Pharmacologic properties of AG-012986, a pan-cyclin-dependent kinase inhibitor with antitumor efficacy

Cyclin-dependent kinases (CDK) and their regulatory cyclin partners play critical roles in cell cycle control and the regulation of cell transcription. Progression through the different stages of cell cycle is governed by the activities of the CDK1, CDK2, CDK4, CDK6, and possibly CDK3. CDK4/cyclin D, CDK6/cyclin D, and CDK2/cyclin E phosphorylate the retinoblastoma (Rb) protein at multiple sites, which results in activation of the E2F family of transcription factors and serves as a trigger for cells to advance beyond the G₁ checkpoint into S phase (1–3). During S phase, CDK2/cyclin A phosphorylates several proteins, including E2F, to regulate progression through S phase. In fact, it is the phosphorylation and inactivation of the E2F/DP1 transcription factor by CDK2 that is partially responsible for orchestrating an orderly exit from S phase, and interference with this process can lead to apoptosis (4). CDK1/cyclin B functions during G₂-M to initiate mitosis, and it also phosphorylates members of the antiapoptotic protein family, including survivin and Mcl-1. Inhibition of this phosphorylation can result in apoptosis and anticancer activity in vivo (5). CDK7, CDK8, CDK9, and CDK11 are involved with regulation of transcription, and CDK7 also phosphorylates and directly regulates other CDKs. The inhibition of CDK9 has been reported to cause down-regulation of Mcl-1 protein through RNA polymerase II–dependent transcription, resulting in tumor cell apoptosis (6, 7).

Several CDK pathways, particularly those involved in cell cycle control, are deregulated and activated in most human cancers in a variety of ways and this activation contributes to tumorigenesis. Therefore, CDKs have been extensively pursued as cancer pharmacology targets, but many of the first-generation CDK inhibitors have suffered from poor potency, selectivity, or efficacy (8). Flavopiridol was the first CDK inhibitor to advance to the clinic. It is a broad-spectrum moderately potent CDK inhibitor and inhibits CDK1, CDK2, CDK4, CDK6, CDK7, and CDK9 with IC₅₀s in the 0.1 to 0.4 μmol/L range (9). R-roscovitine (seliciclib, currently in phase II trials), a purine-based CDK inhibitor, appears to be more selective toward CDK1, CDK2, CDK7, and CDK9 than CDK4, CDK6, and CDK8 (10), although the lack of potency of this compound makes broad-spectrum selectivity difficult to assess. The efficacies of flavopiridol (9) and seliciclib (11) in solid tumor models have been shown to correlate with the inhibition of cell cycle progression through apparent cytostatic growth arrest. Both of these drugs also inhibit CDK9, and at least part of the mechanism for the induction of apoptosis in hematopoietic tumor cell lines has been attributed to the inhibition of CDK9-mediated cellular transcription (7, 12). Indolinone-based CDK inhibitors (e.g., SU9516) follow a similar inhibitory pattern as purine-based inhibitors, with increased potency (6). Greater potency and selectivity toward CDK2 has recently been achieved with compounds such as PHA533533 and BMS-387032 (8). PD332991 is unique among the CDK inhibitors for its potency and high selectivity for CDK4/6 over other CDKs and kinases (13), and this compound is currently in phase I/II clinical trials.

Our understanding of the complexity of the roles of the CDKs has continued to grow in recent years. Published data have suggested that cells can recover from targeted inhibition of a single CDK likely by compensatory activity from other CDKs (14, 15). Therefore, a multitargeted or pan-CDK inhibitor may be preferred for more robust antitumor efficacy. Based on this principle, the multitargeted CDK inhibitor, AG-012986, was designed and assessed for the antiproliferative activity in tumor cells and in vivo tumor models.

AG-012986, 4-[4-amino-5-(2,6-difluoro-benzoyl)-thiazol-2-ylamino]-N-(2-dimethylamino-1R-methyl-ethyl)-benzamide dihydrochloride was synthesized by the Pfizer Chemist (16). For in vitro assays, 10 mmol/L stock solutions of the compound were prepared in 100% DMSO. The routine vehicle of AG-012986 for in vivo administration was 30% polyethylene glycol 400 in aqueous 5% dextrose.

CDK activity was measured by monitoring the incorporation of [γ-³²P]ATP into a fragment of the Rb protein or histone H1 in 96-well plates using microfiltration and phosphoimaging as described previously (17). The K_i was measured by fitting the dose-dependent inhibition of CDKs to an equation for tight-binding competitive inhibition. Kinase counterscreening was conducted via in-house kinase assays, via commercial services, or in collaboration with the University of Dundee, and IC₅₀s were determined from four-variable fits of the dose-dependent inhibition of enzymes in the presence of ATP concentrations = K_m.

All cell lines used in our study, except MV522, were obtained as frozen stocks from American Type Culture Collection and cultured in RPMI 1640 supplemented with 10% FCS. Human MV522 tumor cells were provided by Dr. Michael Kelner (University of California-San Diego). For tumor cell proliferation studies, cells from log-phase growth cultures in 96-well plates were treated with various concentrations of AG-012986 at 37°C. After 72 h, cell viability was measured by the 3-(4,5-dimethylthiazol-2-yl)-2,5-2H-diphenyltetrazolium bromide assay. Tumor cell survival (colony-forming assay) was done by seeding 150 cells into 60-mm-diameter dishes for 4 h followed by treatment with AG-012986 at various concentrations for different periods before drug removal. Colonies were visible in 2 to 3 weeks for counting. IC₅₀s were calculated by using nonlinear four-variable curve fitting.

Cells in mid-log-phase growth were treated with AG-012986, and at termination, both detached and adherent cells were harvested. Cell cycle analysis was done with a FACSCalibur flow cytometry system (Becton Dickinson) using the CycleTEST Plus kit (Becton Dickinson).

Terminal deoxynucleotide transferase–mediated dUTP nick end labeling (TUNEL)/4,6-diamidino-2-phenylindole staining (the DeadEnd Colorimetric TUNEL System; Promega) was done to detect apoptosis using immunofluorescence microscopy. Slides were scored based on the percentage of positive FITC-stained cells (apoptotic) versus 4,6-diamidino-2-phenylindole–stained cell counts (total). The final score was obtained from the average count in five representative fields per slide.

Cell or tissue lysates were subjected to Western blot analysis as described previously (18). Primary antibody for phospho-Rb was rabbit anti-human phospho-Rb Ser⁷⁹⁵ (Cell Signaling; 1:300 dilution) and secondary antibody was conjugated goat anti-rabbit (Cell Signaling; 1:1,000 dilution). Poly(ADP-ribose) polymerase (PARP) cleavage was assessed by using rabbit anti-PARP IgG (Santa Cruz Biotechnology; 1:1,000 dilution). β-Actin protein expression was determined as an internal standard. The protein level of phospho-Rb Ser⁷⁹⁵ was quantified using the FluorChem 8800 digital image system (Alpha Innotech).

All animal husbandry and experimental procedures conducted complied with the Guide for the Care and Use of Laboratory Animals (ILAR, 1996) and were approved by the Pfizer Global Research and Development La Jolla Institutional Animal Care and Use Committee. The pharmacokinetics and antitumor efficacy of AG-012986 in mice were evaluated following i.p. injection in 5 mL/kg or s.c. infusion via implanted minipumps at 0.5 μL/h for a period of 7 days (Alzet).

Whole blood was collected retro-orbitally from three to five mice per treatment group for each time point. Plasma was prepared immediately and an internal standard was added in the plasma before protein precipitation with a mixture of acetonitrile and methanol (50:50, v/v). AG-012986 concentrations were analyzed using Quattro Ultima triple-quad mass spectrometry (Waters). Typically, 50 μL supernatant was injected onto a 50 × 2.1 mm/XDB-C18 column (Agilent) and the compound was eluted with a gradient of 35% to 93% acetonitrile in water over 3 min at a flow rate of 0.4 mL/min. Pharmacokinetic variables were calculated using the noncompartmental model of WinNonlin Professional Software version 4.0.1 (Pharsight).

Severe combined immunodeficient or athymic NCr-nu/nu mice were obtained from Charles River Breeding Laboratories. Tumor models were chosen based on in vitro response to CDK inhibitors, genetic status, and robustness. Two million cells were suspended in 30% (v/v) Matrigel (VWR) at a final volume of 200 μL and implanted in the dorsal region of mice. When tumor volumes reached 100 to 150 mm³, mice were randomized and 10 to 12 mice were placed in each group. Treatments for efficacy studies were done using doses at or below the maximum tolerated dose. The maximum tolerated dose in mice, defined as the dose that induced <10% drug-related weight loss with no other adverse effect, was assessed before efficacy studies. Tumors were measured two to three times weekly using calipers and tumor volume was calculated as 0.5 × [length × width²]. Additional details for each experiment are given in the table or figure legends.

Tumor samples for pharmacodynamic analysis were collected from mice in efficacy studies. BrdUrd pulse was done by i.p. administration of 30 mg/kg BrdUrd to tumor-bearing mice 2 h before tumor harvest. Excised tumor pieces were either fixed in 10% formalin for immunohistochemical analysis or snap frozen in liquid nitrogen for Western analysis.

Immunohistochemistry was done on paraffin-embedded tumor sections. The slide preparation and antigen retrieval were done as described previously (19, 20). For Ki-67 staining, rabbit monoclonal antibody (Lab Vision; 1:800 dilution) and goat anti-rabbit IgG (DAKO; 1:200 dilution) were used as primary and secondary antibodies, respectively. BrdUrd immunohistochemical staining was done with a biotin-conjugated monoclonal antibody (Zymed/Invitrogen; 1:20 dilution) followed by incubation with an avidin-biotin-peroxidase complex (Vectastain ABC kit, Vector Laboratory). Quantitative measurement of Ki-67- or BrdUrd-positive cells was determined using the Chromovision automated cell imaging system.

Statistical analysis was conducted using Prism software (GraphPad). One-way ANOVA followed by Dunnet's t test was done to assess the significant difference between the tumor volumes of the vehicle and AG12986-treated groups when the mean tumor volume in the vehicle-treated group reached 700 to 800 mm³.

As an ATP-competitive inhibitor, AG-012986 displays nanomolar potency (K_i) against the cell cycle CDKs, CDK4/cyclin (9.2 nmol/L), CDK2/cyclin A (94 nmol/L), and CDK1/cyclin B (44 nmol/L). CDK9/cyclin T and CDK5/p35 were also inhibited, with an IC₅₀ of 4 and 22 nmol/L, respectively. The selectivity of AG-012986 against a diverse panel of kinases is shown in Table 1A. AG-012986 showed at least 100-fold selectivity for CDK4 over the kinases profiled, with the exception of calmodulin-dependent kinase II (∼50-fold selective). AG-012986 was also profiled against 64 receptors, ion channels, and transporters in radioligand binding assays (Table 1B). Only three of these potential off-target interactions showed IC₅₀s of <10 μmol/L that translated into functional activity in subsequent cell- or tissue-based assays. AG-012986 displayed antagonism toward the calcium type L ion channel and the serotonin transporter with K_is of 2.44 and 2.26 μmol/L, respectively, and agonism for the histamine H3 receptor with a K_i of 0.837 μmol/L.

The antiproliferative activity of AG-012986 was tested against a panel of human tumor cell lines in culture (Table 2A). The average IC₅₀ of AG-012986 against the 18 cell lines was 120 nmol/L (55 ng/mL). In 13 of 18 cell lines tested, AG-012986 displayed IC₅₀s of <100 nmol/L, showing its broad-spectrum antiproliferative activity. The antiproliferative activities of AG-012986 were independent of the genetic status of p53 and Rb.

As one of the known natural substrates for CDKs, Rb would be expected to become hypophosphorylated upon drug-induced inhibition of CDKs (21–23). Of the 16 known phosphorylation sites on Rb, Ser⁷⁹⁵ can be phosphorylated by CDK2/cyclin A, CDK2/cyclin E, and CDK4/cyclin D1. Thus, a phosphospecific antibody to Ser⁷⁹⁵ of Rb was used to investigate the effect of AG-012986 on the phosphorylation status of Rb Ser⁷⁹⁵. As determined by Western blot, treatment of asynchronous HCT116 human colon cancer cells with AG12986 resulted in a concentration-dependent loss of Rb phosphorylation at the Ser⁷⁹⁵ site. Treatment of HCT116 tumor cells with up to 240 nmol/L AG-012986 showed minimal effects on the status of phospho-Rb Ser⁷⁹⁵ at 8 h after treatment (Fig. 1A). However, after 24-h treatment, 60 to 240 nmol/L AG-012986 induced dose-dependent Rb Ser⁷⁹⁵ hypophosphorylation, with maximal effects at >120 nmol/L. These results indicate that treatment with AG-012986 for >8 h is required for achieving the hypophosphorylation.

The cell cycle response of AG-012986 on tumor cells was evaluated using flow cytometry (Fig. 1B). When HCT116 cells were treated with concentrations of AG-012986 between 30 and 120 nmol/L for 24 h, cell cycle distribution analysis exhibited accumulation of cells in G₁ phase of the cell cycle, consistent with CDK4 inhibition. However, when cells were treated with AG-012986 at ≥240 nmol/L, G₂-phase accumulation was observed after 24 h, consistent with CDK1 inhibition. Cell cycle analysis results also indicated that transient exposure (<8 h) of AG-012986 at concentrations up to 1 μmol/L failed to induce any sign of cell cycle arrest (data not shown). To further investigate the downstream effect of CDK inhibition by AG-012986 in vitro, we assessed apoptosis post-treatment by the TUNEL assay using immunofluorescence microscopy. HCT116 cells were treated with AG-012986 at concentrations ranging from 30 to 240 nmol/L for 8 or 24 h. At 8 h, no apoptosis was observed (data not shown). At 24 h, AG-012986 at ≥120 nmol/L induced a greater proportion of apoptotic cells (Fig. 1C) than control conditions, with an IC₅₀ of ∼160 nmol/L (curve not shown). Apoptosis induced by AG-012986 (240 nmol/L) in HCT116 cells was confirmed by assessing PARP cleavage and electron microscopy (Supplementary Figs. S1 and S2).⁶

The colony-forming assay assessed tumor cell survival under varying treatment durations to assess net cell kill (Fig. 1D). SW620 human colon carcinoma cells were treated for different periods to assess tumor cell survival. At the desired time points (8, 24, 72, and 320 h), the treatment was stopped by removing medium containing AG-012986 before continuing growth in fresh medium. At time points ≥72 h, substantial cytotoxicity (IC₅₀ <100 nmol/L) of AG-012986 was shown. Moderate activity with an IC₅₀ of 300 nmol/L was observed at 24 h, whereas minimal cytotoxic activity was observed at the 8-h time point. These results suggest that a minimum of 24-h treatment duration is desired for AG-012986 to show efficacy. Similar results were observed in other cell lines.

The in vivo efficacy of AG-012986 was evaluated at or under maximum tolerated dose (35-40 mg/kg/d; 8-12 days) in s.c. implanted human tumor xenograft models. The dosing period varied depending on the tumor growth rate and generally covered the tumor exponential growth period. The results of representative studies of different tumor models are summarized in Table 2B. AG-012986 showed >83% tumor growth inhibition (TGI) in 10 of the 11 tumor models tested. Figure 2 shows the dose-dependent efficacy of AG-012986 against COLO205 colon carcinoma (Fig. 2A) and NCI-H522 non-small cell lung carcinoma (Fig. 2B) when mice were treated at different doses daily for 12 days. In mice bearing COLO205 tumors, AG-012986 was able to induce net log tumor cell kill of 1.20 and 0.64 (Fig. 2C) when dosed at 40 and 20 mg/kg, respectively. In the NCI-H522 tumor model, AG-012986 dosed at 35 and 17.5 mg/kg produced 83.1% and 48% TGI, respectively, but without net log tumor cell kill (Fig. 2D).

In parallel with the in vivo efficacy evaluations, tumors from satellite groups were harvested for pharmacodynamic analysis. COLO205 tumors were harvested at 8 and 24 h after mice received a single i.p. injection of efficacious (20 and 40 mg/kg) or nonefficacious (10 mg/kg) doses of AG-012986. Phospho-Rb Ser⁷⁹⁵ status and PARP cleavage in tumor extracts was assessed by Western blot (Fig. 3A). After 24 h, efficacious doses (20 and 40 mg/kg) of AG-012986 reduced Rb Ser⁷⁹⁵ phosphorylation by >80% relative to controls and induced concomitant PARP cleavage. A nonefficacious dose (10 mg/kg) of AG-012986 only reduced Rb Ser⁷⁹⁵ phosphorylation by 50% and failed to cause PARP cleavage. A decrease in phospho-Rb Ser⁷⁹⁵ and cleaved PARP (data not shown) was not observed at any dose at the 8-h time point. In addition to Rb Ser⁷⁹⁵ hypophosphorylation, time-dependent inhibition of cell cycle arrest was observed in COLO205 tumors following i.p. administration of AG-012986. S phase, as estimated by BrdUrd uptake in COLO205 tumors (Fig. 3B), significantly decreased between 12 and 24 h after mice were administered a single dose of AG-012986 (20 mg/kg). Dose-dependent Rb Ser⁷⁹⁵ hypophosphorylation was also shown in MV522 tumor-bearing mice after a single administration of AG-012986. Moreover, there was a concurrent elimination of Ki-67-stained cells assayed by immunohistochemical staining. The decrease of phospho-Rb Ser⁷⁹⁵ and Ki-67 proliferation index in MV522 tumor correlates with the TGI (done in a separate study) induced by AG-012986 (Fig. 3C), suggestive of target-associated efficacy.

The pharmacokinetic profile of AG-012986 in mice was evaluated following i.p. or s.c. implanted minipump administrations. These dosing regimens provided the opportunity to test the effects of plasma concentration or exposure profile on efficacy. In the tested dose range (12-75 mg/kg), AG-012986 via i.p. administration produced slightly greater than dose-proportional increases in C_max and AUC (Fig. 4A and C) with an average half-life of 1 h. In mice receiving AG-012986 via s.c. minipumps (Fig. 4B and D), the steady-state plasma levels were achieved within 6 to 24 h and remained relatively constant for up to 120 h post-implant. The daily AUC_0-24, estimated from average of the total AUC_0-168, was dose proportional in the range from 10 to 40 mg/kg/d. Mice receiving AG-012986 via a s.c. implanted pump displayed substantially (∼30-fold) lower plasma levels in comparison with the C_max values achieved after i.p. bolus administration at similar doses.

Dosing optimization was done by comparing the efficacies of AG-012986 in COLO205 tumor-bearing mice using various dosing regimens via i.p. administrations or s.c. implanted pump infusion (Table 3). Due to dose proportionality, some AUC and C_max values in Table 3 were estimated based on the linear fit from the data in pharmacokinetic section (Fig. 4C). After a 7-day infusion at 20 mg/kg/d, mice in group 2 (Table 3) received a 3-day rest and the pump was replaced with a new pump infusing AG-012986 at the same dose. Significant efficacy (TGI, 77.1%) was observed in this group. Under a different dosing schedule, AG-012986 at 20 mg/kg/d via i.p. administration for 12 days (Table 3, group 1) induced 71.3% TGI. Although the total AUC and total dose were comparable between these two groups, the plasma C_max in group 2 (s.c. infusion pump) was significantly lower than that of group 1 dosed i.p. (0.273 versus 5.18 μg/mL). Higher unsustained C_max values did not correlate with better efficacy. This relationship was further shown by comparing i.p. administration of AG-012986 at 40 mg/kg/d (group 4) with 20 mg/kg twice daily (group 5) for 8 days. The C_max value at 40 mg/kg was nearly twice as high as the C_max resulting from administration at 20 mg/kg twice daily; however, no proportional increase of efficacy was observed, suggesting that the efficacy of AG-012986 was not related to the C_max.

The effect of dosing interval on the efficacy of AG-012986 was also evaluated by comparing an intermittent dosing schedule with the schedule of daily administration. On a schedule of daily i.p. administration at 37.5 mg/kg, AG-012986 displayed 92.3% TGI (Table 3, group 7). Increasing the dosing interval from 24 to 48 h (group 8) resulted in less efficacy (TGI, 72.9%).

The effects of total dose or total exposure on efficacy of AG-012986 were evaluated by adjusting the ratio of daily dose to treatment period. AG-012986 administration at 40 mg/kg/d for 7 days (280 mg/kg total) via s.c. pump (Table 3, group 3) delivery displayed moderate efficacy (TGI, 47.9%). However, better efficacy (77.1%) was seen when the same total dose was delivered by expanding the dosing period to 2 weeks at a lower daily dose (20 mg/kg/d) via two sequentially implanted pumps (Table 3, group 2).

AG-012986 showed the characteristics expected of a pan-CDK inhibitor: the mechanism of action and pharmacodynamic studies show a correlation between pan-CDK inhibition and inhibition of cell growth, cell cycle effects, and induction of apoptosis. The molecule displayed potent inhibition of several cell cycle CDK enzymes, as well as inhibition of Rb phosphorylation in vitro and in vivo, in both a time- and dose-dependent manner. The inhibition of Rb phosphorylation correlated with cell growth inhibition in cell cultures and implanted tumors. The arrest of cells at both G₁ and G₂ phases, as well as hypophosphorylation of Ser⁷⁸⁰, Ser⁸²¹ (data not shown), and Ser⁷⁹⁵, show the pan-CDK inhibitory activity of AG-012986. The cell cycle arrest, shown by fluorescence-activated cell sorting analysis in vitro and BrdUrd uptake in vivo, was consistent with the decrease in the clinically relevant proliferation biomarker, Ki-67. Another CDK inhibitor, PD0332991, has been shown to decrease Ki-67 in tumor sections (13), showing a common link between CDK inhibition and this biomarker.

The antiproliferative effects of CDK inhibitors could be attributed to a decrease in cell division or an increase in cell death (24). In HCT116 cells, immunofluorescent TUNEL staining showed apoptosis induced by AG-012986 with an IC₅₀ of 160 nmol/L after 24 h of treatment (Fig. 1C). PARP cleavage and electron microscopy assessment (Supplementary Figs. S1 and S2)⁶ further confirmed that AG-012986 (at 240 nmol/L) induced apoptosis. In vivo, apoptosis was also seen in COLO205 tumors via PARP cleavage after mice were treated with efficacious concentrations (20 and 40 mg/kg) of AG-012986 (24 h). The apoptosis induced by AG-012986 under both in vitro and in vivo settings occurred concomitantly with the hypophosphorylation of Rb Ser⁷⁹⁵ and cell cycle arrest, suggesting that the apoptosis was a downstream effect of cell cycle CDK inhibition. This finding agrees with literature reports that tumor cells treated with other cell cycle CDK inhibitors show an overall increase in apoptosis in a cell line–specific manner (25, 26). The fact that COLO205 tumors showed some net log cell kill after AG-012986 treatment for 12 days at 40 mg/kg shows the ability of the compound to induce tumor cell death in vivo with repeated dosing. Possible mechanisms of apoptosis could include deregulation of the feedback inhibition of CDK2/cyclin A, which could cause aberrant exit from S phase, or inhibition of CDK1-induced phosphorylation of the antiapoptotic proteins survivin or Mcl-1. Cell death could also arise from the depletion of the Mcl-1 antiapoptotic protein induced by CDK9 inhibition as was reported for flavopiridol (12), seliciclib (7), and SU9516 (6) in hematopoietic tumor cells. However, the onset of apoptosis in those studies was significantly more rapid (3-5 h) than observed for AG-012986. At efficacious concentrations/doses, AG-012986 did not induce apoptosis within 8 h. The sum of these data suggests that the major mechanism of antitumor activity of AG-012986 is via inhibition of the cell cycle CDKs, although we cannot exclude the possibility that some cytotoxicity may be attributed to off-target effects.

The duration of treatment of tumor cells with a cell cycle CDK inhibitor would be expected to be critical. In vitro, tumor cells treated with AG-012986 at efficacious concentrations showed dose-dependent Rb Ser⁷⁹⁵ hypophosphorylation in 24 h but not 8 h (Fig. 1A). The colony-forming assay also emphasized the value of treatment duration on in vitro efficacy; minimal cytotoxic effects were observed following 8-h exposure of AG-012986, whereas better efficacy was observed when SW620 cells were treated for ≥24 h (IC₅₀ <300 nmol/L). In order for AG-012986 to inflict target-associated cytotoxicity in vivo, maintenance of plasma levels at or above efficacious concentrations (C_eff) was required for extended periods. The exact efficacious concentration (C_eff) of AG-012986 in the mouse model was not defined in this report and was estimated from the in vitro antiproliferation assay (IC₅₀ = 55 ng/mL, total). The unbound fraction of AG-012986 in mouse plasma is ∼20%, whereas ∼60% is unbound in the in vitro antiproliferation assay. Taking into account the differences in plasma protein binding, the estimated in vivo C_eff is ∼165 ng/mL.

The pharmacokinetic and pharmacodynamic data suggest that maintaining plasma levels of the CDK inhibitor at or above C_eff over a dosing period is a better predictor of antitumor activity than acutely achieved C_max. The in vivo antitumor efficacy of AG-012986 was independent of C_max, as comparable efficacies were observed between groups administered 20 mg/kg via i.p. bolus daily for 12 days and s.c. implanted minipump for 14 days (Table 3). In addition, AG-012986 dosed (i.p.) at 40 mg/kg/d for 8 days produced similar efficacy (TGI, 45.8%) to 20 mg/kg twice daily (TGI, 47.1%); the higher C_max value did not yield increased efficacy. The in vivo efficacy evaluation results are consistent with the in vitro pharmacodynamic data showing that transient exposure (<8 h) of AG-012986 even at high concentrations (1,000 nmol/L) failed to induce Rb Ser⁷⁹⁵ hypophosphorylation and cell cycle arrest (data not shown). Further evidence of the correlation of efficacy with treatment duration rather than peak plasma levels was shown by comparing two groups of mice that were given the same total dose via s.c. implanted minipumps (Table 3, groups 2 and 3). AG-012986 displayed better efficacy (77.1%) at 20 mg/kg for 14 days compared with 40 mg/kg for 7 days (47.9%). Not surprisingly, these results suggest that, for the same total AUC, better efficacy is achieved if the plasma concentration is maintained at or above the C_eff (165 ng/mL) for extended periods. A similar relationship between pharmacokinetics and efficacy has been observed with other CDK inhibitors. For the initial clinical development of flavopiridol, a long continuous infusion schedule (24-72 h) was implemented based on preclinical studies demonstrating that longer period of drug exposure time improved the efficacy (27, 28). Seliciclib was also reported to have better target-associated efficacy as exposure time increased to 16 h (29).

Dosing optimization was further done by comparing between dosing intervals of 24 and 48 h for 12 i.p. injections of AG-012986 at 37.5 mg/kg. AG-012986 appeared less efficacious when the dosing interval was increased from 24 to 48 h (Table 3). An extended duration of subefficacious plasma concentrations of AG-012986 could allow tumor cells to escape cell cycle arrest and begin proliferating. Pharmacodynamic endpoint assessment indicated that the largest decrease of BrdUrd uptake occurred between 12 and 24 h after i.p. administration of AG-012986. After 24 h, tumor cells started to proliferate, as indicated by increasing BrdUrd uptake, again emphasizing the importance of extended maintenance of drug plasma levels above C_eff for extended duration for cell cycle inhibition.

The goal of dosing optimization is to maximize the therapeutic potential of a drug while minimizing the risks for toxicity. AG-012986 has been reported to show toxicities that are likely specific to this compound and are inconsistent with its expected pharmacologic mechanism. Acute peripheral leukocyte toxicity, retinal toxicity, and peripheral nerve toxicity (17, 30) were shown in toxicology studies in which mice were dosed i.v. with AG-012986. The C_max levels (>10 μmol/L) in those studies were higher than those achieved in the efficacy studies reported herein. Clearly, achieving a high plasma concentration (C_max) for a short duration (<8 h) is not desired for maximizing the therapeutic window of AG-012986, because it would not yield better efficacy but rather would cause potential safety issues. Therefore, dosing optimization of this type of cell cycle inhibitor is important not only to maximize efficacy by extending the duration of minimally efficacious plasma levels but also to minimize toxicities that are more likely to occur from high plasma levels.

The efficacies of flavopiridol (31, 32) and seliciclib (29), the most advanced CDK inhibitors under clinical development (33), were also reported to be dependent on dose and duration of exposure in solid tumors. Seliciclib displayed limited efficacy in phase I trials (10) possibly because the plasma concentrations could not be maintained at the micromolar concentrations for extended periods that are required for efficacy (>16 h; ref. 33). Flavopiridol has also displayed disappointing efficacy in the clinic, again possibly due to the challenges of maintaining sufficiently high free plasma concentrations of the drug (27, 28). These findings indicate that an ideal therapeutic CDK inhibitor should have sufficient potency and selectivity against the primary targets as well as a pharmacokinetic profile that enables in vivo cytoreductive activity with manageable safety risks. A newer generation of CDK inhibitors, PD0332991 (13) and R547 (34), are reported to have more desirable features, such as high oral bioavailability and longer plasma half-lives, as well as robust antitumor efficacies in preclinical studies. Although the question remains whether CDK inhibitors can be developed as single-agent therapeutics, some clinical investigators have recently focused on combination studies with CDK inhibitors, based on the hypothesis that CDK inhibitors can induce specific perturbations in the cell cycle and signal transduction pathways that lower the threshold for cytotoxic agent-induced lethality against tumor cells without increasing safety risks (8, 9). For example, flavopiridol displayed sequence-dependent cytotoxic synergy with chemotherapy agents in clinical trials (9).

In summary, we have outlined the broad-spectrum antitumor activity of AG-012986 both in vitro and in vivo. The results from pharmacodynamic endpoint assessments in solid tumor lines are consistent with the antiproliferative mechanism expected for a pan-CDK inhibitor. A clear dose- and time-dependent connection between CDK inhibition, Rb hypophosphorylation, cell cycle arrest, suppression of tumor cell proliferation, and tumor cell apoptosis was established both in vitro and in vivo for AG-012986. In addition, dosing schedule optimization with AG-012986 showed that the plasma levels should be maintained at or above the C_eff (165 ng/mL) for >8 h daily to achieve antitumor activity and at least 72 h to maximize the potential for tumor cytotoxicity. Higher C_max plasma levels offer little or no added antitumor efficacy and may lead to dose-limiting toxicities; thus, a dosing regimen of the CDK inhibitor should be managed such that a minimal difference between C_max and C_trough levels is maintained throughout each treatment cycle. The lessons learned from AG-012986 may apply to pan-CDK inhibitors in general and help guide development of future CDK drug candidates.

We thank all of the Pfizer La Jolla CDK Project team members for the intellectual contribution to this work.