The genetic instruments for proteins.
Figure 1 presented the schematic diagram of the datasets and analyses. The pQTL summary-level statistics were curated from nine studies: 734 proteins from Zheng et al. (N total = 20,013, Study 1–5)[12, 8, 10, 7, 9, 11], 1,561 proteins from Pietzner et al. (N = 10,708, Study 6)[13], 1,769 proteins from Ferkingstad et al. (N = 35,559, Study 7)[14], 75 proteins from Folkersen et al. (N = 21,758, Study 8)[15], 1,573 proteins from Zhang et al. (N = 7,213 Study 9)[16]. After being matched with AF GWAS (Cases N = 60,620, Controls N = 970,216)[6] and selected, a single cis-SNP was obtained for each protein. Finally, genetic variants for 1,949 proteins were obtained as proposed instruments for two-sample MR (Supplementary Table 1).
The causal effect of proteins on atrial fibrillation.
Two-sample MR analysis
The causal effects of 1,949 plasma proteins on AF were systematically evaluated using the two-sample MR analysis (Supplementary Table 2). The MR analysis yielded 30 causal protein-AF relationships at a Benjamini-Hochberg corrected threshold (PB-H adjusted < 0.05). In the sensitivity analyses using possible alternative variants, 22 protein-AF associations were validated (P < 0.05). Those who failed the sensitivity analysis indicated potential bias due to instrument selection (Supplementary Table 3).
Among these 30 proteins, 12 proteins were associated with increased risks of AF. Ordered by the effect size, they were: TES, CFL2, MTHFD1, RAB1A, DUSP13, SRL, ANXA4, NEO1, FKBP7, SPON1, LPA, and MANBA.
Other 18 proteins that decreased the risk of AF included: PMVK, UBE2F, SYT11, CHMP3, PFKM, FBP1, TNFSF12, CTSZ, QSOX2, EFEMP1, ALAD, FLRT2, LRIG1, OLA1, SH3BGRL3, IL6R, B3GNT8, and FCGR2A (Fig. 2).
Colocalization analysis of cis-pQTLs.
The MR analysis identified 30 proteins that had causal relationships with AF. To distinguish causality from LD, colocalization analyses were conducted. In the primary colocalization analysis, the summary statistics of SNPs surrounding the lead variants from Study 7 were utilized. There was no summary data available for LPA. Of the 29 cis-pQTLs with evidence from MR, the PPH4 in 13 proteins were༞0.5 (ALAD, B3GNT8, CFL2, CTSZ, DUSP13, FCGR2A, FLRT2, LRIG1, PFKM, SH3BGRL3, SPON1, SRL, TNFSF12), in 6 proteins were༞0.75 (CFL2, CTSZ, DUSP13, FLRT2, SPON1, TNFSF12), and in DUSP13 was༞0.95. The 14 cis-pQTLs that did not satisfy the assumption H4 may have distinct variants for the proteins and AF, implying they may be confounded by LD. As evidence, the PPH3 of these 14 proteins is greater than 0.95.
The secondary colocalization analysis used the SNPs from Study 9 and obtained similar results. There were no summary data available for CFL2 and LPA. The PPH4 in 11/28 cis-pQTL were > 0.5 (ALAD, B3GNT8, DUSP13, EFEMP1, FCGR2A, FLRT2, LRIG1, SH3BGRL3, SPON1, SRL, TNFSF12), in 4/28 cis-pQTL were > 0.75 (DUSP13, FLRT2, SPON1, TNFSF12), and were > 0.95 in DUSP13 and TNFSF12 (Fig. 3; Supplementary Table 4).
Drug target validation and repurposing.
Phenome-wide MR
To further explore the broader indications and concomitant side effects of the 30 proteins, phenome-wide MR (phe-MR) was performed. Phe-MR finding demonstrates the safety and efficacy outcomes relevant for the treatment of AF (Supplementary Table 5).
Most of the identified proteins possessed causal effects on established AF risk factors, including anthropometry traits (weight, height, body mass index, fat mass, waist or hip circumference), as well as metabolic traits (diastolic/systolic blood pressure, triglycerides, cholesterol, fasting glucose, fasting insulin, urate). In addition to these well-known risk factors, some proteins also affected other traits, such as basal metabolic rate, respiratory traits [forced expiratory volume in 1 s (FEV1) and forced vital capacity (FVC)], level of 25-hydroxyvitamin D, testosterone, and estradiol. Of note, genetically determined basal metabolic rate and respiratory traits have been reported to be linked to AF risk[21].
Furthermore, phe-MR suggested that some proteins were promising drug targets for other indications in the same direction as AF. Proteins affecting AF also mediate the risk of multiple cardiovascular and non-cardiovascular diseases. Genetic determined LPA, RAB1A, SPON1, SRL, and TES increased the risk of AF and other cardiovascular diseases. For instance, LPA disturbed blood lipid profile and increased the risk of coronary artery disease, ischemic heart disease, peripheral artery disease, and so on. In non-cardiovascular diseases, genetically determined ALAD, FCGR2A, and IL6R decreased rheumatoid arthritis risk. Genetic determined FCGP2A, EFEMP1, and IL6R decreased inflammatory bowel disease risk. As for cancer, FCGR2A decreased biliary tract cancer risk, FBP1 decreased ovarian cancer risk, LPA increased prostate cancer risk, and PMVK decreased malignant neoplasm of thyroid gland. Genetic determined ANXA4 increased breast cancer risk and SH3BGRL3 decreased its risk.
When considering proteins as targets for AF, safety also needs to be assessed. For instance, B3GNT8 decreased AF but increased prostate cancer risk. CHMP3 decreased AF but increased mood disorders risk, including neuroticism, irritability, and nervous feeling. IL6R increased risk of allergic diseases (asthma, hay fever, and eczema). These adverse side effects should be considered in gauging their preventive utility for AF.
Drug target database
MR analyses identified genetic associations between 30 proteins and AF. Given their potential as drug targets, the proteins were validated as targets already approved or investigated in ongoing clinical trials in drug target databases.
The records on past or present clinical drug development programs for 7/30 (23.3%) proteins have been identified. The 7 proteins included ALAD, CTSZ, FBP1, FCGR2A, IL6R, LPA, and TNFSF12. Their indications included acne vulgaris, epidermal dysplasias (ALAD), low bone mass disorder (CTSZ), type 2 diabetes mellitus (FBP1), thrombocytopenia (FCGR2A), rheumatoid arthritis (IL6R, TNFSF12), vasculitis (IL6R), solid tumor/cancer (LPA, TNFSF12).
Three proteins are targets for already approved drugs: aminolevulinic acid hci, an inhibitor of ALAD for acne vulgaris and actinic keratosis; SM-101, an antagonist of FCGR2A for idiopathic thrombocytopenic purpura; sarilumab and tocilizumab, modulators of IL6R for rheumatoid arthritis. None of them has been targeted for AF implicated in our MR results, but provide novel and promising target candidates for AF (Fig. 2; Supplementary Table 6).