Cell Specific eQTL Analysis without Sorting Cells

Abstract

The functional consequences of trait associated SNPs are often investigated using expression quantitative trait locus (eQTL) mapping. While trait-associated variants may operate in a cell-type specific manner, eQTL datasets for such cell-types may not always be available. We performed a genome-environment interaction (GxE) meta-analysis on data from 5,683 samples to infer the cell type specificity of whole blood cis-eQTLs. We demonstrate that this method is able to predict neutrophil and lymphocyte specific cis-eQTLs and replicate these predictions in independent cell-type specific datasets. Finally, we show that SNPs associated with Crohn's disease preferentially affect gene expression within neutrophils, including the archetypal NOD2 locus.

Fig 1. Method overview.

I) Starting with a dataset that has cell count measurements, determine a set of probes that have a strong positive correlation to the cell count measurements. Calculate the correlation between these specific probes in the other datasets, and apply principal component analysis to combine them into a single proxy for the cell count measurement. II) Apply the prediction to other datasets lacking cell count measurements. III) Use the proxy as a covariate in a linear model with an interaction term in order to distinguish cell-type-mediated from non-cell-type-mediated eQTL effects.

Fig 2. Validation of neutrophil proxy.

There is a strong correlation between the neutrophil proxy and the actual neutrophil percentage measurements in the training dataset (EGCUT, Spearman R = 0.75). Validation of neutrophil prediction in the SHIP-TREND cohort shows a strong correlation (Spearman R = 0.81) between the neutrophil proxy and actual neutrophil percentage measurements in this dataset.

Fig 3. Validation of neutrophil and lymphoid specific cis-eQTLs in purified cell type eQTL datasets.

A) We validated the neutrophil- and lymphoid-mediated cis-eQTL effects in four purified cell type datasets from the lymphoid lineage (B-cells, CD4+ T-cells, CD8+ T-cells and lymphoblastoid cell lines) and in two datasets from the myeloid lineage (monocytes and neutrophils). Compared to generic cis-eQTLs, large effect sizes were observed for neutrophil-mediated cis-eQTLs in myeloid lineage cell types, and small effect sizes in the lymphoid datasets. Conversely, lymphoid-mediated cis-eQTL effects had large effect sizes specifically in the lymphoid lineage datasets, while having smaller effect sizes in myeloid lineage datasets. These results indicate that our method is able to reliably predict whether a specific cis-eQTL is mediated by cell type. B) Comparison between average gene expression levels between different purified cell type eQTL datasets shows that neutrophil mediated cis-eQTLs have, on average a lower expression in cell types derived from the lymphoid lineage, and a high expression in myeloid cell types, while the opposite is true for lymphocyte mediated cis-eQTLs.

Fig 4. Effect of sample size on power to detect cell type specific cis-eQTLs.

We systematically excluded datasets from our meta-analysis in order to determine the effect of sample size on our ability to detect significant interaction effects. The number of significant interaction effects was rapidly reduced when the sample size was decreased (the number of unique significant probes given a Bonferroni corrected P-value < 8.1 x 10^–6 is shown). In general, due to their low abundance in whole blood, lymphoid-mediated cis-eQTL effects are harder to detect than neutrophil-mediated cis-eQTL effects.