Nine NB cell lines were used for microarray analyses: five N-type (neuroblastic phenotype: IMR-32, LA1-55n, SHSY-5Y, SK-N-BE(2)-M17, and SH-IN [FSK-treated]), two S-type (Schwannian stromal phenotype: SH-EP and SMS-KCNs), and two I-type (intermediate phenotypes, with features of both N- and S-type cells: SH-IN and SK-N-BE(2)c-T38) cell lines. Twelve cell lines (SHSY-5Y, IMR-32, GOTO, SK-N-BE(2)-M17, LA1-55n, SMS-KCN-69n, SH-EP, LA1-5s, WSN, CA-2E, SK-N-SH310, and SK-N-AS) were used in Western blot analysis. Cell lines were grown in minimal essential medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 100 units/ml penicillin, and 100 µg/ml streptomycin in a humidified 5% CO2 incubator at 37°C. IMR-32 medium was further supplemented with 1 mM pyruvate and 0.075% NaHCO3.
RNA Extraction and cDNA Synthesis
The isolation of RNA from NB cell lines was carried out using the Qiagen RNAeasy kit (Qiagen, Valencia, CA) according to the manufacturer's recommended protocols. The isolated RNA was purified with DNase (RNase-Free DNase Set, Qiagen) to digest genomic DNA according to the manufacturer's instructions. The integrity of RNA was visually assessed by conventional agarose gel electrophoresis. mRNA was reverse-transcribed into cDNA using TaqMan Reverse Transcription Reagents (Applied Biosystems, Branchburg, NJ) according to the manufacturer's protocol. Parallel control reactions without the addition of reverse transcriptase were performed for each sample. The gene-specific reverse transcription-polymerase chain reaction (RTPCR) products were visualized with ethidium bromide staining after agarose gel electrophoresis.
Real-time Quantitative RT-PCR
Quantitative RT-PCR (QRT-PCR) analysis was performed using an ABI 7700 sequence detector (Applied Biosystems). For the amplification of Bcl6 and the internal control GAPDH, we used TaqMan probe labeled with 6-carboxyfluorescein from Applied Biosystems. The standard curves for Bcl6 and GAPDH were prepared using serial dilution of known concentrations of RNA (0.5,2.5,12.5, 62.5, 312.5, and 1562.5 pg) from Ramos cells (a Bcl6-overexpressing human Burkitt's lymphoma line), which was also reverse-transcribed and amplified in a similar manner as that of RNA from NB cell lines (previously mentioned). The quantity of Bcl6 mRNA was calculated after normalizing for the internal control GAPDH.
Nuclear protein (20 µg) from NB cell lines was resolved by 8% SDS-PAGE and transferred to polyvinylidene fluoride membrane (Invitrogen, Grand Island, NY). Membranes were blocked with 5% milk and immunoblotted with anti-Bcl6 polyclonal antibody (Santa Cruz Biotechnologies, Santa Cruz, CA) at a 1:1000 dilution. The blots were visualized by chemiluminescence using a horseradish peroxidase-linked anti-rabbit immunoglobulin G secondary antibody (GE Healthcare, Buchinghamshire, United Kingdom) diluted at 1:5000. Anti-cAMP response element-binding protein (Upstate Biotechnology, Lake Placid, NY) was used as a loading control. For experiments evaluating Bcl6 posttranslational processing, cells were pretreated for 18 hours with specific proteosome inhibitors MG132 (10 mM; Calbiochem, La Jolla, CA) dissolved in dimethylsulfoxide or bortezomib (50 nM; Velcade, Millenium, Boston, MA).
Oligonucleotide Microarray Hybridization and Analysis
Preparation of cRNA from total RNA, hybridization, scanning, and image analysis were performed according to the manufacturer's protocol and as previously described [23
]. We used commercially available oligonucleotide arrays (HG_U133A; Affymetrix, Santa Clara, CA) containing 22,283 probe sets. Each probe set typically contains 11 perfectly matched 25-base-long probes (PMs) as well as 11 mismatch probes (MMs) that differ from the PM probes at the central base. Using publicly available software [24
], we computed the average of the PM - MM differences for each probe set on each array after discarding the largest and smallest 20% of the PM - MM differences, and scaled the data for each array to give a mean of 1500 units. Data were then quantile-normalized to make the values of 99 evenly spaced quantiles (at 0.01, 0.02, …, 0.99) agree across all of the arrays.We log-transformed values using log [max(x
+ 50; 0) + 50]. Conservative average fold differences were computed based on the differences in log-transformed data.
Tissue Microarray Construction
Tissue specimens were obtained from the University of Michigan Department of Pathology archives under the approval of the Institutional Review Board. Tissue arrays were constructed using three 1.0-mm cores taken from 48 paraffin-embedded, formalin-fixed neuroblastic tumors (32 NBs, 10 intermixed and nodular ganglioneuroblastomas (GNBs), and 6 ganglioneuromas (GNs)), 5 normal adrenal glands, and 7 other neural crest-derived (melanoma, pheochromocytoma, paraganglioma, and schwannoma). The histologic classification of archived tumor specimens was available, but all specimens were blindly reviewed by a single pathologist (J.A.J.) and classified according to current standards using the International Neuroblastoma Pathology Committee's classification [10,25
]. Core sampling was performed to ensure that optimal representation of both neuroblastic and stromal components was present. Completed array blocks were cured at 37°C for at least 24 hours, then heated at 52°C for 15 minutes, and rapidly cooled in an ice water bath. Multiple 5-µm tissue sections were cut from each block, with every 10th section stained with hematoxylin/eosin formorphologic assessment and comparison to adjacent immunostained levels.
Protein Expression Analysis
Immunohistochemistry was performed using commercially available monoclonal antibodies and standard manufacturer's protocols on a Ventana automated stainer. Briefly, slides were incubated with anti-Bcl6 (PG-B6p; Dako Cytomation, Carpinteria, CA) at a 1:20 dilution after a 20-minute pretreatment in a high-pH buffer at 95°C. Signals were evaluated for tumoral distribution (neuroblastic cells vs Schwannian stromal cells), cellular distribution (nuclear vs cytoplasmic), and intensity (semiquantitatively graded as 0, 1+, 2+, or 3+). The neuroblastic component was defined as any or all of the following: neuroblasts, immature gangliocytoid cells, and mature ganglion cells; the stromal component was defined as spindled cells with scant clear to eosinophilic cytoplasm embedded in a loose connective tissue matrix resembling Schwann cells or perineurium. Specimens excluded from analysis included those tumor array cores with insufficient tissue or missing sections, those that were necrotic, and those with insufficient neuroblastic or stromal components for evaluation. For each evaluated specimen, the mean neuroblastic component staining, the mean stromal component staining, and the mean component staining difference were calculated for all three cores.
Tumor array specimens from 48 patients were evaluable and included in the analysis. Patient characteristics included tumor specimens from 23 females and 25 males, ranging in age from 0 to 16 years. The mean age at diagnosis was 3.0 years, and the median age was 2.5 years.Twenty-five tumor specimens were from stage I or II patients, 5 from stage III patients, and 17 from stage IV patients. There was one stage IVS patient included in this analysis. Length of available follow-up data ranged from 1 to 17 years from the time of diagnosis. Thirty-seven patients were alive at the time of the study with mean follow-up of 6.7 years. Of the 32 NB patients, 13 had favorable histologic diagnosis and 19 had unfavorable histologic diagnosis. Of the 10 GNB patients, 7 had favorable histologic diagnosis, 2 had unfavorable histologic diagnosis, and 1 patient had missing data precluding proper classification. The GN group included six patients, all with favorable histologic diagnosis.
The goal of the analysis was exploratory in nature. Clinical outcomes were correlated to the intensity of the expression and the presence of the expression in neuroblastic and stromal regions. For Bcl6 intensity analysis, the subject-level average was used in correlation with the outcomes. Subject-level average was computed by averaging the intensity of Bcl6 expression from cores on tissue microarray that were from the same subject. The intensity in the neuroblastic and stromal regions and the differential expression between neuroblastic and stromal regions were evaluated. For the analysis on the presence, Bcl6 expression was grouped into four categories based on the presence (+) or absence (-) of expression in S-type cells and N-type cells. The four possible groupings were as follows: S+/N+ (both stromal and neuroblastic cell Bcl6 expression), S+/N- (stromal cell expression only), S-/N+ (neuroblastic expression only), or S-/N- (both stromal and neuroblastic cells negative). Covariates of interest included: MYCN (amplified, indeterminate, and normal), clinical stage (I, II, III, and IV), and tumor histologic diagnosis (NB, GNB, or GN). Outcomes analysis included Bcl6 expression in relation to overall survival and time to recurrence from end of therapy (event-free survival). Spearman rank correlation coefficient was used to assess univariate association between Bcl6 expression intensity and covariates of interest, whereas Fisher exact test was used to assess univariate associations between four-group N and S regions presence of Bcl6 expression and covariates of interest. The Kaplan-Meier method and log-rank test were used to compare the homogeneity of survival rates between categories of Bcl6 expression, the intensity of Bcl6 expression, as well as discrete covariates. Time to recurrence was defined as the time when NB relapse was documented. Subjects who did not recur at last follow-up or who died for reasons other than NB recurrence were censored from the analysis. All statistical analyses were done using SAS v9.0 (Carey, North Carolina). A 2-tailed P value of 0.05 or less was considered to be statistically significant.