Expression of p16 and Retinoblastoma Determines Response to CDK4/6 Inhibition in Ovarian Cancer

In ovarian cancer, 75% of patients present with advanced (stage III or IV) disease and, although more than 80% of these women benefit from first-line therapy, tumor recurrence occurs in almost all these patients at a median of 15 months from diagnosis (1). Advances in understanding of the molecular pathogenesis of ovarian cancer coupled with the development of novel-targeted therapies are needed to improve outcomes. As such, cell cycle dysregulation is a common molecular finding in ovarian cancer and the cyclin-dependent kinases (CDKs) represent attractive targets in this pathway (2). Under normal control the cell cycle functions as a tightly regulated process consisting of several distinct phases. Progression through the G1-S phase requires phosphorylation of the retinoblastoma (Rb) protein by CDK4 (3, 4) or the highly homologous enzyme CDK6 (5) in complex with their activating subunits, the D-type cyclins, D1, D2, or D3 (6). Hyperphosphorylation of Rb diminishes its ability to repress gene transcription through the E2F family of transcription factors and consequently allows synthesis of several genes that encode proteins, which are necessary for DNA replication (7). Assembly of active cyclin D–CDK4/6 complexes is negatively regulated by INK4 protein family members p16, p15, p18, and p19 and by a second group of proteins including p21 and p27 (8).

Deregulation of the CDK4/6–cyclin D/p16–Rb signaling pathway is among the most common aberrations found in human cancer (9). In the case of ovarian cancer, p16 expression is most commonly altered due to promoter methylation, and less commonly by homozygous deletion or mutation (10–12). A recent report indicates that of 249 ovarian cancer patients, 100 (40%) tested positive for p16 promoter methylation (10). Homozygous deletions of the p16 gene (CDKN2A) were detected in 16/115 (14%) or 8/45 (18%) (11, 12), and mutations in 53/673 (8%) of ovarian cancers (www.sanger.ac.uk/genetics/CGP/cosmic). Moreover, overexpression of cyclin D1 has been described in 25/135 (19%) ovarian cancer tumors and has been associated with a more aggressive tumor phenotype and poor prognosis (13). Mutations of the Rb gene (RB) have been reported to occur in 4/42 (10%) of ovarian cancers (www.sanger.ac.uk/genetics/CGP/cosmic) and hemizygous deletions at the RB locus have been described in 8/34 (24%) or even up to 25/48 (52%) of ovarian cancers (14, 15). Taken together these reports suggest that deregulation of the CDK4/6–cyclin-D/p16–Rb signaling pathway is commonly found in ovarian cancer, thus CDK4/6 inhibition may represent a promising new treatment strategy for ovarian cancer.

PD-0332991 is an orally active, potent, and highly selective inhibitor of the CDK4 (IC₅₀, 0.011 μmol/L) and CDK6 (IC₅₀, 0.016 μmol/L) kinases with the ability to block Rb phosphorylation at serine 780 and 795 in the low nanomolar range (16). Initial reports demonstrate antiproliferative activity in luminal breast cancer (17), myeloma (18), and glioblastoma multiforme cell lines (19, 20), as well as multiple Rb-positive subcutaneous human tumor xenografts (16). PD-0332991 is currently undergoing phase I/II clinical testing in mantle cell lymphoma, multiple myeloma, and breast cancer (www.clinicaltrials.gov/ct2/results?term=PD+0332991). Given its effectiveness in a variety of tumor types and its specificity for activated CDK4/6, we hypothesized that PD-0332991 may be useful for the treatment of ovarian cancer. To evaluate this, we tested the in vitro effects of PD-0332991 against a panel of 40 ovarian cancer cell lines, representing all histological subtypes of the disease. To more fully understand the antiproliferative activity we studied the effects of PD-0332991 on Rb phosphorylation, as well as cell cycle and apoptosis. Furthermore, we sought to identify response markers known to be implicated in CDK4/6–cyclin-D/p16–Rb signaling. We also performed multiple drug effect/combination index (CI) isobologram analysis to study the interactions between PD-0332991 and chemotherapy. Finally, we examined the expression pattern of 2 key preclinical response markers p16, and Rb using immunohistochemistry in a clinical cohort of 263 ovarian cancer patients to help characterize a subset of ovarian cancer patients that would most likely benefit from treatment with PD-0332991.

The effects of PD-0332991 on growth inhibition were studied in a panel of 40 established human ovarian cancer cell lines. Individuality of each cell line was checked by mitochondrial DNA sequencing. Cell lines were passaged for fewer than 3 months after authentication. Additional information on the cell lines is provided in the Supplementary Material (Supplemental Table S1).

Cells were plated into 24-well tissue culture plates at a density of 2 × 10⁵ to 5 × 10⁵ and grown without or with increasing concentrations of PD-0332991 (ranging between 0.001 and 10 μmol/L). Cells were harvested by trypsinisation on day 7 and counted using a particle counter (Z1; Beckman Coulter Inc.). Experiments were performed at least 3 times in duplicate for each cell line.

Effects of PD-0332991 on the cell cycle were assessed using Nim-DAPI staining (NPE Systems, Pembroke Pines, FL). Cells were allowed to grow to log phase and were then treated with 0.5 μM PD-0332991 for 72 hours. Samples were analyzed using a Cell Lab Quanta SC flow cytometer (Beckman-Coulter Inc.) according to the manufacturer's protocol.

Effects of PD-0332991 on apoptosis were performed using an Annexin V-FITC apoptosis detection kit (MBL) and flow cytometry. Cells were exposed to 0.5 μM PD-0332991 for 5 days. Samples were analyzed using the Cell Lab Quanta SC flow cytometer (Beckman-Coulter Inc.).

Western blot and immunoprecipitation were performed as previously described (21). Total Rb expression was detected using a mouse monoclonal antibody to Rb (Cell Signaling). Rb phosphorylation was detected using rabbit polyclonal antibody to phospho-serine 780 (Cell Signaling). Total p16 was detected using a rabbit polyclonal antibody to p16 (Cell Signaling).

Multiple drug effect analysis was performed as described previously (22). CI values were derived from variables of the median effect plots and statistical tests were applied (unpaired, 2-tailed Student's t test) to determine whether the mean CI values at multiple effect levels were significantly different from CI = 1. In this analysis, synergy is defined as CI values significantly lower than 1.0, antagonism as CI values significantly higher than 1.0, and additivity as CI values equal to 1.0.

Microarray hybridizations were performed of 40 ovarian cell lines at baseline using the Agilent Human 44K array chip. The techniques used have been described in detail elsewhere (17). The mixed reference pool consisted of equal amounts of RNA from the 40 ovarian cancer cell lines. Microarray slides were read using an Agilent Scanner and the Agilent Feature Extraction software version 7.5 was used to calculate gene expression values. The feature extracted files were imported into the Rosetta Resolver system for gene expression data analysis version 7.2 (Rosetta Biosoftware). The intensity ratios between the cell line sample and mixed reference were calculated for each sequence and were computed according to the Agilent error model. A particular sequence was considered differentially expressed if the calculated P-value was 0.01 or less. The original data are available online with the accession number (GEO accession number GSE26805).

Genomic DNA was extracted from frozen cells using the DNeasy Blood and Tissue Kit (Qiagen). Labeling and hybridization of Agilent 105K oligonucleotide CGH arrays were performed according to the manufacturer's protocol for Human Genome CGH 105A Oligo Microarray Kit, Version 5.0 (Agilent Technologies). Files were extracted using Agilent Feature Extraction software v9.5 with the default CGH protocol. Extracted arrays with a DRL Spread <0.3 were included in the analysis. CGH Analytics software v.4.0 (Agilent Technologies) was used for copy number analysis, employing the ADM2 algorithm (Threshold 5), with Fuzzy Zero and Centralization corrections to minimize background noise. All map positions were based on the March 2006 NCBI36/hg18 genome assembly. A minimum of 3 consecutive probes were required to define a region as amplified or deleted. All data were inspected visually using the interactive view. Log₂ ratios larger than 1 were considered amplified (2-fold increase) and log₂ ratios larger than 2 highly amplified, log₂ ratios smaller than −1 were considered hemizygous and log₂ ratios smaller than −2 homozygous deletions.

The relevant exons of the p53 gene in each tumor sample were PCR amplified, sequenced, and assessed for potential sequence alterations using approaches previously described (23). The nucleotide sequences were analyzed using the Mutation Surveyor program (Soft Genetics LLC) and through visual inspection. All somatic mutations were confirmed by independent PCR and sequencing reactions.

On approval from the Institutional Review Board at Mayo Clinic, we selected a representative group of 263 primary ovarian patients who underwent surgery for primary ovarian cancer at Mayo Clinic between June 1999 and April 2007 and who had archived paraffin embedded tissue available for analysis. Demographics, tumor characteristics, distribution of disease at the beginning of surgery, and residual disease were all retrospectively collected from the clinical records.

For tissue microarray construction, a hematoxylin and eosin (H&E) stained histology slide from each patient's archival tumor block was reviewed to identify and mark the location of tumor components. The markings were transferred to the corresponding tissue block. Marked donor blocks were cored into the recipient master block according to a grid map by use of an automated Beecher ATA 27 Tissue Arrayer (Beecher Instruments Inc.). The tissue microarray blocks constructed from the study tumors incorporated 3 tumor cores from an archival block for each subject. Tissue microarrays were labeled with monoclonal antibodies against Rb protein (Cat. No. 9309, Cell Signaling), and p16 protein (clone 16P07; NeoMarkers). Immunostaining was performed with the avidin–biotin complex method (Vector). For tissue microarray analysis images of H&E, Rb, and p16 stains were scanned with the Bliss Imaging System (Bacus Laboratories Inc.). Immunoreactivity was qualitatively scored by interpreting the staining intensity (absent, weak, or strong staining) and the percentage of positive tumor cells per core (<25%, 25–50%, >50%).

Associations between biomarkers and in vitro sensitivity were analyzed using Spearman's rho correlation or the χ² statistic. Differences between subgroups were compared using the Student's t test. The Resolver system ANOVA was performed on the ovarian cancer cell lines classified by response to PD-0332991 across a cell cycle gene set of 711 sequences on the Agilent Whole Human Genome platform that is classified as “Cell Cycle” or GO:0007049 according to the gene ontology database of December 2009. All ANOVAs were performed using the Benjamini-Hochberg False Discovery Rate (FDR) multiple test correction and a statistical cutoff value for sequences of a 1.75 change in at least 3 experiments, which reduced the number of sequences to 601. Of these 601 sequences, 117 had a significant P-value of P ≤ 0.05. Unadjusted survival was assessed by the Kaplan–Meier method and log-rank statistic was used for outcome comparison. Cox regression analysis was used to estimate hazard ratios and their 95% CIs in multivariate analysis adjusted for tumor grade, histology type, FIGO stage, and postoperative residual tumor. All reported P-values and CIs are from 2-sided tests. Because well-established and replicated cutoffs for the expression status of Rb, and p16 in ovarian cancer were not available, we made the a priori choice to analyze and report the scores as dichotomized values: absence of Rb expression versus low or high Rb expression (defined as Rb-proficiency), and absent or low p16 expression (defined as low p16 expression) versus high p16 expression (strong immunostaining in at last 25% of tumor cells per core).

The effects of PD-0332991 on human ovarian cancer cells were evaluated using a panel of 40 established human ovarian cancer cell lines. These cells lines were selected to be representative of a range of ovarian cancer subtypes. Nineteen cell lines were obtained from patients with serous papillary ovarian cancer, 10 cell lines from patients with clear cell ovarian cancer, and 6 cell lines from patients with undiffentiated ovarian cancer. The remaining 5 cell lines were obtained from patients with endometrioid or mucinous ovarian cancer.

The effective dose range (IC₁₀–IC₈₀) was identified using a wide range of PD-0332991 concentrations (0.001–10 μM). PD-0332991 inhibited the proliferation of all ovarian cancer cell lines investigated in a concentration-dependent fashion; however, the IC₅₀ values varied significantly between individual cell lines with up to a 3 log-fold difference in the IC₅₀ values and ranged between 0.021 μM in RMG1 ovarian cancer cells to over 20 μM in cell lines such as OV2008, OVCAR3, or CAOV3 (Table 1, Fig. 1A). There was no statistically significant correlation between the traditional histological subtypes and sensitivity to PD-0332991 (data not shown).

To classify cell lines as either sensitive or resistant to PD-0332991, the mean of the log₁₀(IC₅₀) across the 40 cell lines was calculated and used as a line of demarcation. The cell lines with log₁₀(IC₅₀) below the mean were defined as sensitive to the drug, whereas those with log₁₀(IC₅₀) above the mean were considered to be resistant. As shown in Figure 1A, 22 and 18 cell lines were classified as either sensitive or resistant to PD-0332991, respectively. The sensitive/resistant demarcation was 1.2 μmol/L. Although the demarcation is arbitrary, data from phase I solid tumor clinical trials showed that 1.2 μmol/L was within the peak range of PD-0332991 plasma concentrations in patients treated with the doses near the maximum tolerated dose. Moreover, using this cut off resistant cell lines exhibited IC₅₀ values greater 3 μmol/L, a level at which PD-0332991 is known to affect other targets (16).

Gene expression profiles for these 40 cell lines were generated using the Agilent 44K chips. First, we specifically explored the correlation between cell cycle biomarkers implicated in CDK4/6 signaling and in vitro response to PD-0332991. Cell lines with higher expression of Rb were associated with lower IC₅₀ values and thus more sensitive to PD-0332991 (RB, r = −0.42, P = 0.007). In contrast, cell lines with high expression of p16 (CDKN2A, r = 0.69, P < 0.001), or high expression of cyclin E1 (CCNE1, r = 0.58, P < 0.001) were associated with higher IC₅₀ values and thus more resistant to PD-0332991 (Table 2, Fig. 1B). A cluster diagram of the 40 ovarian cancer cell lines was developed using the markers RB, CDKN2A, and CCNE1 (Fig. 1B). When cell lines were ordered from low IC₅₀ values to high IC₅₀ values it becomes apparent that sensitive ovarian cancer cell lines (IC₅₀ < 1 μmol/L) demonstrated higher Rb expression, but lower p16 or cyclin E1 expression. We also studied additional markers implicated in the CDK4/6 signaling pathway, such as other members of the INK4 family (p15, p18, p19), the Cip family (p21, p27), the Rb protein family (Rb2/p130, Rb3/p107), the D-cyclins (D1, D2, and D3), and the E2F transcription factor and downstream targets (E2F1, TK1, CDK1), none of which were as strongly correlated with in vitro sensitivity to PD-0332991 as Rb, p16, or cyclin E1 (Table 2, Supplementary Fig. S1).

To validate the association between p16 deficiency, Rb-proficiency, and response to PD-0332991, we assessed p16 and Rb protein expression in all 40 ovarian cancer cell lines using Western blot analysis. The protein expression levels of p16 and Rb correlated well with the RNA levels (Rb, r = 0.48, P = 0.002; p16, r = 0.63, P < 0.001, data not shown), and an association between low expression of p16 or Rb-proficiency with in vitro sensitivity to PD-0332991 was confirmed at the protein level (p16, r = 0.57, P < 0.001; Rb, r = −0.48, P = 0.002, data not shown).

Next, array CGH profiles were generated for the 40 ovarian cancer cell lines using an Agilent platform (Table 1). Hemizygous or homozygous deletions of the p16 gene were found in 11/22 (50%) of the sensitive but only 3/18 (17%) of the resistant lines (P = 0.028, Table 1). In contrast, hemizygous or homozygous deletions of the RB gene were found in 9/18 (50%) of the resistant lines but only 2/22 (10%) of the sensitive lines (P = 0.004). Consistent with the expression data, a copy number gain of CCNE1 (>1.5 copies, log₂ >0.58) was found in 9/18 (50%) of the resistant lines but in only 2/22 (10%) of the sensitive lines (P = 0.004). Of note, we also found an association between CCND1 copy number gain and resistance to PD-0332991, as CCND1 copy number gain was found in 6/18 (33%) of the resistant lines and in only 1/22 (5%) of the sensitive lines (P = 0.017).

Activated p53 binds DNA and activates expression of several genes including WAF1/CIP1 encoding for p21. When p21 is complexed with CDK2, the cell cannot continue to the next stage of cell division. A mutant p53 however will no longer bind DNA in an effective way, and, as a consequence, the p21 protein will not be available to act as the “stop signal” for cell division. We found a significant association between the p53 mutations and in vitro resistance to PD-0332991, in that p53 mutations were found in 12/18 (67%) of the resistant lines and in only 7/22 (32%) of the sensitive lines (P = 0.028, Table 1).

ANOVA and hierarchical cluster analysis of the expression profiles were performed on the ovarian cancer cell lines classified by response to PD-0332991 (sensitive cell lines, n = 22, IC₅₀ < 1 μmol/L; resistant cell lines, n = 18, IC₅₀ > 3 μmol/L) across a cell cycle gene set of 711 sequences. A set of 117 differentially expressed genes was identified and the most significant gene was CDKN2A (p = 4.25E−10). An image of the complete matrix and a list of the 117 differentially expressed genes ordered by F-values are provided in the supplemental online material (Supplementary Fig. S2, Table S2).

It is known that CDK4/6 complexes with cyclin D1 to phosphorylate and inactivate Rb, thus allowing cell cycle progression. As a measure of functional activity of CDK4/6 the phosphorylation state of Rb was assessed using immunoblotting techniques. Exposure of sensitive cell lines to PD-0332991 resulted in a time- and concentration-dependent inhibition of Rb phosphorylation (Fig. 1C). Moreover, we found a time- and concentration-dependent decrease in expression of total Rb following treatment with the CDK4/6 inhibitor for up to 24 hours (Fig. 1C).

Earlier reports have suggested that PD-0332991 may have antiproliferative activity by inducing cell cycle arrest and senescence (19). Here we confirm and extend these findings to ovarian cancer cells in that PD-0332991 induced a G0/G1 cell cycle arrest in PD-0332991 sensitive ovarian cancer cell lines (Fig. 2A). Treatment with PD-0332991 for 5 days lead to a modest albeit significant increase in the fraction of cells undergoing apoptosis when compared with the untreated controls in sensitive cell lines (Fig. 2B). Nevertheless, PD-0332991 did appear to primarily inhibit growth across the entire ovarian cancer cell line panel in a cytostatic manner rather than a cytotoxic manner as no lethality was observed in any of the 40 examined ovarian cancer cell lines (data not shown). Taken together, these data support the proposed cytostatic mechanism of action of this CDK4/6 inhibitor involving prevention of cell cycle progression by blocking hyperphosphorylation of Rb.

Multiple drug effect analyses were performed to determine the nature of the interactions between PD-0332991 and the 2 chemotherapeutic agents most commonly used for the treatment of ovarian cancer: carboplatin and paclitaxel. Cells were either treated first with PD-0332991 followed by chemotherapy 24 hours later, or treated first with chemotherapy followed by PD-0332991, or treated with both agents concomitantly. PD-0332991, carboplatin, and paclitaxel concentrations used for these experiments were below the reported peak plasma concentrations achievable in humans (22). Additive or synergistic interactions were observed when PD-0332991 was given concomitantly with carboplatin (mean CI values ranged between 0.80 [95% CI 0.62–0.99, P = 0.024] and 0.82 [95% CI 0.58–1.07, P = 0.154]) or concomitantly with paclitaxel (0.84 [95% CI 0.70–0.99, P = 0.028] – 0.83 [95% CI 0.59–1.07, P = 0.146], Fig. 2C). Similar interactions were observed when PD-0332991 was given 24 hours after carboplatin (1.29 [95% CI 0.84–1.75, P = 0.157] – 1.60 [95% CI 1.19–2.01, P = 0.004]) or paclitaxel (0.74 [95% CI 0.51–0.98, P = 0.022] – 0.85 [95% CI 0.74–0.97, p = 0.009]). In contrast, giving PD-0332991 24 hours before chemotherapy resulted in antagonistic interactions for carboplatin (1.72 [95% CI 1.45–2.00, P < 0.001] and 2.98 [95% CI 2.26–3.70, P < 0.001]) and paclitaxel (1.30 [95% CI 1.02–1.59, P = 0.023] and 1.48 [95% CI 1.25–1.72, P < 0.001], Fig. 2C).

Because Rb-proficiency and low expression of p16 correlate with in vitro response to PD-0332991, we wanted to study the expression pattern of both markers in a large clinical cohort of ovarian cancer patients. The study group consisted of 263 patients who had been diagnosed with primary ovarian cancer at Mayo Clinic. Median progression-free survival was 23 months and median overall survival was not yet reached. Evaluation of patient demographic information and disease characteristics of this study cohort demonstrates that the group is representative of the general primary ovarian cancer population (Table 3). Representative results for immunohistochemical staining of Rb and p16 are shown in Figure 3A and B. Absence of Rb expression in tumors was seen in 16/263 (6%) of the patients and was associated with significantly worse progression-free survival (PFS) when compared with those whose tumors demonstrated Rb expression (log rank, P = 0.047, Fig. 3C). Only absence of Rb expression in tumors was associated with adverse clinical outcome, whereas patients with tumors with low expression of Rb (44/263; 17%) had a PFS similar to that seen in patients with tumors with high Rb expression (203/263; 77%, Supplementary Fig. 3A). Absence of Rb expression was associated with high nuclear grade (P = 0.141), advanced FIGO stage (P = 0.052), suboptimal debulking (P = 0.034), and high p16 expression (P = 0.032).

On the other hand, low p16 expression in tumors was seen in 99/263 (38%) of the patients, and was more commonly found in mucinous, endometrioid, and clear cell histologies when compared with serous or mixed-type histologies (P < 0.001). Moreover, low p16 expression was more commonly seen in tumors with low nuclear grade when compared with tumors with high nuclear grade (P = 0.004). When analyzing all patients, including those with low- and high-grade tumors, p16 expression was not associated with clinical outcome (Supplementary Fig. S3B). However, in multivariate analysis, when adjusting for nuclear grade, FIGO stage, histology, and residual tumor, low p16 expression in tumors obtained independent prognostic relevance and was associated with worse PFS when compared with patients with tumors with high p16 expression (adjusted relative risk 1.49, 95% CI 0.99, 2.22, P = 0.054). Accordingly, when analyzing only patients with tumors of high nuclear grade, low p16 expression in tumors was significantly associated with adverse clinical outcome when compared with patients with tumors with high p16 expression (P = 0.025, Fig. 3D).

In the present clinical cohort Rb-proficiency together with low p16 expression was found in tumors of 97/263 (37%) primary ovarian cancer patients and 60/182 (33%) of advanced ovarian cancer patients and was similarly associated with adverse clinical outcome in multivariate analysis (PFS, adjusted relative risk 1.49, 95% CI 1.00, 2.24, P = 0.052, and 1.51, 95% CI 1.00, 2.28, P = 0.053, respectively, Supplementary Table S3 and S4). In patients with tumors of high nuclear grade or patients with advanced FIGO stage and high nuclear grade, Rb-proficiency with low p16 expression was significantly associated with adverse clinical outcome when compared with patients with tumors demonstrating high p16 expression and/or absence of Rb expression (P = 0.036 and 0.043, respectively, Supplementary Fig. 4A and B). Importantly, patients with advanced primary ovarian cancer that demonstrates Rb-proficiency but low p16 expression have the worst clinical outcome, but may be those most likely to benefit from CDK4/6 inhibition.

These data represent the most comprehensive preclinical evaluation of a CDK4/6 inhibitor in ovarian cancer cell lines to date, and build the case for its clinical development in specific molecular subgroups of ovarian cancer.

Here, we are able to show that approximately half of the ovarian cancer cell lines examined showed high in vitro sensitivity to CDK4/6 inhibition. We therefore sought to validate biomarkers that predict in vitro response to PD-0332991. In doing so, we focused primarily on biomarkers that are known to be implicated in cell cycle signaling.

Rb-proficient human ovarian cancer cell lines with low p16 expression were most responsive to CDK4/6 inhibition. The effectiveness of CDK4/6 inhibition in this genetic context was recently underscored by functional data using genetic inhibition by RNAi in glioblastoma multiforme (GBM) cell lines. Only CDKN2A/C-deleted GBM cell lines were sensitive to RNAi-mediated depletion of either CDK4 or CDK6, whereas Rb-null and cell lines without apparent Rb pathway alterations were unaffected by RNAi targeting CDK4 or CDK6 (20).

Our studies also suggest that a gain of cyclin E1 or cyclin D1 gene copy number, or the presence of p53 mutations with subsequent low expression of p21 may confer resistance to CDK4/6 inhibition in ovarian cancer cells. Additional clinical studies, however, will be necessary to know whether assessment of these additional markers will help to refine response prediction or whether their expression/mutation correlates well with a regulatory-induced overexpression of p16.

Frequent genetic alteration affecting the CDK4/6-cyclin D-p16-Rb growth-regulatory pathway in ovarian cancer is well documented (10–15). However, differences may exist regarding the frequency of these alterations in cancer cell lines versus clinical samples. The most common alteration of this pathway in the ovarian cancer cell line panel used for this study was hemizygous and homozygous deletion of CDKN2A encoding p16, which was present in 14/40 (35%) of the cell lines examined. In contrast, the most likely cause of low p16 expression in clinical ovarian cancer tumors is likely to be promoter methylation of p16 that has been described to occur in approximately 40% of ovarian cancers (10). Importantly, currently available data suggest that deletions of CDKN2A appear to be less common in clinical samples of ovarian cancer compared with ovarian cancer cell lines (11).

Recent studies in breast cancer demonstrate that the Rb pathway is a key regulator in response to stress induced senescence (24). Mitogenic signals all engage cellular stress response programs in normal cells. If the level of damage is substantial, an apoptotic or senescent program limits the propagation of damaged cells. Activated p16 signaling drives premature and replicative senescence by inhibiting CDK4/6 and ultimately blocking cell cycle progression (25, 26). High expression of p16 can represent 2 different biological processes; a response to cellular stress or abrogation of functional Rb signaling (24). A cell with functional p16/Rb signaling will initiate stress-induced overexpression of p16 resulting in a proliferative arrest. On the other hand, a cell with a compromised Rb pathway will initiate a regulatory-induced overexpression of p16 due to unobstructed negative feedback but proliferate unimpeded (24). Vice versa, low expression of p16 can also represent 2 different biological processes: for example, low p16 was more common in low grade tumors and less aggressive mucinous or endometrioid ovarian tumors when compared with high-grade tumors or those of serous type histology. Moreover, all of the 25 women with low grade, early stage ovarian cancers in this study had low p16 and none of them recurred. Thus, p16 may be low due to lack of mitogenic cellular stress. In contrast, low p16 among FIGO stage III/IV and high-grade tumors was associated with significantly worse clinical outcome, likely because they were not able to initiate a regulatory-induced overexpression of p16 and cellular arrest.

In the current cohort, we found low or absent expression of Rb to be a rare event in ovarian cancer. This observation is consistent with earlier reports, indicating that whereas molecular genetic studies reveal frequent hemizygous deletions at the Rb locus in ovarian cancer, normal Rb protein expression was present in the majority (96%) of these tumors (14), suggesting that hemizygous deletion by itself without concomitant mutation may not be associated with reduced protein expression. Moreover, our own results of a preliminary study using array CGH in 128 primary ovarian cancer specimens indicate that homozygous or hemizygous deletions may be less common than originally suggested. In that study, we found homozygous or hemizygous deletions of Rb to occur in 2% and 5% of the cases, respectively.

In recent years, p16 has emerged as a valuable surrogate marker for high-risk human papillomavirus infection and shows increased immunoexpression with worsening grades of cervical intraepithelial neoplasia. Numerous studies support its role in the detection of high-grade dysplasia and have lead to the use of p16 immunohistochemistry in many laboratories. Importantly, this assay would be immediately available for clinical use for patient selection in future clinical trials with CDK4/6 inhibitors in ovarian cancer.

Taken together, our findings support further clinical evaluation of PD-0332991 as a single agent or in combination with chemotherapy in patients with ovarian cancer. Moreover, the assessment of functionally implicated response predictors in these clinical trials may help to identify the patient subgroup most likely to benefit from treatment with PD-0332991. Of the primary ovarian cancer patients included in this study, 37% demonstrated Rb-proficiency with low p16 expression. Among patients with advanced primary ovarian cancer, that subgroup had poor clinical outcome; but importantly, may have been identified as the group of patients most likely to benefit from CDK4/6 inhibition.

No potential conflicts of interest were disclosed.

This work has been supported in part by the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation, the Thelma L. Culverson Endowed Cancer Research Fund, and the Stranahan Foundation for Translational Cancer Research and Advanced Clinical Cancer Research.

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.