Science - USA (2022-04-08)

on gene expression and identified more than
26,000 independent cis-acting eQTLs and 990
trans-eQTLs outside the MHC locus. Summary
statistics for our cis-eQTLs can be found in
table S10 and are available for browsing from
http://www.onek1k.org/. We show that most of these
eQTLs have an allelic effect on gene expression,
which is largely cell typeÐspecific, yet replicate
in two independent cohorts. We identify ex-

amples of how genetic loci contribute to key

immune function pathways. Lineage-dynamic

analyses applied to B cells demonstrated ex-

pected changes in markers of B cell matura-

tion. They facilitated the identification of

dynamic eQTLs, many of which had not been

identified through our primary cis-eQTL anal-

ysis. By integrating scRNA-seq eQTL data

with autoimmune risk loci identified through

GWASs, we uncovered both the causal gene at

these loci and resolved the cells through which

these genes exert their pathogenetic effects.

Mendelian randomization and colocalization

analysis of our eQTLs and disease-associated

SNPs were performed, providing complemen-

tary insights into the relationship between

eQTLs and disease risk loci. The colocalization

analysis provides evidence that the same

Yazaret al.,Science 376 , eabf3041 (2022) 8 April 2022 11 of 14

Fig. 7. Dissection of autoimmune
disease loci using eQTL mapping
at single-cell resolution.
(A) Breakdown of cis-eGenes colocal-
ized with GWAS associations for MS
using Bayes factors. (B) Mendelian
randomization was used to establish
causation between overlapping GWAS
loci for MS and identified eSNPs.
Significant results were identified for
39 MS-related genes (FDR < 0.05),
with the 12 outside the MHC locus
(dashed box) displaying highly
cell typeÐspecific effects. Colored
symbols depict cell types. Differences
were identified between the direct
NHGRI-EBI GWAS Catalog ( 9 ) overlap
and Mendelian randomization analysis
for eGene and cell-specific profiles.
(C) The effect sizes of OneK1K eQTL
SNPs plotted against the allelic effects
from the MS GWAS forFCRL3,EAF2,
andMAST3in BIN,BMem, and CD8ET
cells, respectively, are displayed.
pvalues are from the heterogeneity in
dependent instruments (HEIDI) test.

BC

CD4NCCD4ETCD4SOX4CD8ETCD8NCCD8S100BNK NKRPlasmaBMem BINMonoCMonoNCDC

DC

MonoNC

MonoC

BIN

BMem

Plasma

NKR

NK

CD8S100B

CD8NC

CD8ET

CD4SOX4

CD4ET

CD4NC

Number of colocalized eGenes

0 100 200 300

A

−0.02 0 0.02

eQTL effect sizes

−0.09

−0.04

0

0.04

0.09

GWAS effect sizes

top cis−eQTL

cis−eQTL

p=8.49x10-7

eQTL effect sizes

GWAS effect sizes

−0.01 0 0.01

−0.11

−0.05

0

0.05

0.11 p=1.06x10-6

eQTL effect sizes

GWAS effect sizes

−0.05 0 0.05

−0.09

−0.04

0

0.04

0.09

p=7.11x10-8

FCRL3

PLEK

EAF2

HLA-F

HLA-G

HLA-A

ZNRD1

VARS2

HLA-C

HLA-B

MICA

DDX39B

LST1

AIF1

HSPA1B

C2

SKIV2L

HLA-DRB5

HLA-DRB1

HLA-DQA1

HLA-DQB1

HLA-DQA2

HLA-DQB2

HLA-DOB

PSMB9

HLA-DMA

HLA-DPA1

RPS18

AHI1

CD6

DDX6

CLEC2D

CLECL1

MAST3

MPV17L2

TYMP

MMEL1

1q23

2p14

3q13

6p22

6p22

6p22

6p22

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6p21

6q23

11q12

11q23

12p13

12p13

19p13

19p13

22q13

1p36

top cis−eQTL

cis−eQTL

top cis−eQTL

cis−eQTL

RESEARCH | RESEARCH ARTICLE