Nature Medicine10, 643 - 648 (2004)
Published online: 2 May 2004; Corrected online: 21 2004 | doi:10.1038/nm1047
Bioluminescent imaging of Cdk2 inhibition in vivo
Guo-Jun Zhang1, 6, Michal Safran1, Wenyi Wei1, Erik Sorensen2, Peter Lassota2, Nikolai Zhelev3, 4, Donna S Neuberg5, Geoffrey Shapiro1
& William G Kaelin Jr1, 6
1
Department of Adult Oncology, Dana-Farber Cancer Institute and Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts
02115, USA.
2
Novartis Institutes for Biomedical Research, Cambridge, Massachusetts
02139, USA.
3
Cyclacel Ltd James Lindsay Pl, Dundee, Scotland
DD1 5JJ.
4
CMCBR, SIMBIOS, School of Contemporary Sciences, University of Abertay, Bell Str., Dundee, Scotland
DD1 1HG.
5
Department of Biostatistical Science, Dana-Farber Cancer Institute, Boston, Massachusetts
02115, USA.
6
Howard Hughes Medical Institute, Chevy Chase, Maryland
20815, USA.
Many proteins and pathways of pharmaceutical interest impinge on ubiquitin ligases or their substrates. The cyclin-dependent kinase (Cdk) inhibitor p27, for example, is polyubiquitylated in a cell cycle−dependent manner by a ubiquitin ligase complex containing the F-box protein Skp2. Regulated turnover of p27 is due, at least partly, to its phosphorylation by Cdk2 on threonine 187, which generates a Skp2-binding site. We made a p27-luciferase (p27Luc) fusion protein and show here that its abundance, like that of p27, is regulated by Skp2 in a cell cycle−dependent manner. As predicted, p27Luc levels increased after blocking Cdk2 activity with inhibitory proteins, peptides or small interfering RNA (siRNA). Accumulation of p27Luc in response to Cdk2 inhibitory drugs (flavopiridol and R-roscovitine) was demonstrable in human tumor cells in vivo using noninvasive bioluminescent imaging. In theory, the approach described here could be used to develop bioluminescent reporters for any drug target that directly or indirectly affects the turnover of a ubiquitin ligase substrate.
Establishing that a potential drug modulates its intended target in vivo is a crucial step in drug development. Quantitative assays that allow for repetitive, noninvasive assessment of molecular targets would facilitate this process and accelerate efforts to optimize lead compounds with respect to issues such as optimal structure, formulation, dose and schedule. A number of modalities are available for molecular imaging, including positron emission tomography, magnetic resonance imaging and optical imaging1. Optical imaging is based on quantitative or qualitative changes in light emission by fluorescent or bioluminescent probes, which may be protein based (for example, green fluorescent protein and firefly luciferase, respectively). Fluorescent probes require the delivery of an excitatory photon that produces a background signal resulting from autofluorescence. In contrast, bioluminescent imaging does not require an excitation light and is typically associated with low background1.
At least two non-mutually exclusive strategies can be used to examine a molecular target or pathway with a bioluminescent or fluorescent reporter protein (hereafter called a reporter, for simplicity). One would be to place the reporter under the control of a transcriptional promoter that is responsive to the molecule or pathway of interest. The status of p53, for example, can be monitored by placing luciferase under the control of a p53-responsive promoter2, and tissue hypoxia can be monitored by placing green fluorescent protein under the control of the hypoxia-inducible vascular endothelial growth factor promoter3. Another strategy would be to discover or reengineer reporter proteins such that their activity is quantitatively or qualitatively altered by the molecule or pathway under study. For example, aequorin is intrinsically calcium dependent and has been used to monitor calcium availability in vivo4. Genetically engineered reporters have been made that respond to phosphorylation by particular kinases or cleavage by specific proteases5,
6,
7,
8,
9,
10. These efforts have been enhanced by a growing understanding of the peptidic determinants recognized by enzymes within their protein substrates. In many cases, these determinants are modular and will function in cis when fused to a heterologous reporter.
Many molecular pathways related to cell proliferation and cancer directly or indirectly impinge on the activity of E3 ubiquitin ligases, which target proteins for ubiquitin-dependent proteolysis. For example, the SCF (Skp1/Cullin/F-box) and APC (anaphase-promoting complex) ubiquitin ligase complexes are responsible for coordinated waves of protein destruction that drive unidirectional cell cycle progression11,
12. A number of oncoproteins and tumor suppressor proteins, including MDM2, Skp2, pVHL, APC and Cdc4, are components of ubiquitin ligase complexes. Ubiquitin ligase recognition motifs are often transportable and can be used to target foreign proteins for destruction in cis13. Many ubiquitin ligase recognition motifs are regulated by post-translational modifications, such as phosphorylation or hydroxylation, and hence might be exploited to investigate such modifications.
The Cdk inhibitor p27 is polyubiquitinated in a cell cycle−dependent manner by an SCF complex containing Skp1, Cul1 and Skp2 (ref. 14). Recognition of p27 by SCF Skp2 is enhanced after phosphorylation of Thr187 on p27 by Cdk2 in late G1 and S phases14. A number of pharmaceutical companies are currently developing Cdk2 inhibitors as potential anticancer agents15,
16,
17. We reasoned that a p27Luc fusion protein could be used to monitor Cdk2 activity in vitro and in vivo.
Results p27Luc fusion protein is cell cycle regulated We made an expression plasmid encoding full-length p27 fused, in frame, to the N terminus of firefly luciferase (p27Luc). As expected, p27Luc retained cyclin-binding activity, inhibited Cdk2-associated kinase activity and induced an acute G1/S block when overexpressed in susceptible cells (Supplementary Fig. 1 online). We confirmed that the abundance of p27Luc in mammalian cells, like that of p27 itself, was regulated by Skp2 and the cell cycle. In transient transfection assays, production of Skp2 and p27Luc diminished p27Luc and endogenous p27 (Fig. 1a). Conversely, knockdown of endogenous Skp2 with siRNA increased exogenous p27Luc and endogenous p27 (Fig. 1b). Induction of p27Luc was demonstrable in immunoblot assays using antibody to luciferase (Fig. 1b), as well as in luciferase-based light assays (Fig. 1c). In cell synchronization experiments, exogenous p27Luc and endogenous p27 were low in cells arrested in late G1 by mimosine, and increased steadily as cells exited S phase after mimosine release (Fig. 1d−f). As expected, the accumulation of p27 and p27Luc coincided with a fall in cyclin E protein and cyclin E−associated kinase activity (Fig. 1f and Supplementary Fig. 2 online). Notably, the amount of p27Luc produced by stable, polyclonal transfectants did not exceed that of endogenous p27 (Fig. 1f and data not shown), presumably because individual clones producing supraphysiolological levels of p27 were at a growth disadvantage and were therefore underrepresented.
Figure 1. p27Luc mimics p27 with respect to regulation by Skp2 and the cell cycle.
(a) Immunoblot (IB) analysis of HeLa cells transiently transfected with luciferase (Luc) or p27Luc fusion protein, in the presence or absence of Myc-tagged Skp2. (b,c) Immunoblot analysis (b) or luciferase activity (c) in HeLa cells stably transfected to produce luciferase or p27Luc, then mock-transfected or transfected with Skp2 or scrambled (control) siRNA. (d−f) U2OS cells stably transfected to produce luciferase or p27Luc were arrested in late G1 by growth in mimosine. At various time points after removal of mimosine, cells were analyzed for DNA content by FACS after propidium iodide staining (d) or lysed. Cell extracts were assayed for luciferase activity (e) or analyzed by immunoblotting (f). * indicates nonspecific binding. Error bars (c,e) indicate 1 s.d.
p27Luc accumulates in response to Cdk2 inactivation Control of p27 by Skp2 during the cell cycle is partly the result of Cdk2-dependent phosphorylation, which targets it for destruction. The Cdk inhibitor p21, but not a loss-of-function p21 mutant18, induced the accumulation of p27Luc in transient transfection assays (Fig. 2a), as did a dominant-negative version of Cdk2 (ref. 19) (Fig. 2b). Neither p21 nor dominant-negative Cdk2 affected wild-type luciferase. We next treated cell lines stably producing p27Luc and cell lines producing wild-type luciferase with membrane-permeable (penetratin fusion) peptides that inhibit cyclins A and E by binding to their respective substrate docking sites20,
21,
22. Luciferase activity was only increased in the cells producing p27Luc (Fig. 2c,d). siRNA targeting Cdk2, cyclin A or cyclin E also induced the accumulation of p27Luc, which mirrored the behavior of endogenous p27 (Fig. 2e,f and Supplementary Fig. 3 online). Cdk2 siRNA had no effect on a p27Luc reporter in which Thr187 of p27, which serves as the Cdk2 phosphoacceptor site, was converted to alanine (Supplementary Fig. 4 online). p27Luc activity is therefore responsive to changes in Cdk2 activity in mammalian cells. In time-course experiments, cells treated with cyclin A siRNA or with the cyclin A/E inhibitory peptides described above ultimately died, in keeping with earlier studies showing that cyclin A inhibitory peptides induce apoptosis in transformed cells20,
23 (data not shown).
(a,b) HeLa cells were transiently transfected with plasmids encoding luciferase (Luc) or p27Luc, in the presence or absence of plasmids encoding the Cdk inhibitor p21 (a) or dominant-negative Cdk2 (Cdk2-DN; b). p21N+C is a p21 mutant that cannot block Cdk activity. Each transfection mix also contained a Renilla luciferase reporter plasmid as an internal control. Cell extracts were prepared, and normalized firefly luciferase activity was determined. (c−f) HeLa cells stably transfected to produce luciferase or p27Luc were incubated with wild-type or mutant (control) cyclin inhibitory peptides fused to penetratin (c,d) or the indicated siRNAs (e,f). Cell extracts were immunoblotted with the indicated antibodies (c,e) or assayed for luciferase activity (d,f). * indicates nonspecific binding. Error bars indicate 1 s.d.
Imaging Cdk2 inhibitor pharmacodynamics We next asked whether p27Luc activity could be used to monitor the activity of Cdk2 inhibitors with drug-like properties in cellulo and in vivo. Flavopiridol and R-roscovitine, although not monospecific, are Cdk2 inhibitors currently undergoing clinical testing. Both compounds caused a dose-dependent decrease in Cdk2 kinase activity in vitro (Supplementary Fig. 5 online), as expected. They also caused a dose-dependent increase in luciferase activity in various tumor cells (U2OS osteosarcoma, HeLa cervical carcinoma and H1299 lung carcinoma cells) stably producing p27Luc, but not in cells producing wild-type luciferase (Fig. 3 and data not shown). Induction of p27Luc by these two agents was not caused by decreased Skp2 levels (Supplementary Fig. 5 online). In contrast to flavopiridol and R-roscovitine, no induction of p27Luc was observed with cytotoxic agents such as cis-platinum and camptothecin, both of which induce a G2/M block and apoptosis (data not shown). An approximately twofold induction of p27Luc by flavopiridol and roscovitine was demonstrable under standard monolayer cell culture conditions (Fig. 3a) and in cells grown in transparent hollow fibers (Fig. 3b). In the latter experiments, luciferin was added directly to the cell culture medium and luciferase activity was monitored using a CCD (charged-coupled device) camera. The observed induction of p27Luc is unlikely to simply reflect a cell cycle block, as induction of p27Luc was apparent at doses of flavopiridol and roscovitine (50 nM and 10 M, respectively) that did not alter cell cycle distribution assessed by propidium iodide staining followed by FACS (data not shown). At the highest doses tested, flavopiridol (200 nM) caused an increase in S- and G2/M-phase U2OS cells, but had no demonstrable effect on the cell cycle of HeLa or H1299 cells (data not shown). Roscovitine (25 M), however, led to a partial G2/M block in HeLa cells and an increase in S- and G2/M-phase cells in U2OS cells, and had no demonstrable cell cycle effect in H1299 cells (data not shown).
Figure 3. Induction of p27Luc by Cdk2 inhibitory drugs in vitro.
(a,d) Luciferase activity of polyclonal HeLa and U2OS cells producing luciferase (Luc) or p27Luc, after treatment with flavopiridol (50 or 200 nM), roscovitine (10 or 25 M), compound A (0.2 or 1 M), compound B (50 or 200 nM) or compound C (0.2 or 1 M) for 24 h (indicated by triangles). Luciferase and p27Luc activity was normalized to a value of 1 for untreated cells. (b) Bioluminescent imaging of hollow fibers filled with polyclonal U2OS cells producing luciferase or p27Luc, and treated with 200 nM flavopiridol or 25 M roscovitine for 24 h. Fold induction refers to induction of p27Luc with treatment, after normalization to luciferase signal. (c) Cdk2 immunoprecipitates derived from asynchronously growing HeLa cells were used to phosphorylate GST-RB in vitro. Compound A (0.2 or 1 M), compound B (50 or 200 nM) or compound C (0.2 or 1 M) was added before adding GST-RB, where indicated by triangles. Immunoblotting (IB) was carried out with the indicated antibodies. Radiolabeled GST-RB was detected by autoradiography. Error bars indicate 1 s.d.
A number of new, hopefully more specific, Cdk2 inhibitors are currently being developed as possible anticancer drugs. We obtained three such compounds, labeled 'A', 'B' and 'C', from three different pharmaceutical companies. As expected, all three compounds inhibited Cdk2 activity in vitro using immunopurified (Fig. 3c) or recombinant (data not shown) Cdk2. Compound A, however, did not induce p27Luc under any conditions tested (Fig. 3d and data not shown). At concentrations above 2.5 M, compound A caused overt toxicity and parallel reductions in both Luc and p27Luc activity. Compound B induced p27Luc in U2OS cells, but not in HeLa cells. Given our reporter validation assays (Figs. 1 and 2), we concluded that compounds A and B do not act as direct Cdk2 antagonists in intact cells, and that compound B can indirectly affect p27 turnover in a cell line−specific manner. Among many possibilities, the selective induction of p27Luc by compound B in U2OS cells might be caused by the fact that they contain wild-type retinoblastoma protein. HeLa cells are functionally null for retinoblastoma because of the presence of the human papillomavirus E7 protein.
Next, hollow fibers containing cells expressing p27Luc or wild-type luciferase were implanted subcutaneously into the opposite flanks of nude mice. Seven days later, the animals were treated with flavopiridol or roscovitine and subsequently imaged after intraperitoneal administration of luciferin. Both flavopiridol and roscovitine, as expected, induced p27Luc activity in vivo (Fig. 4a−c and Supplementary Fig. 6 online). Although modest, the induction of p27Luc after normalization was highly reproducible and statistically significant (in Fig. 4b, for example, P = 0.004 by Student t test and P = 0.016 by Wilcoxon rank test, with only seven mice treated). Likewise, flavopiridol specifically induced p27Luc activity in tumor xenograft assays (Fig. 4d). The induction of p27Luc by the Cdk2 inhibitors was, however, slightly less marked in vivo, possibly because the percentage of untreated tumor cells in G0 or early G1 is higher when cells are grown in vivo rather than in serum-rich medium in vitro.
Figure 4. Monitoring Cdk2 inhibition in vivo using bioluminescent imaging.
(a) Hollow fibers filled with polyclonal U2OS cells producing luciferase (Luc; left flank) or p27Luc (right flank) were implanted subcutaneously into nude mice. Left, baseline bioluminescent images taken 7 d later. Right, repeat images obtained after two doses of flavopiridol (5 mg/kg, once a day by intraperitoneal injection). (b) Normalized fold induction (p27Luc/Lucpost-treatment ÷ p27Luc/Lucpretreatment) of p27Luc in seven mice after treatment with flavopiridol. (c) Normalized fold induction of p27Luc in mice treated with the indicated doses of flavopiridol (eight mice per treatment group) and imaged as in a. (d) Polyclonal H1299 cells producing luciferase (left flank) or p27Luc (right flank) were injected subcutaneously into nude mice. Left, bioluminescent images were obtained 6 weeks later, when tumors of comparable size (5 mm) had formed bilaterally. Right, repeat images were obtained after two doses of flavopiridol (5 mg/kg, once a day by intraperitoneal injection). Fold induction was calculated as p27Luc/Lucpost-treatment ÷ p27Luc/Lucpretreatment. Error bars indicate standard error (s.e.m.).
Discussion We found that a reporter consisting of p27 fused to luciferase was responsive to changes in Cdk2 activity in cellulo and in vivo. Responsiveness to Cdk2 activity was validated using a variety of approaches, including Cdk2 inhibitory proteins, Cdk2 inhibitory peptides and siRNA targeting Cdk2 or its partners, cyclin A and cyclin E. We found that the p27Luc reporter can provide a pharmacodynamic readout of Cdk2 inhibitor action in animal models.
Acute inhibition of Cdk2 activity by various means induced the accumulation of the p27Luc reporter, as well as endogenous p27. Surprisingly, however, p27 levels are not increased in Cdk2-deficient mouse embryonic fibroblasts24. This apparent paradox might relate to compensatory changes that can occur after germline, rather than somatic, gene disruption as a result of developmental 'plasticity'25.
We had difficulty generating stable cell lines producing high levels of a p27Luc chimera lacking the Thr187 Cdk2 phosphoacceptor site (p27LucT187A), presumably because phosphorylation of Thr187 by Cdk2 is required for Skp2-dependent destruction of p27 in S phase. The low expression of p27LucT187A in the clones that did emerge was not induced by Cdk2 siRNA, but was induced after inhibition of Skp2 with siRNA (Supplementary Fig. 2 online), in keeping with an earlier conclusion that regulation of p27 by Skp2 involves Cdk2-dependent and Cdk2-independent pathways26. In addition, the abundance of p27T187A in mouse embryonic fibroblasts that exclusively produce this p27 variant as a result of homologous recombination (knockin) remains cell cycle dependent, with low expression in late G1 and peak expression in G0 and S phases26. For these reasons, our reporter can be viewed specifically as an indicator of Cdk2 activity, and more generally as an indicator of cell cycle progression.
A recurrent challenge in drug discovery is to determine whether a compound capable of modulating a target in a test tube can do so in an intact cell. The p27Luc reporter should be most valuable in determining whether compounds that inhibit Cdk2 in biochemical assays are actually capable of inhibiting Cdk2 in vivo. In this setting, failure to induce p27Luc (such as we observed with compound A) would provide potentially valuable information. A particularly useful feature of bioluminescent imaging is the ability to take repeated measurements in the same animal over time, which might help to identify a temporal window of drug action. In some cases, Cdk2 inhibition might represent an 'off-target' effect, but might still be exploited to assess a compound's bioavailability. On the other hand, induction of p27Luc cannot be taken as proof that a drug directly or indirectly inhibits Cdk2, as other cytostatic agents acting at G0 or S, as well as drugs that interfere with protein turnover, would presumably also induce p27Luc. This problem could, in theory, be mitigated by analyzing additional cell cycle− or SCF-dependent luciferase fusion proteins in parallel.
A number of Cdk2 inhibitors are currently being developed as anticancer agents, although it is not known whether the antitumor effects of these compounds are specifically a result of Cdk2 inhibition15,
16,
17. A recent study concluded that Cdk2 is not likely to be a useful anticancer drug target because elimination of Cdk2 by siRNA or antisense RNA did not affect cell growth27. A caveat, however, is that the absence of Cdk2 might not have the same biological consequences as the presence of a drug-bound, catalytically inactive form of Cdk2 that sequesters cyclins A and E. Peptides that inhibit these cyclins induce apoptosis selectively in transformed cells, compared with normal cells20,
23. Likewise, siRNA targeting cyclin A, but not Cdk2, can induce apoptosis in cancer cells (W.W. and W.G.K., unpublished data). It remains to be determined whether Cdk2 is a viable drug target in cancer.
Many biological pathways, including pathways relevant to cancer, impinge on the action of ubiquitin ligases. In theory, the approach described here should be generalizable to other ubiquitylation targets, especially in situations where a discrete, modular ubiquitin ligase substrate recognition motif has been identified. An earlier study showed that a chimera consisting of a fragment of cyclin B fused to luciferase could be used to monitor APC activity in cellulo13. Likewise, we have used a fragment of HIF (hypoxia-inducible factor) fused to luciferase to monitor the status of the von Hippel−Lindau tumor suppressor protein in cellulo and in vivo (M.S. and W.G.K., unpublished data). In many cases, it should be possible to configure such reporters so that inhibition of the target of interest leads to increased, rather than decreased, signal. In this way, 'false positives' caused by nonspecific effects (such as global inhibition of transcription or translation) might be minimized. Bioluminescent reporters for targets and pathways of pharmaceutical interest should expedite preclinical drug discovery. In time, arrays of such reporters probing orthogonal aspects of biological space might be used to monitor the actions of drug-like molecules on biological systems.
Methods Plasmids. To make pGL3-p27Luc, we PCR-amplified human p27 cDNA with the primers 5'-GCGCAAGCTTGCCACCATGTCAAACGTGCGAGTGTCTAAC-3' and 5'-GCGCCCATGGTCGTTTGACGTCTTCTGAGGCCAGG-3'. The resulting PCR product was cut with HindIII and NcoI and ligated into the pGL3-control plasmid (Promega), in which firefly luciferase cDNA is under the control of the SV40 early promoter and is cut with these two enzymes. pGL3- p27LucT187A was made by site-directed mutagenesis using a QuickChange XL kit (Stratagene) according to the manufacturer's protocol, as well as the primers 5'-GGATCCGTGGAGCAGGCGCCCAAGAAGCCTGGCCTC-3' and 5'-GAGGCCAGGCTTCTTGGGCGCCTGCTCCACGGATCC-3'. pRcCMV-HAp21 and pRcCMV-HAp21N+C were described previously18. pCMV-CDK2-DN19 and pcDNA3-myc-SKP2 were gifts from E. Harlow (Harvard Medical School) and J. DeCaprio (Dana-Farber Cancer Institute and Harvard Medical School), respectively.
Antibodies and chemicals. Rabbit polyclonal antibody to luciferase was purchased from Sigma, and mouse monoclonal antibody to p27 (clone 57) was purchased from Transduction Laboratories. Rabbit polyclonal antibodies to Cdk2 (M2), cyclin A (HE12), hemagglutinin epitope (Y11) and Skp2 (H435), and mouse monoclonal antibodies to cyclin E (HE12) and c-Myc (9E10) were purchased from Santa Cruz Biotechnology.
Cdk2 inhibitory peptides were synthesized by Cyclacel. The cyclin inhibitory peptide contained the sequence AAURSLI(pFF), where U = -aminobutyric acid and pFF = p-fluorophenylalanine. The control peptide contained the sequence HAKARAIA. The C termini of both peptides were tagged with the penetratin sequence RQIKWFQNRRMKWKK and blocked with an amide group28. Flavopiridol and R-roscovitine were provided by the National Cancer Institute (Drug Synthesis & Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis) and by Cyclacel, respectively. D-luciferin was purchased from Sigma.
siRNA. siRNAs were purchased from Dharmacon. siRNA sequences were unique to their intended targets, based on BLAST searches. The siRNA sequences (sense strand) were 5'-AAGGUGGUGGCGCUUAAGAAA-3' (Cdk2), 5'-AAGGCAGCGCCCGUCCAACAA-3' (cyclin A), 5'-AAUAAUGCAGUCUGUGCAGAC-3' (cyclin E), 5'-AACAUCCCCCAGGAACUGCUC-3' (Skp2) and 5'-AACAACCUGCCGCGACGGAA-3' (control).
Tissue culture and transfection. Cells were maintained in DMEM supplemented with 10% FBS. Plasmid transfections were performed with LipofectaminePLUS (Invitrogen). To make stable cell lines, cells were cotransfected with 5 g of pGL3-Luc or pGL3-p27Luc and 0.5 g of empty pcDNA3 (Invitrogen). Forty-eight hours later, cells were placed in medium containing G418 (1 mg/ml). siRNA transfection was carried out using Oligofectamine (Invitrogen) according to the manufacturer's instructions. Cells were analyzed 48−72 h later.
Cell cycle analysis. Cells were blocked in late G1 by incubation in medium containing 0.2 mM mimosine for 18 h, then released into mimosine-free medium29. At various time points thereafter, cells were lysed or fixed with ice-cold 70% ethanol. Fixed cells were incubated in PBS containing 69 M propidium iodide and 20 g/ml RNAse A for 30 min at 37 °C, and analyzed using a FACScan flow cytometer (Becton Dickinson).
Cell extract luciferase assay. A dual luciferase reporter system (Promega) was used according to the manufacturer's instructions. Luciferase values for stable cell lines were normalized to total protein concentration.
Immunoprecipitation (IP) kinase assays. IP kinase assays were performed as described previously18. Briefly, cells were lysed in EBC buffer supplemented with protease inhibitors (Complete Mini, Roche) and phosphatase inhibitors (phosphatase inhibitor cocktail set I and II, Calbiochem). Cell extract (100 g) was incubated with antibodies to cyclin E (HE111) or Cdk2 (M2) for 1 h. Immune complexes were captured with protein A−Sepharose, washed five times with NETN buffer (20 mM Tris (pH 8.0), 100 mM NaCl, 1 mM EDTA and 0.5% NP-40) and twice with protein kinase inhibitor buffer (50 mM Tris (pH 7.5), 10 mM MgCl2 and 1 mM dithiothreitol). Kinase reactions were performed in 30 l protein kinase inhibitor buffer containing 0.1 g of GST-RB and 1 l [32P]-ATP (10 Ci) for 15 min at 30 °C.
Hollow fiber assay. Growth of cells in hollow fibers was done essentially as described30. Briefly, a hollow fiber (1.2-mm outer diameter, 1-mm inner diameter; Spectrum Medical) was filled with cells (106 cells/ml) and cut into 1.5-cm pieces that were sealed at both ends. For in vitro studies, hollow fibers were placed in six-well culture dishes containing DMEM with 10% FBS for 3 d, before adding Cdk2 inhibitors. For in vivo studies, hollow fibers were implanted subcutaneously in NCR-Nu mice using a 13-gauge trochar inserted through a neck incision under anesthesia (140 mg/kg ketamine and 12 mg/kg xylazine, given by intraperitoneal injection). All animal experiments described in this paper were approved by the Animal Care and Use Committee of the Dana-Farber Cancer Institute.
Tumor xenografts. Approximately 5 106 cells in 0.2 ml PBS were injected subcutaneously into each site in the flanks of NCR-Nu nude mice under anesthesia (isoflurane (Isoflo), Abbott Laboratories).
Bioluminescent imaging. For in vitro studies, D-luciferin was added to tissue culture medium, to a final concentration of 50 g/ml. Five minutes later, photons were counted using the IVIS imaging system (Xenogen) according to the manufacturer's instructions. Data were analyzed using LivingImage software (version 2.12, Xenogen). For in vivo studies, mice were administered D-luciferin (50 mg/kg), ketamine (140 mg/kg) and xylazene (12 mg/kg) by intraperitoneal injection. Ten minutes later, photons were counted and analyzed as above.
Statistical analysis. For imaging data analysis, we calculated the ratio of the intensity of p27Luc luminescence after treatment compared with that of p27Luc before treatment. To control for mouse-to-mouse variability, the p27Luc ratio for each mouse was normalized by dividing by the before/after-treatment ratio of luciferase intensity for that mouse. Statistical significance was assessed using the Student t test, under the assumption of a normal distribution of the normalized ratios with an estimate of variance. The Wilcoxon signed rank test was also performed to provide an additional distribution-free assessment. All statistical tests were two sided.
Note: Supplementary information is available on the Nature Medicine website. NOTE: In the supplementary information originally attached to this article online, the labels at the bottom of Supplementary Fig. 1c were omitted by the authors. This error has been corrected in the supplementary information now online.
Received 8 September 2003; Accepted 23 March 2004; Published online: 2 May 2004.
REFERENCES
Weissleder, R. & Ntziachristos, V. Shedding light onto live molecular targets. Nat. Med.9, 123–128 (2003). | Article | PubMed | ChemPort |
Wang, W. & El-Deiry, W.S. Bioluminescent molecular imaging of endogenous and exogenous p53-mediated transcription in vitro and in vivo using an HCT116 human colon carcinoma xenograft model. Cancer Biol. Ther.2, 196–202 (2003). | Article | PubMed | ISI | ChemPort |
Fukumura, D. et al. Tumor induction of VEGF promoter activity in stromal cells. Cell94, 715–725 (1998). | Article | PubMed | ISI | ChemPort |
Kendall, J.M., Sala-Newby, G., Ghalaut, V., Dormer, R.L. & Campbell, A.K. Engineering aequorin to measure Ca2+ in defined compartments of living cells. Biochem. Soc. Trans.20, 144S (1992). | PubMed | ChemPort |
Laxman, B. et al. Noninvasive real-time imaging of apoptosis. Proc. Natl. Acad. Sci. USA99, 16551–16555 (2002). | Article | PubMed | ChemPort |
Violin, J.D., Zhang, J., Tsien, R.Y. & Newton, A.C. A genetically encoded fluorescent reporter reveals oscillatory phosphorylation by protein kinase C. J. Cell Biol.161, 899–909 (2003). | Article | PubMed | ISI | ChemPort |
Sala-Newby, G.B. & Campbell, A.K. Engineering a bioluminescent indicator for cyclic AMP-dependent protein kinase. Biochem. J.279 (part 3), 727–732 (1991). | PubMed | ChemPort |
Ting, A.Y., Kain, K.H., Klemke, R.L. & Tsien, R.Y. Genetically encoded fluorescent reporters of protein tyrosine kinase activities in living cells. Proc. Natl. Acad. Sci. USA98, 15003–15008 (2001). | Article | PubMed | ChemPort |
Zhang, J., Ma, Y., Taylor, S.S. & Tsien, R.Y. Genetically encoded reporters of protein kinase A activity reveal impact of substrate tethering. Proc. Natl. Acad. Sci. USA98, 14997–5002 (2001). | Article | PubMed | ChemPort |
Jenkins, T.M., Sala-Newby, G. & Campbell, A.K. Measurement of protein phosphorylation by covalent modification of firefly luciferase. Biochem. Soc. Trans.18, 463–464 (1990). | PubMed | ISI | ChemPort |
Harper, J.W., Burton, J.L. & Solomon, M.J. The anaphase-promoting complex: it's not just for mitosis any more. Genes Dev.16, 2179–2206 (2002). | Article | PubMed | ISI | ChemPort |
Lukas, C. et al. Accumulation of cyclin B1 requires E2F and cyclin-A-dependent rearrangement of the anaphase-promoting complex. Nature401, 815–818 (1999). | Article | PubMed | ISI | ChemPort |
Bloom, J. & Pagano, M. Deregulated degradation of the Cdk inhibitor p27 and malignant transformation. Semin. Cancer Biol.13, 41–47 (2003). | Article | PubMed | ISI | ChemPort |
Senderowicz, A.M. Cyclin-dependent kinases as new targets for the prevention and treatment of cancer. Hematol. Oncol. Clin. North Am.16, 1229–1253 (2002). | PubMed | ISI |
McClue, S.J. et al. In vitro and in vivo antitumor properties of the cyclin dependent kinase inhibitor CYC202 (R-roscovitine). Int. J. Cancer102, 463–468 (2002). | Article | PubMed | ISI | ChemPort |
Adams, P.D. et al. Identification of a cyclin-Cdk2 recognition motif present in substrates and p21-like Cdk inhibitors. Mol. Cell. Biol.16, 6623–6633 (1996). | PubMed | ISI | ChemPort |
van den Heuvel, S. & Harlow, E. Distinct roles for cyclin-dependent kinases in cell cycle control. Science262, 2050–2053 (1993). | PubMed | ChemPort |
Chen, Y. et al. Selective killing of transformed cells by cyclin/cyclin-dependent kinase 2 antagonists. Proc. Natl. Acad. Sci. USA96, 4325–4329 (1999). | Article | PubMed | ChemPort |
McInnes, C., Andrews, M.J., Zheleva, D.I., Lane, D.P. & Fischer, P.M. Peptidomimetic design of CDK inhibitors targeting the recruitment site of the cyclin subunit. Curr. Med. Chem. Anti-Canc. Agents3, 57–69 (2003). | PubMed | ChemPort |
Zheleva, D.I. et al. Highly potent p21(WAF1)-derived peptide inhibitors of CDK-mediated pRb phosphorylation: delineation and structural insight into their interactions with cyclin A. J. Pept. Res.60, 257–270 (2002). | Article | PubMed | ISI | ChemPort |
Mendoza, N. et al. Selective cyclin-dependent kinase 2/cyclin A antagonists that differ from ATP site inhibitors block tumor growth. Cancer Res.63, 1020–1024 (2003). | PubMed | ISI | ChemPort |
Ortega, S. et al. Cyclin-dependent kinase 2 is essential for meiosis but not for mitotic cell division in mice. Nat. Genet.35, 25–31 (2003). | Article | PubMed | ISI | ChemPort |
Sage, J., Miller, A.L., Perez-Mancera, P.A., Wysocki, J.M. & Jacks, T. Acute mutation of retinoblastoma gene function is sufficient for cell cycle re-entry. Nature424, 223–228 (2003). | Article | PubMed | ISI | ChemPort |
Malek, N.P. et al. A mouse knock-in model exposes sequential proteolytic pathways that regulate p27Kip1 in G1 and S phase. Nature413, 323–327 (2001). | Article | PubMed | ISI | ChemPort |
Tetsu, O. & McCormick, F. Proliferation of cancer cells despite CDK2 inhibition. Cancer Cell3, 233–245 (2003). | Article | PubMed | ISI | ChemPort |
Fischer, P.M. et al. Structure-activity relationship of truncated and substituted analogues of the intracellular delivery vector Penetratin. J. Pept. Res.55, 163–172 (2000). | Article | PubMed | ISI | ChemPort |
Krek, W. & DeCaprio, J.A. Cell synchronization. Methods Enzymol.254, 114–124 (1995). | Article | PubMed | ISI | ChemPort |
Hollingshead, M.G. et al. In vivo cultivation of tumor cells in hollow fibers. Life Sci.57, 131–141 (1995). | Article | PubMed | ISI | ChemPort |
Acknowledgments We thank A. Kung for advice, S. Ren for help with the Cdk2 kinase assays, G. Enders and J. Roberts for critical reading of the manuscript, and members of the Kaelin laboratory for useful discussions. This work was supported by a National Institutes of Health RO1 to W.G.K. W.G.K. is a Howard Hughes Medical Institute Investigator.
1 s.d.
N+C is a p21 mutant that cannot block Cdk activity. Each transfection mix also contained a Renilla luciferase reporter plasmid as an internal control. Cell extracts were prepared, and normalized firefly luciferase activity was determined. (c−f) HeLa cells stably transfected to produce luciferase or p27Luc were incubated with wild-type or mutant (control) cyclin inhibitory peptides fused to penetratin (c,d) or the indicated siRNAs (e,f). Cell extracts were immunoblotted with the indicated antibodies (c,e) or assayed for luciferase activity (d,f). * indicates nonspecific binding. Error bars indicate 
M, respectively) that did not alter cell cycle distribution assessed by propidium iodide staining followed by FACS (data not shown). At the highest doses tested, flavopiridol (200 nM) caused an increase in S- and G2/M-phase U2OS cells, but had no demonstrable effect on the cell cycle of HeLa or H1299 cells (data not shown). Roscovitine (25 
5 mm) had formed bilaterally. Right, repeat images were obtained after two doses of flavopiridol (5 mg/kg, once a day by intraperitoneal injection). Fold induction was calculated as p27Luc/Lucpost-treatment ÷ p27Luc/Lucpretreatment. Error bars indicate standard error (s.e.m.).
-aminobutyric acid and pFF = p-fluorophenylalanine. The control peptide contained the sequence HAKARAIA. The C termini of both peptides were tagged with the penetratin sequence RQIKWFQNRRMKWKK and blocked with an amide group
-ATP (10 
106 cells in 0.2 ml PBS were injected subcutaneously into each site in the flanks of NCR-Nu nude mice under anesthesia (isoflurane (Isoflo), Abbott Laboratories).
