Specific inhibition of cyclin-dependent kinase 4/6 by PD 0332991 and associated antitumor activity in human tumor xenografts

The proliferation of eukaryotic cells typically involves an orderly progression through four distinct phases of the cell cycle: G₁, S, G₂, and M (1–3). Cyclin-dependent kinases (Cdks) play a key role in regulating cell cycle progression and to a large degree govern cellular transitions from growth phases (G₁ and G₂) into phases associated with DNA replication (S) and mitosis (M; refs. 4, 5). Regulation of Cdk catalytic activity occurs at multiple levels including phosphorylation and dephosphorylation of the Cdk itself; cyclin synthesis and degradation; and expression, degradation, and availability of naturally occurring protein inhibitors and subcellular localization of these various regulatory components (2, 5, 6).

Progression through the G₁-S phase requires phosphorylation of the retinoblastoma (Rb) protein by Cdk4 (7, 8) or the highly homologous enzyme Cdk6 (9, 10) in complex with their activating subunits, the D-type cyclins, D1, D2, or D3 (11). Hyperphosphorylation of Rb diminishes its ability to repress gene transcription through the E2F family of transcription factors and consequently allows synthesis of several genes, the protein products of which are necessary for DNA replication (12–15). Thus, the catalytic activity of Cdk4 or Cdk6 regulates a critical checkpoint for the G₁-S transition and the commitment to cell division (16).

More than 90% of human tumors abandon the control mechanisms for this transition point through a variety of genetic and biochemical adaptations (1, 17). Examples of these abnormalities include up-regulation of Cdk4 itself; amplification of the D-type cyclins; down-regulation of a naturally occurring inhibitor of Cdk4, called p16^INK4A; mutations in Cdk4 that prevent p16^INK4A binding to the enzyme; and deletion or mutation of Rb itself (17–21). All of these aberrations can lead to loss of proliferative controls either through elimination of the checkpoint altogether or through inappropriate or enhanced Cdk4 activity resulting in hyperphosphorylation of Rb. The frequency of these alterations alone clearly implies that abrogation of the G₁ checkpoint or acceleration of the Cdk4/cyclin D pathway provides a distinct advantage to cancer cells in terms of proliferation and perhaps survival.

Based on these observations, cyclin D–dependent kinases have been considered for many years a prime target for cancer chemotherapy (22, 23). Experimental evidence suggests that inhibition of cyclin D–dependent kinase activity may prevent tumor growth and/or at least partially revert the transformed phenotype. For example, reduction of cyclin D expression through antisense technology causes a concomitant decline in cyclin D–dependent kinase activity and results in inhibition of tumor growth, abolition of tumorigenicity, or, in some instances, tumor cell death (24–27). Other reports indicate that exogenous expression of p16^INK4A in tumor cells using adenoviral gene delivery systems or inducible promoters blocks proliferation and tumorigenic potency both in vitro and in vivo (28–31). These observations lend credence to Cdk4/6 as a target for cancer treatment and allow a reasonable expectation that a specific inhibitor of these enzymes would produce a meaningful therapeutic response.

Several drug discovery programs have produced potent small molecule Cdk inhibitors (22, 32–35) from a variety of chemical classes, which include purine analogues (36–40), pyrimidine analogues (41–43), indenopyrazoles (44, 45), pyridopyrimidines (46–48), pyrazolopyridines (49, 50), indolocarbazoles (51), pyrrolocarbazoles (52, 53), oxindoles (54, 55), and aminothiazoles (56). Several compounds are currently in clinical trials including flavopiridol, R-roscovitine (CYC202), UCN-01 (7-hydroxystaurosporine), and BMS-387032 (57, 58). Most of these compounds, however, inhibit multiple Cdks, with Cdk2 being a particularly common target in drug discovery programs because this enzyme is easily crystallized with inhibitors of varying molecular structure (59, 60). Interestingly, several recent reports have provided evidence that mammalian cells can continue to proliferate in the absence of Cdk2/cyclin E activity (61–63) possibly due to compensation by Cdk4 and/or Cdk6. These reports suggest that Cdk2 may be less attractive than Cdk4 as an anticancer target. Although some reports in the literature claim Cdk4-specific agents, these compounds are generally weakly selective for Cdk4 versus Cdk2 in isolated enzyme assays (10–20-fold), are weakly potent (64, 65), or produce a G₂ cell cycle arrest at elevated concentrations, which is inconsistent with Cdk4/6 inhibition (42, 43, 47, 51–53, 66). Several groups have reported significant Cdk4 selectivity versus Cdk2 without supporting biological data (67). Thus, the biological effects of selective inhibition of Cdk4 by small, drug-like molecules have not been shown previously.

The present report discloses PD 0332991 as a potent inhibitor of Cdk4/6 and provides evidence that the sole mechanism of action for this compound is inhibition of these enzymes.⁶

All Cdk-cyclin complexes were expressed in insect cells through baculovirus infection and purified as described previously (68). The substrate for the Cdk assays was amino acids 792 to 928 of pRb fused to glutathione S-transferase (10). Cdk assays and IC₅₀ determinations were done as described previously (47). Enzyme assays for other protein kinases were done as described previously (69–71).

All cell lines were obtained from American Type Culture Collection (Manassas, VA) and maintained at 37°C at 5% CO₂ in DMEM containing 10% fetal bovine serum (FBS; Life Technologies, Inc., Rockville, MD).

Cells were seeded at 2 × 10⁴ per well in a 96-well Cytostar T plate (Amersham Biosciences, Piscataway, NJ) and incubated overnight to allow cells to attach. Varying concentrations of PD 0332991 were added to the wells and incubated for 24 hours at 37°C. [¹⁴C]thymidine (0.1 μCi) was added to each well and incorporation of the radiolabel was allowed to proceed for 72 hours. Incorporated radioactivity was determined with a β plate counter (Wallac, Inc., Gaithersburg, MD).

Cell or tissue samples were lysed and Western blot analysis was done as described previously (47). Rb phosphorylation status was assayed with Ser⁷⁸⁰ and Ser⁷⁹⁵ phosphospecific antibodies (Cell Signaling Technology, Beverly, MA) and total Rb expression was monitored using Rb 4H1 monoclonal antibody (Cell Signaling Technology).

Rb 4H1, phospho-Rb (Ser⁷⁸⁰), and Ki-67 immunohistochemical staining were done using the Ventana Discovery autostainer (Ventana Medical Systems, Tucson, AZ) with heat at 37°C to 46°C. Slides were deparaffinized using Ventana EZ Prep buffer and antigen retrieval was done using the Ventana CC1 mild reagent (containing EDTA) at 100°C for 20 minutes. Endogenous peroxidase was inactivated using the Ventana Enhanced Inhibitor solution and nonspecific antibody binding was blocked using a 3% blocking solution (Roche Diagnostics, Indianapolis, IN) for 30 minutes. Primary antibodies [Rb 4H1 (1:100 dilution, catalogue no. 9309, Cell Signaling Technology), phospho-Rb (Ser⁷⁸⁰, 1:100 dilution, catalogue no. 9307, Cell Signaling Technology), and Ki-67 (MIB-1, 1:50 dilution, catalogue no. M7240, DAKO, Carpinteria, CA)] were manually applied separately for 30-minute incubation(s) followed by application of Ventana Endogenous Blocker kit solutions (4 minutes each) to block endogenous biotin and avidin. Secondary antibodies [goat anti-rabbit IgG (for pRb) or horse anti-mouse IgG (for Rb 4H1 and Ki-67) from Vector Laboratories (Burlingame, CA)] were incubated for 30 minutes. Following SA-HRPO conjugate (18 minutes), Ventana enhanced 3,3′-diaminobenzidine tetrahydrochloride (8 minutes), and copper sulfate solution (4 minutes) incubations, sections were counterstained with hematoxylin and bluing reagent for 4 minutes each. Slides were manually rinsed in diluted detergent, dehydrated through graded alcohols, cleared in xylene, and mounted with Permount.

Cells were harvested and washed in PBS containing EDTA (5 mmol/L). They were then washed in PBS containing 1% FBS (1% FBS/PBS), fixed in 85% ethanol, and stored at 4°C for at least 16 hours and up to 5 days. Cells were then washed again in 1% FBS/PBS and incubated at 37°C for 30 minutes in 1% FBS/PBS containing propidium iodide (40 mg/mL, Molecular Probes, Eugene, OR) and RNase A (250 mg/mL, Roche Diagnostics). Data were collected using a Coulter EPICS Elite ESP (Miami, FL) equipped with a Spectraphysics argon ion laser and analyzed using ModFit (Verity Software House, Inc., Topsham, ME). Results represent a minimum of 15,000 cells assayed for each sample.

Solid human tumor models Colo-205 colon, SW-620 colon, MDA-MB-435 breast, SF-295 glioblastoma, ZR-75-1 breast, PC-3 prostate, H125 non–small cell lung, H23 non–small cell lung, and MDA-MB-468 breast were developed from cell lines and maintained in severe combined immunodeficient mice. All tumor models were serially passaged as s.c. implants of tumor fragments (∼30 mg) from tumors weighing ∼1,000 mg. Tumor models were passaged ≤10 generations in vivo before returning to cryopreserved early passage stock material. The mice used in these studies were obtained from Charles River Breeding Laboratories (Wilmington, MA) and Taconic Farms (Germantown, NY; National Cancer Institute colonies). All animals were examined prior to the initiation of studies to ensure that they were healthy and acclimated to the laboratory environment. Mice were housed in barrier facilities with food and water provided ad libitum on a 12-hour light/dark cycle. Animal care was provided in accordance with Association for Assessment and Accreditation of Laboratory Animal Care International guidelines. All protocols involving animals were reviewed and approved by the Pfizer Laboratories animal care and use committee (Ann Arbor, MI).

Mice (18–22 g) were randomized and then implanted s.c. with tumor fragments (∼30 mg) into the region of the right axilla. Treatment was initiated when tumors reached 100 to 150 mg. PD 0332991 was given according to the schedule and dose indicated in the table and figure legends by gavage as a solution in sodium lactate buffer (50 mmol/L, pH 4.0) based on mean group body weight. In all experiments, there were 12 mice in the control group and 8 mice each in the treated groups. Additional details for each experiment are given in the table legends.

RNA was isolated from frozen tumor powders from three separate mice for each dose using the Trizol method (Invitrogen, Carlsbad, CA) and further purified with DNase treatment and Qiagen column according to Qiagen RNA cleanup protocols (Qiagen, Valencia, CA). Quantitative real-time PCR was carried out using the Taqman 5′ nuclease assay system with the signal from a gene-specific probe that is normalized to the signal for the reference gene of β₂-microglobulin (72–74). Primers and probes were ordered from Assay-on-Demand Gene expression products (Applied Biosystems, Foster City, CA). Each assay consisted of two unlabeled PCR primers and a FAM dye-labeled Taqman MGB probe. The first-strand cDNA was generated using total RNA (5 μg) and High-Capacity cDNA Archive kit in a 100 μL reaction (Applied Biosystems). PCR was done in 384-well plates (20 μL) containing equivalent to cDNA (11 ng), primer/probe sets (1 μL) for the gene of interest, and 2× Taqman Universal PCR Master mix (10 μL, Applied Biosystems). Duplicates were done for each cDNA sample. Thermal cycling was carried out in ABI PRISM 7900 with default cycling conditions and real-time fluorescence detection. Expression fold change was calculated using the manufacturer's suggested ΔΔC_t method.

Statistical analysis was done on the times for individual tumors (both treated and control) to reach 750 mg. Where tumors did not reach 750 mg, the regrowth curves were extrapolated to 750 mg. Time to 750 mg was selected as the examined variable because it is independent of the evaluation size chosen after completion of therapy (75). All data from each experiment were analyzed by ANOVA and, where significant differences were identified, analyzed using the Bonferroni t test for multiple comparisons versus control. Levels of statistical significance are indicated for individual curves in each figure. Statistical analyses were done with SigmaStat 3.0 for Windows.

Inhibition of Cdks by pyridopyrimidines has been reported previously (46–48). Optimization of this class of compounds for selective inhibition of Cdk4 was achieved by testing compounds against a small panel of four enzymes, including Cdk4/cyclin D1, Cdk2/cyclin A, fibroblast growth factor receptor, and platelet-derived growth factor receptor. Compounds that possessed high selectivity for Cdk4 versus the other kinases were further evaluated in an expanded panel of kinases and assessed for antiproliferative potential against MDA-MB-435 human breast carcinoma cells. Those compounds with potent antiproliferative properties were evaluated by flow cytometry for their ability to produce a clean G₁ arrest in tumor cells across a range of concentrations. This strategy led to the identification of a subset of pyridopyrimidines that displayed unprecedented levels of selectivity for Cdk4, and from these compounds, PD 0332991 (Fig. 1) was selected for its superior physical and pharmaceutical properties.

PD 0332991 is a potent and highly selective inhibitor of Cdk4/cyclin D1 kinase activity; it exhibits an IC₅₀ value of 0.011 μmol/L against this enzyme under the conditions described in Materials and Methods (Table 1). Another enzyme, Cdk6, which also complexes with cyclin D subunits and is highly homologous to Cdk4, can perform identical functions to Cdk4 by phosphorylating Rb at the same sites (9, 10). Consequently, inhibition of both Cdk4 and Cdk6 is necessary to ensure complete suppression of Rb phosphorylation and to produce the maximum therapeutic response with the greatest spectrum of antitumor activity. PD 0332991 inhibited Cdk6 with equivalent potency to Cdk4 (Table 1). Within a panel of enzymes, PD 0332991 exhibited absolute selectivity for Cdk4/6 with little or no activity against 36 additional protein kinases including other Cdks and a wide variety of tyrosine and serine, threonine kinases (Table 1). The structure of PD 0332991 and its relationship to PD 0183812 (47) suggest that PD 0332991 is likely to inhibit Cdk4 by binding to the ATP site; however, it was not possible to show competitive inhibition of Cdk4 by PD 0332991 using a kinetic analysis due to the near equivalence of the K_I and the enzyme concentration required to perform this assay. Similarly, it has not been possible yet to obtain crystal structures of Cdk4 either with or without bound ligands.

The only known natural substrates for Cdk4/6 are the Rb family of gene products, p110(Rb), p107, and p130 (76). Of the 16 known phosphorylation sites on Rb, 2 are specifically phosphorylated by Cdk4/6: Ser⁷⁸⁰ and Ser⁷⁹⁵ (77–79). The phosphorylation status of Rb at these specific sites in treated tumors therefore represents an appropriate biomarker for inhibition of Cdk4/6 by PD 0332991 in tumor cells and tissue. The IC₅₀ for reduction of Rb phosphorylation at Ser⁷⁸⁰ in MDA-MB-435 breast carcinoma cells was 0.066 μmol/L (Fig. 2A and B). PD 0332991 was equally effective at reducing Rb phosphorylation at Ser⁷⁹⁵ in this tumor with an IC₅₀ of 0.063 μmol/L, and similar effects on both Ser⁷⁸⁰ and Ser⁷⁹⁵ phosphorylation were obtained in the Colo-205 colon carcinoma (data not shown). In vitro time course experiments indicated that the reduction of Rb phosphorylation began to occur as soon as 4 hours after exposure to PD 0332991 and reached a maximum at 16 hours. The inhibition was completely reversible because phosphorylation on Ser⁷⁸⁰ and Ser⁷⁹⁵ began to return 2 hours after removal of the drug and was complete within 16 hours, by which time the cells were actively proliferating (data not shown).

PD 0332991 is a potent inhibitor of cell growth and suppresses DNA replication by preventing cells from entering S phase. The compound inhibited thymidine incorporation into the DNA of Rb-positive human breast, colon, and lung carcinomas as well as human leukemias, with IC₅₀ values ranging from 0.040 to 0.17 μmol/L (Table 2). PD 0332991 had no activity against Rb-negative cells, indicating that it does not interact with antiproliferative targets other than Cdk4/6. The compound was tested against the MDA-MB-468 human breast carcinoma and the H2009 human non–small cell lung carcinoma, both of which have deleted Rb and exhibited no antiproliferative activity in these cells at concentrations as high as 3 μmol/L (Table 2), which is 1 to 2 orders of magnitude higher than the concentration necessary to inhibit proliferation in Rb-positive tumor cells.

A selective Cdk4/6 inhibitor should cause a specific accumulation of cells in G₁ but have no effect on other phases of the cell cycle in which cells should continue to progress and eventually decline in number. MDA-MB-453 breast carcinoma cells exposed to varying concentrations of PD 0332991 for 24 hours showed a significant increase in the percentage of cells in G₁ in the presence of as little as 0.04 μmol/L PD 0332991 with a concomitant decline in other phases of the cell cycle (Fig. 3). Maximum effects were attained at 0.08 μmol/L and an exclusive G₁ arrest was maintained even at concentrations as high as 10 μmol/L, consistent with the complete absence of any other effects on the cell cycle. Similar results were obtained with Colo-205 cells (data not shown).

PD 0332991 exhibits significant antitumor efficacy against multiple human tumor xenograft models. In mice bearing Colo-205 colon carcinoma xenografts (p16 deleted), daily p.o. dosing for 14 days with PD 0332991 (150 or 75 mg/kg) produced rapid tumor regressions (Fig. 4A) and a corresponding tumor growth delay of ∼50 days with >1 log of tumor cell kill at the highest dose tested (Table 3). At 37.5 mg/kg, the tumor slowly regressed during treatment. Even at doses as low as 12.5 mg/kg, a 13-day growth delay was obtained indicating a 90% inhibition of tumor growth rate. Likewise, robust antitumor activity was seen in the MDA-MB-435 breast carcinoma (p16 deleted) where complete tumor stasis was apparent at 150 mg/kg (Fig. 4B) and some cell kill was evident at the highest dose (Table 3).

PD 0332991 was efficacious against a variety of human tumor xenografts (Table 3). Similar to the Colo-205, significant tumor regression was observed in mice bearing the SF-295 glioblastoma xenografts at doses of 150 mg/kg, and responses nearing stasis (complete suppression of tumor growth) were attained in the ZR-75-1 breast and PC-3 prostate tumor models (Table 3). Modest activity was achieved in H125 non–small cell lung carcinoma. The compound was completely inactive, however, against the Rb-negative MDA-MB-468 breast carcinoma, which is consistent with inhibition of Cdk4/6 as the sole mechanism by which PD 0332991 produces antitumor activity. All doses reported in Table 3 were well tolerated by the animals, with no obvious clinical signs and either minimal weight loss or some weight gain. Treated animals were maintained for up to 50 days after dosing to facilitate the observation of any delayed clinical toxicities. At all doses reported, the animals succumbed to tumor-related death or were sacrificed once their tumor burden reached ∼1,000 mg. Based on the data shown, 150 mg/kg is defined as a maximum tolerated dose of PD 0332991 given to mice daily for 14 or 28 days.

Despite complete regression of certain tumors in response to PD 0332991, the tumors eventually grew back after these short treatment periods. Since in recent years some in vitro evidence has suggested that tumors may have the potential to circumvent a Cdk4 block through activation or elevation of downstream control elements of the cell cycle (80–82), we wanted to determine if the new tumors that appeared in treated animals remained sensitive to further treatment with PD 0332991 or whether a tumor variant had evolved exhibiting acquired resistance to the compound. To address this possibility, Colo-205 colon tumors that had emerged after substantial regression from initial treatment with PD 0332991 in the experiment described in Fig. 4A were harvested and reimplanted into naïve mice. After the tumors grew to 100 to 150 mg, these tumor-bearing mice were treated with PD 0332991 using a schedule identical to the original experiment. The tumors responded with equal sensitivity to the drug and fully regressed, indicating that no resistance had developed during the initial treatment (Fig. 4C). Similar results were obtained with MDA-MB-435 tumors (data not shown).

In parallel with the in vivo efficacy tests, additional tumors were harvested for pharmacodynamic analysis to test if antitumor activity correlated with modulation of the target and to investigate whether the proposed biomarker might predict for efficacy. Efficacious and nonefficacious doses of PD 0332991 were given to mice bearing the MDA-MB-435 breast carcinoma and the phosphorylation status of Ser⁷⁸⁰ on Rb in tumor tissue was monitored over time. The results show that all doses caused a reduction in the biomarker shortly after drug administration but that phosphorylation returned at nonefficacious doses (12.5 and 37.5 mg/kg) over the 24-hour interval before the next dose (Fig. 5A). The highly efficacious dose of 150 mg/kg suppressed Rb Ser⁷⁸⁰ phosphorylation over the full 24-hour period, implying that, with this particular breast tumor, complete suppression of the biomarker needs to be maintained between drug doses to achieve maximum efficacy. Target modulation in this tumor by PD 0332991 given p.o. was also established by immunohistochemistry where animals treated with an efficacious dose of PD 0332991 were characterized by complete elimination of phospho-Ser⁷⁸⁰ from their tumor (Fig. 5B). This effect also was associated with a substantial reduction in staining for Ki-67, a well-known marker of proliferation (83). Similar experiments with the highly sensitive Colo-205 colon carcinoma showed that complete suppression of the biomarker between doses is not necessary to produce growth inhibition in this tumor. However, total inhibition must be maintained between doses to achieve tumor regression (Fig. 5C). Comparing the results of efficacy experiments using the highly sensitive colon tumor and the moderately sensitive breast tumor reveals a 7- to 8-fold difference in the dose necessary to produce comparable efficacy. The pharmacodynamic results show a similar difference between the two tumors in the dose required to suppress Rb Ser⁷⁸⁰ phosphorylation, indicating that modulation of this biomarker is strongly associated with efficacy.

Hypophosphorylated Rb has been proposed to inhibit cellular proliferation partly by suppressing the transcription of genes under the control of the E2F transcription factors (12–15). Four E2F-regulated genes were selected and their expression levels were monitored by reverse transcription-PCR in Colo-205 tumors from mice treated p.o. with PD 0332991. The genes include CDC2 that codes for Cdk1, CCNE2 that codes for cyclin E2, TK1 that codes for thymidine kinase, and TOP2A that codes for topoisomerase 2A. All four genes were down-regulated in a dose-dependent manner, with maximum reductions ranging from 13- to 87-fold (Fig. 5D). The extent of the gene changes paralleled the magnitude of the therapeutic response in this tumor, providing additional evidence that PD 0332991 functions through inhibition of Cdk4/6.

In this report, we describe PD 0332991 as a potent and highly selective inhibitor of Cdk4 and Cdk6 and show that suppression of these enzymes in human tumor xenografts results in significant antitumor activity. Given that a major obstacle to establishing the usefulness of a Cdk4/6 inhibitor has been the difficulty in obtaining a molecule with complete specificity for these enzymes versus other Cdks and protein kinases, considerable effort was taken to establish the selectivity of this compound. PD 0332991 was tested against 39 individual serine, threonine, and tyrosine kinases, representing most of the primary protein kinase families (84). Other than Cdk4 and Cdk6, the compound had little or no activity against any of these enzymes. Based on the understood role of Cdk4/6 in cell cycle progression, a specific Cdk4/6 inhibitor is predicted to produce an exclusive G₁ arrest. Consistent with this expectation, cells treated with concentrations of PD 0332991 as high as 200-fold above the IC₅₀ for growth inhibition maintained an unequivocal G₁ block as the sole perturbation in their DNA histograms. In contrast, previously reported Cdk4/D inhibitors have produced a G₂-M block at high concentrations (47, 52). PD 0332991 had no effect on proliferation in tumor cells that were Rb deficient at concentrations >50 times the levels that cause growth inhibition in Rb-positive cells, consistent with selective inhibition of Cdk4/6. Initiation of the G₁ block, emergence of antiproliferative activity, and dephosphorylation of Rb at the Cdk4/6-specific site of Ser⁷⁸⁰ all occurred at similar concentrations of PD 0332991. Collectively, these data show that the biochemical activities and biological effects of PD 0332991 are in accord with exclusive inhibition of its target enzymes and suggest a high degree of selectivity in the cellular environment.

Oral administration of PD 0332991 to mice bearing human tumor xenografts clearly showed that inhibition of Cdk4/6 can result in a significant antitumor effect. The spectrum of responses ranged from partial growth inhibition and tumor stasis to complete regressions, depending on the tumor, with a therapeutic index as high as 12 in the Colo-205 model. [The therapeutic index is defined as the maximum tolerated dose divided by the lowest dose to give a tumor growth delay of >50%.] The anticancer activity was associated with biochemical changes that are consistent with the mechanism of action for PD 0332991. Thus, significant antitumor activity was detected only at doses where there was a sustained and significant reduction of phospho-Ser⁷⁸⁰ on Rb. The decline in Rb Ser⁷⁸⁰ phosphorylation in tumor tissue was confirmed by immunohistochemistry techniques, which showed substantial reduction of this marker in tumors from treated animals. Moreover, there was a concomitant elimination of staining for Ki-67, a commonly measured marker of proliferation (83). Furthermore, reverse transcription-PCR of selected genes in tumor tissue known to be under the control of E2F and down-regulated in cells expressing a nonphosphorylatable Rb (85) showed a 13- to 87-fold reduction in expression in tumors from animals treated with therapeutically active doses of PD 0332991. Finally, PD 0332991 was completely inactive against Rb-negative tumor xenografts, which is expected based on its proposed mechanism of action. Taken together, these in vivo data are consistent with inhibition of Cdk4/6 as the sole basis for the observed antitumor activity.

The results of this study address two issues that have been raised in the past concerning inhibitors of Cdk4/6 as therapeutic agents for oncology, the first being an assumption that inhibition of Cdk4/6 may result in a cytostatic phenotype, causing little more than temporary tumor growth arrest. Indeed, cytostasis was the phenotype observed in tissue culture experiments in vitro, wherein cells exposed to PD 0332991 ceased proliferating but did not die even after prolonged exposure to the compound. In vivo, however, we show that inhibition of Cdk4/6 not only produces a robust growth suppressive effect in many human tumor xenografts but also is capable of causing complete regressions in some tumors. Secondly, based on results with in vitro culture models, it has been suggested that tumors might be able to circumvent the antiproliferative effects of Cdk4/6 inhibition through activation or elevation of downstream control elements of the cell cycle such as Cdk2/cyclin E activity or c-myc (80–82). The results of this study not only show a robust and sustained therapeutic response to Cdk4/6 inhibition but also show that tumors do not readily develop an ability to overcome a cell cycle block incurred through inhibition of Cdk4/6. Colo-205 human tumor xenografts treated with PD 0332991 for 14 days and then allowed to regrow over several weeks do not develop resistance and remain sensitive to inhibition of Cdk4/6. These results complement recent data suggesting that Cdk2 may not be generally required for proliferation in mammalian cells and that Cdk4/6 alone might be sufficient to regulate the G₁-S phase transition (61). In fact, Cdk2 knockout mice have been reported to be entirely viable (86). Taken together, the results support Cdk4/6 as a relevant target for cancer chemotherapy.

We can only speculate on how an inhibitor that appears cytostatic in vitro can cause tumor regressions in vivo. One possibility is that Cdk4/6 activity is required for tumor survival. The fact that Cdk4 disruption renders MEF cells resistant to transformation by certain oncogenes (87) implies that the transformed phenotype requires Cdk4 activity. Furthermore, cyclin D1 is a true oncogene capable of transforming cells itself, and p16, a natural inhibitor of Cdk4 that is commonly absent or inactivated in tumors, has all the characteristics of a tumor suppressor. Both of these observations imply that Cdk4 activity is a critical component of oncogenesis in certain tumors. Another potential explanation for the observed tumor regressions might be that most solid tumors contain a significant population of spontaneously dying cells in addition to a proliferating fraction, with the latter having a slight advantage in an expanding tumor mass. If proliferation were slowed to the point that the naturally dying population was dominant, the net result would be regression of the tumor. Because Rb interacts with well over 100 proteins of various function (88) and because the response to PD 0332991 is tumor specific, sensitivity is likely governed by multiple factors.

In conclusion, the striking therapeutic response to inhibition of Cdk4/6 observed in the current study makes this approach attractive for clinical application because an inhibitor of Cdk4/6 could have the potential to cause tumor regressions or cures as a single agent. Pragmatically, conventional clinical means of assessing antitumor activity should be appropriate for PD 0332991 and mechanistically related compounds. Furthermore, we have shown that several robust biomarkers are measurable by immunohistochemistry and could possibly be used to detect drug-related target modulation in patients or perhaps eventually to predict therapeutic response. PD 0332991 is scheduled for clinical trials in 2004 and should help provide an answer to whether selective inhibitors of Cdk4/6 can provide a therapeutic benefit in cancer patients.

We thank Joseph Repine, Hairong Zhou, Dennis McNamara, John Quin, Cathlin Flamme, Michael Waldo, and Vladimir Beylin for contributions and Alexander Bridges, Seth Sadis, and Ellen Dobrusin for support.