Gene expressions in liver cells associated with LDL-C and CAD
We carried out an SMR analysis to find gene expressions in liver cells that were associated with circulating LDL-C level and CAD risk by integrating eQTL data from the GTEx project with large-scale GWAS data. The mRNA expression levels of a total of 21,032 genes were analyzed, and expression levels of 15 genes (RHD, RHCE, ANGPTL3, CELSR2, PSRC1, SORT1, SYPL2, ATXN7L2, DNAH11, FADS3, ST3GAL4, NYNRIN, CETP, EFCAB13 and SPTLC3) in the liver were significantly associated with LDL-C (P < 5.0 × 10− 6) (Table 1, Supplemental Table S1). Ten of them pass the HEIDI test. These genes were enriched in gene ontology (GO) biological process of cellular lipid metabolic process (P = 5.0 × 10− 3).
Table 1
Liver cell gene expressions associated with circulating LDL-C level
Gene
|
CHR
|
Probe ID
|
Beta
|
SE
|
p_SMR
|
p_HEIDI
|
RHD
|
1
|
ENSG00000187010.14
|
-0.0260
|
0.0053
|
7.68E-07
|
6.51E-04
|
RHCE
|
1
|
ENSG00000188672.12
|
0.0515
|
0.0113
|
5.08E-06
|
8.17E-02
|
ANGPTL3
|
1
|
ENSG00000132855.4
|
0.1423
|
0.0305
|
3.05E-06
|
1.79E-01
|
CELSR2
|
1
|
ENSG00000143126.7
|
-0.1714
|
0.0179
|
8.39E-22
|
6.12E-02
|
PSRC1
|
1
|
ENSG00000134222.12
|
-0.1380
|
0.0128
|
2.88E-27
|
1.48E-01
|
SORT1
|
1
|
ENSG00000134243.7
|
-0.1420
|
0.0120
|
3.80E-32
|
9.62E-03
|
SYPL2
|
1
|
ENSG00000143028.7
|
-0.0335
|
0.0054
|
4.59E-10
|
7.39E-05
|
ATXN7L2
|
1
|
ENSG00000162650.11
|
-0.0510
|
0.0094
|
6.00E-08
|
1.58E-03
|
DNAH11
|
7
|
ENSG00000105877.13
|
0.0396
|
0.0082
|
1.45E-06
|
2.33E-01
|
FADS3
|
11
|
ENSG00000221968.4
|
0.1049
|
0.0223
|
2.44E-06
|
2.35E-01
|
ST3GAL4
|
11
|
ENSG00000110080.14
|
0.0870
|
0.0180
|
1.39E-06
|
6.85E-02
|
NYNRIN
|
14
|
ENSG00000205978.5
|
0.0672
|
0.0141
|
1.82E-06
|
1.00E-01
|
CETP
|
16
|
ENSG00000087237.6
|
0.1815
|
0.0398
|
4.99E-06
|
NA
|
EFCAB13
|
17
|
ENSG00000178852.11
|
0.0448
|
0.0083
|
5.83E-08
|
8.05E-01
|
SPTLC3
|
20
|
ENSG00000172296.8
|
-0.0542
|
0.0107
|
3.75E-07
|
3.81E-01
|
CHR: Chromosome; SE: Standard error; SMR: summary data–based Mendelian randomization; HEIDI: Heterogeneity in dependent instruments. |
The mRNA expression of 3 genes (CELSR2, PSRC1 and SORT1) in the liver were significantly associated with CAD risk (P < 5.0 × 10− 6) (Table 2, Supplemental Table S2). Therefore, the associations of mRNA expression levels of SORT1, PSRC1 and CELSR2 in liver with both LDL-C and CAD passed the threshold of 5.0 × 10− 8 (Fig. 2). Higher expression levels of SORT1, PSRC1 and CELSR2 in liver cells were associated with lower LDL-C levels (beta= -0.142, -0.138 and − 0.171, respectively) and CAD risk (beta= -0.10, -0.097 and − 0.121, respectively).
Table 2
Liver cell gene expressions associated with CAD risk
Gene
|
CHR
|
Probe ID
|
Beta
|
SE
|
p_SMR
|
p_HEIDI
|
CELSR2
|
1
|
ENSG00000143126.7
|
-0.1209
|
0.0172
|
2.35E-12
|
8.48E-01
|
PSRC1
|
1
|
ENSG00000134222.12
|
-0.0973
|
0.0131
|
9.46E-14
|
3.68E-02
|
SORT1
|
1
|
ENSG00000134243.7
|
-0.1001
|
0.0129
|
9.32E-15
|
3.32E-02
|
CHR: Chromosome; SE: Standard error; SMR: summary data–based Mendelian randomization; HEIDI: Heterogeneity in dependent instruments. |
Gene expressions in blood cells associated with LDL-C and CAD
By integrating LDL-C GWAS data with eQTL data, we found that expression levels of 31 genes were significantly associated with LDL-C level, including expression level of PSRC1 (Fig. 2), without significant heterogeneity (Supplemental Table S3). We then performed SMR analysis for CAD. By integrating CAD GWAS data with data from eQTL studies, we found that mRNA levels of 17 probes of 11 genes (PSRC1, IL6R, GGCX, VAMP8, LIPA, NT5C2, SWAP70, EIF2B2, FURIN, FES and ATP5G1) were significantly associated with CAD (Table 3), including expression level of PSRC1 (Fig. 2), without significant heterogeneity (PHEIDI > 0.05). These genes were enriched in GO biological process of myeloid leukocyte mediated immunity (P = 9.8 × 10− 4). We also found additional 88 genes that were nominally associated with CAD in blood cells (Supplemental Table S4).
Table 3
Blood cell gene expressions associated with CAD risk
Gene
|
CHR
|
Studies
|
Probe ID
|
Beta
|
SE
|
p_SMR
|
p_HEIDI
|
PSRC1
|
1
|
CAGE
|
ILMN_1671843
|
-0.2078
|
0.0242
|
8.53E-18
|
3.57E-01
|
PSRC1
|
1
|
CAGE
|
ILMN_2315964
|
-0.4152
|
0.0644
|
1.15E-10
|
6.57E-01
|
PSRC1
|
1
|
WhB
|
ENSG00000134222.12
|
-0.3423
|
0.0711
|
1.46E-06
|
1.68E-01
|
IL6R
|
1
|
Westra
|
ILMN_1696394
|
0.2204
|
0.0472
|
2.98E-06
|
4.09E-01
|
GGCX
|
2
|
WhB
|
ENSG00000115486.7
|
0.1279
|
0.0233
|
3.84E-08
|
6.91E-02
|
VAMP8
|
2
|
Westra
|
ILMN_2190084
|
-0.0763
|
0.0126
|
1.42E-09
|
5.90E-02
|
VAMP8
|
2
|
CAGE
|
ILMN_2190084
|
-0.0707
|
0.0117
|
1.68E-09
|
5.95E-02
|
LIPA
|
10
|
CAGE
|
ILMN_1718063
|
0.0657
|
0.0097
|
1.25E-11
|
9.51E-01
|
LIPA
|
10
|
WhB
|
ENSG00000107798.13
|
0.1386
|
0.0227
|
1.07E-09
|
9.09E-01
|
NT5C2
|
10
|
Westra
|
ILMN_1682165
|
-0.1418
|
0.0282
|
4.97E-07
|
1.84E-01
|
NT5C2
|
10
|
CAGE
|
ILMN_1682165
|
-0.1080
|
0.0219
|
8.14E-07
|
8.67E-01
|
SWAP70
|
11
|
Westra
|
ILMN_1785175
|
-0.1477
|
0.0287
|
2.74E-07
|
6.39E-01
|
SWAP70
|
11
|
CAGE
|
ILMN_1785175
|
-0.1058
|
0.0201
|
1.43E-07
|
7.81E-01
|
EIF2B2
|
14
|
CAGE
|
ILMN_1713380
|
0.0593
|
0.0129
|
3.94E-06
|
6.76E-01
|
FURIN
|
15
|
Westra
|
ILMN_1790228
|
-0.2324
|
0.0498
|
3.06E-06
|
2.40E-01
|
FES
|
15
|
Westra
|
ILMN_1693650
|
-0.1309
|
0.0247
|
1.11E-07
|
7.41E-02
|
ATP5G1
|
17
|
Westra
|
ILMN_1712430
|
0.1695
|
0.0363
|
3.03E-06
|
3.29E-01
|
CHR: Chromosome; SE: Standard error; SMR: summary data–based Mendelian randomization; HEIDI: Heterogeneity in dependent instruments. |
pQTLs in the identified genes
We selected 45 genes of interest according to the shared genes. In UCSC database we found 7,290 SNPs in these genes. In data from the KORA and INTERVAL pQTL studies we found 1,238 pQTLs (Supplemental table S5). Many of these pQTLs were significantly associated with LDL-C or CAD. For example, the pQTL rs599839 in PSRC1 was significantly associated with LDL-C (P = 2.75 × 10− 268) and CAD (P = 5.20 × 10− 17). This SNP was strongly associated with circulating levels of granulins in both of the two pQTL studies (beta = -0.7453 and 0.8071, P = 3.32 × 10− 51 and 1.00 × 10− 200, respectively). Indeed, twenty SNPs in PSRC1 were found to be associated with circulating level of granulins. Fourteen SNPs in APOB and five SNPs in PSRC1 were strongly associated with circulating level of apolipoprotein B. Among these SNPs, rs1367117 was associated with both LDL-C (P = 1.45 × 10− 282) and CAD (P = 1.10 × 10− 4); rs693 (P = 3.16 × 10− 5) and rs10199768 (P = 3.10 × 10− 3) were associated with LDL-C; rs2337383 was associated with CAD (P = 3.70 × 10− 4). A total of 81 SNPs in ERGIC3 were significantly associated with circulating level of copine 1, and 19 and 5 of them were associated with LDL-C and CAD, respectively. In addition, we found that the associations of some SNPs with circulating levels of granulins, apolipoprotein B and copine 1 were replicated in the two pQTL studies.
Proteins causally associated with CAD
We tested that if the three proteins, i.e., copine 1, granulins and apolipoprotein B, were genetically associated with CAD using several MR methods. The results were presented in Table 4. We first examined the associations between circulating levels of these proteins and LDL-C and found that the associations between circulating levels of granulins and apolipoprotein B and LDL-C were significant in every MR analyses using data from KORA and INTERVAL studies. In datasets available from the NMR_GWAS, we only found data for apolipoprotein B among the tested metabolites. By using this data, the association between circulating level of apolipoprotein B and LDL-C was validated. Circulating copine 1 was found to be significantly associated with LDL-C by using KORA data, but not in MR-PRESSO analysis using the INTERVAL data.
Table 4
The causation between circulating protein levels and LDL-C and CAD
Protein levels
|
Trait/
Disease
|
IVW/MR-Egger
|
MR-PRESSO
|
Study
|
Protein
|
Beta
|
SE
|
P value
|
Beta
|
SE
|
P value
|
1. KORA
|
Copine 1
|
LDL-C
|
0.022
|
0.005
|
1.08E-05
|
0.014
|
0.003
|
2.60E-03
|
2. INTERVAL
|
Copine 1
|
LDL-C
|
0.038
|
0.012
|
1.54E-03
|
0.015
|
0.011
|
0.213
|
1. KORA
|
Granulins
|
LDL-C
|
0.262
|
0.019
|
2.95E-43
|
0.123
|
0.027
|
4.49E-04
|
2. INTERVAL
|
Granulins
|
LDL-C
|
0.217
|
0.020
|
1.99E-27
|
0.127
|
0.022
|
3.87E-05
|
1. KORA
|
Apolipoprotein B
|
LDL-C
|
0.526
|
0.202
|
9.22E-03
|
0.261
|
0.113
|
4.37E-02
|
2. INTERVAL
|
Apolipoprotein B
|
LDL-C
|
1.401
|
0.229
|
4.89E-10
|
0.676
|
0.123
|
3.89E-08
|
3. NMR_GWAS
|
Apolipoprotein B
|
LDL-C
|
1.348
|
0.163
|
1.34E-16
|
1.023
|
0.049
|
1.40E-09
|
1. KORA
|
Copine 1
|
CAD
|
0.004
|
0.004
|
0.359
|
0.004
|
0.004
|
0.360
|
2. INTERVAL
|
Copine 1
|
CAD
|
0.013
|
0.015
|
0.378
|
0.002
|
0.008
|
0.821
|
1. KORA
|
Granulins
|
CAD
|
0.122
|
0.028
|
1.32E-05
|
0.052
|
0.015
|
1.57E-03
|
2. INTERVAL
|
Granulins
|
CAD
|
0.138
|
0.020
|
5.20E-12
|
0.079
|
0.016
|
4.60E-05
|
1. KORA
|
Apolipoprotein B
|
CAD
|
0.010
|
0.018
|
0.574
|
0.010
|
0.018
|
0.604
|
2. INTERVAL
|
Apolipoprotein B
|
CAD
|
0.289
|
0.105
|
5.92E-03
|
0.225
|
0.089
|
0.052
|
3. NMR_GWAS
|
Apolipoprotein B
|
CAD
|
0.373
|
0.093
|
6.05E-05
|
0.339
|
0.034
|
5.36E-09
|
CAD, Coronary artery disease; IVW, inverse-variance weighted; LDL-C, Low-density lipoprotein cholesterol; MR, Mendelian randomization; MR-PRESSO, MR pleiotropy residual sum and outlier; SE: Standard error. |
Because circulating levels of granulins and apolipoprotein B were significantly associated with LDL-C, we thought that these proteins may be risk factors of CAD. As expected, significant associations between granulins and apolipoprotein B levels and CAD were found (Table 4). These associations were validated in several MR analyses on data from two pQTL studies. The effect of protein granulins levels on CAD risk was relatively small, with an odds ratio (OR) of about 1.15 (95% confidence interval: 1.10–1.19). The effect of protein apolipoprotein B levels on CAD risk was larger than that of granulins, with an OR of about 1.45 (95% confidence interval: 1.21–1.74). As it is known, LDL-C level was causally associated with CAD. Therefore, granulins, apolipoprotein B, LDL-C and CAD were genetically correlated with each other.