Expression of a MYCN-interacting Isoform of the Tumor Suppressor BIN1 Is Reduced in Neuroblastomas with Unfavorable Biological Features¹

NB³ is the most common solid tumor of childhood and is responsible for 15% of childhood cancer-related deaths (1, 2). The hallmark of NB is clinical heterogeneity with some tumors regressing spontaneously, especially in infants, whereas others differentiate into benign ganglioneuromas with or without treatment. Unfortunately, the majority of patients presenting with NB are over 1 year of age with locally aggressive and/or disseminated disease that is rapidly progressive and resistant to therapy. These disparate tumor behaviors may in part be predicted by clinical variables such as age (3), extent of disease (4), and histopathologic classification (5); as well as by tumor-specific genetic characteristics such as amplification of MYCN (6, 7), TRK receptor expression (8), ploidy (9), loss of the distal short arm of chromosome 1 (10, 11), and unbalanced gain of 17q (12). Of these, MYCN amplification is the most robust predictor of disease behavior and the only specific genetic alteration used currently in the majority of risk-stratification schemas.

MYCN amplification occurs in ∼25% of primary NBs and is invariably accompanied by elevated MycN protein expression (13). It is strongly associated with the presence of metastatic disease (6) and independently predictive of death from tumor progression (7). The malignant phenotype MYCN amplification portends has been presumed to be a result of unencumbered cell proliferation driven by the transcriptional activity of deregulated MycN. Yet it is increasingly clear that deregulated Myc proteins (including MycN) are potent inducers of programmed cell death when expressed in the absence of mitogenic signals or in the presence of genotoxic stress (14). This apoptotic signal may be blocked by constitutive survival signals or rendered ineffectual by loss of apoptotic pathways, as has been shown frequently in malignant cells with deregulated Myc expression. We similarly hypothesize that loss of engagement or execution of this apoptotic signal is a requirement for neoplastic neuroblasts with deregulated MYCN to survive and maintain a proliferative state after genotoxic stress or in a mitogen-poor environment. The mechanisms whereby deregulated MYC primes cells for apoptosis remain poorly characterized but may involve direct transcriptional activation of proapoptotic genes such as BAX, as well as indirect activation of BIM and/or p14/ARF (15, 16). These critical pathways have not been fully characterized with deregulated MYC and are less well understood for MYCN.

BIN1 maps to the long arm of human chromosome 2 (2q14) and encodes multiple tissue-specific isoforms of a Myc-interacting adaptor protein implicated in tumor suppression and cell death processes in malignant human cells. BIN1 has homology to amphiphysin (a breast carcinoma-related autoantigen responsible for “stiff man” syndrome) and RVS167, a cell-cycle regulator in yeast (17, 18). This homology is restricted to the NH₂-terminal BAR domain (Fig. 1). Additionally, BIN1 contains a neuron-specific endocytic-function domain (NTS) encoded by exons 12A, 12B, 12C, and 12D; a MBD encoded by exons 13 and 14; and a COOH-terminal SH3 domain encoded by exons 15 and 16 (18, 19). Multiple ubiquitous and tissue-specific alternatively spliced isoforms of BIN1 have been identified. The ubiquitously expressed isoforms omit exon 10 (which is expressed only within muscle-derived tissues) and exons 12A–D (which encode a domain important in mediating synaptic vesicle function and are found mainly in brain; also called amphiphysin-II isoforms; Refs. 18, 20, 21, 22, 23). Alternate splicing also contributes to the transcription of isoforms omitting exon 13 predicted to abrogate direct Myc or MycN interactions, because it encodes approximately half of the MBD. This may be relevant to tumorigenesis, because Myc-interacting isoforms have been demonstrated to have tumor suppressor function, and may mediate cell differentiation and death decisions (17, 21, 24).

BIN1 isoforms, which include the MBD, have been shown to physically interact with Myc in vitro and in vivo (17, 21). BIN1 potently inhibits transformation when coexpressed with MYC and activated RAS in rat embryo fibroblasts (17). BIN1 expression is reduced or absent in many tumor-derived cell lines and primary tumors (including melanoma, breast cancer, and prostate cancer), and forced re-expression of BIN1 in these cell lines results in apoptosis and reduced clonogenicity (20, 25, 26). Together these data suggest that BIN1 may function as a tumor suppressor through inhibition or alteration of specific Myc functions. Because BIN1 is highly expressed in neural tissues, functionally interacts with Myc within a domain with 100% identity to MycN, and inhibits Myc-mediated transformation, we hypothesized that BIN1 might function as a NB suppressor through interactions with MycN. We have shown previously that BIN1 expression is reduced in NB cell lines with MYCN amplification compared with NB cell lines without MYCN amplification (27). This deficit was of functional significance as forced expression of full length BIN1 inhibited colony formation selectively in MYCN amplified cell lines through the induction of programmed cell death.

In the present study we have identified the BIN1 isoforms that are expressed in primary NBs. The expression level of BIN1 isoforms has been determined in 56 primary NBs diagnosed both clinically and through a urine catecholamine-screening program. We have correlated BIN1 expression with MYCN expression, and with clinical and biological features of the tumors (including INRG classification, INSS tumor stage, age at diagnosis, and MYCN gene status and expression level). The functional role of these isoforms was investigated after transient transfection into NB cell lines with and without amplification of the MYCN proto-oncogene, and initial mutation analyses were performed. We present data correlating reduced expression of MycN-interacting BIN1 isoforms with unfavorable features in primary NB including amplification of the MYCN proto-oncogene, INRG high-risk disease, age >1-year, and metastatic disease. This reduction in BIN1 expression may be of functional significance, as forced ectopic expression of these isoforms (containing an intact MBD but not those deleting this domain) potently and selectively inhibited colony formation in NB cell lines with MYCN amplification.

Fifty-six tumor samples were obtained from previously untreated patients with NB (Department of Pediatric Surgery, Kyushu University). The clinical features of the 56 NB patients are shown in Table 1. Ten tumors (18%) demonstrated amplification of the MYCN proto-oncogene by Southern blot methodology as described previously (28). Twenty-nine cases (52%) were detected through mass screening at 6 months of age. All of the studies were approved by the Institutional Review Board at The Children’s Hospital of Philadelphia. Tumors were snap-frozen at the time of surgery, and total RNA was extracted using Isogen LS (Nippon Gene, Osaka, Japan) according to the manufacturer’s instructions. Human fetal brain RNA (Clontech, Palo Alto, CA) was used as a control for expression analyses. Reverse-transcribed cDNA was generated from 1 μg of total RNA in a final volume of 20 μl using the First-Strand cDNA Synthesis kit with random hexamer priming according to the manufacturer’s instructions (Amersham Pharmacia, Piscataway, NJ).

Expression of total BIN1 (all of the isoforms together) and MYCN were determined using real-time PCR (TaqMan; PE Biosystems, Foster, CA). Primers and probes for BIN1 and MYCN were designed using Primer Express v2.0 (PE Biosystems). We targeted sequences within exon 9 and 11 to analyze total BIN1 expression, because this region shares no homology with amphiphysin. The forward BIN1 primer (5′-AAG GCC CAG CCC AGT GAC-3′) is complimentary to sequences at the boundary of exons 9 and 11, whereas the reverse BIN1 primer (5′-GAG CCA TCT GGA GGC GAA G-3′) is complimentary to sequences within exon 11. The BIN1 probe is carboxyfluorescein-5′-CGC GCC TGC AAA AGG GAA CAA GA-3′-carboxytetramethylrhodamine. These primers and probe were designed to quantitate all of the BIN1 isoforms except those containing exon 10 (spiking experiments using pCMV-BIN1 containing exon 10 confirmed that this assay did not detect exon 10 containing isoforms). Of note, no exon 10 containing isoforms were detected in any NB cell line or primary tumor analyzed previously using RT-PCR with primers flanking or within exon 10 (data not shown).

The MYCN forward primer (5′-GAC CAC AAG GCC CTC AGT ACC-3′) is complimentary to sequences within MYCN exon 2, whereas the reverse MYCN primer (5′-TGA CCA CGT CGA TTT CTT CCT-3′) is complimentary to sequences within exon 3. The MYCN probe (carboxyfluorescein-5′-CCG GAG AGG ACA CCC TGA GCG A-3′-carboxytetramethylrhodamine) hybridizes within exon 2. GAPDH was used as an internal control for these analyses. PCR primers and probe for GAPDH (TaqMan GAPDH Control Reagent) were purchased from ABI and are available from the manufacturer.

TaqMan PCR was carried out in a reaction volume of 25 μl. Each reaction mixture contained 0.1 pmol/μl fluorogenic probe, 0.2 pmol/μl of each primer, 1× TaqMan Universal PCR Mastermix (PE Biosystems), and 1.0 μl of cDNA. Each sample was analyzed at least in duplicate with good reproducibility. Thermal cycling was started with a 2-min incubation at 50°C, followed by a first denaturation step of 10 min at 95°C, and then 40 cycles of two-step PCR consisting of 95°C for 15 s and 60°C for 1 min. A quantitative analysis was performed using the ABI PRISM 7700 Sequence Detection System and software (PE Biosystems). For analysis of total BIN1 expression, pCMV-BIN1(−10) a plasmid encoding the ubiquitous BIN1 isoform that excludes exon 10, was serially diluted to establish the calibration curve. For MYCN and GAPDH, the cDNA from the NB cell line SMS-KAN was serially diluted to establish the calibration curve. The expression level of BIN1 and MYCN were defined as the ratio of the signal obtained for target gene (BIN1 or MYCN) divided by that for GAPDH, then normalized to a reference cDNA.

To determine the relative expression of individual BIN1 isoforms in primary NBs we designed PCR primers that span exons 11 to 14 and reliably amplify all of the alternatively spliced BIN1 isoforms. The forward primer (5′-TCC CCA AGT CCC CAT CTC AG-3′) is complimentary to sequences within exon 11 of BIN1, whereas the reverse primer (5′-CTC CAC GGT GCC ATT CAC AG-3′) is complimentary within exon 14. These primers were biotinylated at their 5′ ends. Assays using BIN1 primers flanking exons 9–14 as well as exons 11–16 demonstrated no isoforms not identified with the BIN1 exon 11–14 primers (data not shown). PCR was carried out in a final volume of 20 μl, containing 1 unit of Taq Gold DNA polymerase (PE Biosystems), 100 μm deoxynucleoside triphosphates, 1.5 mm MgCl₂, 0.4 μm of each primer, and 1 μl of cDNA. The samples were first denatured at 95°C for 10 min, followed by 30 cycles of 95°C for 45 s, 55°C for 45 s, and 72°C for 45 s. The final cycle was followed by a 5-min extension step at 72°C. Each PCR sample was electrophoresed through a nondenaturing 7% polyacrylamide gel, transferred to a nylon membrane (Hybond N+; Amersham Pharmacia), and immobilized by UV cross-linking. Chemiluminescent detection of biotin-labeled PCR products was performed using the Southern Light Detection System (Tropix, Bedford, MA) and captured on X-Omat-AR film (Eastman Kodak, Rochester, NY). Band images were captured with a Model 4910 high-performance CCD camera and analyzed using NIH Image v1.55 software. For each isoform, expression was calculated as the relative proportion of that isoform determined by densitometry using RT-PCR as above (intensity of specific isoform divided by the sum of the intensities of all present BIN1 isoforms) times the total BIN1 expression value for that sample determined by the real-time PCR assay.

The NB cell lines used have been described previously (29, 30, 31, 32). All of the cell lines were grown in RPMI 1640 (Life Technologies, Inc., Gaithersburg, MD) supplemented with 10% fetal bovine serum, 2 mm l-glutamine, 1% OPI media supplement (Sigma-Aldrich, USA) 100 units/ml of penicillin, and 100 μg/ml streptomycin. Tissue culture conditions were 37°C in a humidified atmosphere of 5% CO₂. The BIN1 expression vectors pCMV-BIN1, pCMV-BIN1(−10), pCMV-BIN1(−10 + 12A), and pCMV-BIN1(ΔMBD) have been described previously (17). The pCMV-BIN1(−10) vector contains the coding sequence of the most abundant BIN1 isoform (which excludes exons 10 and 12A-12D; GenBank accession no. U68485) in pcDNA3 (Invitrogen, Carlsbad, CA). Similarly, pCMV-BIN1(−10 + 12A) excludes exons 10 and 12B-D but includes exon 12A. pCMV-BIN1ΔMBD is an artificial construct encoding a Bin1 protein with a deletion in the MBD (amino acid residues 270–377). Empty vector pcDNA3 was used to control for transfection efficiency.

For clonogenic growth assays, NB cells were seeded into T25 culture flasks and grown to ∼60% confluence. Transfection with 8 μg of pCMV-BIN1, pCMV-BIN1(−10), pCMV-BIN1(−10 + 12A), pCMV-BIN1(ΔMBD), or pcDNA3 was performed using FuGENE 6 Transfection Reagent (Roche Molecular Biochemicals) at 2 mcl transfection reagent/μg plasmid DNA in serum-free medium according to the manufacturer’s recommendations. Approximately 1 × 10⁵ cells were plated into 35-mm wells, and selection for neomycin resistance using 0.5 mg/ml G418 was initiated 24 h later. Medium with G418 was changed every 2–3 days, and cell colonies were scored after 100% methanol fixation and 0.005% crystal violet staining after ∼21 days. Each colony formation assay was performed in triplicate and was repeated from at least three independent transfection reactions. Colony formation was expressed relative to colonies obtained with control vector transfection.

Constitutional DNA was not obtained routinely from patients in our study group. Therefore, BIN1 LOH studies were performed using a separate group of 67 paired primary NB and constitutional-tissue derived DNA samples and seven NB cell lines (KCN, CHP-904, CHP-901, LA-N-5, CHLA-118, CHLA-123, and CHLA-150) with matched constitutional DNA available from the Pediatric Oncology Group, The Children’s Hospital of Philadelphia Tumor bank and a generous gift from Dr. C. Patrick Reynolds. PCR was performed to amplify an ∼125 bp fragment containing a polymorphic (GT)_n microsatellite within intron 5 of the BIN1 locus to assess for LOH. The forward primer BIN1(GT)-F is 5′-CCA TTC CTT TCC TGT CCC TTT G-3′, and the reverse primer BIN1(GT)-R is 5′-CCT TCC ACA ATC ATA TAG CCA GCC-3′. PCR reactions were performed in 20-μl volumes using conditions as described above for the RT-PCR. Forward primers for each reaction were γ-³³P end-labeled using T4 polynucleotide kinase (New England Biolabs, Inc., Beverly, MA). PCR products were separated by electrophoresis through an 8% polyacrylamide denaturing sequencing gel (Burst-Pak; Owl Scientific, Inc., Woburn, MA), and analyzed by autoradiography and densitometry as described. The threshold for LOH was defined by an allelic intensity reduction of >60% for one allele, and all of the samples with LOH were analyzed in triplicate for confirmation.

To assess for homozygous deletion within the BIN1 locus a set of 46 unique NB cell lines demonstrating unique genotypes at three highly polymorphic microsatellite markers were used and have been described previously (33). The 46 cell lines, a normal DNA control, and a negative control containing no added template were amplified by PCR essentially as above using the BIN1(GT)-F and -R primer pair, and 14.4 ng template DNA. Twenty μl of reaction solution was analyzed by electrophoresis on an enhanced sensitivity gel system (Visigel; Stratagene, La Jolla, CA), with products detected by ethidium bromide staining.

Southern analysis to assess for genomic rearrangements within the BIN1 locus was performed on 21 NB cell lines (11 with MYCN amplification and 10 without) and 11 primary NBs (all with MYCN amplification). Ten μg of genomic DNA was restriction digested with PstI (Promega, Madison, WI), and fractionated on 0.6% agarose and transferred to Hybond-N+ membranes followed by UV cross-linking and hybridized with two BIN1 probes sequentially. Probes used were an ∼1.5 kb EcoRI fragment from pCMV-BIN1 including most of the coding region of BIN1, and a PCR-generated 1.2 kb fragment encompassing ∼0.8 kb of the BIN1 promoter region and exon 1 and intron 1 sequences (primer sequences available by request). On the basis of the organization of the BIN1 gene this latter probe is >20 kb from the coding sequence probe (18). Probe radiolabeling was performed with the Rediprime Random Labeling kit using Redivue [α-³²P]dCTP (Amersham Pharmacia). Hybridization conditions were as described previously (33), and detection was by autoradiography on Biomax-MS film (Eastman Kodak) at −80°C from 1 to 3 days.

Analyses correlating MYCN expression with clinical and biological variables, as well as results for colony formation assays, were performed using the two-tailed Student’s t test. BIN1 expression levels in the tumor samples were skewed (non-normal distribution) and, therefore, all of the statistical analyses for BIN1 were done on the natural logarithm of the expression level using two-sample t tests. Two-sample Wilcoxon rank-sum tests were used to confirm these results. Means and 95% confidence intervals were calculated based on log scale values transformed back to expression levels. The χ² test was used to test for significance in comparisons of BIN1 alternative splice isoforms and biological features.

All 56 of the primary NBs expressed detectable levels of MYCN RNA with a mean of 72.7 ± 290.9 units (range, 0.1–2034). As expected, expression levels for tumors with MYCN amplification (n = 10) were higher (325 ± 653; range, 0.1–2034) than for MYCN single-copy tumors (18 ± 31; range, 0.3–189; P < 0.01). For purposes of comparison, MYCN expression in human fetal brain cDNA was 35 units using this assay. MYCN expression from INRG high-risk NBs (n = 19) was 175 ± 490 and from low- or intermediate risk NBs was 20 ± 34 (P = 0.06). There was likewise a trend but not a statistically significant difference in MYCN expression comparing localized NB (INSS stages 1, 2, or 3) to metastatic NB (INSS stages 4 or 4S; 22 ± 37 and 157 ± 467; P = 0.09). Eleven NBs had MYCN expression levels greater than that of fetal brain (>35 units): 6 arising in children >1-year-old, and 5 arising in infants. Of the 6 high MYCN expressing NBs in children, all demonstrated MYCN amplification and were high risk using INRG criteria. In contrast, of the 5 highly MYCN expressing NBs arising in infants, none had MYCN amplification or was high-risk (P < 0.01), and 4 were stage 1 NBs detected through the screening program. This is consistent with prior reports that NBs in infants frequently have higher levels of MYCN expression in the absence of gene amplification or other high-risk NB features, possibly reflecting normal developmental regulation of the gene, rather than deregulated overexpression (34, 35).

Of the 56 patients, 27 were clinically diagnosed (17 after having had a negative urine catecholamine screen at 6 months of age, and 10 diagnosed before the 6-month screening), and 29 were diagnosed through the screening program. Nine (33%) of the 27 clinically diagnosed tumors had MYCN amplification compared with only 1 (3%) of the tumors diagnosed through the screening program, although MYCN expression was not statistically different between the two groups (132 ± 414 and 18 ± 21, respectively; P = 0.15). A similar low incidence of MYCN amplification in NBs detected during mass screening programs has been described in multiple prior studies as a result of ascertainment bias.

BIN1 isoforms were detected in all 56 of the NBs assayed. Values for total BIN1 expression ranged from 1.4 to 1304 units (mean 149 ± 275). BIN1 expression for fetal brain cDNA was 8.1 units with this assay. RT-PCR demonstrated four predominant BIN1 isoforms in NB: BIN1(−10−13), BIN1(−10), BIN1(−10 + 12A-13), and BIN1(−10 + 12A); Fig. 1. These same four isoforms were also detected in all 12 of the NB-derived cell lines assayed (Fig. 1 C; data not shown). Of interest, isoforms containing exons 12B-12D were not detectable in either primary NBs or human fetal brain cDNA, although they are expressed in murine brain cDNA (18) and are considered to be neural-tissue restricted in their expression. Nineteen primary NBs (34%) expressed all four of the BIN1 isoforms, whereas 37 NBs had undetectable expression (≤1.0 unit) of at least one isoform. Six tumors (11%) did not express BIN1(−10−13); 9 (16%) did not express BIN1(−10); 23 (41%) did not express BIN1(−10 + 12A-13); and 32 (57%) did not express BIN1(−10 + 12A). Expression of these isoforms was 89 ± 172 units (range, 0.6–717) for BIN1(−10−13); 37 ± 68 (range, 0.1–295 units) for BIN1(−10); 15 ± 38 units (range, 0–220) for BIN1(−10 + 12A-13); and 8 ± 31 units (range, 0–220) for BIN1(−10 + 12A). Thus, the BIN1(−10−13) and BIN1(−10) isoforms were the most commonly and highly expressed BIN1 isoforms in primary NBs.

Because of the skewed distribution of BIN1 expression values (as demonstrated in Fig. 2, box plots) the data were log transformed to achieve a normal distribution for statistical analyses. Expression of each identified BIN1 isoform was then correlated with specific clinical and biological features of NB (Table 2). Reduced expression of BIN1(−10) was correlated with unfavorable biological features of NB including the presence of MYCN amplification (P < 0.03), high-risk disease based on INRG classification (P < 0.01), and age at diagnosis >1 year (P < 0.04). There was a strong trend toward lower BIN1(−10) expression in metastatic NB compared with localized NB as well (P = 0.058). In addition, BIN1(−10) expression was significantly lower in NBs diagnosed clinically compared with those diagnosed through a screening program (P = 0.02). BIN1(−10) was not detectable (defined as expression ≤1 unit) in 9 primary NBs. Of these, 5 (56%) had MYCN amplification, and 7 (78%) were INRG high-risk tumors [compared with 5 of 47 (11%) with MYCN amplification and 12 of 47 (26%) high-risk in NBs with detectable BIN1(−10); P < 0.01]. Thus, reductions in the major MycN-interacting BIN1 isoform were demonstrated in NBs with multiple unfavorable biological features. An association of reduced BIN1(10–13) expression was apparent only for INRG high-risk versus low- or intermediate-risk disease (Table 2). Of the less abundant BIN1(−10 + 12A-13) and BIN1(−10 + 12A), there was a correlation between lower BIN1(−10 + 12A-13) levels and MYCN amplification, as well as INRG high-risk disease (P < 0.05).

We next assessed whether the pattern of alternative splicing could be correlated with biological features of NB. The least highly expressed isoforms, BIN1(−10 + 12A) and BIN1(−10 + 12A-13), which both contain the neural-tissue restricted exon 12A, were not detectable in 57% and 39% of the primary NBs examined, respectively, and were not the predominant BIN1 isoform in any tumor. BIN1(−10 + 12A) accounted for <20% of BIN1 in all of the samples, and BIN1(−10 + 12A-13) accounted for <20% in 40 of 56 NBs. No correlations between the proportion of these exon 12A containing BIN1 isoforms and biological features of NB were found. In contrast, there was more variability in the proportion of the abundantly expressed BIN1(−10) and BIN1(−10−13). BIN1(−10) was detectable in 47 of 56 NBs (84%) and ranged from 0 to 68% of all of the BIN1 isoforms present. BIN1(−10−13) was present in 50 of 56 NBs (89%) and accounted for a range from 10 and 90% of all of the BIN1 isoforms present. Fifteen of 30 (50%) NBs with >50% of BIN1 as BIN1(−10−13), were INRG high-risk tumors (7 with MYCN amplification), whereas only 4 of 26 (15%) tumors with <50% BIN1(−10−13)isoforms were high-risk (3 with MYCN amplification; P < 0.01) suggesting that a possible mechanism for down-regulation of MYCN-interacting BIN1 isoforms in unfavorable NBs is through alternate exon splicing to exclude exon 13 (encoding a portion of the MBD), in addition to transcriptional regulation. An increase in isoforms omitting exon 13 (with a reciprocal decrease in exon 13 containing isoforms through alternative splicing at this junction) could effectively limit MycN-Bin1 interactions through omission of the MBD of Bin1.

Despite having demonstrated reduced BIN1 expression in NBs with MYCN amplification (among other unfavorable biological features), we found no correlation between MYCN and BIN1 expression in the study cohort overall (correlation coefficient r = 0.14; P = 0.32). In fact, after excluding the 10 tumors with MYCN amplification (with marked deregulated overexpression) there was a moderate direct correlation between MYCN and BIN1 expression (r = 0.67; P < 0.001). This correlation between MYCN and BIN1 expression in single-copy tumors results from a strong correlation in tumors arising in children <1 year of age alone. In these cases (n = 33), BIN1 expression was correlated with MYCN expression (r = 0.68; P < 0.001), whereas in children >1 year of age (n = 13) there was no correlation (r = 0.32; P = 0.28). To exclude a direct transcriptional effect of MYCN on BIN1 expression we used the human neuroectoderm-derived cell line hTERT-RPE1 retrovirally expressing an inducible MYCN-ER construct. BIN1 expression was invariant in the presence or absence of induced MYCN activity refuting a direct transcriptional effect (data not shown).

Thus, BIN1 expression is higher in NBs with high levels of MYCN expression (in the absence of MYCN amplification) in infants but is reduced in NBs with unfavorable biological features such as MYCN amplification and deregulated expression (only 1 patient in our cohort had MYCN amplification and age <1 year so we could not assess the relationship between MYCN and BIN1 expression in this subset). Because MYCN is more highly expressed in the developing peripheral nervous system than in mature tissues, we hypothesize that these findings in infant NBs reflect higher levels of MYCN regulated in a developmentally appropriate manner with no impetus to inactivate BIN1 (which may also be developmentally regulated). In contrast, tumors with MYCN amplification have deregulated MYCN expression, and inactivation of BIN1 may provide a survival advantage by abrogating MycN-mediated apoptosis after genotoxic stressors.

As described previously, BIN1 is subject to tissue-specific splicing in normal tissues (18) as well as aberrant tumor-specific splicing (20). The NB-derived cell line SK-N-AS does not have amplification of the MYCN locus and expresses BIN1 at levels comparable with fetal brain, whereas IMR-5 has MYCN amplification and markedly reduced BIN1 expression (27). Although total BIN1 expression differs, both SK-N-AS and IMR-5 express all four of the isoforms of BIN1 seen in the majority of primary NBs (Fig. 1 C; data not shown).

To determine the functional effect of ectopic expression of BIN1 isoforms on these cell lines, transient transfection studies using pCMV-BIN1, pCMV-BIN1(−10), pCMV-BIN1(−10 + 12A), pCMV-BIN1ΔMBD, or pcDNA3 (transfection control) were performed, and in vitro colony formation was assessed. Forced expression of BIN1, BIN1(−10), or BIN1(−10 + 12A), each of which includes the unique BAR domain as well as an intact MBD, resulted in markedly reduced colony formation in IMR-5 but had no effect on colony formation of SK-N-AS (Fig. 3). Because we hypothesized that BIN1 acts as an adapter protein mediating MycN function, we repeated these assays with pCMV-BIN1 or pcDNA3 (transfection control) in two additional NB-derived cell lines with MYCN amplification, NGP and SMS-KAN. Similar marked reductions in colony formation (to 11% and 42% of control transfectants, respectively) were seen (P < 0.05). Additionally supporting a role for the requirement of direct Bin1-MycN interactions in inhibiting colony formation is the inability of pCMV-BIN1ΔMBD to similarly inhibit colony formation in either cell line tested (Fig. 3). It should be noted that pCMV-BIN1ΔMBD is an artificial BIN1 construct lacking the entire MBD through the exclusion of amino acid residues 270–377 resulting in an isoform with a larger internal deletion than the naturally occurring splice isoforms BIN1(−10−13) and BIN1(−10 + 12A-13), which omit amino acid residues 315–345 from the MBD.

In the present study we have correlated reduced BIN1 expression with MYCN amplification, INRG high-risk disease, and the presence of metastatic disease in primary NBs. As BIN1 maps to the long arm of human chromosome 2 (2q14; Ref. 36) that has been reported to be hemizygously deleted in up to 30% of primary NBs (37), we assessed for LOH within the BIN1 locus that might account for reduced BIN1 expression. A polymorphic dinucleotide repeat within intron 5 (defined in Ref. 25) was amplified from 74 paired samples representing tumor and constitutional DNA.

Sixty-seven samples were from primary tumor specimens, and an additional 7 were from NB-derived cell lines. Twenty-one of the 67 primary NBs (31%) and 6 of the 7 NB-derived cell lines (86%) had MYCN amplification. Forty-eight of the 74 total constitutional DNA samples tested (65%) were heterozygous at this intragenic polymorphism. Of these informative cases, 3 (6%) demonstrated LOH with reductions in allelic intensity of 61%, 63%, and 76%, and all were from primary tumor specimens. Two of these were stage 1, and 1 was stage 4 (none had MYCN amplification). Information on the DNA content (or ploidy) of these tumor specimens was not available, nor was RNA/cDNA available to assess expression of BIN1 isoforms in those samples demonstrating LOH. As 2 of the 3 primary NBs demonstrating LOH were stage 1, and there is an association with low-stage NB and near-triploidy, it is possible that in these cases the allelic imbalance seen could be accounted for by differences in whole chromosome number, i.e. trisomy for chromosome 2 rather than chromosome 2q deletion events.

We next assessed for homozygous deletion at the BIN1 locus in a panel of 46 unique NB-derived cell lines (a full listing of the panel of cell lines used is found in Ref 33). Homozygous deletion has been observed occasionally in cell lines or primary tumors from a variety of malignancies, and, although rare, their characterization has been instrumental in the identification of several tumor suppressor genes, including RB1, WT1, and CDKN2A (38, 39, 40). DNA was amplified via PCR using the same primers and conditions described above to amplify an intragenic BIN1 marker. A product of the predicted size was visible for each of the 46 NB cell lines tested but not in negative control lanes (data not shown).

To assess for genomic rearrangements or larger genomic deletions elsewhere within the BIN1 locus, Southern analysis was performed using DNA from 21 NB-derived cell lines and 11 primary NBs (all primary tumors had MYCN amplification and were INSS stage 3 or 4). Two BIN1 probes were sequentially hybridized, one encompassing a portion of the promoter region and exon 1, and a second including the majority of the BIN1 coding region (this latter probe is ∼20 kb from the former). All of the tumors and cell lines demonstrated the anticipated single hybridization band of ∼3.4 kb when hybridized with the BIN1 exon1/promoter probe, and multiple identical bands at ∼2.8 kb, 2.0 kb, 1.6 kb, 1.0 kb, and 0.8 kb when hybridized with BIN1 coding sequences (data not shown). No evidence for genomic rearrangement or deletion was obtained. Taken together, these experimental results suggest that BIN1 loss occurs in NB through epigenetic rather than genetic alterations.

We reported previously that expression of BIN1 is reduced in NB cell lines with MYCN amplification compared with cell lines with single-copy MYCN and hypothesized that BIN1 may function as a tumor suppressor through MycN interactions (20, 25, 26, 27). However, multiple alternatively spliced BIN1 transcripts have been described, including isoforms that demonstrate neural-tissue restricted expression (inclusion of exons 12A-D) or that lack the putative MBD (exon 13), and it is likely that these alterations have functional consequences. We have now determined the isoforms of BIN1 expressed in primary NBs, and have demonstrated that reduced expression of the predominant MycN-interacting BIN1 isoform [BIN1(−10)] is correlated with unfavorable biological features including amplification of the MYCN proto-oncogene, INRG high-risk disease, age >1 year, and the presence of metastases at diagnosis. Furthermore, we have shown that forced expression of BIN1 isoforms that contain the MBD potently inhibit colony formation in NB cell lines with MYCN amplification. No effect on colony formation was seen with forced expression of a MBD-deleted construct nor when BIN1 isoforms were expressed in NB cell lines without MYCN amplification. However, it will be necessary to confirm that deletion of that portion of the MBD encoded within exon 13 alone results in a similar inability to inhibit colony formation in MYCN-amplified NB cell lines, as the MBD deleted construct used in these experiments lacked the entire MBD. We hypothesize that BIN1 isoforms with an intact MBD interact with MYCN to induce apoptosis when MYCN remains forcibly overexpressed (as occurs in NBs with MYCN amplification). A similar finding of reduced BIN1 expression in MYC-overexpressing neural tumors has been reported previously. Gene expression profiling of 11 primary diffuse astrocytomas demonstrated 8 with MYC overexpression, and all of these had >50% reductions in BIN1 expression (41).

As shown in other tissues as well, the most abundantly expressed BIN1 isoforms in NB were BIN1(−10) and BIN1(−10−13), rather than those isoforms specific to neural tissues. These isoforms localize predominantly to the nuclear and cytosolic compartments, respectively. Differential cytolocalization may be responsible for governing the differential functional effects of these isoforms, particularly in mediating MycN interactions (21). Inclusion of the neural tissue-specific exons 12A-12D was rarely observed in NBs despite the genesis of this tumor within the peripheral nervous system. BIN1 transcripts containing exons 12B-12D were not detected, and isoforms containing exon 12A were in low abundance, which may reflect the primitive differentiation state of neuroblasts. The 12A-12D exons encode determinants necessary for synaptic vessicle endocytosis (42, 43), and these isoforms are located preferentially within the cytosol. It has been reported previously that aberrant splicing to include exon 12A in melanoma cells may be a mechanism of BIN1 inactivation, as this isoform was incapable of inhibiting Myc-mediated transformation or cell viability (20) in these cells, possibly through aberrant localization outside the nuclear compartment. In contrast, we have found that forced expression of BIN1(−10 + 12A) in the NB cell line IMR-5 did inhibit colony formation, although immunohistochemical studies to assess localization of the protein were not performed. In most neural cells exon 12A is not included in BIN1 in the absence of 12B-12D (which encode clathrin binding determinants); it may be that exon 12A is insufficient for cytosolic localization in neuroblasts.

Overall there was a positive correlation between BIN1 and MYCN expression levels in primary NBs without MYCN amplification, although there is no evidence that BIN1 is a MYCN (or MYC) target gene. Therefore, it is more striking that BIN1 is reduced significantly in NBs with MYCN amplification and deregulated overexpression. These findings may be explained by the large proportion of tumors in our cohort diagnosed during infancy through a mass screening program. BIN1 isoforms were more highly expressed in these tumors, as was MYCN (though in the absence of gene amplification). However, there is no evidence that MYCN expression is deregulated in these tumors but instead may be indicative of the normal temporal expression pattern for this gene within the developing peripheral nervous system. In contrast, high levels of MYCN persist in a deregulated fashion after MYCN amplification, and it is the inability to down-regulate MycN that generates an apoptosis signal after genotoxic stress or in the setting of limited mitogenic factors. In these cells, the inability to engage MYCN-primed apoptosis, as may occur through loss of MycN-Bin1 signaling, would provide a survival advantage (17, 21, 24, 27). Therefore, BIN1 loss of function would be selected for in-tumor cells with deregulated MYCN, but not necessarily in those cells with higher levels of MYCN that are normally regulated, consistent with our findings.

In the present study we have not explored the mechanistic aspects of Bin1 function, although we have shown previously that re-expression of Bin1 in MYCN-amplified cell lines with reduced endogenous expression results in an increase in programmed cell death (27). This apoptotic effect was more pronounced in the setting of genotoxic stress (low-dose doxorubicin exposure) or mitogen deprivation (serum-free culture conditions), and we hypothesize that Bin1 signaling plays a role in the engagement of MYCN-primed apoptosis when MycN expression is aberrantly regulated. Although the relative proportion of individual BIN1 isoforms generally trended together in tumor samples, suggesting transcriptional regulation, we provide evidence that altered splicing (to omit a MycN-interacting functional domain encoded partially within exon 13) may contribute to BIN1 loss of function. Fifteen of 19 high-risk NBs (79%) had >50% of their BIN1 transcripts as BIN1(−10−13) and a reciprocal decrease in the putative MycN-interacting BIN1(−10) isoform. That aberrant splicing may play a role in tumor suppressor inactivation has been suggested previously for DCC, FHIT, and BIN1 (20, 44, 45). We found no evidence for genomic rearrangements to account for loss of BIN1 expression, and only infrequently was there evidence for BIN1 deletion using LOH analysis (6%). This is in contrast to prostate cancer in which LOH within BIN1 is seen in 40% of tumors (25), but consistent with prior studies of breast carcinoma in which loss of BIN1 expression was frequent (60% of cases) in the absence of gross genomic rearrangements (26). The finding that reductions in BIN1 expression occur frequently in the absence of demonstrable genomic alterations suggests that epigenetic factors may play a role in regulating BIN1 expression.

We thank Dr. Garrett Brodeur for support and helpful discussion in the preparation of this manuscript. We also thank the Pediatric Oncology Group, The Children’s Hospital of Philadelphia Tumor Bank, and Dr. C. Patrick Reynolds for providing tumor, cell line, and/or constitutional DNA, as well as Dr. Dai Sakamuro for the generation and provision of BIN1 plasmid clones for these studies.