Flavopiridol is the first potent inhibitor of cyclin-dependent kinases (cdks) to reach clinical trial. In the majority of solid tumor cell lines and xenografts, flavopiridol induces cell cycle arrest and tumor growth inhibition. This is reflected in clinical outcomes: across multiple Phase II trials there are subsets of patients with prolonged stable disease, although few responses have been observed. Flavopiridol displays sequence-dependent cytotoxic synergy with chemotherapy agents. These effects are most marked when chemotherapy precedes flavopiridol. In the case of DNA-damaging agents that impose S-phase delay, flavopiridol-mediated cdk inhibition disrupts the phosphorylation of E2F-1, leading to inappropriate persistence of its activity, inducing apoptotic pathways. This mechanism has been exploited in a Phase I trial of sequential gemcitabine and flavopiridol that has produced promising results. Flavopiridol is also synergistic with taxanes. Inhibition of cyclin B-cdk1 by flavopiridol accelerates exit from an abnormal mitosis associated with taxane-induced cell death and reduces the phosphorylation of survivin, preventing its stabilization and the cellular protection it affords after taxane exposure. The sequential combination of docetaxel and flavopiridol has been investigated in a Phase I trial in patients with advanced non-small cell lung cancer, and a randomized Phase II study is under way. Initial schedules of flavopiridol used prolonged continuous infusions that produced nanomolar levels of drug thought to be capable of achieving cdk inhibition based on results in tumor cell lines. Recently, it has been discovered that micromolar concentrations are likely to be more effective, and shorter infusions that achieve a higher Cmax have now been adopted. Loading followed by maintenance infusions are also under development, designed to achieve sustained micromolar drug levels. Clinical trials remain complicated by the absence of pharmacodynamic end points to confirm target inhibition.
CYCLIN-DEPENDENT KINASES (cdks)
Cdks comprise a family of enzymes that are the core components of the cell cycle machinery (1) . Cyclin D-dependent kinases 4 and 6, as well as cyclin E-cdk2 complexes, sequentially phosphorylate the retinoblastoma protein (Rb) to facilitate the G1-S transition (2) . Cyclin A-cdk2 and cyclin B-cdk1 complexes are required for orderly S-phase progression and the G2-M transition, respectively (3) . In addition, two groups of inhibitors, known as the Cip/Kip proteins and the INK4 proteins, also regulate cdk activity (4 , 5) . In human cancers, genetic and epigenetic events result in overexpression of cyclins or absent or diminished levels of cdk inhibitors, which provides tumor cells with a selective growth advantage (6 , 7) . Ectopic expression of cdk inhibitors in tumor cell lines restores cell cycle control, usually causing G1 and G2 arrest. In addition, in some cell systems, expression of cdk inhibitors leads to apoptosis (reviewed in ref. 8 ). These observations have prompted the development of pharmacological cdk inhibitors that could potentially produce similar antitumor effects (9, 10, 11) . Of note, the continued proliferation of nontransformed and transformed cells that lack cyclin E (12) or cdk2 (13 , 14) raises the possibility that the specific targeting of cdk2 may not be useful therapeutically.
In addition to their role in cell cycle progression, cyclin-cdk complexes also participate in the regulation of RNA transcription (15) . For example, cyclin H-cdk7, which has cdk-activating kinase activity and is responsible for the phosphorylation events that activate other cdks, also phosphorylates the COOH-terminal domain of RNA polymerase II. Cyclin C-cdk8 and cyclin T-cdk9 (pTEFb) also possess COOH-terminal domain-directed kinase activity, and together these cyclin-cdk holoenzymes control efficient transcriptional elongation.
Flavopiridol is a novel flavone that directly competes with the ATP substrate and inhibits multiple cdks, including cdk1, cdk2, cdk4, cdk6, and cdk7, because it has been shown to directly inhibit the kinase activity of cdk immune precipitates from exponentially growing cells with IC50 values of 100–400 nm (16, 17, 18, 19) . Deschloro-flavopiridol has been cocrystallized with cdk2 (20) . Inhibition of cdk7 prevents the phosphorylation events necessary for activating the other cdks (21) , whereas inhibition of cdk9 (22) globally affects cellular transcription (23) , with the most profound effects on the synthesis of mRNAs with short half-lives, including those encoding growth and apoptosis regulators (24) . Effects on cdk9 may underlie transcriptional repression of cyclin D1 (25) by flavopiridol. The effects of a flavopiridol as a pan-cdk inhibitor are shown in Fig. 1<$REFLINK> .
In a large proportion of solid tumor cell lines, concentrations of 200–300 nm flavopiridol cause arrest at both the G1 and G2 phases of the cell cycle, consistent with inhibition of cdks 2, 4, 6, and 1 (11 , 26) . While p53-independent apoptosis has been observed, cell death usually follows cell cycle arrest and is delayed, maximally occurring at 72 h after the initiation of treatment and requiring concentrations higher than those required to inhibit cdk activity (27 , 28) . Therefore, the primary response of many solid tumor cell lines to flavopiridol-mediated cdk inhibition is cytostatic growth arrest. Consistent with this concept, subsets of patients across multiple Phase II trials, including trials in non-small cell lung cancer [NSCLC (29)] , have achieved stable disease for prolonged periods.
SELECTIVE SENSITIZATION OF TRANSFORMED CELLS TO FLAVOPIRIDOL-INDUCED APOPTOSIS AFTER RECRUITMENT TO S PHASE
Recent work has shown that cells are sensitized to flavopiridol if they are first recruited to S phase before drug exposure (30) . In these experiments, S-phase recruitment was achieved by (a) release after hydroxyurea-induced synchronization at the G1-S boundary or (b) treatment with noncytotoxic concentrations of chemotherapy agents, including gemcitabine and cisplatin, capable of imposing an S-phase delay. Combinations of these chemotherapy drugs, followed by flavopiridol at concentrations that correlate with cdk inhibition, produce sequence-dependent cytotoxic synergy (31 , 32) . In addition, a survey of paired cell lines, including WI38 diploid fibroblasts or normal human bronchial epithelial cells, along with their SV40-transformed counterparts, demonstrated that treatment with flavopiridol during S phase is selectively cytotoxic to transformed cells (30) .
The known consequences of cdk inhibition during S phase suggest a mechanism for these observations, shown in Fig. 2<$REFLINK> . After cdk-mediated phosphorylation of Rb during G1, E2F-1 activity is derepressed, and E2F-1 is released. E2F-1, bound to its heterodimeric partner, DP-1, directs transcription of genes required for S phase. However, this transcription is activated only transiently. Orderly S-phase progression requires the down-regulation of E2F-1 activity, accomplished in part by cdk-mediated phosphorylation (33, 34, 35) . Cdk2 phosphorylates both E2F-1 (most likely at Ser307) and DP-1, abrogating DNA binding of the transcription factor (36) . Cdk1 targets the Ser375 site and promotes reassociation of E2F-1 with Rb during G2 and M, inhibiting E2F-1 activity (37) . Cyclin H-cdk7 phosphorylates E2F-1 at Ser403 and Thr433, facilitating its efficient ubiquitination and degradation (38) . Therefore, inhibition of cdk activity during S phase is expected to result in inappropriately persistent E2F-1 (39) . In normal cells, inappropriate persistence of E2F-1 is tolerable and ultimately drives cell cycle progression. However, in transformed cells, in which the Rb axis is disrupted and E2F-1 activity is already high, reduction in cdk activity during S phase promotes the persistence of enough active E2F-1 to surpass the threshold required to induce apoptosis (40) , which occurs by both p53-dependent and p53-independent pathways. This mechanism may explain the sensitivity of S-phase cells to flavopiridol and may also account for the selective cytotoxicity to transformed cells observed in vitro. Consistent with this model, flavopiridol-induced apoptosis has been shown to be E2F-1 dependent (41 , 42) .
These principles have formed the basis for ongoing studies of sequential gemcitabine followed by flavopiridol. In these trials, gemcitabine is administered at 10 mg/m2/min to maximize incorporation of the active metabolite difluorodeoxycytidine triphosphate (dFdCTP) into DNA with the goal of achieving maximal retardation of S-phase progression. After 24 h, flavopiridol is then administered, with a variety of schedules under investigation. In the first study, in which flavopiridol was administered as a 24-h continuous infusion, three of nine previously treated subjects with NSCLC achieved a partial response (43) .
COMBINATIONS OF FLAVOPIRIDOL WITH TAXANES
Because flavopiridol alone induces G2 arrest and prevents entry into mitosis, its administration before or concomitantly with taxanes is antagonistic (31 , 44) . In contrast, administration of flavopiridol after a taxane-induced mitotic block results in synergistic cytotoxicity. Mechanistically, it has been proposed that cell death after taxane exposure occurs as cells exit an abnormal mitosis. Because reduction of cyclin B-cdk1 activity is required for exit from mitosis, its inhibition by flavopiridol after taxane facilitates mitotic exit and hastens cell death (44) . It has also been shown that microtubule stabilization induces a survival pathway that depends on elevated activity of cyclin B-cdk1, which directs the phosphorylation and stabilization of survivin, an apoptosis inhibitor and mitotic regulator. Inhibition of cyclin B-cdk1 by flavopiridol prevents the phosphorylation and accumulation of survivin, effectively removing a viability checkpoint and enhancing apoptosis (45 , 46) . The interactions of flavopiridol and taxanes are shown in Fig. 3<$REFLINK> .
Several Phase I trials have been designed to exploit the synergism between taxanes and flavopiridol, and a wide variety of schedules have been explored (47, 48, 49) . These studies have demonstrated that combinations are tolerable and have led to the design of randomized trials (Table 1)<$REFLINK> . Whereas the preferred sequence is clearly taxane followed by flavopiridol, the optimal interval between drugs is less clear (50 , 51) . Currently, in patients with advanced NSCLC previously treated with platinum-based therapy, docetaxel alone (administered at 75 mg/m2 every 3 weeks) is randomized to one of two schedules of docetaxel and flavopiridol (administered at 60 mg/m2 over 1 h). In the first schedule, the interval between drugs is 3–4 h, and in the second schedule, the interval is increased to 16–24 h. In each combination arm, flavopiridol is administered weekly if not precluded by neutropenia, with the goal of achieving improved disease control from a potential disease-stabilizing effect of the cdk inhibitor alone. Unfortunately, a paucity of responses have been observed thus far among patients with advanced lung cancer who have received docetaxel/flavopiridol combinations, although disease stabilization has occurred (49) .
The potency of flavopiridol in preclinical in vitro and in vivo models has not been reflected in the clinical setting to date. Monotherapy trials have not been designed to distinguish prolonged disease stabilization from indolent tumor growth, and enhanced cytotoxicity after chemotherapy/flavopiridol combinations has yet to be confirmed in the randomized setting.
Initial trials of flavopiridol used long continuous infusions (24–72 h) to reflect the effect of repeated low-concentration drug treatment that showed antitumor activity in animal models (52 , 53) . These schedules produce nanomolar plasma levels of drug; primary toxicities of flavopiridol in this setting include secretory diarrhea, asthenia, and proinflammatory events. In some trials, a higher-than-expected incidence of clotting events has also been observed (29 , 54) . Subsequent data suggested the superiority of bolus administration designed to achieve high micromolar concentrations, which led to cures in lymphoma xenografts (55 , 56) . This prompted the development of 1-h infusions, which have achieved micromolar Cmax plasma concentrations, albeit with short half-life (57) . Of note, the dose-limiting toxicity of 1-h infusions is neutropenia (58) . Importantly, however, significant inhibition of colony formation in A549 NSCLC cells requires sustained exposure to high micromolar concentrations of drug (27) , suggesting that schedules developed to date may be inadequate to achieve tumor growth inhibition. Currently, 30-min to 1-h loading infusions, followed by maintenance infusions of variable duration designed to achieve sustained micromolar concentrations, are being explored both alone and in chemotherapy combinations.
PHARMACODYNAMIC END POINTS
The absence of pharmacodynamic end points has complicated flavopiridol trials. Current efforts to document cdk inhibition in patients’ tumors involve immunohistochemical staining of pre- and posttreatment samples with phospho-specific antibodies to demonstrate reduced phosphorylation of cdk targets, including Rb (59) , p27Kip1, E2F-1, and survivin. The development of fluorothymidine positron emission tomography scanning may provide a noninvasive method of measuring cell cycle inhibition by flavopiridol and other cdk inhibitors (60) . Such pharmacodynamic measures will ultimately contribute to the interpretation of stable disease and may help justify the commitment to randomized trials involving large patient populations.
Dr. Ramaswamy Govindan: What happens to peripheral blood mononuclear cells (PBMCs) in humans when you give this?
Dr. Geoffrey I. Shapiro: PBMCs are not cycling cells, so not much happens to them. At baseline, we can’t find phosphorylated Rb or phosphorylated cdk targets in noncycling PBMCs. Also, by the time you have isolated them, the drug is washed away, so it is not possible to discern changes in PBMCs isolated pre- and posttreatment. In our assay, we stimulate them in vitro and then add back the plasma with or without flavopiridol to see if there is an effect. However, all that this indicates is that there is biologically active flavopiridol in posttreatment plasma. It provides no information on what is happening in the tumor. Although it appears that posttreatment plasma can inhibit the proliferation of stimulated PBMCs, the inhibition is incomplete. There is, at most, a 50–60% reduction in incorporation of labeled thymidine, which sounds good, but that probably does not represent potent enough inhibition of cdk activity or cellular proliferation to translate to a clinically meaningful outcome.
Dr. Alan Sandler: You have some nice science and the compound looks like it does what it’s supposed to do, but therein lies the potential trap. We utilize the science to study agents that may be totally inactive. Because of the fact that we can do some science, tweak a little thing here or there, we keep these research programs alive. Farnesyl transferase inhibitors, for example—how many more negative studies do we need with farnesyl transferase inhibitors? But the fact is that you can do some incredible studies.
Dr. Shapiro: With flavopiridol, we’re coming to the recognition that the better schedules should be used where there has been some hint of success. I mentioned that in solid tumor lines, you may not only need to get the dose up to achieve micromolar levels of drug, but that these high plasma levels must also be maintained. None of our current schedules achieve this. However, in hematopoietic cell lines, which are more apoptotically bent, a short exposure to flavopiridol may be adequate to induce cell death. So, for example, one area where there has been some hint of a response is mantle cell lymphoma, which is very intriguing because it is a cyclin D1-dependent tumor. So it is in this setting where new schedules need to be tried. We are trying to refine the use of flavopiridol, rather than expand it. Flavopiridol is the primary cdk inhibitor out there right now that is easy to do work with, so our studies are at least defining the assays that will be important.
Dr. Bruce Johnson: Dr. Shapiro’s point is that there are compounds in this class coming forward through the pipeline that are between 10 and 100 times more potent inhibitors of cdks than flavopiridol, so the principles can be worked out here and validated in the newer molecules. The other important thing about lung cancer is that you have both small cell and non-small cell, which have different cell cycle regulation. The hope is that you will find one compound that will work with a p16 inactivation in non-small cell lung cancer and one that will work with a Rb inactivation in small cell lung cancer.
Dr. Sandler: Right, the science I understand, and the concept wasn’t so much that the class was bad. I agree with you that if you have something that is much more potent and can actually do what you thought this would do, that would make this research of value.
Dr. Shapiro: Obviously, it has been very hard for people to let go of the concept of the cell cycle as a target in human cancer. After all, a Nobel Prize was awarded just a few years ago for the discovery of cdks. The small cell question is very interesting because small cell lung cancer is nearly universally Rb negative, so you really don’t have a very potent arrest in G1 phase at all. Everything I mentioned about cell death during S phase is relevant to small cell.
Presented at the First International Conference on Novel Agents in the Treatment of Lung Cancer, October 17–18, 2003, Cambridge, Massachusetts.
Grant support: G. Shapiro is supported by NIH Grants R01 CA90687 and P20 CA90578.
Requests for reprints: Geoffrey I. Shapiro, Dana 810A, Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 02115. Phone: (617) 632-4942; Fax: (617) 632-1977; E-mail: