors via regulation of Mcl-1 expression in tumor cell lines

Cyclin-dependent kinases (CDK) are serine/threonine kinases that play a pivotal role in regulating cell cycle progression. In the G₁-S phase of the cell cycle, external mitogenic stimuli induce the expression of cyclin D and activate CDK4 and CDK6. The activated CDK4/6-cyclin D complex phosphorylates RB protein, which results in the dissociation of E2F from the RB-mediated repressor complex, leading to transactivation of the E2F regulatory genes required for S-phase entry (1). For the initiation of DNA replication, CDK2 is required for recruiting DNA polymerase and triggering centrosome duplication (2, 3). In the G₂-M phase, CDK1 is involved in each phase of mitosis progression, such as nuclear envelope breakdown, spindle formation, and chromosome segregation, by phosphorylating various mitotic proteins (4, 5). This coordinated regulation of CDK activities is achieved by the carefully timed association of positive regulators (cyclins A–J) and two classes of negative regulators: the INK4 family genes (p16^Ink4a, p15^Ink4b, p18^Ink4c, and p19^Ink4d) and the Cip/Kip families (p21^Cip1, p27^Kipl, and p57^Kip2; ref. 6). In addition, CDKs have been implicated in the activation of transcriptional initiation and elongation. CDK7 and CDK9, as well as CDK1/2, phosphorylate the COOH-terminal domain of the large subunit of RNA polymerase II, which is crucial for RNA polymerase II–dependent transcriptional activity (7).

Alterations in the CDK/RB signaling pathway are seen in almost all types of human tumors and have been shown to contribute to tumor development and progression. Germ-line mutation of RB protein is a causative event in hereditary retinoblastoma (8). Gene amplification of cyclin D1, located in 11q13, occurs in >10% of breast cancer cases and in up to 30% of non–small cell lung cancer cases (9, 10). Alterations in p16 gene expression are caused by methylation in its promoter region or the deletion of the 9p21 locus and have been observed in most non–small cell lung cancer cases (11). Abnormal expression of CDK2 regulatory proteins p27 and cyclin E has been observed with a high frequency in breast cancer cases and is strongly correlated with poor prognosis (12). In addition to human clinical data, various mouse genetic models have shown that alteration in CDK pathways is a causative event in tumorigenesis and provides advantageous growth conditions for tumor cells. For example, mice that are deficient in CDK inhibitory genes, such as p18 or p27, exhibited a broader range of tumors with concomitant mutations in tumor suppressor genes (13, 14).

Due to the pivotal roles of CDKs in tumor development, the inhibition of CDK family members has attracted particular attention in the development of antitumor therapy. Numerous studies have shown that inhibition of CDK is an effective option for suppressing abnormal tumor cell proliferation. The established U-2OS osteosarcoma cell line, which expresses short hairpin RNAs for both CDK1 and CDK2, exhibits cell cycle arrest in the S and G₂-M phases (15). Membrane-permeable inhibitory peptides that antagonize the binding of CDK2 to its substrates induced apoptosis selectively in T antigen–transformed fibroblast cells with deregulated E2F activity but not in nontransformed parental cells (16). In addition, numerous CDK inhibitors have been developed pharmacologically and their antitumor effects have been well characterized in tumor cell lines. For example, R-roscovitine (CYC202) suppresses both cell cycle progression and transcriptional activity by inhibiting the phosphorylation of RB protein and RNA polymerase II (17, 18). This transcriptional repression attenuates the expression of several short-lived proteins, including Mcl-1, an apoptosis inhibitor of the Bcl-2 family, and this attenuation contributes to CDK inhibitor–mediated apoptosis.

Clinical responses to chemotherapeutic agents differ depending on the genetic context of the given tumor cells. The development of biomarkers, such as those for gene expression and mutations that predict drug sensitivity, has facilitated more effective use of anticancer drugs by selecting individuals most likely to respond. The availability of biomarkers to stratify patients is best illustrated by the development of molecule-targeted agents. The survival benefits of gefitinib were improved when patients with activating epidermal growth factor receptor mutations were selected (19). Trastuzumab elicits higher rates of overall response in patients with breast cancer with Her2 overexpression when trastuzumab is administered as a single agent or concomitantly with standard chemotherapy (20). Furthermore, using a cDNA microarray, Potti et al. (21) developed a gene expression signature consisting of 50 genes and further validated its ability to accurately extrapolate the response of patients with breast cancer to docetaxel treatment in clinical settings with an overall accuracy exceeding 90%. In the case of CDK inhibitors, despite significant preclinical and clinical developments, biomarkers to predict efficacy against certain cancers in individual patients have not yet been identified. In the present study, we characterized a novel macrocyclic quinoxalin-2-one CDK inhibitor, compound A, and identified a gene biomarker for predicting its efficacy. We found that the expression levels of p18 alter sensitivity to CDK inhibitors through regulation of Mcl-1 expression and may be used as a methodology for stratifying patients to predict the chemotherapeutic efficacy of CDK inhibition.

IC₅₀ values of compound A for the inhibition of CDK1, CDK2, CDK4, CDK6, CDK7, and CDK9 and other kinase activities were determined using recombinant human full-length or glutathione S-transferase (GST)–fused kinase domain proteins expressed in insect cell line Sf9 with appropriate peptide substrates, as described previously (22).

Compound A is an analogue of macrocyclic quinoxalin-2-one CDK inhibitor (designated compound 13 in the previous report), which is substituted at the 7 position of quinoxalin-2-one by an aniline methyl group to increase cell potency (23).

Total RNA was extracted using the RNeasy kit (Qiagen). Reverse-transcribed cDNA was subjected to Taqman PCR for quantification of mRNA expression using an ABI PRISM 7700 sequence detector system (Applied Biosystems). Relative mRNA expression data were normalized against β-actin expression. Pre-Developed Taqman Assay Reagents (Applied Biosystems) were used for Taqman probe and primers for β-actin, p18, cyclin A, cyclin E, and Mcl-1.

Each cell was transfected with E2F reporter plasmid and luciferase expression plasmid with FuGENE 6 (Roche Diagnostics), and 24 h later, medium was replaced with fresh medium including compound A (24). Medium was collected after 24 h of treatment and SEAP activity was measured by the Reporter Assay Kit-SEAP (TOYOBO) and normalized against luciferase activity.

Equal amounts of protein were resolved on NuPAGE 4% to 12% Bis-Tris polyacrylamide gels (Invitrogen) and then electrophoretically transferred onto polyvinylidene difluoride membranes (Millipore). Membranes were incubated with anti-RB (MBL), anti-RB phospho-Ser²⁴⁹/Thr²⁵² (Invitrogen), anti-RB phospho-Thr³⁵⁶ (Invitrogen), anti–RNA polymerase II (pSer²; Covance), anti–RNA polymerase II (pSer⁵; Covance), anti-RNA polymerase II (Covance), anti-p16 (Santa Cruz Biotechnology), anti-p18 (Santa Cruz Biotechnology), anti-p27 (Santa Cruz Biotechnology), anti-CDK4 (Cell Signaling Technology), anti-CDK6 (Cell Signaling Technology), anti–Mcl-1 (Santa Cruz Biotechnology), or anti–β-actin (Santa Cruz Biotechnology) antibodies followed by horseradish peroxidase–conjugated secondary antibody. Antibody binding was detected using the enhanced chemiluminescence system (GE Healthcare). Relative protein expressions were quantified by densitometry.

Total protein lysates (400 μg) were incubated at 4°C for 12 h with 1 μg of the anti-CDK6 antibody (Santa Cruz Biotechnology) followed by incubation with protein G-Sepharose (GE Healthcare) for an additional 1 h. Immune complexes were precipitated by centrifugation and beads were rinsed twice with lysis buffer and then twice with kinase assay buffer [20 mmol/L Tris-HCl (pH 7.4), 10 mmol/L MgCl₂, 4.5 mmol/L 2-mercaptoethanol, 1 mmol/L EGTA]. Purified CDK6 was incubated in 30 μL kinase buffer with 50 μmol/L ATP, 0.5 μCi [γ-³³P]ATP, and 1 μg GST-RB (379–928) substrate for 20 min at 30°C. The reaction products were resolved on NuPAGE 4% to 12% Bis-Tris polyacrylamide gels.

Each cell line was treated with compound A, flavopiridol (Sigma-Aldrich), roscovitine (Calbiochem), camptothecin (Sigma-Aldrich), and paclitaxel (Sigma-Aldrich) for 72 h. Multiple myeloma cells were treated with compound A and roscovitine for 8 h. In the case of small interfering RNA (siRNA) transfection, cells were transfected with siRNA for p18, p16, or luciferase (control) using siLentFect (Bio-Rad) and, 24 h later, treated with 10 nmol/L compound A or 30 μmol/L roscovitine for 96 h. After treatment, cells were stained using Cycle TEST PLUS DNA Reagent kit (Becton Dickinson) or Annexin V-FLUOS Staining kit (Roche Diagnostics) and analyzed by flow cytometry. The siRNAs for p18, p16, and luciferase were purchased from Dharmacon. To estimate the effect of compound A on cell cycle progression, T98G cells were treated with compound A for 24 h and then labeled with bromodeoxyuridine (BrdUrd) for 1 h. Harvested cells were fixed with ethanol and incubated with a monoclonal antibody to BrdUrd conjugated with a fluorochrome followed by DNA counterstaining with propidium iodide.

The growth inhibition effects of compound A in each cell line were determined using the sulforhodamine B dye-staining method, as previously described (25). Caspase-3/7 activation was determined by the Caspase-Glo Assays (Promega). S-phase progression was assessed based on the measurement of BrdUrd incorporation during DNA synthesis (Roche Diagnostics).

Total RNA (10 μg) was used for microarray analysis of ∼14,000 genes with the HG-U133A chip (Affymetrix). To extract differentially expressed genes between the two groups, the following criteria were established: (a) error-weighted ANOVA P value between the two groups is ≤0.001, (b) minimum intensity in sensitive cell lines is larger than the maximum intensity in resistant cells or vice versa, and (c) median intensity in sensitive cell lines is >2-fold or <0.5-fold of that in the other group. Statistical analyses were done using Resolver and R statistical software package (Rosetta Biosoftware).

To identify novel potent CDK inhibitors, we screened a series of compounds for the ability to inhibit the enzymatic activity of CDK proteins (CDK1, CDK2, CDK4, CDK6, CDK7, and CDK9). Macrocyclic quinoxalin-2-one compounds were found to show potent CDK inhibitory effects and were suitable for i.v. administration in vivo, as reported recently (23). In the present study, we characterized one macrocyclic quinoxalin-2-one compound (compound A) for anticancer efficacy. In vitro kinase assays showed that compound A strongly inhibited all CDK proteins examined in an ATP-competitive manner with an IC₅₀ of ∼10 nmol/L (Fig. 1A). When we tested six representative tyrosine kinases and other structurally similar serine/threonine kinases for inhibitory activity, compound A had no inhibitory effects up to 1,000 nmol/L. To examine whether compound A inhibits cellular CDK activities, we analyzed its effects on the RB/E2F pathway. When human glioblastoma T98G cells were treated with compound A for 24 hours, phosphorylation of the RB protein at Ser²⁴⁹/Thr²⁵² (CDK4 phosphorylation site) and Thr³⁵⁶ (CDK2 phosphorylation site) was inhibited with IC₅₀ values of 5.4 and 9.2 nmol/L, respectively (Fig. 1B; refs. 15, 26). This inhibition accompanied the suppression of E2F-dependent transcriptional activity with an IC₅₀ of 12 nmol/L and subsequent cell cycle arrest in the G₁ and G₂-M phases and apoptosis induction (Fig. 1C and D; Supplementary Fig. S1)¹

The antitumor effects of CDK inhibition were examined by treating 30 cell lines derived from various tissue origins with compound A. Three days after treatment, compound A inhibited cell proliferation with IC₅₀ values ranging from 4.1 to 33 nmol/L, although these effects were not dependent on genetic background factors, such as p53 mutation and RB inactivation (Fig. 2A). We then examined the apoptosis induction rate following treatment with increasing concentrations of compound A by measuring the sub-G₁ phase population by flow cytometry. Apoptosis induction was evident with compound A at 30 nmol/L (IC₈₀ of E2F transcriptional activity in Fig. 1C) in most of the cell lines, but the extent of apoptosis induction clearly differed among the cell lines tested (Fig. 2B). This was in contrast to cell growth inhibition, which occurred at a similar rate in all cell lines. Based on the results of the apoptosis induction rate from sub-G₁ and Annexin V analyses, we classified PC13, SCC-25, PA-1, MSTO-211H, and BxPC-3 as sensitive cell lines (>60% cell death), whereas J82, U-118MG, NCI-H69, and SK-OV-3 were classified as resistant cell lines (<10% cell death; Fig. 2C and D). To confirm whether this sensitivity spectrum of compound A is correlated with those of other CDK inhibitors, we examined the sensitivity of these sensitive/resistant cell lines to effective concentrations of structurally different CDK inhibitors, such as flavopiridol and roscovitine, and also anticancer agents with different mechanisms of action, such as DNA-damaging agents (camptothecin) and antimicrotubule agents (paclitaxel; refs. 18, 27, 28). Flavopiridol and roscovitine treatments showed a similar apoptosis-inducing effect with compound A, but camptothecin and paclitaxel treatments exhibited a different spectrum of anticancer activities from CDK inhibitors, indicating that the sensitive/resistant phenotype of the examined cell lines is characteristic of CDK inhibitors (Supplementary Fig. S2A–D).¹

To identify the molecular factors affecting CDK-mediated apoptosis, we developed a gene expression signature correlated with the susceptibility to CDK inhibition–mediated apoptosis by comparing expression profiles between sensitive cells and resistant cells by microarray. The expression profiles of ∼14,000 genes in the cell lines listed in Fig. 2C were analyzed using the GeneChip oligonucleotide microarray HG-U133A system, and genes differentially expressed between sensitive and resistant cell lines were extracted based on the criteria given in Materials and Methods. Based on these criteria, we identified 17 genes that were differentially expressed between sensitive and resistant cell lines; their expression patterns are shown as a hierarchical clustering in Fig. 3A.

Among the genes identified to have a negative correlation with compound A sensitivity, p18 is known to suppress CDK4/6 function and malignant alteration in many primary human tumors, including medulloblastomas, glioblastoma multiforme, pituitary adenomas, and multiple myelomas (Supplementary Table S1; refs. 29-35).¹ Cells with lower p18 expression might be susceptible to apoptosis following CDK inhibition due to aberrantly high CDK activities and possible dependence on CDK activities for proliferation (30, 36, 37). As indicated by microarray data, sensitive cell lines showed lower expression of p18 than resistant cell lines at both the mRNA and protein levels (Fig. 3B and C). Because MSTO-211H cells have relatively higher expression of CDK4, they exhibited the increased stability of p18 protein despite low p18 mRNA expression, as recently reported (38). We also analyzed the expression levels of other mediators in CDK pathways: RB, p16, p27, CDK4, and CDK6. However, no correlations were observed between their expression and sensitivity to compound A (Fig. 3C). Additionally, the S-phase population in each cell line was not correlated with the sensitivity, although previous reports indicated that cells in S phase are susceptible to CDK inhibition (Fig. 3D; ref. 39).

To evaluate the validity of p18 status as a biomarker for CDK inhibitors, we sought to address whether expression levels of p18 could predict the sensitivity of cells to CDK inhibitors in an independent set of tumors. In multiple myeloma cells, p18 has been shown to function as a key determinant in faithful cell cycle regulation and its inactivation, which occurs in >40% of multiple myeloma cases through deletion or hypermethylation, gives rise to the most proliferative myeloma tumors with the worse overall survival (31, 40). Thus, we assessed the effect of CDK inhibitors on apoptosis induction in a panel of multiple myeloma cell lines with different p18 statuses. First, we quantified p18 mRNA and protein expression levels by real-time reverse transcription-PCR (RT-PCR) and Western blot analysis (Fig. 4A and B). KMS-28PE, KMM-1, KMS-28BM, and U266 cells showed higher p18 expression, whereas 8226, OPM-2, and IM9 cells showed very low p18 expression at both the mRNA and protein levels. KMS12-BM did not display intact p18 protein expression because of a missense mutation, as previously reported (40). From these analyses, we classified KMS-28PE, KMM-1, KMS-28BM, and U266 cells as p18-positive cells and 8226, OPM-2, IM9, and KMS12-BM cells as p18-negative cell lines. Next, cells were treated with 30 nmol/L compound A or 30 μmol/L roscovitine for 8 h, because of rapid response to CDK inhibitor, and apoptosis induction was examined by measuring the sub-G₁-phase population (Fig. 4C; ref. 17). Treatment of p18-positive cells with CDK inhibitors showed little increase in apoptosis induction, with a maximal effect in U266 cells with 11% increase of apoptosis induction. In contrast, treatment of p18-negative cells with CDK inhibitors resulted in marked increase in apoptosis induction, ranging from 30% to 67%. Together, the response to CDK inhibitor treatment exhibited a significant distinction between p18-positive cells and p18-negative cells. These results support the evidence that CDK inhibitors preferentially exert antitumor effects in the context of different p18 statuses.

To further investigate whether the expression levels of p18 are important in determining sensitivity to CDK inhibition, the resistant cell line J82 was transfected with p18 siRNA before treatment with compound A or roscovitine. Apoptosis induction was then compared between the control and p18-silenced cells. After 48 hours, >70% suppression of p18 mRNA and protein expression was seen in p18 siRNA-transfected cells (Fig. 5A). Under these conditions, control and p18-silenced cells were treated with 10 nmol/L compound A or 30 μmol/L roscovitine for 48 hours. Caspase activation was increased in p18-silenced cells treated with compound A by ∼2-fold and in p18-silenced cells treated with roscovitine by >3-fold compared with the control (Fig. 5B). These results were also confirmed by the reduced viability of p18-silenced cells after 96 hours of treatment with compound A or roscovitine (Fig. 5C). Additionally, with flow cytometry analysis, p18-silenced cells also showed increases in the sub-G₁ population from 7.8% to 33% by compound A treatment and from 18% to 42% by roscovitine treatment compared with the control (Fig. 5D). In contrast, p16-silenced cells did not show the increased sensitivity (Supplementary Fig. S3).¹ Similar results were obtained in another resistant cell line, U118-MG (Fig. 5A–D). These results show that expression levels of p18 are able to alter apoptosis induction by CDK inhibition.

To elucidate the molecular basis by which CDK inhibitors exert a more pronounced anticancer effect in p18-negative cells, we examined the functional significance of p18 silencing in each cell line with p16 silencing as a control (Fig. 6A). In U118-MG cells (p18⁺, p16⁻), immunoprecipitated cellular CDK6 activity was approximately twice as elevated in p18-silenced cells than in the control cells (Fig. 6A). Subsequently, E2F transcriptional activity and the S-phase population were increased (Fig. 6B). E2F is reported to regulate gene expression both positively and negatively. In fact, on CDK6 activation, the expression of both cyclin A and cyclin E, which are positively regulated by E2F, was increased in p18-silenced cells (Fig. 6C). Notably, the expression of Mcl-1, which is negatively regulated by E2F, was decreased in p18-silenced cells (Fig. 6C; ref. 41). Similarly, in J82 cells (p18⁺, p16⁺), activation of CDK6 and elevation of cyclin expression occurred in p18-silenced cells and in p16-silenced cells, but down-regulation of Mcl-1 expression was observed only in p18-silenced cells (Fig. 6A–C). Previous studies have shown that CDK inhibitors exert a cytotoxic effect by down-regulating antiapoptotic protein Mcl-1, so expression levels of Mcl-1 after treatment were compared among cells. As expected, compound A treatment decreased the amount of Mcl-1 expression in all cell lines. In p18-silenced cells in which the basal amount of Mcl-1 expression was lowered, the remaining Mcl-1 expression after treatment was significantly compromised compared with that of the control cells (Fig. 6D). According to these observations, p18-silenced cells are supposed to have increased CDK and E2F activities, thereby leading to a lowered threshold for Mcl-1–mediated mitochondrial apoptosis and to an increased susceptibility to CDK inhibition. Indeed, the sensitive cell line MSTO-211H exhibited lower expression of Mcl-1 than the resistant cell line SK-OV-3, which would account for different sensitivities to CDK inhibition (Supplementary Fig. S4).¹

In various types of tumors, numerous molecules involved in the CDK/RB pathway are often dysregulated. In light of this information, pharmacologic CDK inhibitors have been extensively investigated and some are under evaluation in clinical trials. However, the genetic or cellular background factors that affect sensitivity to CDK inhibitors remain to be elucidated. We here examined the effects of a highly potent and selective CDK inhibitor having a macrocyclic quinoxalin-2-one structure, compound A, on the proliferation of cancer cells. Compound A treatment caused apoptosis induction in a cell line–specific manner. By comparing the expression profiles of sensitive and resistant cell lines, we found that the expression level of p18 showed a strong negative correlation to the sensitivity. In fact, the treatment of CDK inhibitors exhibited antitumor efficacy in a p18 status-dependent manner in the separate collections of multiple myeloma cell lines, supporting the predictive validity and robustness of p18 status. By unraveling the intracellular mechanisms in p18-silenced cells, we showed that p18-negative cancer cells maintain a coordinated balance between high proliferation and apoptosis activities and then become dependent on CDK activity for their cell survival. These findings indicate that p18, a key regulator of CDK4/6, may be used as a predictive marker for sensitivity to CDK inhibitors.

This study showed that the expression level of p18 modifies the cellular efficacy of CDK inhibitors. We revealed by a comprehensive gene expression analysis that p18 expression level showed significant correlation with the sensitivity to CDK inhibitor in a panel of 30 tumor cell lines and further validated the capacity of p18 status to accurately predict the drug sensitivity in an independent data set of multiple myeloma cell lines. Our analysis of intracellular events in p18-negative cancers unveiled a critical role of p18 for the maintenance of cancer cell proliferation; p18 ablation led to increased CDK6 activities and subsequently caused deregulated expression of E2F target genes such as reduced Mcl-1 and increased cyclins. Consistently, several studies have pointed out that CDK inhibitor selectively kills tumor cells with inappropriate E2F activity induced by the recruitment to the S phase or oncogenic alterations, such as SV40 T antigen transformation or cyclin D1 overexpression (16, 42). Furthermore, increased E2F1 transcriptional activity is shown to directly repress Mcl-1 expression and the mechanism underlying CDK inhibitor–mediated apoptosis has been attributed to the down-regulation of Mcl-1 mRNA (41, 43). It has been reported that the inhibition of transcription-related CDKs suppressed the activity of RNA polymerase II and compromised the mRNA expressions with short half-lives, including that of Mcl-1, triggering mitochondrial apoptosis. In fact, reduced Mcl-1 expression predisposes human leukemia U-937 cells to be susceptible to CDK inhibitor–mediated apoptosis, whereas overexpression of Mcl-1 protects NCI-H1299 lung carcinoma cells from flavopiridol-induced apoptosis (43, 44). Given this functional relevance of p18 to proliferation- and survival-related genes, the basal expression level of p18 could determine the efficacy of CDK inhibitor by affecting diverse biological and gene expression phenotypes. Because predictive biomarkers are usually measured before exposure to antitumor agents, pretreatment expression signatures, including that of p18, would also be applied to identify potentially responsive patients.

Different alterations in CDK inhibitory proteins, such as the loss of p16 and p27 expressions, have been linked to tumorigenesis, but in our analysis, we only identified p18 gene expression as a predictive marker of CDK inhibition. Although all INK4 proteins have analogous structures and functions to cooperatively constrain proliferation, recent reports have raised the possibility of distinct roles or each INK4 protein. Most notably, p16 has been suggested to function in the senescence response. Although the levels of p16 mRNA do not apparently change during the cell cycle, p16 expression has a close connection to the aging process in mice, in which p16 expression increases with age (45). The involvement of p16 in senescence is also supported by the antagonistic effects on Ras-induced senescence by a p16 mutation in human fibroblast cells (46, 47). On the other hand, p18 is tightly regulated by cell cycle phase in cultured cell lines, being elevated from the G₁ phase and peaking during the S phase (29). In a mouse study, p18 showed more abundant expression than p16 in prenatal tissue and highly proliferating tissues, such as thymus, testis, and spleen, suggesting a more crucial function for p18 in the control of the cell cycle (45).

Mice lacking each of the INK4 inhibitory genes provided insight into the distinct roles of the genes in carcinogenesis. Mice lacking p19 did not show obvious tumor phenotypes, whereas those lacking p15 showed a low incidence of tumorigenesis; the frequency of angiosarcoma was 8.2% (30, 48). In two reports involving the ablation of p16 in mice, little or no tumor development was seen; no statistically significant spontaneous tumors were reported in one study and a 10% rate of sarcoma or lymphoma was reported in the other study (49). Interestingly, p18-deficient mice exhibited larger body size and widespread organomegaly shortly after birth and had a higher incidence of spontaneous tumors in a variety of tissues (40% pituitary hyperplasia, 12% testicular tumors, and 10% adrenal gland tumors), thus indicating that p18 rather than p16 has tumor suppressor properties in a variety of tissues (30, 36). Taken together with the present data, these reports suggest that p18 plays an important role in CDK-mediated cell cycle regulation and it is likely that the unique characteristics of p18 represent a key determinant factor in CDK-mediated cell death.

Delineation of the mechanism by which cells with dysfunctional p18 are susceptible to CDK inhibition–mediated apoptosis suggests that p18-deficient tumors are mainly dependent on oncogenic CDK activity for their sustained survival. This phenomenon is known as “oncogene addiction,” which was proposed to describe the acquired dependence of cancer cells on an aberrant oncogene or tumor suppressor for their proliferation (50). For example, recent studies showed a linear correlation between the deregulation of a particular oncogenic pathway and the effects of therapeutic agents that target the corresponding pathways, including Ras and phosphatidylinositol 3-kinase (21). The present results and those of other studies have shown that p18 ablation leads to aberrant activation of the CDK/E2F pathway to drive the pathogenesis of tumors (37). Notably, our study showed that CDK activities are necessary for the survival of p18-deficient multiple myeloma cells. This dependence is supported by the previous finding that genetically engineered mice cells that lack p16 and p18 expression had aberrant CDK4 activity that resulted in increased cell cycle progression, thus becoming dependent on CDK activity for cell proliferation, and thereby increasing the sensitivity to acute pharmacologic CDK inhibition. These data indicate that CDK plays a key role in maintaining a state of cellular homeostasis in p18-negative cells. Because attenuation of a CDK-mediated proliferation signal would impair the proper balance between prosurvival and proapoptosis and induce cell death, it is of great interest to exploit this weakness arising from p18 dysfunction for therapeutic intervention with CDK inhibitors in clinical settings.

In the present study, we characterized a novel macrocyclic quinoxalin-2-one CDK inhibitor, compound A, and showed that p18 expression has a strong negative correlation with CDK inhibitor–mediated apoptosis induction. Although various CDK inhibitors have been developed and have shown strong anticancer effects in vitro and in vivo, biomarkers to predict the efficacy of CDK inhibition in various cancer types and patients have not yet been identified. The fact that p18 expression levels predict sensitivity to CDK inhibitors including flavopiridol and roscovitine, as revealed by the present study, provides the opportunity for a better methodology for stratifying patients undergoing anti-CDK therapy. Because our analysis is based on in vitro experiments using cancer cell lines, however, the significance of p18 as a biomarker for CDK inhibitor therapy should be further examined in a clinical study.

No potential conflicts of interest were disclosed.