Genetic analysis of interferon induced thyroiditis (IIT): evidence for a key role for MHC and apoptosis related genes and pathways

Abstract

Autoimmune thyroid diseases (AITD) have become increasingly recognized as a complication of interferon-alpha (IFNα) therapy in patients with chronic Hepatitis C virus (HCV) infection. Interferon-induced thyroiditis (IIT) can manifest as clinical thyroiditis in approximately 15% of HCV patients receiving IFNα and subclinical thyroiditis in up to 40% of patients, possibly resulting in either dose reduction or discontinuation of IFNα treatment. However, the exact mechanisms that lead to the development of IIT are unknown and may include IFNα-mediated immune-recruitment as well as direct toxic effects on thyroid follicular cells. We hypothesized that IIT develops in genetically predisposed individuals whose threshold for developing thyroiditis is lowered by IFNα. Therefore, our aim was to identify the susceptibility genes for IIT. We used a genomic convergence approach combining genetic association data with transcriptome analysis of genes upregulated by IFNα. Integrating results of genetic association, transcriptome data, pathway, and haplotype analyses enabled the identification of 3 putative loci, SP100/110/140 (2q37.1), HLA (6p21.3), and TAP1 (6p21.3) that may be involved in the pathogenesis of IIT. Immune-regulation and apoptosis emerged as the predominant mechanisms underlying the etiology of IIT.

Keywords: Autoimmunity; Interferon; Thyroid; Thyroiditis.

Published by Elsevier Ltd.

Figures

Figure 1

Principal Component Analysis (PCA) was performed to extrapolate ethnicity based on genotype information of our study population. Although our two HCV datasets were ethnically heterogeneous, both IIT cases (dots) and HCV controls (crosses) showed similar distributions and while they did not cluster at any one corner of the graph (as would happen in an ethnically homogenous sample), their compatibility enabled us to compare them in our association analyses. Moreover, the PCA showed that the lambda score, or genomic inflation factor, was 1 which implies that the data is not confounded by effects of population stratification.

Figure 2

Quantile-quantile (QQ) plots for genetic association analysis comparing all IIT cases (n = 53) to HCV controls (n = 190) that passed quality controls. The expected values (x-axis), from a theoretical x2 distribution, are plotted against the observed p-values (y-axis) for each SNP. If no SNP showed association all points would remain on or very near to the baseline (dotted line) which represents the null distribution. Panel A shows results of association analysis for all SNPs that were tested on the immunochip. The QQ plot demonstrates deviation from the null towards the upper extreme end of the line implying that multiple SNPs show association with IIT when compared to controls. Panel B shows QQ plot after removal of all SNPs in the HLA region. After removing the HLA SNPs the QQ plot still demonstrates deviation from the null towards the upper extreme end of the line suggesting that non-HLA genes are also likely to be associated with IIT.

Figure 3

Manhattan plot of immunochip association analysis data comparing all IIT cases (n = 53) to HCV controls (n = 190). 182,832 SNPs that passed the quality control testing are sorted by chromosomal location [x- axis] and are plotted against the −log10(p value) [y-axis], with the height of each point corresponding to the strength of association with disease. The dotted line indicates a p-value < 1 × 10−3 (the threshold we chose as our cutoff; see text) and the SNPs that show association with IIT with p < 1 × 10−3 are displayed above this line.

Figure 4

Manhattan plot of immunochip association analysis data comparing only Caucasian IIT cases (n = 44) to healthy, Caucasian controls not infected with HCV (n = 851). 185,425 SNPs that passed the quality control testing are sorted by chromosomal location [x- axis] and are plotted against the −log10(p value) [y-axis], with the height of each point corresponding to the strength of association with disease. The dotted line indicates a p-value < 1 × 10−3 (the threshold we chose as our cutoff; see text), and the SNPs that show association with IIT with p < 1 × 10−3 are displayed above this line.

Figure 5

Ingenuity pathway analysis (IPA) of PBMC RNAseq expression data. Shown are 7 pathways that belonged to groups of pathways identified by genomic convergence to be both genetically associated with IIT and significantly upregulated by IFNα in PBMCs. The percentage of genes that were upregulated are shown in light gray and genes that were downregulated are shown in dark gray, with the −log(p-value) shown at the right side of each bar. Interferon signaling pathway was found to be the most significantly upregulated. Other groups of pathways found to be associated with IIT and upregulated by IFNα in PBMCs included pattern recognition receptor related, apoptosis, complement, NFκB signaling, IL-10 signaling, and IL-6 signaling pathways. Four of these groups of pathways were found to be strongly associated with IIT in both datasets (all IIT cases vs. HCV controls and Caucasian IIT cases vs. healthy, Caucasian controls). These include interferon signaling, apoptosis related pathways, NFκB signaling, and IL-10 signaling (underlined).