ABSTRACT: A causal role of mutations in genes encoding for multiple general transcription factors in neurodevelopmental disorders including autism suggested that alterations at the global level of gene expression regulation might also relate to disease risk in sporadic cases of autism. This premise can be tested by evaluating for global changes in the overall distribution of gene expression levels. For instance, in mice, we recently showed that variability in hippocampal-dependent behaviors was associated with variability in the pattern of the overall distribution of gene expression levels, as assessed by variance in the distribution of gene expression levels in the hippocampus. We hypothesized that a similar change in the variance in gene expression levels might be found in children with autism. Gene expression microarrays covering greater than 47,000 unique RNA transcripts were done on purified RNA from peripheral blood lymphocytes of children with autism (n=82) and controls (n=64). The variance in the distribution of gene expression levels from each microarray was compared between groups of children. Also tested was whether a risk factor for autism, increased paternal age, was associated with variance in the overall distribution of gene expression levels. A decrease in the variance in the distribution of gene expression levels in peripheral blood lymphocytes (PBL) was associated with the diagnosis of autism and a risk factor for autism, increased paternal age. Traditional approaches to microarray analysis of gene expression suggested a possible mechanism for decreased variance in gene expression. Gene expression pathways involved in transcriptional regulation were down-regulated in the blood of children with autism and children of older fathers. Thus, results from global and gene specific approaches to studying microarray data were complimentary and supported the hypothesis that alterations at the global level of gene expression regulation are related to autism and increased paternal age. Regulation of transcription, thus, represents a possible point of convergence for multiple etiologies of autism and other neurodevelopmental disorders. The study was designed to compare gene expression profiles in peripheral blood lymphocytes of children with autism (n=82) and controls(n=64). Expression profiling: Expression profiling was performed at Translational Genomics (TGen), a member of the NIMH Neuroscience Microarray Consortium. Total RNA was extracted from peripheral blood lymphocytes (PBL) within 30 minutes of the blood draw using the Qiagen Qiaquick kit (Germantown, MD). Isolated total RNA was double round amplified, cleaned, and biotin-labeled using Affymetrix’s GeneChip Two-Cycle Target Labeling kit (Santa Clara, CA) with a T7 promoter and Ambion’s MEGAscript T7 High Yield Transcription kit (Austin, TX) as per manufacturer’s protocol. Amplified and labeled cRNA was quantified on a spectrophotometer and run on a 1% TAE gel to check for an evenly distributed range of transcript sizes. Twenty micrograms of cRNA was fragmented to approximately 35-200bp by alkaline treatment (200 mM Tris-acetate, pH 8.2, 500 mM KOAc, 150 mM MgOAc) and run on a 1% TAE gell to verify fragmentation. Separate hybridization cocktails were made using 15 micrograms of fragmented cRNA from each sample as per Affymetrix’s protocol. Two hundred microliters (containing 10 micrograms of fragmented cRNA) of each cocktail was separately hybridized to an Affymetrix Human Genome U133 Plus 2.0 Array for 16h at 45 degree Celsius in the Hybridization Oven 640. The Affymetrix Human Genome Arrays measure the expression of over 47,000 transcripts and variants, including 38,500 characterized human genes. Arrays were washed on Affymetrix’s GeneChip Fluidics Station 450 using a primary streptavidin phycoerythrin (SAPE) stain, subsequent biotinylated antibody stain, and secondary SAPE stain. Arrays were scanned on Affymetrix’s GeneChip Scanner 3000 7G with AutoLoader. Scanned images obtained by the Affymetrix GeneChip Operating Software (GCOS) v1.2 were used to extract raw signal intensity values per probe set on the array. A scaling factor of 150 was used to normalize array signal intensity in MAS 5.0. Arrays were scanned over 1 day on 2 different machines. Arrays scanned on the same machine and in the same day were considered to be from the same scan batch. Rescanning of a limited number of samples indicated that there were no significant differences between machines, nonetheless, for all comparisons groups were balanced for the scan batch. Gene expression levels were not adjusted for possible batch effects as algorithms that attempt to adjust for batch effects also alter the gene expression distribution. When samples could not be prepped simultaneously they were balanced for group membership (autism vs. control). To statistically control for possible confounds related to scan batches in our analysis of gene expression variance, batch number was entered into an analysis of covariance. For traditional analysis of gene expression, experimental groups were balanced with respect to batch membership. Microarray data analysis: Affymetrix .cel files were imported into Affymetrix Expression Console version 1.1. Data was pre-processed and summarized by Microarray Analysis Suite (MAS) 5.0 and Robust Multiarray Analysis (RMA). For the analysis of gene expression distributions, MAS 5.0 was used because the algorithm does not alter the gene expression distribution, whereas, RMA utilizes quantile normalization of probes prior to summarization and, therefore, has the potential to remove group level differences in gene expression distributions. Because of the numerous advantages in its handling of noise in gene expression and background subtraction, RMA was used for traditional gene expression analyses looking for specific gene expression differences between groups. Because, we found group level differences in the distribution of gene expression levels between groups, for traditional gene expression analyses summarized gene expression levels were also quantile normalized after the summarization step. Quantile normalization adjusts all data sets such that they have identical distribution patterns. Probesets were then filtered for those that were called present in at least 50 out of the 146 subjects (n = 25,146 probesets). A p-value of .05 was used as a threshold for significance. A fold-change of 1.1 was used as a cut off for magnitude of change. All microarrays met manufacturers recommended quality control criteria. Present calls ranged from 37.4% to 49%, mean 43.7%, SD 2.7%. Actin 3’to5’ ratios ranged from .726 to 5.15, mean1.37, SD 0.5. There were no significant group level differences in quality control measures.
