Cyclin-dependent kinases (cdks) represent potentially promising molecular targets for cancer therapeutic strategies. To evaluate the antitumor activity of selective cyclin/cdk inhibition, we constructed a chimeric protein composed of a F-box protein (TrCP) fused to a peptide comprising the cyclin/cdk2 binding motif in p21-like cdk inhibitors (TrCP-LFG). We now demonstrate that endogenous cyclin A and its binding substrate, cdk2, can be tethered to β-TrCP, ubiquitinated, and effectively degraded. Degradation of cdk2 and cyclin A together, but not cdk2 alone, results in massive tumor cell apoptosis in vitro and in vivo in a proteasome-dependent manner with no toxicity to normal tissue. These data demonstrate that cyclin A and/or the cyclin A/cdk2 complex is a promising anticancer target with a high therapeutic index.
Most cyclin-dependent kinases (cdks) are involved in the orderly progression of the cell cycle. Dysregulation of cell cycle progression can result in disordered and uncontrolled cell growth, which is characteristic of neoplasia (1, 2, 3, 4, 5, 6) . A number of cdks (i.e., cdk4/6, cdk2) target the retinoblastoma protein (pRb), leading to its inactivation through phosphorylation (7, 8, 9, 10, 11, 12) . One critical set of downstream targets of pRb are the family of cell-cycle regulatory transcription factors, E2F (13, 14, 15) . In mitotically quiescent normal cells, hypophosphorylated pRb complexed to E2F forms a transcriptional repressor, whereas phosphorylation of pRb leads to reduced levels of functional pRb/E2F repressor complexes and activation of E2F-responsive promoters important for moving the cell through the G1-S phase of the cell cycle (16, 17, 18, 19, 20, 21, 22) .
In addition to pRb, E2F is also negatively regulated by cdk activity, particularly cyclinA/cdk2 (19 , 23, 24, 25, 26, 27) . E2F-1, E2F-2, and E2F-3 each contain a short, collinear cyclin A/cdk2 binding motif that is required for the phosphorylation-mediated neutralization of E2F as a cell prepares to exit S phase (26 , 28) . Mutation of this motif abrogates phosphorylation of E2F, leading to unopposed E2F activity, which in the proper cellular context can lead to apoptosis (25) . Previous studies suggested that in the absence of functional pRb, as is found in most human cancer (29, 30, 31) , inhibition of cyclin A/cdk2 should additionally increase unphosphorylated E2F-1 and thus E2F-1 activity (32 , 33) . Furthermore, because some E2F family members, including E2F-1, are themselves transcribed from E2F-responsible promoters, unopposed activation of E2F activity should establish a positive feedback loop, leading to very high intracellular E2F activity and ultimately apoptotic cell death (34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45) .
Thus, cdk2 inhibition, particularly in the setting of functional pRb inactivation, has the potential as a selective cancer therapeutic strategy. Small molecule cdk inhibitors such as flavopiridol have been developed, although their specificity for any particular cdk (i.e., cdk2) is limited (33 , 46 , 47) . In the search for a more specific cdk2 inhibitor, a short peptide (PVKRRLFG) (“LFG”) was found on the cyclin/cdk2 binding motif in p21-like cdk inhibitors (28 , 48 , 49) . LFG specifically blocks the phosphorylation of substrate by cyclin A/cdk2 and cyclin E/cdk2 and preferentially induces apoptosis in transformed cells (24 , 50) . As a small peptide, however, the clinical use of LFG may be limited by pharmacological constraints, leading to inefficient delivery in vivo. Furthermore, a recent study questions whether selective cdk2 inhibition is a useful cancer strategy (51 , 52) . Additionally, the Barbacid laboratory has shown no cell cycle abnormalities in either a cdk2 null mouse or following acute ablation of cdk2 in primary cells by Cre-loxP-mediated recombination, thereby raising the question of the true importance of cdk2 in cellular proliferation (53) . Thus, there is an urgent need for the evaluation of other potentially more efficient strategies aimed at selectively targeting cdk2 and its binding partners in tumor cells.
One of the potential reasons for the inefficiency of the LFG peptide approach may be that the peptide merely competes with substrate for cdk2 binding sites, thus requiring high, continuous intracellular concentration of the peptide. We reasoned that destruction of the cdk2 protein, with or without its binding partners, might represent a more efficient approach. A major pathway used by eukaryotic cells to degrade specific proteins is ubiquitin-dependent proteolysis, which involves a cascade of enzymatic reactions catalyzed by the E1 ubiquitin-activating enzyme, the E2 ubiquitin-conjugating enzymes, and the E3 ubiquitin-protein ligases. The substrate specificity of the ubiquitin pathway is conferred by the E3. One of the best characterized ubiquitin ligase complexes is the Skp1-Cullin-F-box (SCF) complex (reviewed in Ref. 54 ). Within the SCF complex, the F-box-containing proteins serve as the receptors for specific substrates (55) . The F-box protein contains two essential modular domains: the F-box that is required for binding to Skp1 and a protein-protein interaction domain for binding distinct substrates (56) . It has been previously reported that specific proteins can be degraded when a F-box protein is engineered to contain a specific protein interaction domain without affecting degradation of normal endogenous substrates (57 , 58) . We now report the construction of a chimeric protein composed of a F-box protein (TrCP) fused to the LFG peptide and demonstrate that endogenous cdk2 and its binding substrate, cyclin A, can be tethered to β-TrCP, ubiquitinated, and degraded. Degradation of cdk2 and cyclin A together, but not cdk2 alone, results in massive tumor cell apoptosis in vitro and in vivo in a proteasome-dependent manner.
MATERIALS AND METHODS
N-Acetyl-l-leucinyl-l-leucinyl-l-norleucinal (ALLN) were purchased from Sigma (St. Louis, MO). All other reagents were purchased either from Invitrogen (Gaithersburg, MD) or from the sources stated in the text. Monoclonal antibody specific to Flag was purchased from Sigma. All of the other antibodies used in immunoblotting and immunoprecipitation were purchased from Pharmingen (San Jose, CA).
U87, 293, U2OS, SAOS, HeLa, HCT116, and SW480 cells were purchased from American Type Culture Collection (Manassas, VA). Cells were cultured in the media as described in the manufacturer’s data sheet.
Plasmids and Adenovirus.
The Flag-tagged β-TrCP-LFG (TrCP-LFG), Flag-tagged ΔTrCP-LFG (ΔTrCP-LFG), and Flag-tagged TrCP (TrCP) were generated by PCR using TrCP/PcDNA3 or ΔF-TrCP/PcDNA3 (a gift from Dr. Klaus Strebel) as a template. The final PCR fragments were cloned into the BglII/NotI sites of pAdTrack plasmid. All PCR products were verified by automated sequencing at the National Institute of Neurological Disorders and Stroke core facility at the NIH. These plasmids were recombined with pAdEasy1 and transfected into 293 cells to make recombinant adenoviruses (59) . Ad/p73dd was generated by PCR using PcDNA3-p73dd as a template (a gift from Dr. William G. Kaelin). Recombinant adenovirus for p73dd was generated using the same strategy as recombinant adenovirus for TrCP.
Western Blots and Immunoprecipitation.
For Western blots, protein extracts of cells were prepared in lysis buffer [20 mm HEPES (pH 7.4), 50 mm β-glycerol phosphate, 2 mm EGTA, 1 mm DTT, 10 mm NaF, 1 mm sodium orthovanadate, 1% Triton X-100, and 10% glycerol with proteinase and phosphorylation inhibitors]. The protein concentration of the cell lysates was measured by the Bradford assay (Bio-Rad, Hercules, CA), and 20–30 μg of protein/sample (70 μg for p73) were loaded onto SDS/PAGE gels or NuPAGE gels (Invitrogen). Proteins were transferred to polyvinylidene difluoride membranes and immunoblotted with specific antibodies. For immunoprecipitation, cells were lysed using immunoprecipitation buffer [10 mm Tris (pH 7.4), 150 mm NaCl, 1% Triton X-100, 1 mm EDTA, 1 mm EGTA (pH 8.0), 0.2 mm sodium orthovanadate, 0.2 mm phenylmethylsulfonyl fluoride, and 0.5% NP40]. Anti-Flag or cdk2 antibodies were added to 500-μg lysates and incubated at 4°C for 1 h. Protein A- or G-agarose beads were added to the antigen-antibody mixture and additionally incubated for 30 min at 4°C. Beads were washed (twice) with immunoprecipitation buffer and subjected to Western blot analysis using anti-cyclin A, cdk2, and Flag antibodies as described above.
Transient Transfection in HeLa Cells.
Various plasmids were introduced into HeLa cells using Lipofectamine Plus reagent following manufacturer’s instructions (Invitrogen). Forty-eight h after transfection, cells were harvested, and protein lysates were resolved via gel-electrophoresis and blotted onto polyvinylidene difluoride membrane. The expression of Flag-tagged TrCP-conjugated protein was confirmed by immunoblotting using the anti-Flag monoclonal antibody.
Small Interfering (si)RNA Transfection.
The siRNA sequences used for targeted silencing of cdk2 (5′-aagctgctggatgtcattcac-3′), cyclin A (5′-aaccattggtccctcttgatt-3′), and p27 (5′-aagtacgagtggcaagaggtg-3′) were chosen as described by Elbashir et al. (60) and recommended by the siRNA supplier (Xeragon, Germantown, MD). Human genome database (BLAST) searches were carried out to ensure that the sequences would not target other gene transcripts. siRNAs were introduced to the cells using Oligofectamine according to the protocol of Elbashir et al. (60) and manufacturer’s instructions (Invitrogen). Briefly, two siRNA transfections were sequentially performed at 24 and 48 h after cells were plated. Four h after the second transfection, cells were infected with the recombinant adenovirus. 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assays (Roche, Germany) were carried out 96 h after adenovirus treatment. Two days after treatment, cells were harvested for Western blotting analysis.
Apoptosis and Cell Cycle Analysis.
Cells (3 × 105/well) were seeded in 6-well plates and infected 16–18 h later with adenoviruses at a multiplicity of infection of 200. Forty-eight h after infection, both attached cells and floating cells were harvested, washed in PBS, and fixed with 70% ethanol. Apoptotic cells were detected by using ApopTag peroxidase in situ apoptosis detection kit (Intergen, Norcross, GA). Cell cycle analysis was prepared by first fixing cells and treating with RNase A, followed by staining with 5 μg/ml propidium iodide in PBS. Cell cycle analysis was performed on a FACS brand flow cytometer (Becton Dickinson, San Jose, CA); the data were analyzed by MULTI-FIT software.
Determination of Cell Growth Rate.
Cell growth was assessed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (Roche). U87 cells were plated on 96-well plates at a density of 10,000 cells/well; all other cells were plated at a density of 5000 cells/well. One day after plating, cells were treated with adenovirus at a multiplicity of infection of 200–500 or transfected with siRNA. Cell viability was determined 96 h after adenovirus infection or siRNA transfection. Each experiment was repeated at least three times with each experiment done in a set of 6 wells. The student t test was used to assess statistical significance. Ps of <0.05 were considered significant.
Cdk Kinase Activity Assay.
Two hundred μg of protein were immunoprecipitated for each sample by anti-cdk2 or anti-cdk4 antibodies. The immunoprecipitated protein was resuspended in 50 μl of kinase buffer [20 mm HEPES (pH 7.2), 25 mm β-glycerol phosphate, 5 mm EGTA, 1 mm sodium orthovanadate, 1 mm DTT, 7.5 mm MgCl2, and 50 mm ATP] containing 10 μCi of 32P-dATP (3000 Ci/mmol; Amersham, Piscataway, NJ) and 5 μg of Rb-C fusion protein (Cell Signaling Technology, Beverly, MA), and incubated for 2 h at 30°C. Reaction was stopped by adding 15 μl of 4× NuPAGE SDS sample buffer. Samples were boiled for 5 min, and incorporation of radioactive phosphate was determined by 10% NuPAGE. Analysis of the dried gel was performed using a PhosphoImager (Molecular Dynamics, Piscataway, NJ).
For the s.c. tumor model, U87 cells (1 × 107) were s.c. injected into the right flank of 4-week-old female nude mice (National Cancer Institute-Frederick Cancer Research & Development Center, Frederick, MD). Once tumor size had reached 200 mm3, the mice were treated with adenovirus expressing either green fluorescent protein (GFP) or TrCP-LFG or PBS. The mice were injected with 100 μl of Ad/CMV-GFP (109 pfu) or Ad/TrCP-LFG (109 pfu) every 3–4 days for a total of six treatments. Tumor size was measured every 3–4 days, and tumor volume was calculated from the formula (length) × (width)2, where length and width are perpendicular. The student t test was used to assess statistical significance. Ps of <0.05 were considered significant.
For the intracranial U87 glioma model, 4-week old nude mice were anesthetized with ketamine and xylazine and immobilized in a rodent stereotactic frame (Stoetling Co., Wood Dale, IL). Through a small scalp incision, an intracranial opening was made using an 18-gauge needle tip in a position 2 mm right and 1 mm anterior from the bregma. A single guide cannula (with a 4-mm pedestal and a 1-mm subpedestal; Plastic-One, Roanoke, VA) was installed and secured onto the scalp at the stereotactic opening using medical grade cyanoacrylate gel. The following day, U87 cells (5 × 104) suspended in 5 μl of PBS were injected stereotactically into the right parietal region to a depth of 2.5 mm using a Hamilton syringe(Hamilton Co, Reno, NV) guided through the secured cannula. Three days after tumor implantation, animals were randomly chosen to receive recombinant adenoviruses expressing either GFP or TrCP-LFG through the installed cannula. Adenovirus (5 × 109 pfu) in 5 μl of PBS was injected into the substance of the intracranial tumor using the guide cannula. The mice received three total intracranial viral injections at 3–4-day intervals. These mice were examined daily and after death, and the brains were examined to confirm that the cause of death was tumor progression, not infection or treatment toxicity in each case. Survival for the treated and control animals were compared using a log-rank analyses from the Kaplan-Meier survival curves.
The Engineered β-TrCP-LFG Targets Cdk2 for Degradation.
To explore the effects of inducing proteasome-mediated degradation of cdk2, we engineered chimeric proteins with the full-length flag-tagged β-TrCP (TrCP) protein fused at the protein-protein interaction domain to the cdk2 binding 8 amino acid peptide LFG (TrCP-LFG; Fig. 1A ⇓ ). As shown in Fig. 1B ⇓ , expression of TrCP-LFG in HeLa cells resulted in a decrease in the endogenous level of cyclin A and cdk2. The expression of TrCP or F-box-truncated β-TrCP fused to LFG (ΔTrCP-LFG) did not reduce cyclin A protein levels nor cdk2 protein levels. The faster migrating phosphorylated (Thr160) active form of cdk2 (lower band; Refs. 48 , 61, 62, 63 ) showed a preferential and marked reduction. Cyclin E protein levels remained constant (Fig. 1B) ⇓ . To confirm that the TrCP-LFG-mediated decrease in the endogenous cyclin A and cdk2 protein levels was through a ubiquitin-dependent proteolysis pathway, we performed a series of experiments where TrCP-LFG treated cells were exposed to N-acetyl-l-leucinyl-l-leucinyl-l-norleucinal, a peptide inhibitor of proteolysis by the 26S proteasome (Fig. 1C) ⇓ . N-Acetyl-l-leucinyl-l-leucinyl-l-norleucinal inhibited the degradation of TrCP-LFG in a dose-dependent manner and prevented the TrCP-LFG-mediated down-regulation of cyclin A and cdk2 protein levels. The specificity of this effect for cdk2 and its binding partners was demonstrated by the fact that cdk4 protein levels remained unchanged under identical experimental conditions (Fig. 1C) ⇓ . Specific binding of chimeric TrCP-LFG protein to cyclin A and cdk2 was shown by coimmunoprecipitation in that anti-Flag antibodies precipitated TrCP-LFG and coprecipitated cyclin A and cdk2 (Fig. 1D) ⇓ . The faster migrating active form of cdk2 was preferentially precipitated when anti-Flag antibodies were used for immunoprecipitation. Coimmunoprecipitation with anti-cdk2 antibody also showed cyclin A and TrCP-LFG (Fig. 1D) ⇓ . These results indicate that β-TrCP fused with the cdk2-specific binding motif, LFG, can selectively interact with cyclin A/cdk2 complex and result in the degradation of cyclin A and cdk2 through a proteasome-dependent mechanism.
Inhibition of Tumor Cell Growth and Induction of Cell Death by Adenovirus-Mediated Transduction of TrCP-LFG.
Recombinant adenoviral vectors are a commonly used delivery system for cancer gene therapy. We constructed a replication defective recombinant adenovirus that expresses TrCP-LFG (Ad/TrCP-LFG). Western blot analysis was performed to examine the effect of Ad/TrCP-LFG on cyclin A and cdk2 protein level in different cell lines. In all tested cell lines, both cyclin A and cdk2 protein levels were significantly reduced following Ad/TrCP-LFG transduction (Fig. 2A) ⇓ . Consistent with the effects on protein levels, transduction by Ad/TrCP-LFG resulted in significantly reduced intracellular cdk2 activity (Fig. 2B) ⇓ . Ad/TrCP-LFG transduction of various cell lines in vitro mediated a significant cytotoxic effect in most tumor cell lines examined, whereas there were minimal effects on the survival and growth of normal human fibroblasts and astrocytes (Fig. 2, B and C) ⇓ . Control adenoviral vectors expressing TrCP or ΔTrCP-LFG caused no significant change in cdk2 protein level nor cytotoxicity in any of the cell lines (Fig. 2D) ⇓ . These data suggest a differential cytotoxic response between cancer and normal cells to cyclin A/cdk2 down-regulation.
Induction of Apoptosis in Tumor Cells Transduced with Adenovirus-Expressing TrCP-LFG.
We performed experiments to elucidate the mechanisms responsible for Ad/TrCP-LFG-mediated death in tumor cells. Tumor cells treated with Ad/TrCP-LFG exhibited membrane blebbing and nuclear condensation (data not shown), suggestive of apo-ptotic cell death. To confirm that apoptosis had occurred in Ad/TrCP-LFG-transduced cells, we performed terminal deoxynucleotidyl transferase dUTP nick end-labeling assay. Two days after Ad/TrCP-LFG transduction, U2OS, SAOS, and U87 tumor cells all demonstrated significant terminal deoxynucleotidyl transferase dUTP nick end-labeling staining, whereas there were little or no terminal deoxynucleotidyl transferase dUTP nick end-labeling-positive normal cells (Fig. 3A) ⇓ . We quantitated the degree of apoptosis induced by adenoviral-mediated transduction of TrCP-LFG by flow cytometric analysis (fluorescence-activated cell sorting). The percentage of apoptotic cells (sub-G1 fraction) increased steadily over the first 96 h after Ad/TrCP-LFG transduction in all tumor cell lines, reaching a high of 40–90% in all transduced cells (Fig. 3B) ⇓ . By contrast, Ad/TrCP-LFG transduction caused essentially no apoptosis in normal human fibroblasts or astrocytes (Fig. 3B ⇓ and data not shown).
Induction of p73 Is Required for Ad/TrCP-LFG-Induced Cell Death in Rb-Negative Cell Lines.
Cdk2 kinase activity has been shown to be important for phosphorylating and inactivating E2F activity. High E2F activity can induce apoptosis by a p73-dependent mechanism. Thus, one might predict that inactivation of the Cdk2 complex would lead to unopposed E2F activity resulting in apoptosis. To test this hypothesis, we evaluated the levels of p73 in each of our cell lines (Fig. 4A) ⇓ . Basal expression of p73 was much higher in U2OS cells compared with the other cell lines. Ad/TrCP-LFG treatment induced expression of p73 in SAOS cells and, to a much lesser extent, in U87 cells but failed to induce expression in other cell lines, including U20S (Fig. 4A ⇓ , top panel). The increased expression level of p73 in the Rb-negative cell line SAOS and Rb-dysregulated U87 in response to the treatment of Ad/TrCP-LFG suggested that the induction of p73 might contribute to Ad/TrCP-LFG-mediated cell death. To test this hypothesis, we constructed a recombinant adenoviral vectors expressing a dominant-negative p73 construct (Ad/p73dd). Preinfecting SAOS cells with Ad/p73dd partially protected cells from apoptotic death induced by Ad/TrCP-LFG (from 42.6% cytotoxicity to 11% cytotoxicity; P < 0.01; Fig. 4B ⇓ ). Ad/p73dd also partially protected U87 cells (Fig. 4B ⇓ , from 82.3% cytotoxicity to 37.5% cytotoxicity; P < 0.01). Interestingly, Ad/p73dd also protected against Ad/TrCP-LFG-induced cell death in another Rb-negative cell line, H596 (Fig. 4B ⇓ , from 62.35 cytotoxicity to 39.3% cytotoxicity; P < 0.01), despite the fact that no p73 induction was noted in these cells after infection by Ad/TrCP-LFG. By contrast, Ad/p73dd offered no protection against Ad/TrCP-LFG-induced cell death in the two Rb-positive cell lines we tested (U2OS and A549; Fig. 4B ⇓ ). These data suggest that p73 may play a preferential role in Ad/TrCP-LFG-induced apoptotic cell death in Rb-negative cell lines. To explore other potential mechanisms involved in TrCP-LFG-mediated cell death, we examined the changes of protein expression of the cdk inhibitors p27 and p21 in response to Ad/TrCP-LFG treatment. Induction of p27 was evident in all cell lines 48 h after transduction by Ad/TrCP-LFG (Fig. 4A) ⇓ . By contrast, p21 levels were either not or only minimally increased after exposure to TrCP-LFG, except for a large increase seen in SAOS cells. Knockdown of p27 induction by p27 siRNA treatment did not protect the cell death induced by Ad/TrCP-LFG in the Rb-negative cell lines, whereas it afforded some protection from Ad/TrCP-LFG-induced cell death in Rb-positive cells (data not shown). Taken together, these data suggest that the molecular mechanisms responsible for TrCP-LFG-mediated cell death may be different in Rb-positive cells compared with Rb-negative cells.
Both Cyclin A and Cdk2 Degradation Are Necessary for Maximal Tumor Cell Apoptosis.
A recent study (51) suggests that cdk2 is not required for tumor cell survival and growth and cdk2-knockout mice and cells do not show abnormality in cell cycle progression. Because TrCP-LFG mediates destruction of both cdk2 and cyclin A, we performed a series of experiments designed to individually evaluate the effects of isolated cdk2 and cyclin A down-regulation. siRNA-mediated down-regulation of cdk2 alone did not result in cytotoxicity in any of the tumor cell lines evaluated (Fig. 5) ⇓ . siRNA down-regulation of cyclin A did, however, have a small but statistically significant cytotoxic effect in U87 glioma cells and HCT116 colon carcinoma cells lines (Fig. 5B) ⇓ . By contrast, simultaneous down-regulation of cyclin A and cdk2 by siRNAs caused a greater effect on cell killing than seen with down-regulation of either protein alone (Fig. 5B) ⇓ . Of note, SW480 were totally resistant to TrCP-LFG- and siRNA-mediated cell killing despite the effective down-regulation of cyclin A and cdk2 protein levels (Fig. 5A) ⇓ and diminished cdk2 kinase activity to levels seen in all of the other sensitive cell lines we examined (Fig. 2B) ⇓ . These data demonstrate that the cytotoxic effects of simultaneous cdk2/cyclin A down-regulation are more profound than down-regulation of either protein alone in all cell lines tested with one notable exception.
Suppression of Tumor Growth by Intratumoral Injection of Ad/TrCP-LFG.
Given the pronounced and selective cytotoxic effects of Ad/TrCP-LFG on tumor cells but not normal cells in vitro, we evaluated the efficacy of Ad/TrCP-LFG administration on glioma tumor growth in vivo. Subcutaneous U87 gliomas were treated with intratumoral injections of Ad/TrCP-LFG or a control adenoviral vector starting day 3 after tumor implantation and then every 3–4 days for a total of six treatments. All tumors in the mice treated with Ad/TrCP-LFG showed significant suppression of growth compared with tumors in PBS-treated and Ad/GFP-treated mice (Fig. 6A) ⇓ . In a separate experiment, we euthanized a subgroup of animals 24 h after viral injection to examine the effects of Ad/TrCP-LFG on tumor cells in vivo. The transduction efficiency of the different adenoviral vectors was roughly equivalent as evidenced by the fact that the intensity and distribution of GFP expression was similar in both the control GFP-expressing and GFP/TrCP-LFG-expressing adenoviral vector-treated tumors. By contrast, there were significantly more terminal deoxynucleotidyl transferase dUTP nick end-labeling-labeled apoptotic nuclei in the Ad/TrCP-LFG-treated group compared with the control group (Fig. 6B) ⇓ .
We further assessed the potential of Ad/TrCP-LFG as a therapeutic strategy for human brain cancers using an orthotopic intracranial U87 glioma model. To evaluate the effects of the TrCP-LFG construct on normal cerebral tissue, Ad/TrCP-LFG or the control Ad/GFP virus was stereotactically injected into the right caudate of nontumor-bearing mice. All animals remained clinically well with no signs of neurological toxicity. The brains of the injected animals were examined at 1, 7, and 28 days after viral transduction. There was no histological evidence of neurotoxicity in any of the examined brains with only a minimal mononuclear cell infiltrate within the injected area, characteristic of the effect seen after adenoviral injection into the brain (data not shown). There was no difference between the histological appearance of the brains injected with the either the Ad/GFP or Ad/TrCP-LFG virus. We next stereotactically inoculated U87 glioma cells into the right parietal region of the brain of nude mice after 3 days later by stereotactic administration of recombinant adenovirus expressing either GFP or TrCP-LFG. Vector administration was repeated three times at 3–4-day intervals, and animals were followed for survival. Kaplan-Meier survival curves for the control animals and those treated with Ad/TrCP-LFG are shown in Fig. 7A ⇓ . The median survival of control vector treated animals was 34 days, whereas the median survival of the Ad/TrCP-LFG-treated animals was 50 days (P < 0.01). Log-rank analysis of Kaplan-Meier survival curves demonstrated a significant survival advantage for the Ad/TrCP-LFG treated animals (P < 0.01). More impressively, 50% of the animals receiving Ad/TrCP-LFG were ultimately cured of their tumors (>120-day survival) with no neurological or physical evidence of tumor nor toxicity. The long-term survivors were ultimately euthanized, and microscopic examination of their brains revealed no evidence of either residual tumor or normal tissue toxicity (Fig. 7B) ⇓ .
We demonstrate that a cyclin A/cdk2 inhibitor binding peptide fused to a recombinant TrCP protein can selectively direct cyclin A and its binding partner, cdk2, toward proteasome-mediated degradation resulting in tumor cell apoptosis in vitro and in vivo. This strategy may represent a more efficient way of inhibiting the cyclin A/cdk2 pathway than the LFG peptide approach because, instead of only competitively inhibiting the kinase activity, the TrCP-LFG fusion protein targets the kinase complex for destruction.
Not only was the TrCP-LFG strategy highly efficient for cell killing, but the strategy also appears to be specific for targeting the cyclin A/cdk2 complex given that we did not see effects of TrCP-LFG on cyclin B or D protein levels or associated kinase activity (Figs. 1 ⇓ and 2 ⇓ and data not shown). Consistent with these observations is the fact that previous work has demonstrated that LFG and similar peptides can inhibit cyclin A-associated kinase activity but not cyclin B or D-associated kinase activity (28 , 42 , 48, 49, 50) . Furthermore, we did not see effects of the transected TrCP-LFG on endogenous substrates for the β-TrCP E3 ligase such as nuclear factor-κβ (data not shown). At first glance, the highly specific nature of the TrCP fusion peptide approach might appear surprising. Given the central role of proteasome-mediated proteolysis for processes as important and sensitive as cell cycle regulation and cell survival, however, E3 ligases are by necessity highly specific for their targeted binding proteins. Thus, it is not all that unexpected that a TrCP-peptide fusion approach might so specific for its targeted protein binding partner.
Much attention has recently focused on interfering RNAs as a revolutionary strategy for uniquely down-regulating-specific target mRNAs. Recent experience, however, has demonstrated siRNAs to be problematic relative to intracellular transduction inefficiencies, their unpredictable effectiveness depending on the specific sequence that is used, and the difficulties in efficiently delivering these molecules in vivo. Additionally, recent data has shown that siRNAs may not be as specific as previously thought resulting in interference in a number of off-target mRNAs (64) . By contrast, an approach such as our TrCP-LFG strategy that targets the specific protein of interest, rather than an intermediate RNA species, is theoretically attractive. Thus, we believe our data supports the premise that the TrCP-fusion peptide approach represents a potentially powerful strategy for targeting specific proteins both in vitro and in vivo (57) .
Given the abundance of data suggesting that cdk2 is a potentially promising cancer therapeutic target, we initially devised the TrCP-LFG strategy with the expectation that a cytotoxic affect would be a consequence of cdk2 down-regulation. McCormick et al. (51) , however, have recently published data using principally colon cancer cell lines, questioning the need for cdk2 in tumor cell proliferation and survival. Barbacid et al. (53) also showed that cdk2 may not be necessary for normal cell cycle progression by using cdk2-knockout mice and Cre-loxP-mediated somatic cell knockout. Because cdk2 binds with several different protein partners, however, loss of cdk2 function may not be synonymous with targeted destruction of the cdk2/cyclin complex. Because TrCP-LFG down-regulates both cyclin A and cdk2, it was possible that the observed cytotoxic effects were a consequence of the destruction of either cdk2, cyclin A, or both. We, therefore, performed experiments using siRNAs to specifically target cdk2 and cyclin A mRNA and demonstrate that, indeed, down-regulation of cdk2 alone had no effects on cell viability, whereas cyclin A down-regulation did mediate a cytotoxic effect. Interestingly, however, it appeared that the combination of both cyclin A and cdk2 inhibition led to the most potent cytotoxic effect in sensitive cell lines. Although more in-depth studies will be required to fully understand these observations, it appears that the cyclin A or the cyclin A/cdk2 complex may be a preferable to cdk2 as a therapeutic tumor target.
The downstream mediators of cytotoxicity after TrCP-LFG-mediated down-regulation of cyclin A/cdk2 remain to be fully elucidated, although our data suggest that those mechanisms may differ depending on the pRb status of the cell. TrCP-LFG-treated tumor cells, which have deregulated Rb function, appear to undergo apoptosis in a p73-dependent manner. This is consistent with previous studies (41 , 65 , 66) demonstrating that unopposed E2F activity results in p73-dependent apoptosis. Because cyclin A/cdk2 phosphorylates E2F-1, thereby inhibiting its activity, TrCP-LFG-mediated down-regulation of cyclin A/cdk2 would be expected to result in even higher E2F activity in cells with baseline deregulated pRb activity compared with those that have intact pRb pathways. Our data does in fact demonstrate that Rb-deregulated cells generate high E2F transcriptional activity after TrCP-LFG-mediated cyclin A/cdk2 down-regulation and that the resultant apoptosis can be partially inhibited by a dominant-negative p73 mutant. By contrast, tumor cells with intact Rb are not rescued by the dominant-negative p73 after TrCP-LFG transduction. Although our siRNA experiments suggest that inhibiting the p27 induction that occurs after TrCP-LFG treatment in pRb intact tumor cells can reduce the amount of apoptotic cell death, these effects are modest and suggest another prominent mechanism for TrCP-LFG-mediated apoptosis in pRb intact cells. Whether this mechanism involves an E2F-mediated, p73-independent apoptotic pathway remains to be elucidated.
The efficiency with which TrCP-LFG-induced tumor cell death appeared to be cell type dependent, although all cell lines tested (n = 10) were sensitive except for SW480. One of the most impressive features of TrCP-LFG-mediated cell death was the apparent tumor cell selectivity. There was no significant toxicity to normal cell lines in vitro or normal brain tissue in vivo after transduction by TrCP-LFG-expressing vectors. The fact that cyclin A/cdk2 inhibition did not cause toxicity to mitotically quiescent normal cells in the brain is perhaps not surprising given the dominant effect of the cdk inhibitors on G1 cyclins, thereby maintaining pRb in a predominantly hypophosphorylated state. The fact that cyclin A/cdk2 inhibition did not cause cytotoxicity in normal proliferating cells in vitro, however, is more surprising and may reflect the baseline altered state of E2F activity in tumor versus normal cells. Regardless of the mechanisms, the efficient cell killing and lack of normal tissue toxicity from TrCP-LFG suggests that cyclin A and/or the cyclin A/cdk2 complex may represent a promising molecular target with a high therapeutic index for many types of tumors. Future strategies aimed at targeting other intracellular proteins using the TrCP-peptide binding approach and strategies aimed at further inhibiting the cyclin A or the cyclin A/cdk2 complex appear warranted.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: W. Chen and J. Lee contributed equally to the work.
Requests for reprints: Howard A. Fine, Neuro-Oncology Branch, NIH, 9000 Rockville Pike, Rockville, MD 20892.
- Received December 14, 2003.
- Revision received February 24, 2004.
- Accepted March 24, 2004.
- ©2004 American Association for Cancer Research.