Dual CDK4/CDK6 Inhibition Induces Cell-Cycle Arrest and Senescence in Neuroblastoma

Neuroblastoma is a pediatric cancer that originates in tissues of the developing sympathetic nervous system. It is typically diagnosed in very young children and has a highly variable clinical presentation, with tumors displaying significant genomic and biological heterogeneity (1–3). Although patients with favorable clinical and biological features (low-risk disease) can often be cured by surgery alone, patients with high-risk disease, particularly those harboring amplification of the MYCN oncogene, have survival rates of less than 40% despite intensive multimodal therapy (1, 4). Such high mortality thus mandates a need for the development of novel therapies that will significantly improve high-risk patient survival.

The identification of molecular abnormalities driving neuroblastoma tumorigenesis and disease progression followed by their targeted treatment is a pragmatic strategy for meeting this need. Currently, converging evidence points to members of the Cyclin D/CDK4/CDK6/RB cell-cycle regulatory pathway as potential candidates for such therapeutic exploitation. Both CDK4 and CDK6 (CDK4/6) encode cyclin-dependent serine–threonine kinases that, in response to mitogenic or pro-proliferative stimuli, complex with D-type cyclins to phosphorylate the RB tumor suppressor protein. This phosphorylation induces the release of RB from E2F transcription factors, and thus enables E2F to transcribe the genes that are required for G₁–S phase cell-cycle progression and ultimately cellular proliferation (5–7). In addition to cell-cycle regulation, CDK4/6 signaling has also been linked to senescence suppression via regulation of the FOXM1 transcription factor (8). Given its demonstrated ability to override suppressive cues in favor of cellular proliferation, it is not surprising that deregulation of the Cyclin D/CDK4/CDK6/RB pathway is associated with unrestricted growth and is a hallmark of nearly every tumor histotype.

With respect to neuroblastoma, we and others have identified several genetic aberrations that increase CDK4/6 kinase activity. Genomic amplification of CCND1 (Cyclin D1) and CDK4, as well as homozygous deletion of CDKN2A, have been reported in a subset of neuroblastomas (9–14). In addition, CCND1, CDK4, and CDK6 have not only been shown to be overexpressed in almost all cases of neuroblastoma, but also their expression was found to be higher in neuroblastoma in comparison to other tumors (15). More recently, a synthetic lethality screen of the protein kinome identified CDK4 as a potential candidate for therapeutic targeting in neuroblastoma (16). Taken together, these findings suggest a dependency on CDK4/6 activity for neuroblastoma survival, and thus highlight their potential as molecular targets for pharmacologic inhibition.

LEE011 (Novartis Oncology) is an orally bioavailable, small molecule inhibitor of both CDK4 and CDK6. Here, we report on the preclinical evaluation of LEE011 as a neuroblastoma therapy. Our results show that a subset of neuroblastomas are highly sensitive to LEE011, and therefore support the clinical development of CDK4/6 inhibition strategies in this disease.

All cell lines were obtained from the Children's Hospital of Philadelphia cell line bank and were cultured in RPMI-1640 media containing 10% FBS, 1% l-glutamine, and 1% penicillin/streptomycin at 37°C and 5% CO₂. Annual genotyping (AmpFISTR Identifier Kit) of these lines and a single-nucleotide polymorphism (SNP) array analysis (Illumina H550) were performed to ensure maintenance of cell identity using methods as previously described (17). All annotated but de-identified patient tumor samples were obtained from the Children's Oncology Group neuroblastoma biorepository. The Illumina H550 SNP arrays were used to determine DNA copy number status of 375 high-risk primary neuroblastoma tumors, and gene expression profiling of 251 tumors (30 low-risk, 221 high-risk) was performed using Affymetrix Human Exon 1.0 ST microarrays.

A neuroblastoma tissue microarray of duplicate paraffin-embedded tumor cores from 106 diagnostic patients, as well as control tissues, was used for this study (18). Using the Bond Refine polymer staining kit (Leica Microsystems), slides from each tumor core were stained with an antibody against endogenous RB (Cell Signaling #9309, 1:300 dilution). Antigen retrieval was performed with E2 retrieval solution (Leica Microsystems), and developed slides were imaged at 20× magnification on an Aperio OS slide scanner (Aperio Technologies). Positive staining was described by staining intensity (negative, weak, moderate, or strong), and a staining score was also calculated as the product of staining intensity and the percentage of neuroblasts stained.

Approximately 40 μg of cell line or patient tumor lysate was prepared as previously described (19), separated by electrophoresis on 4% to 12% polyacrylamide gels (Lonza), transferred to PVDF membranes (Millipore), and probed with primary antibodies at the indicated dilutions: RB, 1:2,000; pRB^S780, 1:2,000; pRB^S795, 1:2,000; pRB^S807/811, 1:2,000; Cyclin D1, 1:1,000; Cyclin D3, 1:1,000; MYCN, 1:1,000; FOXM1, 1:1,000 (Cell Signaling); CDK4, 1:2,000; CDK6, 1:3,000; and β-Actin, 1:3,000 (Santa Cruz). All blots were quantified with ImageJ (National Institutes of Health).

Cell lines were plated in triplicate in 96-well plates, and knockdown of CDK4 and CDK6 was performed 24 hours later via combined transfection with 25 nmol/L ON-TARGET SMARTpool siRNA targeting CDK4 and 25 nmol/L ON-TARGET SMARTpool siRNA targeting CDK6 (Thermo Scientific). Cell lines were also transfected with siRNAs directed against PLK1 as well as nontargeting oligonucleotides (NTC), representing positive and negative controls, respectively. Cell viability was assayed 72 hours posttransfection using Cell Titer Glo (Promega). Gene knockdown was confirmed to be >85% at this time point by quantitative real-time PCR, and protein-level knockdown was confirmed by Western blot.

LEE011 was provided by Novartis pharmaceuticals. A panel of neuroblastoma cell lines, selected based upon prior demonstration of substrate adherent growth, was plated in triplicate on the Xcelligence Real-Time Cell Electronic Sensing system (ACEA Biosciences) and treated 24 hours later with a four-log dose range of inhibitor or with a dimethyl sulfoxide (DMSO) control. Cell indexes were monitored continuously for ∼100 hours, and IC₅₀ values were determined as follows: growth curves were generated by plotting the cell index as a function of time and were normalized to the cell index at the time of treatment for a baseline cell index of 1. The area under the normalized growth curve from the time of treatment to 96 hours posttreatment was then calculated using a baseline area of 1 (the cell index at the time of treatment). Areas were normalized to the DMSO control, and the resulting data were analyzed using a nonlinear log inhibitor versus normalized response function (GraphPad Prism). All experiments were repeated at least once.

Cell lines were plated in duplicate in 35 mm plates and treated 24 hours later with the indicated concentrations of LEE011 or with a DMSO control. At 96 hours posttreatment, cells were gently harvested and fixed overnight in 70% ethanol. Cells were then washed in PBS, stained with 1 μg/μL FxViolet Stain (Invitrogen), and assayed for DNA content on an Attune Acoustic Focusing Cytometer (Invitrogen). Analysis was carried out using VenturiOne software (Applied Cytometry).

Cellular senescence was assayed via measurement of senescence-associated β-galactosidase activity (SA-β-gal). Cells were grown for 24 hours in 35 mm plates, treated with 500 nmol/L LEE011 for 6 days, and then fixed and stained overnight according to the manufacturer's protocol (Cell Signaling #9860). Cells were then imaged for SA-β-gal using an Axio Observer D.1 phase contrast microscope (Zeiss). The percentage of SA-β-gal positive cells was determined by counting the number of positive cells present in 3 separate microscope frames, and then normalizing to the control. To assess apoptotic activity, cells were plated in triplicate in 96-well plates, treated with LEE011, and assayed for caspase 3/7 activation 16 hours after treatment with Caspase-Glo 3/7 (Promega). Cells treated with SN-38 were used as a positive control (20).

The BE2C, NB-1643, or EBC1 cell line–derived xenografts were implanted subcutaneously into the right flank of CB17 SCID^−/− mice. Animals bearing engrafted tumors of 200 to 600 mm³ were then randomized to oral treatment with 200 mg/kg LEE011 in 0.5% methylcellulose (n = 10) or vehicle (n = 10) daily for a total of 21 days. Tumor burden was determined periodically throughout treatment according to the formula (π/6) × d², where d represents the mean tumor diameter obtained by caliper measurement. In accordance with the Children's Hospital of Philadelphia Institutional Animal Care and Use Committee, animals were euthanized as soon as tumor volume exceeded 3 cm³. A linear mixed effects model was used to analyze differences in the rate of tumor growth between the LEE011 and vehicle-treated groups.

pRB^S807/811 (Cell Signaling) or Ki67 (Abcam #16667) antibodies were used to stain slides of formalin-fixed, paraffin-embedded tumor that had been excised from mice following 7 days of treatment with vehicle or LEE011. Staining for pRB^S807/811 was performed in accordance with the manufacturer's protocol (Cell Signaling) using a 1:200 dilution of primary antibody followed by a 1:500 dilution of anti-rabbit immunoglobulin G (IgG) secondary antibody (Abcam # 6721). Slides pretreated with alkaline-phosphatase (New England Biolabs) were used to confirm specificity of phosphorylation detection. For Ki67 staining, slides were incubated in a pressure cooker (Biocare Medical) with antigen unmasking solution (Vector Labs H3300), blocked with 2% FBS and endogenous biotin (Vector Labs SP2001), and stained at a 1:400 dilution of primary antibody for 1 hour and a 1:200 dilution of biotinylated anti-rabbit IgG secondary antibody (Vector Labs) for 30 minutes. Slides were then developed via incubation with ABC (Vector Labs) for 30 minutes followed by DAB (DAKO Cytomation) for 10 minutes. All slides were counterstained with Harris Hematoxylin (Fisher Scientific), dehydrated, mounted, and imaged at 20× magnification with an Aperio OS slide scanner (Aperio Technologies).

Copy number gain and overexpression of CDK4, CDK6, and CCND1 have been reported in many neuroblastoma cell lines (refs. 10–15; Supplementary Table S1). To confirm that these genomic aberrations translate to constitutive CDK4/6 signaling within the Cyclin D/CDK4/CDK6/RB pathway, we examined the activation status of RB in a comprehensive panel of highly characterized human neuroblastoma-derived cell lines. As shown in Fig. 1A, robust phosphorylation of RB at serines 780 and 807/811—residues directly targeted by CDK4 and CDK6 (21–23)—was observed in all neuroblastoma cell lines, and protein-level expression of CDK4, CDK6, and Cyclin D1 occurred in the majority of lines. Comparatively, RB phosphorylation together with protein-level expression of CDK4, CDK6, and Cyclin D1 was substantially lower in several other representative tumor types as well as in immortalized, nontransformed retinal pigmented epithelial (RPE1) cells (Fig. 1B). Thus, it seems that in neuroblastoma cell lines, aberrant overexpression of CDK4, CDK6, and CCND1 does indeed facilitate hyperactive CDK4/6 signaling within the Cyclin D/CDK4/CDK6/RB network.

We next analyzed several neuroblastoma patient samples to verify that the pathway activation observed in neuroblastoma cell lines is also a characteristic of patient tumors at diagnosis and is not simply an artifact of the in vitro setting. We found that CDK4 mRNA was highly expressed in high-risk patients in comparison to low-risk patients, and we observed copy number gain of CDK4 (5.1%), CDK6 (15.7%), and CCND1 (19.5%) in a cohort of 375 high-risk patients (Supplementary Fig. S1 and Table S2). RB was also expressed in the majority of patients, as a tissue microarray comprised of 106 diagnostic tumors revealed that 100 (94%) stained positively for total endogenous RB, with 90 (85%) showing moderate to strong staining (Fig. 1C). A significant increase in RB staining intensity, however, was observed in high-risk, MYCN amplified samples (P = 0.03; Supplementary Fig. S2). Western blot analysis of several diagnostic tumor samples confirmed the expression of CDK4, CDK6, and CCND1 protein, and also indicated the presence of active, phosphorylated RB (Fig. 1D). These data therefore demonstrate that CDK4/6 signaling is indeed hyperactive in both neuroblastoma cell lines and tumors.

Due to our observation that CDK4/6 signaling is highly active in neuroblastoma (15, 16), thus maintaining hyperphosphorylated RB and supporting cell-cycle progression through the G₁–S checkpoint, we chose to interrogate the effect of dual CDK4/6 depletion on neuroblastoma cell lines. Targeted depletion of CDK4/6 by siRNA resulted in differential decreases in cell viability, where some lines responded robustly to CDK4/6 depletion whereas little to no effect was observed in other lines (Fig. 2A). This phenotypic stratification of cell lines into CDK4/6 “sensitive” or “resistant” was not due to knockdown efficiency, as we achieved significant knockdown of CDK4 and CDK6 mRNA and protein in all cell lines (Fig. 2B). Sensitive cell lines, however, were more likely to harbor amplification of MYCN (P = 0.03; Fig. 2A).

We next investigated whether pharmacologic inhibition of CDK4/6 phenocopied siRNA-mediated protein depletion by treating a panel of 17 neuroblastoma cell lines with LEE011 across a four-log dose range (10–10,000 nmol/L). Treatment with LEE011 significantly inhibited substrate adherent growth relative to the control in 12 of the 17 neuroblastoma cell lines examined (mean IC₅₀ = 306 ± 68 nmol/L, considering sensitive lines only, where sensitivity was defined as an IC₅₀ of less than 1 μmol/L; Fig. 3A). This differential sensitivity to pharmacologic CDK4/6 inhibition largely reflected that of CDK4/6 depletion by siRNA, in that MYCN amplified cell lines were more sensitive to LEE011 than nonamplified lines (P = 0.01; Fig. 3A) and cell line MYCN expression was inversely correlated with sensitivity (r = −0.55, P = 0.03; Supplementary Fig. S3). To confirm that the growth inhibition observed in sensitive cell lines was indeed because of a targeted impairment of CDK4/6 signaling, we analyzed the levels of phosphorylated RB following treatment with LEE011. Depletion of pRB^S780 was observed as early as 6 hours posttreatment in the BE2C and IMR5 cell lines, both of which respond to LEE011 with growth inhibition at nanomolar IC₅₀ values (data not shown). This effect was sustained at 96 hours, with depletion of pRB^S780 beginning at 250 nmol/L. Decreased pRB^S780 was also seen in the EBC1- and SKNAS-resistant cell lines, however only at higher inhibitor concentrations (Fig. 3B).

Analysis of the real-time substrate adherent growth curves generated by LEE011 treatment of neuroblastoma cell lines showed that growth inhibition in sensitive cell lines was consistent with a cytostatic effect (data not shown). However, because responses to targeted inhibition of cyclin-dependent kinase pathways are not always strictly a result of cell-cycle arrest (24–26), we sought to fully characterize the mechanism of neuroblastoma growth inhibition in response to pharmacologic CDK4/6 inhibition. LEE011 treatment of 2 neuroblastoma cell lines (BE2C and IMR5) with demonstrated sensitivity to CDK4/6 inhibition resulted in a dose-dependent accumulation of cells in the G₀/G₁ phase of the cell cycle (Fig. 4A). This G₀/G₁ arrest became significant at inhibitor concentrations of 100 nmol/L (P = 0.007) and 250 nmol/L (P = 0.01), respectively, and was also accompanied by dose-dependent decreases in the percentage of cells in S and G₂–M. As expected, cell lines that were resistant to CDK4/6 inhibition arrested in G₁ only at significantly higher doses of LEE011 (EBC1, 5 μmol/L, P = 0.01; SKNAS, no arrest achieved; Fig. 4A and B).

Recently, a systematic screen for novel CDK4/6 substrates identified the FOXM1 transcription factor as a potential target of CDK4/6 signaling, and implicated CDK4/6-mediated activation of FOXM1 in the prevention of cellular senescence (8, 25, 27–29). These results are corroborated by the fact that FOXM1 inhibition, either by deletion or by CDK4/6 inhibition, impairs the self-renewal capacity of cells (29). We therefore investigated whether or not inhibition of CDK4/6 activity by LEE011 would induce senescence in neuroblastoma via downregulation of FOXM1. There was a significant reduction in FOXM1 mRNA as early as 6 hours following administration of LEE011 to sensitive cell lines, and modest but reproducible decrease in FOXM1 protein levels (Fig. 4C and D). This was associated with the induction of cellular senescence in sensitive lines, as indicated by a significant increase in the percentage of SA-β-gal positive cells (Fig. 4E). By contrast, cell lines resistant to LEE011 showed no reduction of FOXM1 mRNA or protein following LEE011 treatment, and subsequently did not senesce. As we did not observe significant increases in caspase 3/7 activity or PARP cleavage in sensitive lines treated with LEE011 (Supplementary Fig. S5), these results suggest that the growth inhibition of neuroblastoma cell lines following CDK4/6 inhibition is primarily cytostatic and is mediated by a G₁ cell-cycle arrest and cellular senescence.

Given the observed differential sensitivity of neuroblastoma cell lines to CDK4/6 inhibition, we assayed for in vivo efficacy using neuroblastoma cell line–derived xenografts representing the extremes of in vitro sensitivity. CB17 immunodeficient mice bearing BE2C, NB-1643 (MYCN amplified, sensitive in vitro), or EBC1 (nonamplified, resistant in vitro) xenografts were treated once daily for 21 days with LEE011 or with a vehicle control. This dosing strategy was well tolerated, as no weight loss or other signs of toxicity were observed in any of the xenograft models. As shown in Fig. 5A and Supplementary Fig. S6, tumor growth was significantly delayed throughout the 21 days of treatment in mice harboring the BE2C or 1643 xenografts (both, P < 0.0001), although growth resumed posttreatment (data not shown). By contrast, as anticipated by the in vitro data, tumor growth suppression was less robust in the EBC1 xenograft model (P = 0.51). Assessment of the Ki67 proliferation marker by immunohistochemistry confirmed that proliferation was impaired only in the BE2C and 1643 xenograft models, as tumors resected from separate cohorts of BE2C or 1643 xenografted mice demonstrated comparatively weaker staining following 7 days of treatment with LEE011 than with the vehicle control, whereas no Ki67 staining differences were observed in the EBC1 xenografts (Fig. 5B). Phosphorylation of RB was also substantially diminished in the BE2C and 1643 xenografts, whereas only a minimal decrease was detected in the EBC1 model (Fig. 5B and C).

Cure rates for children with high-risk neuroblastoma have not significantly improved over the last decade, and of those children who do achieve remission, half will ultimately suffer a relapse (1). Such unfavorable outcomes are due in part to the fact that the current treatment regimen does not sufficiently leverage the unique biological features of this heterogeneous disease. Indeed, although MYCN amplification is the most common genomic lesion in this disease, strategies to target this oncogene have not yet resulted in a clinical deliverable. In addition, although the discovery that 8% to 10% of neuroblastomas harbor somatic activating mutations in the ALK oncogene provides another therapeutic opportunity (30, 31), most neuroblastoma patients will not have a somatic ALK mutation that is actionable with a targeted inhibitor (32–35). Steps must therefore be taken to identify additional molecular abnormalities that drive neuroblastoma disease progression and to subsequently exploit them with targeted therapy.

The data presented here identify CDK4/6 inhibition as a viable therapeutic strategy in neuroblastoma, with selectivity for patients whose tumors harbor MYCN amplification. Specifically, we show that RB phosphorylation via CDK4/6 signaling is nearly ubiquitous in neuroblastoma cell lines and tumors and is likely the result of high expression of CDK4, CDK6, and CCND1 (ref. 15; Fig. 1), but there may be other as yet undiscovered mechanisms of CDK4/6 hyperactivation. However, despite the fact that CDK4/6 signaling is hyperactive in the majority of neuroblastoma cell lines, not all are sensitive to LEE011. Therefore, although the finding that a CDK4-amplified cell line (NGP) was highly sensitive to LEE011 may be clinically relevant, our data suggest that pRB, CDK4, or CDK6 status alone cannot be used to accurately predict a response to CDK4/6 inhibition. We instead show that sensitivity correlated significantly with MYCN amplification status. Indeed, cell lines displaying sensitivity to CDK4/6 inhibition by either siRNA-mediated depletion or LEE011 treatment were likely to be MYCN amplified (Figs. 2A and 3A) as well as harbor high MYCN mRNA and protein levels (Supplementary Fig. S3). Although MYC-induced replicative stress may be a contributing factor, the precise mechanism for this association is unknown. Nevertheless, the finding has important clinical ramifications, as CDK4/6 inhibition may provide an alternative therapy for the 40% of high-risk neuroblastoma cases harboring amplification at the MYCN locus. Future research, however, should focus on the discovery of additional biomarkers of sensitivity as a means to identify a sensitive patient population beyond MYCN or CDK4 amplification status.

Over the last decade, first-generation CDK inhibitors have been evaluated in clinical trials for the treatment of adult malignancies, and a number of second-generation CDK inhibitors are currently undergoing phase I and phase II testing (7, 36). No clinical trial, however, has been adapted for childhood malignancies. As we show that CDK4/6 inhibition induces a cytostatic as opposed to a cytotoxic effect on neuroblastoma growth, combination strategies with conventional cytotoxic agents that rely on S-phase DNA replication may be antagonistic (37), suggesting that CDK4/6 inhibition may be best placed in the post-chemotherapy maintenance phase of treatment (immunotherapy and retinoids). Novel–novel screens with other agents that do not rely on targeted DNA replication should therefore be explored in order to develop a combinatorial strategy that will maximally inhibit the growth of residual, chemo-resistant cells. Taken together, our data suggest that a subset of neuroblastomas are highly sensitive to CDK4/6 inhibition, and support the clinical development of LEE011 in this disease.

J.M. Maris has commercial research grant in Novartis. He is also a consultant/advisory board member of Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Rader, M. Russell, L.S. Hart, E.L. Carpenter, G. Caponigro, R.W. Schnepp, K.A. Cole, J.M. Maris

Development of methodology: J. Rader, M. Russell, M.S. Nakazawa, E.L. Carpenter, A.C. Wood, J.M. Maris

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Rader, M.S. Nakazawa, L.T. Belcastro, D. Martinez, S. Parasuraman, G. Caponigro, B.R. Pawel, J.M. Maris

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Rader, L.S. Hart, Y. Li, E.L. Carpenter, E.F. Attiyeh, S.J. Diskin, S. Kim, S. Parasuraman, A.C. Wood, J.M. Maris

Writing, review, and/or revision of the manuscript: J. Rader, M. Russell, L.S. Hart, M.S. Nakazawa, E.F. Attiyeh, S. Parasuraman, G. Caponigro, R.W. Schnepp, A.C. Wood, K.A. Cole, J.M. Maris

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Rader, D. Martinez, J.M. Maris

Study supervision: L.S. Hart, K.A. Cole, J.M. Maris

The authors thank investigators in the neuroblastoma Therapeutically Applicable Research to Generate Effective Treatments (TARGET) consortium (ocg.cancer.gov/programs/target/projects/neuroblastoma) and especially Dr. S. Asgarzadeh and Dr. R. Seeger for generation of the patient gene expression data.

This work was generously supported through research grants from the Cookies for Kids Cancer, Arms Wide Open, Rally, and Alex's Lemonade Stand Foundations. R.W. Schnepp was supported by NIH Grant No. T32CA009615.

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.