Cross-Study Homogeneity of Psoriasis Gene Expression in Skin across a Large Expression Range

Abstract

BackgroundIn psoriasis, only limited overlap between sets of genes identified as differentially expressed (psoriatic lesional vs. psoriatic non-lesional) was found using statistical and fold-change cut-offs. To provide a framework for utilizing prior psoriasis data sets we sought to understand the consistency of those sets.Methodology/Principal FindingsMicroarray expression profiling and qRT-PCR were used to characterize gene expression in PP and PN skin from psoriasis patients. cDNA (three new data sets) and cRNA hybridization (four existing data sets) data were compared using a common analysis pipeline. Agreement between data sets was assessed using varying qualitative and quantitative cut-offs to generate a DEG list in a source data set and then using other data sets to validate the list. Concordance increased from 67% across all probe sets to over 99% across more than 10,000 probe sets when statistical filters were employed. The fold-change behavior of individual genes tended to be consistent across the multiple data sets. We found that genes with <2-fold change values were quantitatively reproducible between pairs of data-sets. In a subset of transcripts with a role in inflammation changes detected by microarray were confirmed by qRT-PCR with high concordance. For transcripts with both PN and PP levels within the microarray dynamic range, microarray and qRT-PCR were quantitatively reproducible, including minimal fold-changes in IL13, TNFSF11, and TNFRSF11B and genes with >10-fold changes in either direction such as CHRM3, IL12B and IFNG.Conclusions/SignificanceGene expression changes in psoriatic lesions were consistent across different studies, despite differences in patient selection, sample handling, and microarray platforms but between-study comparisons showed stronger agreement within than between platforms. We could use cut-offs as low as log10(ratio) = 0.1 (fold-change = 1.26), generating larger gene lists that validate on independent data sets. The reproducibility of PP signatures across data sets suggests that different sample sets can be productively compared.

Gene Expression Profiles Normalized in Psoriatic Skin by Treatment with Brodalumab, a Human Anti–IL-17 Receptor Monoclonal Antibody

“…One portion of each biopsy was flash frozen in liquid nitrogen for RNA isolation as described previously (18). Briefly, while still frozen, biopsies were homogenized in lysis buffer with or without manual disruption and RNA was prepared using the mirVana miRNA Isolation kit (Applied Biosystems, Carlsbad, CA) with on-column DNase treatment (Qiagen, Valencia, CA).…”

“…A core set of genes differentially expressed between independently procured nonlesional and lesional skin was identified ("Core PsO") using 14 paired samples from the Asterand subset and 15 paired samples from the ZabaAmgen subset of the publically available microarray data set of paired lesional and nonlesional psoriasis samples (GSE41664) (18). A table of probe sets, log ratios, and p values was created from estimated between group differences in probe-set intensities provided by the AffyPLM:: threestep function using R version 2.15.2 (20,21).…”

A GRHL3-regulated repair pathway suppresses immune-mediated epidermal hyperplasia

“…We identified 4 independent gene expression datasets from clinical trials in psoriasis using anti-TNF (etanercept; 2 independent datasets), anti-IL-23 (guselkumab), or anti-IL-17R (brodalumab) (48)(49)(50)(51). Consistent with the gene expression results in Figure 3, in these clinical trials, GRHL3 was significantly upregulated in lesional versus nonlesional skin, and after treatment, this high level was significantly reduced to levels similar to the nonlesional controls in all but 1 etanercept trial, which still showed the correct trend (Supplemental Figure 6, A-D).…”

Computational Systems Biology of Psoriasis: Are We Ready for the Age of Omics and Systems Biomarkers?

“…Several studies have examined the psoriasis-related transcriptome by comparing lists of differentially expressed genes with inconsistent results (Bigler et al, 2013;Gudjonsson et al, 2010;Suarez-Farinas et al, 2010). The absence of agreement between these studies might be due to threshold affects (e.g., p-value, fold change, false discovery rate) on selection of over-and underexpressed genes and also parameter differences in analyzing microarrays (Pan et al, 2005).…”