Performance assessment of copy number microarray platforms using a spike-in experiment.

MOTIVATION: Changes in the copy number of chromosomal DNA segments [copy number variants (CNVs)] have been implicated in human variation, heritable diseases and cancers. Microarray-based platforms are the current established technology of choice for studies reporting these discoveries and constitute the benchmark against which emergent sequence-based approaches will be evaluated. Research that depends on CNV analysis is rapidly increasing, and systematic platform assessments that distinguish strengths and weaknesses are needed to guide informed choice. RESULTS: We evaluated the sensitivity and specificity of six platforms, provided by four leading vendors, using a spike-in experiment. NimbleGen and Agilent platforms outperformed Illumina and Affymetrix in accuracy and precision of copy number dosage estimates. However, Illumina and Affymetrix algorithms that leverage single nucleotide polymorphism (SNP) information make up for this disadvantage and perform well at variant detection. Overall, the NimbleGen 2.1M platform outperformed others, but only with the use of an alternative data analysis pipeline to the one offered by the manufacturer. AVAILABILITY: The data is available from http://rafalab.jhsph.edu/cnvcomp/. CONTACT: pevsner@jhmi.edu; fspencer@jhmi.edu; rafa@jhu.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Absence of E-cadherin expression distinguishes noncohesive from cohesive pancreatic cancer.

PURPOSE: The role of E-cadherin in carcinogenesis is of great interest, but few studies have examined its relevance to pancreatic carcinoma. EXPERIMENTAL DESIGN: We evaluated E-cadherin protein expression by immunohistochemistry in pancreatobiliary cancers having a noncohesive histologic phenotype (21 undifferentiated adenocarcinomas and 7 signet ring carcinomas), comparing the results with pancreatic cancers having a cohesive phenotype (25 moderately differentiated and 14 poorly differentiated adenocarcinomas). RESULTS: Twenty of 21 undifferentiated cancers had complete absence of E-cadherin expression, as did two signet ring carcinomas. In contrast, cohesive cancers (n = 39) had E-cadherin labeling at the plasma membrane (P < 0.001). Subsets of cancers were also evaluated for beta-catenin expression. All of the cohesive lesions (n = 28) showed a membranous beta-catenin expression pattern, whereas noncohesive foci (n = 7) were characterized by either cytoplasmic labeling or complete absence of beta-catenin protein expression, suggestive of a deficient zonula adherens in noncohesive cancers. E-cadherin promoter hypermethylation was observed in an undifferentiated pancreatic cancer cell line, MiaPaCa-2, whereas two pancreatic cancer cell lines derived from differentiated lesions lacked any evidence of E-cadherin promoter methylation. No pattern of E-cadherin promoter methylation could be determined in three primary cancers having mixed histologic patterns (contained both cohesive and noncohesive foci). No somatic mutations in E-cadherin were identified in noncohesive pancreatic cancers having inactivated E-cadherin. CONCLUSIONS: Noncohesive pancreatic cancers were characterized by the loss of E-cadherin protein expression. Promoter hypermethylation is a possible mechanism of E-cadherin gene silencing in a subset of these cancers.

Tumor-specific down-regulation of the tumor necrosis factor-related apoptosis-inducing ligand decoy receptors DcR1 and DcR2 is associated with dense promoter hypermethylation.

TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) induces apoptosis in a large variety of cancer cells but not in most normal human cells. This feature makes TRAIL, a potential antitumor agent. TRAIL can bind to four different receptors, two pro-apoptotic death receptors (DRs), DR4 and DR5, and two antiapoptotic decoy receptors (DcRs), DcR1 and DcR2. Normal cells express all four of the receptors. The increased TRAIL sensitivity of tumor cells has been postulated to result from the lack of DcR expression. We studied the tumor-specific down-regulation of the TRAIL receptors DcR1 and DcR2, as well as DR4 and DR5, in a group of pediatric tumor cell lines [nine neuroblastoma and three peripheral primitive neuro-ectodermal tumors (PNETs)] and three cell lines from adult tumors. Lack of expression of DcR1 and DcR2 was widespread (13 of the 15 cell lines and 10 of 15, respectively), both in the adult tumor cell lines and in the pediatric tumor lines. DR4 and DR5 were expressed in 8 of 15 and 12 of 15 cell lines, respectively. To understand the tumor-specific down-regulation of the TRAIL receptors, the promoter regions were studied for possible methylation changes of their CpG islands. All normal tissues were completely unmethylated, whereas in the tumor cell lines, we found frequent hypermethylation of the promoter. For DcR1 and DcR2, we found dense hypermethylation in 9 (69%) of 13 and 9 (90%) of 10 of nonexpressing cell lines, respectively. DR4 and DR5 were methylated in 5 (71%) of 7 and 2 (67%) of 3 nonexpressing cell lines, respectively. Treatment with the demethylating agent 5-aza-2'deoxycytidine resulted in partial demethylation and restored mRNA expression. In addition, we performed mutation analysis of the death domains of DR4 and DR5 by sequencing exon 9. Mutations were not present in any of the neuroblastoma or PNET cell lines. A panel of 28 fresh neuroblastoma tumor samples also lacked expression of DcR1 and DcR2 in 85 and 74% of cases, respectively. Hypermethylation was observed in 6 (21%) of 28 for DcR1 and 7 (25%) of 28 for DcR2. DR4 and DR5 were both expressed in 22 of 28 tumors, and no promoter methylation was observed. These data suggest that hypermethylation of the promoters of DcR1 and DcR2 is important in the down-regulation of expression in neuroblastoma and other tumor types.