3.1 Associations between plasma proteins and lung function
Following the exclusion of proteins lacking SNPs or characterized by weak instrumental variables (F-statistic < 20), a proteome-wide MR analysis includes a total of 1393 plasma proteins. Upon applying the Bonferroni correction, genetically predicted levels of 38, 28, and 37 proteins significantly correlate with the pulmonary function parameters FEV1, FVC, and FEV1/FVC, respectively (Fig. 2A-2C, Supplementary Table 1). For each standard deviation increase in genetically inferred protein concentrations, the beta coefficients for lung function demonstrate variability: for FEV1/FVC, a decrease of -0.20 (95% CI = -0.28 to -0.11) is observed for X-prolyl aminopeptidase 1 (XPNPEP1), and an increase of 0.41 (95% CI = 0.31 to 0.50) for vesicle trafficking 1 (VTA1) (Fig. 2D). In the case of FEV1, the beta varies from a decrease of -0.26 (95% CI = -0.34 to -0.17) for serine protease 8 (PRSS8) to an increase of 0.32 (95% CI = 0.23 to 0.41) for TNF receptor (RELT) (Fig. 2E). Similarly, for FVC, beta values range from a decrease of -0.28 (95% CI = -0.38 to -0.18) for cell adhesion molecule 1 (ANTXR1) to an increase of 0.25 (95% CI = 0.16 to 0.35) for complement factor H (CFH) (Fig. 2F). Additionally, it is notable that a causal relationship exists between the proteins NPNT, NUDT5, PDLIM4, PLAU, PPA2, RELT, RGMB, AKR7A2, GM2A, HHIP, HTRA1, and ITGA2 and the three indices of lung function (FEV1, FVC, and FEV1/FVC).
3.2 Sensitivity analysis for lung function-associated proteins
Among the proteins associated with lung functions, we conduct colocalization analyses between these proteins and lung function parameters (FEV1/FVC). Results indicate robust colocalization evidence for seven proteins, each with a posterior probability of hypothesis 4 (PH4) exceeding 0.8, including NPNT, SCARF2, FGFR1, ADAMTS5, MAP1LC3A, ARSB, and BOC. Additionally, two associations show moderate colocalization support with PH4 values between 0.8 and 0.5, specifically for GM2A and XPNPEP1 (Fig. 2G). In the colocalization analyses assessing causality with the lung function parameter FEV1, NPNT, SCARF2, GM2A, SFRP1, IVD, FN1, OGN, and DNAJB4 demonstrate substantial colocalization support, while ANTXR1, COL15A1, CD14, and DNAJB12 exhibit moderate support (Fig. 2H). For the lung function parameter FVC, DNER, RHOC, and GM2A reveal significant colocalization support, with ANTXR1 and DNAJB12 exhibiting moderate support (Fig. 2I). Regrettably, no protein passes the colocalization tests for all three lung function parameters. However, NPNT, GM2A, SCARF2, DNAJB4, and COL15A1 pass the tests for both FEV1/FVC and FEV1, while GM2A also passes for both FEV1/FVC and FVC (Supplementary Table 2).In the reverse MR analysis, after applying Bonferroni correction, no substantive evidence was found regarding the associations between genetic predisposition to three lung function parameters and the levels of related blood proteins. Additionally, Steiger filtering further ensured directionality (Supplementary Table 3 and Supplementary Table 4).
3.3 Replication analyses
For the FEV1/FVC parameter, instrumental variables (IVs) for 15 and 20 proteins are derived from the studies conducted by Zheng et al. and UKB-PPP, respectively. With these IVs, we successfully confirm the associations for all 15 proteins from the Zheng et al. study and for 19 proteins from the UKB-PPP, achieving confirmation rates of 100% and 95%, respectively (Fig. 2D, Supplementary Tables S5 and S6). For the FEV1 parameter, IVs for 14 and 18 proteins are obtained from Zheng et al. and UKB-PPP, respectively. We effectively confirm associations for 12 (85%) proteins utilizing IVs from the Zheng et al. study and for 17 (94%) proteins using IVs from the UKB-PPP. In terms of the FVC parameter, IVs for 10 and 7 proteins are sourced from Zheng et al. and UKB-PPP, respectively. We rigorously validate associations for 9 (90%) proteins using IVs from the Zheng et al. study and for all 7 (100%) proteins using IVs from the UKB-PPP (Fig. 3A-C, Supplementary Tables 5 and 6).
3.4 Impact of Significant Proteins on Pulmonary Function Related Diseases
Subsequent analyses investigating the relationships between proteins involved in lung function and pulmonary diseases—including COPD, Asthma, and ILD—demonstrated causative links (Supplementary Tables 7). In proteins associated with the lung function parameter FEV1/FVC, it was shown that the genetically inferred concentrations of NPNT, FGFR1, AGRP, SPINK7, DPEP2, CNTN2, HHIP, PPA2, and ARFIP1 had causal associations with COPD. Similarly, causative associations of NPNT, SPINK7, TYRO3, PPA2, VTA1, ARFIP1, DNAJB4, PDLIM4, ANXA11, and MET with Asthma were identified. Moreover, FGFR1 and AKR7A2 were confirmed to have causal relationships with ILD (Fig. 3D). Additionally, in proteins linked with the lung function parameter FEV1, NPNT, ANTXR1, LINGO1, SPINK7, IL17RD, PLXNB2, HHIP, DNAJB12, and LEAP2 were found to have causal associations with COPD; NPNT, SPINK7, ATXN3, PRSS8, VTA1, IL17RD, DNAJB4, IVD, and Asthma demonstrated causative links; and ITGA2 and COL15A1 had causal relationships with ILD (Fig. 3E). In proteins related to the lung function parameter FVC, NPNT, ATXN3, PRSS8, ASIP, and IVD were causally associated with Asthma; NPNT, ANTXR1, CFH, CFHR3, TRIL, HHIP, DNAJB12, and LEAP2 had causal relationships with COPD; however, no proteins were found to have causal relationships with ILD (Fig. 3F). Notably, all the aforementioned associations maintained consistency with the directionality observed in their respective lung function parameters.
3.5 Associations between lung function-associated proteins and BMI and smoking
Following MR analyses on the impact of smoking cessation, smok per eday, and BMI on three lung function parameters, causal relationships are observed for all three variables with lung function measures (Supplementary Table 8). Among proteins associated with lung functions, genetically predicted levels of 48 proteins link with BMI, 15 proteins with smok per eday, and 3 proteins with smoking cessation. A two-stage network MR analysis outlines 20 pathways facilitated by proteins from BMI to lung functions and 5 pathways facilitated by proteins from smoking intensity to lung functions. In the pathways from BMI to lung functions, the extent of mediation by these proteins varies significantly, with CA3 mediating 9.6% of the BMI-lung function relationship, while PRSS8 mediates up to 91% of this association. In the pathways from smoking per day to lung functions, the mediation extent also shows variability; PDLIM4 mediates 9.7% of the smoking intensity-lung function relationship, whereas EFEMP1 mediates up to 75% of this association (Fig. 4).
3.6 DrugTarget Database
In our investigation of the protein-protein interaction (PPI) network, we identified significant interactions between the target proteins of ADAMTS5 and the therapeutic agents AVID-200, pirfenidone, FP-020, and Sodium Pyruvate. Additionally, interactions were noted between the target proteins of FGFR1 and the medications Nintedanib, pirfenidone, and Yinfenidone. Further associations were observed between FN1's target proteins and the therapeutic agents dectrekumab + VAK-694, STP-707, famitinib L-malate, and 00L-XWG, and between the target proteins of RHOC and OGN with the pharmacological components of VAL-201 + VAL-301 + complementary active components.Utilizing the STRING database, our analysis revealed robust associations between the aforementioned target proteins and pharmacological targets of lung function therapies. Specifically, strong associations exist between ADAMTS5 and the cytokines TGFB1, IL1, IL6, TIMP1, and MMP12, and between FGFR1 and the tyrosine kinase receptors FLT1, FLT4, PDGFRB, and KDR. FN1 is associated with SRC, TGFB1, CXCR2, FLT1, and IL13, while associations between RHOC and OGN with SRC were also evident.Notably, STRING provided experimental validation for co-expression and protein homology among FGFR1, TGFB family proteins (TGFB1, TGFB2, TGFB3), FLT1, FLT4, PDGFRB, SRC, RET, and KDR. It also confirmed co-expression between RET and SRC, TGFB1 and TGFB3, as well as between RHOC and SRC, and protein homology between FN1 and IL13, SRC, suggesting intricate interactions among these proteins. Furthermore, the co-expression of ADAMTS5 with IL6 and MMP12 underscores significant interactions, indicating potential therapeutic relevance (Fig. 5). This implies that ADAMTS5, RET, RHOC and FN1 may serve as potential targets for these therapeutic agents. Additionally, it is noteworthy that Drugbank recognizes FGFR1 as a validated therapeutic target for Nintedanib and FN1 as a validated therapeutic target for Atorvastatin, Tenecteplase and Alteplase. This unequivocally corroborates our research findings.
3.7 Function and Network Prediction of lung functions-Associated Proteins
The identified lung functions-associated proteins were found to be involved in networks, particularly characterized by co-expression and physical interactions, as depicted in Supplementary Fig. 1. Analysis based on cis-genes associated with these proteins revealed enrichment in several key pathways. These include transmembrane receptor protein kinase activity, protein tyrosine kinase activity, cell-substrate adhesion and transmembrane receptor protein tyrosine kinase activity, as detailed in Supplementary Table 9.
3.8 PheW-MR analysis of the side effects of lung function causal proteins
After adjusting for multiple testing, the MR-PheWAS analysis revealed significant correlations with 20 outcomes linked to proteins associated with lung function. Specifically, an increase in MAP1LC3A levels was associated with elevated risks of two endocrine diseases: diabetes requiring insulin treatment [Kela reimbursement] with additional control exclusions (P = 5.51E-08, OR = 2.083) and combined definitions of Type 2 diabetes (P = 5.96E-08, OR = 2.06). Furthermore, this protein was linked to increased risks for five circulatory diseases: angina pectoris (P = 3.15E-08, OR = 2.677), deep vein thrombosis (DVT) of the lower extremities (P = 1.04E-06, OR = 5.903), ischemic heart disease [broad definition] (P = 2.09E-09, OR = 2.292), and venous thromboembolism (P = 4.46E-08, OR = 3.105). This increase also correlated with rises in common anthropometric risk factors, such as height (inverse-rank normalized, P = 5.28E-25, Beta = 0.366) and weight (inverse-rank normalized, P = 2.42E-15, Beta = 0.363). Conversely, an increase in FN1 levels was linked to a decreased risk of three endocrine diseases: coronary artery bypass grafting (P = 6.83E-11, OR = 0.515), coronary atherosclerosis (P = 3.15E-07, OR = 0.760), and coronary revascularization (ANGIO or CABG) (P = 1.58E-09, OR = 0.641). Additionally, this elevation was associated with a reduced risk of a common anthropometric risk factor (height, inverse-rank normalized, P = 1.00E-08, Beta=-0.069).Moreover, an increase in NPNT levels was linked to an elevated risk of the endocrine disease angina pectoris (P = 4.65E-07, OR = 1.29) and a reduced risk of the pulmonary disease chronic lower respiratory diseases (P = 2.20E-12, OR = 0.786).Further increases in BOC, OGN, and RHOC were correlated with decreased risks of a common anthropometric risk factor (height, inverse-rank normalized, with respective statistics: P = 1.54E-07, Beta=-0.037; P = 7.04E-12, Beta=-0.030; P = 2.15E-07, Beta=-0.090). Meanwhile, an increase in DNAJB4 was linked to a reduced risk of another common risk factor (weight, inverse-rank normalized, P = 1.83E-07, Beta = 0.082)(Supplementary Table 10).