Genetic regulation of gene expression of MIF family members in lung tissue

Abstract

Macrophage migration inhibitory factor (MIF) is a cytokine found to be associated with chronic obstructive pulmonary disease (COPD). However, there is no consensus on how MIF levels differ in COPD compared to control conditions and there are no reports on MIF expression in lung tissue. Here we studied gene expression of members of the MIF family MIF, D-Dopachrome Tautomerase (DDT) and DDT-like (DDTL) in a lung tissue dataset with 1087 subjects and identified single nucleotide polymorphisms (SNPs) regulating their gene expression. We found higher MIF and DDT expression in COPD patients compared to non-COPD subjects and found 71 SNPs significantly influencing gene expression of MIF and DDTL. Furthermore, the platform used to measure MIF (microarray or RNAseq) was found to influence the splice variants detected and subsequently the direction of the SNP effects on MIF expression. Among the SNPs found to regulate MIF expression, the major LD block identified was linked to rs5844572, a SNP previously found to be associated with lower diffusion capacity in COPD. This suggests that MIF may be contributing to the pathogenesis of COPD, as SNPs that influence MIF expression are also associated with symptoms of COPD. Our study shows that MIF levels are affected not only by disease but also by genetic diversity (i.e. SNPs). Since none of our significant eSNPs for MIF or DDTL have been described in GWAS for COPD or lung function, MIF expression in COPD patients is more likely a consequence of disease-related factors rather than a cause of the disease.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1

Schematic representation of the lung tissue dataset and methods used in our study. The total lung tissue dataset (n = 1087) was used for the eQTL analysis and a subset of COPD patients (n = 276) and matched non-COPD subjects (n = 236) from the same dataset was used for the gene expression analysis, comparing expression levels of MIF, DDT and DDTL. *DNA was isolated from blood samples in the Laval cohort and from lung tissue samples in the Groningen and British Columbia cohorts.

Figure 2

MIF, DDT and DDTL expression in lung tissue from COPD and non-COPD patients. Gene expression profiles for MIF (A), DDT (B) and DDTL (C) were obtained using a custom Affymetrix array (see GEO platform GPL10379), using 276 samples of COPD patients and 236 samples of non-COPD subjects, from the lung tissue database. Units of gene expression (y axis) represent Log2(microarray intensity) units. Data are presented as box and whiskers plots of the 5–95 percentile with median. Statistical differences were tested with Mann Whitney test.

Figure 3

eQTL analysis and main results. (A) Schematic representation of the methodology used for the eQTL analysis, subsequent analyses and their corresponding main results. (B) eQTL result for rs5751777. Effect of the rs5751777 genotype on MIF and DDTL expression levels in lung tissue samples from the lung tissue dataset (n = 1087). Data are presented as mean ± standard error of the mean.

Figure 4

MIF splice variants and binding site of Affymetrix MIF probe. (A) Graphic representation of MIF and its splice variants. (B) Sequence and binding site of Affymetrix probes for MIF.

Figure 5

Schematic representation of the airway wall biopsy dataset and methods used in our study. The airway wall biopsy dataset was used for the splice QTL analysis and cis-eQTL analysis in the same dataset. In the current study only results for rs5751777 are shown.

Figure 6

Effect of rs5751777 on expression of MIF splice variants and on total MIF. (A) SpliceQTL results. Split read counts mapping across exon–exon junction according to rs5751777 genotype. The number of split reads of a given junction pair was normalized per sample by correcting for variation in library size and transcript abundance in a gene-wise fashion. (B) Effect of rs5751777 on normalized MIF expression, represented as fragments per kilobase of exon model per million reads mapped (FPKM). Graphs are presented as mean ± standard error of the mean.