Cyclin D1 and CDK4 Activity Contribute to the Undifferentiated Phenotype in Neuroblastoma

Neuroblastic tumors develop from progenitor cell types that originate from the neural crest. These tumors are classified into neuroblastoma and the more differentiated counter parts, ganglioneuroblastoma and ganglioneuroma. This histopathologic classification is based on morphologic features of the tumor. Undifferentiated neuroblastoma at one end of the spectrum contain mainly neuroblasts, whereas on the other end, ganglioneuroma consist of ganglion cells, Schwann cells, and stroma cells (1, 2). This spectrum of neuroblastic tumors is reflected in the development of nonmalignant neuroblasts that differentiate along neural crest cell lineages to neuronal ganglia and chromaffin cells (3–5). The Schwann cells in neuroblastic tumors are believed to be normal cells invading the tumor and not to originate from the malignant neuroblasts, although some researchers have disputed this (6–8). The molecular mechanism underlying the differentiation pattern and the differences between these neuroblastic tumor types have been subject of several studies. Growth signaling pathways and specific neurotrophins and their receptors have been found to determine differentiation patterns in nonmalignant neuroblasts and influence the differentiation state of neuroblastoma (9–11).

Recently, some papers functionally link neuronal differentiation to cell cycle regulation, which frequently involves the G₁ cell cycle entry point (12–15). This is a tightly controlled process by the D-type cyclins and their kinase partners CDK4 and CDK6. In the presence of cyclin D, these kinases phosphorylate the pRb protein, which then releases from the E2F transcription factor. Subsequent transcription of key regulator genes allows further progression of the cell cycle (16). In neuroblastic tumors, several oncogenetic events causing changes in cell cycle regulation have been described. Amplifications of CDK4 and one mutation in CDK6 that disrupts p16 binding have been identified (17–19). We have reported amplification of the Cyclin D1 gene in 5 of 203 neuroblastic tumors and a rearrangement in the 3′ untranslated region of the Cyclin D1 gene in one tumor (20). The sporadic genomic aberrations of CDK4, CDK6, and Cyclin D1, together with the findings of the very high Cyclin D1 expression levels, suggested that dysregulation of the G₁ entry checkpoint is an important cell cycle aberration in neuroblastoma.

In this paper, a detailed analysis of Cyclin D1 overexpression by Affymetrix profiling and immunohistochemical analysis of neuroblastic tumors reveals that the overexpression is specific for malignant neuroblasts. We show that silencing of Cyclin D1 and its kinase partner CDK4 causes an inhibition of the cyclin D1-pRb pathway, a G₁ cell cycle arrest and growth arrest. Moreover, we show a clear differentiation toward a neuronal cell type by immunofluoresence and Affymetrix Microarray analysis after Cyclin D1 and CDK4 silencing.

Patient samples. The neuroblastic tumor panel used for Affymetrix microarray analysis contains 88 neuroblastoma samples, 11 ganglioneuroblastoma samples, and 11 ganglioneuroma samples. All samples were derived from primary tumors of untreated patients. Material was obtained during surgery and immediately frozen in liquid nitrogen. N-Myc amplifications and 1p deletions were all determined using Southern blot analysis of tumor material and lymphocytes of the same patient.

Cell lines. Cell lines were cultured in DMEM supplemented with 10% fetal bovine serum, 20 mmol/L l-glutamine, 10 units/mL penicillin, and 10 μg/mL streptomycin. Cells were maintained at 37°C under 5% CO₂. For primary references of these cell lines, see Cheng et al. (21).

RNA isolation, Northern blotting, and Affymetrix microarray analysis. Total RNA of neuroblastoma tumors was extracted using Trizol reagent (Invitrogen) according to the manufacturer's protocol. RNA concentration was determined using the NanoDrop ND-1000, and quality was determined using the RNA 6000 Nano assay on the Agilent 2100 Bioanalyzer (Agilent Technologies). For Northern blotting, 15 μg of RNA were electrophoresed through a 1% agarose gel containing 6.7% formaldehyde and blotted on Hybond N membrane (Amersham) in 16.9 × SSC and 5.7% formaldehyde. The Cyclin D1 probe was generated by reverse transcription–PCR. We used the following primers: 5′-tcattgaacacttcctctcc-3′ and 5′-gtcacacttgatcactctgg-3′. Probes were sequence verified. Probes were ³²P labeled by random priming. We hybridized Northern blot filters for 16 h at 65°C in 0.5 mol/L Na₂HPO₄, 7% SDS, 1 mmol/L EDTA, and 50 μg/mL herring sperm DNA. Measurement of signal intensity was performed with a Fuji phosphor imager and Aida 2.41 software. Affymetrix microarray analysis, fragmentation of RNA, labeling, hybridization to HG-U133 Plus 2.0 microarrays, and scanning were carried out according to the manufacturer's protocol (Affymetrix, Inc.). The expression data were normalized with the MAS5.0 algorithm within the GCOS program of Affymetrix. Target intensity was set to 100 (α1 = 0.04 and α2 = 0.06). If more then one probe set was available for one gene, the probe set with the highest expression was selected, considered that the probe set was correctly located on the gene of interest. Public available Affymetrix expression data was taken from the National Cancer Institute (NCI) Gene Expression Omnibus database.⁴

Immunohistochemistry. Paraffin-embedded tumors were cut into 4-μm sections, mounted on aminoalkylsaline-coated glass slides, and dried overnight at 37°C. Sections were dewaxed in xylene and graded ethanol, after which they were fixed, and endogenous peroxidase was blocked in a 0.3% H₂O₂ solution in 100% methanol. Subsequently, the slides were rinsed thoroughly in distilled water and incubated in a 0.01 mol/L sodium citrate solution (pH 7.1) in an autoclave. After 10 min at 120°C, the sections were left to cool for at least 5 min. After rinsing in distilled water and PBS, the last step in pretreatment is 10-min incubation in a normal goat serum solution (10% in PBS). For detection of Cyclin D1, we used a mouse monoclonal DCS-6 (Neomarkers) as a primary antibody. Slides were incubated overnight at 4°C in a 1:1,000 dilution in a solution of 1% bovine serum albumin (BSA) in PBS (1% PBSA). Slides were then blocked with a postantibody blocking (Power Vision kit, ImmunoLogic) 1:1 diluted in PBS for 15 min, followed by a 30-min incubation with poly–horseradish peroxidase (HRP)–goat α mouse/rabbit IgG (Power Vision kit, ImmunoLogic) 1:1 diluted in PBS. Chromogen and substrate were 3,3′-diaminobenzidine (DAB) and peroxide (1% DAB and 1% peroxide in distilled water). A 2-min incubation time results in a brown precipitate in Cyclin D1–positive nuclei. Nuclear counterstaining was done with hematoxylin. After dewatering in graded ethanol and xylene, slides were coated with glass and evaluated independently by two observers. As a positive control, mantle cell lymphomas were used, and as negative controls, two Burkitt lymphoma samples were used.

RNA interference. The small interfering RNA (siRNA) oligonucleotides were synthesized by Eurogentec. Three different siRNAs were designed, targeting Cyclin D1 on nucleotides 855 to 875 (CD1-A), 345 to 365 (CD1-B), and 671 to 691 (CD1-C) according to Genbank accession NM_053056. The CDK4 siRNA is targeting nucleotides 1062 to 1082 according to Genbank accession NM_000075. The CDK6 siRNA is targeting nucleotides 1112 to 1132 according to Genbank accession NM_001259. A previously designed siRNA directed against green fluorescent protein (GFP) was used as negative control (sense sequence: GACCCGCGCCGAGGUGAAGTT). Neuroblastoma cell lines were cultured for 24 h in 6-cm plates and transfected with 5.5 μg siRNA using Lipofectamine according to manufacturers' protocol.

Transactivation assays. The following luciferase constructs were used in the transactivation assays: pGL3 TATAbasic-6xE2F (pGL3 containing a TATA box and six E2F binding sites was previous tested for E2F selectivity and was a kind gift of Prof. R. Bernards, Dutch Cancer Institute; ref. 22), Renilla luciferase vector under cytomegalovirus promoter (pRL-CMV). Cells were cultured for 24 h in six-well plates, and transfections were conducted using Lipofectamine 2000 according to manufacturers' protocol. pGL3TATAbasic-6xE2F (0.8 μg) vector was transfected, together with the 0.8 μg pRL-CMV vector and Cyclin D1 siRNA or GFP siRNA. Dual-luciferase assays were performed after 48 h using the Promega dual-luciferase reporter assay system. For each assay, three separate experiments were performed.

Western blotting. The neuroblastoma cell lines were harvested on ice and washed twice with PBS. Cells were lysated in a 20% glycerol, 4% SDS, 100 mmol/L Tris-HCl (pH 6.8) buffer. Protein was quantified with RC-DC protein assay (Bio-Rad). Loading was controlled by Bio-Rad Coomassie staining of a reference SDS-PAGE gel. Lysates were separated on a 10% or 5% SDS-PAGE gel and electroblotted on a transfer membrane (Millipore). Blocking and incubation were performed using standard procedures. DCS-6 mouse monoclonal Cyclin D1 (Neomarkers), CDK4 C22 rabbit polyclonal (Santa Cruz), CDK6 C21 rabbit polyclonal (Santa Cruz), pRb Ser⁷⁸⁰ rabbit polyclonal (Cell Signaling Technology), Cyclin A clone 25 mouse monoclonal (Novocastra), actin C2 mouse monoclonal (Santa Cruz), and tubulin (Boehringer) antibodies were used as primary antibodies. After incubation with a secondary sheep anti-mouse or anti-rabbit HRP-linked antibody (Amersham), proteins were visualized using an enhanced chemiluminescence detection kit (Amersham). Antibodies were stripped from the membrane using a 2% SDS, 100 mmol/L β-mercapto-ethanol, 62.5 mmol/L Tris-HCl (pH 6.7) buffer.

Cell counting. All cell counting experiments were performed in duplo. Cells were collected by trypsinization and diluted in 1 mL of medium. Two samples of 100 μL of this medium were diluted in 100 μL Triton X-100/sapponine and 10 mL isotone II, and duplo counting of these two samples was performed on a Beckman Coulter Counter.

Fluorescence-activated cell sorting analysis. Cells were grown for 24 h in 12-well plates and then transfected with siRNA as described before. At 48 or 72 h after transfection, cells were lysated in the wells with a 3.4 mmol/L trisodiumcitrate, 0.1% Triton X-100 solution containing 50 μg/μL propidium iodide. After 1-h incubation, DNA content of the nuclei was analyzed using a fluorescence-activated cell sorter. A total of 30,000 nuclei per sample was counted. The cell cycle distribution and apoptotic sub-G₁ fraction was determined using WinMDI version 2.8.

TUNEL assay. Apoptotic cells were detected using the in situ cell death detection kit from Roche applied science. Cells with green fluorescent nuclei were indicated as apoptotic cells. The apoptosis index was the number of apoptotic cells divided by the total number of cells. For each experiment, 500 cells were counted.

Fluorescence microscopy. Cells were grown on glass and transfected as described above. Cells were rinsed twice with PBS and fixed on glass using 4% paraformaldehide in PBS for 30 min. One more wash step was performed using PBS permeabilization of the cells using 0.05% Triton X-100 in PBS. After three more wash steps using 0.01% Triton X-100 in PBS, slides were blocked in 1% BSA, 0.01%Triton X-100, and PBS. This was followed by 15-min incubation with 0.1 μg phalloidin-TRITC labeled (Sigma-Aldrich) in 100 μL blocking solution. Slides were washed four times with 0.01% Triton X-100 in PBS and finally with H₂O. After drying, slides were mounted in Vectashield with 25 μg/mL propidium iodide.

Expression of G₁ phase–regulating cyclins and cyclin-dependent kinases. Affymetrix expression profiles of 110 neuroblastic tumors allowed us to analyze expression levels of G₁-regulating genes in vivo. First we compared our data with publicly available Affymetrix HG-U133-plus2 expression profiles of 353 normal samples and 2,047 tumor samples (see Materials and Methods for references). Of the G₁-regulating cyclins, the Cyclin D1 gene shows the most abundant expression (Fig. 1A). The average Cyclin D1 expression in neuroblastic tumors is six times higher compared with normal tissue (n = 184) and 16 times higher compared with normal central nervous system tissue series (n = 169). Also, compared with libraries of 18 common malignancies, the Cyclin D1 expression in neuroblastoma is 2.5-fold to 7.5-fold higher. Even tumor types with known high frequency of genetic aberrations of the Cyclin D1 11q13 locus, such as breast tumors and myeloma, have lower expression of Cyclin D1 compared with neuroblastoma. The partners of Cyclin D1, the G₁-regulating CDK4 and CDK6, show interesting expression patterns. CDK4 and CDK6 are both highly expressed in neuroblastoma compared with various normal tissues, and the CDK4 expression level is comparable with adult tumors. To further analyze the expression of Cyclin D1, CDK4, and CDK6 in neuroblastoma, we generated plots of neuroblastic tumors and cell lines. Figure 1B shows that the Cyclin D1 expression (log2 values shown) in cell lines is within range of the in vivo expression. In our neuroblastoma series, we could identify a weak correlation between Cyclin D1 expression levels and INSS tumor stage and a very significant correlation with the histologic classification of neuroblastic tumors (ganglioneuroma versus neuroblastoma unpaired Student's t test, P = 1.2 × 10⁻¹³). The plot of CDK4 in neuroblastoma tumors shows that CDK4 expression in neuroblastoma cell lines is comparable with tumors with high expression (Fig. 1C). The tumor panel has one sample and the cell line panel has two samples with extremely high expression of CDK4. We therefore screened the tumor and cell line panel for potential genomic amplification of CDK4. The tumor and cell lines with extremely high CDK4 expression are the tumors showing amplification of CDK4 (data not shown). The CDK4 expression shows a correlation with INSS tumor stage, loss of heterozygosity (LOH) of 1p, MYCN amplification, and, most significantly, survival as indicated by the Kaplan Meier curve (log-rank probability of 6.2 × 10⁻¹⁰). The correlation to survival is also significantly independent of histologic classification (log-rank probability of 3.0 × 10⁻⁷ in neuroblastoma only) or MYCN amplification (log-rank probability of 5.0 × 10⁻⁵ in non–N-Myc amplified neuroblastoma only). CDK6 shows a similar pattern to Cyclin D1 (Fig. 1D) with a very significant correlation with the histologic classification of neuroblastic tumors (ganglioneuroma versus neuroblastoma unpaired Student's t test, P = 8.0 × 10⁻¹⁷).

High Cyclin D1 expression is restricted to neuroblasts. The difference in Cyclin D1 expression in ganglioneuroma and neuroblastoma, found in Affymetrix expression profiles, could be the result of the influx of Schwann cells in the tumor and dilution of a possible Cyclin D1 signal. Therefore, we stained paraffin-embedded tumor sections from 63 neuroblastomas, 26 ganglioneuroblastomas, and 8 ganglioneuromas with a Cyclin D1 antibody. Tissue was considered positive if at least 10% of the tumor cells showed specific nuclear staining, strongly positive when in at least 20% of the stained nuclei the hematoxylin counter stain was not visible any more, and focal-positive when only a part of the tumor showed nuclear staining. For 24 tumors, both Affymetrix and protein data were available. There was a clear correlation between the Cyclin D1 mRNA expression by microarray and the protein expression data by immunohistochemistry (data not shown). In the complete panel of 97 tumor samples that were analyzed for Cyclin D1 expression by immunohistochemistry, there was a clear correlation between the histologic classification and the Cyclin D1 expression levels (Table 1). In ganglioneuroma, not only the Schwannian stroma cells have low expression of Cyclin D1 but also the ganglion cells have low expression of Cyclin D1 compared with neuroblasts (Fig. 2). High Cyclin D1 expression that does not leave any hematoxylin visible is restricted to neuroblasts. These results indicate that Cyclin D1 might play a role in maintaining the undifferentiated phenotype of neuroblasts.

Cyclin D1 and CDK4 RNA interference causes pRb pathway inhibition. To study the functional relevance of Cyclin D1 and its kinase partners CDK4 and CDK6 in neuroblastoma, we used RNA interference to silence these genes in neuroblastoma cell lines. We first tested Cyclin D1 siRNAs in the neuroblastoma cell line SK-N-BE, which has a very high expression of Cyclin D1. As a negative control, we used an siRNA targeting the coding region of GFP. From three different siRNAs targeting the coding region of Cyclin D1, the CD1C siRNA showed the best reduction of Cyclin D1 mRNA and protein (85%; Fig. 3A). From our panel of neuroblastoma cell lines, we selected three more cell lines with high expression of Cyclin D1. Transfection of the Cyclin D1 siRNA (CD1-C) in SJ-NB 6, SJ-NB 10, and SK-N-FI showed a 70%, 43%, and 89% reduction of Cyclin D1, respectively (Fig. 3B). The silencing of the Cyclin D1 gene lasted for 4 days. To determine whether silencing of Cyclin D1 causes an effect on the downstream pRb pathway, we used a phosphospecific antibody against CDK4/CDK6 phosphorylated pRb (Ser⁷⁸⁰). All four neuroblastoma cell lines transfected with the Cyclin D1 siRNA showed a decrease in Ser⁷⁸⁰ phosphorylated pRb (Fig. 3B). Hypophosphorylation of pRb causes inhibition of the E2F transcriptional activity. To show that the E2F transcriptional activity was indeed inhibited by the Cyclin D1 protein knockdown, we used a dual luciferase assay with a reporter construct containing six E2F binding sites. In the two tested neuroblastoma cell lines, SK-N-BE and SK-N-FI, a clear down-regulation of E2F transcriptional activity was shown in the Cyclin D1 siRNA-treated cells compared with the GFP siRNA-treated cells (Fig. 3D). E2F induces transcription of Cyclin A. Figure 3B shows that Cyclin A levels also decrease after Cyclin D1 silencing. To determine if silencing of the Cyclin D1 kinase partners results in the same downstream pathway effect, we also developed an effective CDK4 and CDK6 siRNA and transiently transfected the cell lines SK-N-BE and SK-N-FI. Western blot results show an effective knockdown of CDK4 in SK-N-BE and SK-N-FI 72 hours after transfection (Fig. 3C). The CDK4 silencing has no effect on Cyclin D1 protein levels. CDK4 silencing results in evident decrease in pRb phosphorylation and reduction of the E2F target protein Cyclin A. Also CDK6 was very efficiently silenced. However, CDK6 silencing hardly affected the Rb phosphorylation level, and also Cyclin A levels were not or hardly reduced (Fig. 3C). We conclude that knockdown of Cyclin D1 or its kinase partner CDK4 causes inhibition of the pRb-E2F pathway in neuroblastoma cell lines. CDK6 silencing hardly affects the pRb-E2F pathway.

G₁ cell cycle arrest and inhibition of cell growth after Cyclin D1 or CDK4 inhibition. The inhibition of the Cyclin D1-pRb-E2F pathway resulted in marked phenotypical effects in neuroblastoma cell lines. We performed fluorescence-activated cell sorting (FACS) analysis to evaluate the effect of Cyclin D1 inhibition on the cell cycle. The SK-N-FI, SK-N-BE, SJNB-6, and SJNB-10 cell lines were transfected with siRNA-targeting Cyclin D1 and GFP and harvested for FACS analysis after 48 hours. SK-N-BE shows a clear increase in G₀G₁ fraction and a decrease in G₂S fraction in the Cyclin D1 siRNA treatment compared with the GFP negative control (Fig. 4A). This results in a strong increase of the G₀G₁-G₂S ratio, which was also observed in the three other neuroblastoma cell lines tested (Fig. 4B). Also CDK4 inhibition by siRNA showed a strong G₁ arrest. In SK-N-BE, CDK4 silencing caused an even stronger G₁ arrest compared with Cyclin D1 silencing. This is probably due to a more effective knockdown. To analyze whether the G₁ cell cycle arrest results in growth inhibition, we performed growth assays in the SK-N-BE and SK-N-FI cell lines after Cyclin D1 silencing. Cells were trypsinized after 0, 24, 48, and 72 hours and counted with a Coulter counter. At the first 24 hours, no effect is seen on cell growth. At 48 and 72 hours, a significant cell growth reduction is seen in the two cell lines transfected with Cyclin D1 siRNA cells compared with the control cells (Fig. 4C). The silencing of CDK6 did not result in growth inhibition (data not shown).

Cyclin D1 and CDK4 silencing induces neuronal differentiation. The growth inhibition observed in Cyclin D1 siRNA-treated cells can be caused by a growth arrest or cell death. In Cyclin D1 siRNA-treated neuroblastoma cell lines, we did not observe an increased number of detached cells suggesting that no increase in apoptosis has occurred. FACS analysis did not show an increase of the sub-G₁ fraction at 48 and 72 hours after transfection of Cyclin D1 or CDK4 siRNA in cell lines SK-N-BE, SK-N-FI, SJNB-6, and SJNB-10. Also a TUNEL assay performed at 48 hours after Cyclin D1 knockdown showed no increase in the apoptotic index in any of the cell lines. However, analysis of neuroblastoma cells after Cyclin D1 or CDK4 silencing revealed a strong phenotype. Silencing of CDK6 did not induce this differentiated phenotype (data not shown). Cyclin D1 and CDK4 siRNA-treated neuroblastoma cells showed an extensive increase in the number of neurite extensions per cell and the length of those extensions compared with the negative control. In Fig. 4D, this neuronal differentiation is visualized by staining actin with TRITC-labeled phalloidin. This suggests that neuroblasts differentiate toward a neuronal phenotype after inhibition of the G₁ checkpoint.

Cyclin D1 and CDK4 silencing induces a neuronal differentiation pattern on mRNA profiling. To further analyze the characteristics of the differentiation pattern after Cyclin D1 or CDK4 silencing, we performed Affymetrix expression profiling. The SK-N-BE cell line was transfected with Cyclin D1 siRNA, CDK4 siRNA, and GFP siRNA as control. RNA was isolated at time point 0 and 48 hours after siRNA transfection and hybridized on Affymetrix HG-U133-plus2 microarrays. All assays were performed in triplo, and results were analyzed on MAS5 normalized data. Firstly, we excluded all genes without a present call based on the MAS5 algorithm. Secondly, we excluded all genes that were significantly regulated (unpaired t test, P < 0.01) between time points 0 and 48 hours in the GFP transfected cells to avoid selection of genes that are also regulated due to transfection procedures. We subsequently selected genes that were at least 2-fold and significantly (unpaired t test, P < 0.01) regulated both in the Cyclin D1 and CDK4 silenced cells compared with GFP siRNA-treated cells 48 hours after transfection. This resulted in a set of 129 regulated genes, of which 70 were up-regulated and 59 were down-regulated. These genes were grouped according to their function derived from the NCI Gene and OMIM databases (Table 2A and B). For 38 of the up-regulated genes, a functional description was given, and 11 of those genes are involved in neuronal processes as neuronal development, signaling, or neurotransmitter secretion. This indicates that these cells show a change in expression pattern toward a neuronal phenotype which reflects our observations. The majority of down-regulated genes are involved in cell cycle regulation or transcriptional regulation. The genes marked with an asterisk are well-established E2F target genes (23–27), which confirms the reliability of our results, because the Cyclin D1/CDK4 complex functions through phosphorylation of pRb and activation of the E2F transcription factor.

We report a very high mRNA expression of Cyclin D1 in neuroblastoma compared with all other normal tissue libraries and tumor libraries, including tumor libraries with known high frequencies of Cyclin D1 genetic aberrations. The protein expression correlates well with the mRNA expression levels, and Cyclin D1 is mainly located in the nucleus of malignant neuroblasts. CDK4 and CDK6, the kinase partners of Cyclin D1, are also highly expressed and correlate with unfavorable prognosis and histologic classification, respectively. These findings strongly suggest a role for G₁ entry checkpoint dysregulation in the etiology of neuroblastoma. Silencing of Cyclin D1 and CDK4 clearly leads to inactivation of the pRb pathway and E2F transcription. For CDK6, the effect on the pRb pathway is less outspoken. Probably CDK4 plays a more pronounced role in neuroblastoma cell lines compared with CDK6. This does not exclude a role for CDK6 in vivo.

Apart from cell cycle regulation, these G₁ entry checkpoint regulators have been linked to other signal transduction routes. The apoptotic response which has been reported after Cyclin D1 silencing by other authors (28, 29) was not found after Cyclin D1 or CDK4 silencing in neuroblastoma as shown by FACS, TUNNEL assay, and microarray expression analysis. However, we report a change into a more differentiated phenotype after Cyclin D1 and CDK4 inhibition. The involvement of Cyclin D1 in neuronal differentiation processes has been suggested previously (12, 30, 31). In neuroblastoma, several authors linked Cyclin D1 nuclear overexpression to neuronal differentiation, but only as a downstream effect of other regulating genes (14, 15). We now show that Cyclin D1 overexpression itself is a driving event that prevents differentiation in neuroblastoma. First, we show a very high expression of Cyclin D1 in neuroblasts and low Cyclin D1 levels in differentiated ganglion cells in vivo. In vitro we show neuronal outgrowth after Cyclin D1 or CDK4 silencing and induction of a neuronal expression signature by Affymetrix microarray analysis. This is not in contrast with the findings that growth signaling pathways determine differentiation patterns in nonmalignant neuroblasts and influence the differentiation state of neuroblastoma. These signal transduction routes most frequently involve the transcriptional regulation of Cyclin D1, and thus, the effect on neuronal differentiation by these signal transduction routes could partly function through Cyclin D1 regulation (32).

The signal transduction routes that cause the distinct phenotypical effects (cell cycle arrest and neuronal differentiation) on neuroblastoma could both involve E2F activity. Many of the cell cycle progression genes that show up in the microarray profile are E2F target genes. The down-regulation of the cell cycle inhibitor CDKN2C seems in conflict with the cell cycle arrest, but again, this gene is an established E2F target. The phenotypic effect of neuronal differentiation could involve the strongly regulated MELK, which is an established E2F target and a crucial player in maintaining an undifferentiated phenotype in neuronal progenitor cells (27, 33, 34). In our neuroblastoma panel, the Affymetrix expression of MELK is strongly correlated to histologic classification and prognosis (data not shown). Further studies have to identify the exact role of this gene in the pathogenesis of neuroblastoma. The expression profiling after Cyclin D1 and CDK4 silencing also allows us to identify new downstream players involved in neuronal differentiation in neuroblastoma. Zfp423 is a newly identified transcription factor controlling proliferation and differentiation of neural precursors in cerebellar vermis formation. MYT1 is a transcription factor that has a role in the regeneration of oligodendrocyte lineage cells in response to demyelination (35, 36).

We conclude that the high-expression Cyclin D1 and CDK4 that correlates with histologic subtypes and prognosis together with the previous reported genetic aberrations indicate that these G₁-regulating genes are crucial players in neuroblastoma tumorigenesis. The resulting E2F transcriptional activity seems to regulate cell cycle, as well as neuronal differentiation signal transduction.

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.