An Integrative Genomic Strategy Identifies sRAGE as a Causal and Protective Biomarker of Lung Function

Abstract

Background

There are few clinically useful circulating biomarkers of lung function and lung disease. We hypothesized that genome-wide association studies (GWAS) of circulating proteins in conjunction with GWAS of pulmonary traits represents a clinically relevant approach to identifying causal proteins and therapeutically useful insights into mechanisms related to lung function and disease.

Study Question

Can an integrative genomic strategy using GWAS of plasma soluble receptor for advanced glycation end-products (sRAGE) levels in conjunction with GWAS of lung function traits identify putatively causal relations of sRAGE to lung function?

Study Design and Methods

Plasma sRAGE levels were measured in 6,861 Framingham Heart Study participants and GWAS of sRAGE was conducted to identify protein quantitative trait loci (pQTL), including cis-pQTL variants at the sRAGE protein-coding gene locus (AGER). We integrated sRAGE pQTL variants with variants from GWAS of lung traits. Colocalization of sRAGE pQTL variants with lung trait GWAS variants was conducted, and Mendelian randomization was performed using sRAGE cis-pQTL variants to infer causality of sRAGE for pulmonary traits. Cross-sectional and longitudinal protein-trait association analyses were conducted for sRAGE in relation to lung traits.

Results

Colocalization identified shared genetic signals for sRAGE with lung traits. Mendelian randomization analyses suggested protective causal relations of sRAGE to several pulmonary traits. Protein-trait association analyses demonstrated higher sRAGE levels to be cross-sectionally and longitudinally associated with preserved lung function.

Interpretation

sRAGE is produced by type I alveolar cells, and it acts as a decoy receptor to block the inflammatory cascade. Our integrative genomics approach provides evidence for sRAGE as a causal and protective biomarker of lung function, and the pattern of associations is suggestive of a protective role of sRAGE against restrictive lung physiology. We speculate that targeting the AGER/sRAGE axis may be therapeutically beneficial for the treatment and prevention of inflammation-related lung disease.

FOR EDITORIAL COMMENT, SEE PAGE 3

Identifying causal variants, genes, proteins, and biological pathways for complex diseases is a major challenge of translational research,¹ because genome-wide association studies (GWAS) of these phenotypes often yield variants that have little or no apparent biological relevance.² In contrast, GWAS of circulating levels of proteins, as gene products, are more likely to identify genetic associations that are mechanistic in nature and can serve as clinically relevant drug targets.³ As part of the Systems Approach to Biomarker Research in Cardiovascular Disease,⁴ we assayed and conducted a GWAS of 71 plasma proteins in 6,861 Framingham Heart Study (FHS) participants and demonstrated that an integrative genomics approach can identify putatively causal protein biomarkers for cardiovascular disease.⁵ We postulate that this approach can be extended to other complex diseases. For proof-of-concept, we explored the soluble receptor for advanced glycation end-products (sRAGE) as a causally protective protein in relation to lung function.

COPD and its anatomic and radiographic correlate, pulmonary emphysema, constitute the third leading cause of nonaccidental death in the United States.⁶ COPD is diagnosed using spirometry,⁷ and emphysema is identified by lung imaging, including CT, which measures the percentage of emphysema-like lung.⁸ COPD has environmental risk factors such as smoking⁹ and genetic risk factors such as polymorphisms within SERPINA1 that increase the risk of an inherited form of lung disease due to α₁-antitrypsin deficiency.⁸ Here, we focus on the analyses of plasma sRAGE in relation to lung function by spirometry and lung structure by CT imaging. sRAGE has previously been explored as a biomarker of COPD and emphysema.¹⁰ Genetic variants in the AGER gene, which encodes RAGE, have been reported to be associated with circulating sRAGE levels and with impaired lung function, COPD, and emphysema,^11,12 but a causal link in humans between sRAGE and lung function or lung disease has not been proven.

To explore a causal relation of sRAGE to lung function and structure, we carried out a comprehensive integrative genomic study of genetic variants (pQTLs; protein quantitative trait loci) associated with plasma levels of sRAGE⁵ in conjunction with GWAS of lung traits^13,14 and tested for colocalization of these GWAS signals and performed Mendelian randomization (MR) analyses to infer causality of sRAGE in relation to pulmonary traits. Further evidence implicating sRAGE is provided from the results of cross-sectional and longitudinal analyses of sRAGE in relation to pulmonary traits in FHS participants. Our findings, in conjunction with prior in vivo and in vitro functional studies, provide an in-depth exploration of the hypothesis that the advanced glycation end-products receptor gene (AGER)/sRAGE axis is causally related to lung function and structure and that sRAGE is causally protective.

Methods

Study Design

The study consisted of five steps (Fig 1). First, our GWAS results for sRAGE (e-Table 1), performed on 6,861 FHS participants with measured plasma sRAGE levels and genome-wide single-nucleotide polymorphism (SNP) genotyping,⁵ using Affymetrix 500K mapping arrays and imputed to the 1000 Genomes Project reference panel (build 37 phase 1 v3),¹⁵ were interrogated to identify pQTL variants associated with circulating sRAGE levels. Second, all cis-pQTL variants for sRAGE (e-Table 1), defined as within 1 megabase (Mb) of the transcription start site of the RAGE protein coding gene, AGER,⁵ were interrogated to identify those that overlap with SNPs from previously published GWAS of lung traits.^14,15 Third, colocalization analyses were performed to infer shared genetic signals between sRAGE and lung traits. Fourth, using sRAGE cis-pQTL variants as instrumental variables, two-sample MR⁵ was conducted to infer causality of sRAGE in relation to lung traits. Fifth, protein-trait association analyses were performed in FHS participants using plasma sRAGE measurements and cross-sectional and longitudinal spirometry measures of pulmonary function and CT scan measures of pulmonary parenchymal density.

Figure 1.

This study consisted of multiple steps: measurement of plasma sRAGE in ~7,000 FHS participants; identification of single-nucleotide polymorphisms (SNPs) associated with circulating sRAGE (ie, pQTLs; protein quantitative trait loci); identification of sRAGE cis-pQTL variants that overlapped with previously published COPD/emphysema GWAS SNPs; colocalization analysis to infer shared genetic variants for sRAGE and COPD/emphysema; Mendelian randomization to infer putative causality of sRAGE for emphysema; cross-sectional and longitudinal analyses of sRAGE in relation to pulmonary traits in FHS participants.

Colocalization

We sought to determine whether pQTL variants associated with circulating sRAGE levels colocalized with SNPs associated with relevant pulmonary traits, including FEV₁, FVC, the ratio FEV₁/FVC, and the percentage of emphysematous-like lung parenchyma on CT imaging (percentage of low attenuation areas at −950 Hounsfield units [%LAA-950]) from prior GWAS.^14,16 To do this, we first identified all cis-pQTL variants for plasma sRAGE levels. Using GWAS SNPs associated with pulmonary traits from the entire sRAGE cis-pQTL locus (1 Mb upstream and downstream),¹³ we conducted a Bayesian test for colocalization of sRAGE cis-pQTL variants and pulmonary function GWAS SNPs using the COLOC version 3.1 package in R.¹⁷ Briefly, this method calculates Bayes factors for five different hypotheses—no SNPs associated with sRAGE or pulmonary function (H0), SNPs associated with sRAGE only (H1), SNPs associated with pulmonary function only (H2), two distinct SNPs with one for sRAGE only and the other for pulmonary function only (H3), and one distinct SNP for sRAGE and pulmonary function (H4)—within a genomic region by integrating all binary SNP configurations (ie, each SNP is responsible or not responsible for the observed association signal) that support each hypothesis. Prior probabilities were specified for an SNP being associated with plasma sRAGE levels only (p1), pulmonary trait only (p2), and pulmonary trait given its association with sRAGE (p12). p1 and p2 were set to .0001, and p12 was then set to .00001, estimating that 1 in 10,000 SNPs were associated with each trait. Posterior probabilities (PP) were then calculated for each of the five hypotheses, with PPH4 indicating the posterior probability of colocalization (one distinct variant associated with sRAGE and the pulmonary trait).

Mendelian Randomization

sRAGE cis-pQTL variants were pruned at a linkage disequilibrium threshold of r² < 0.1, estimated from the European samples within the 1000 Genomes Project using PLINK.¹⁸ The resulting pruned cis-pQTL variants, along with their associations with plasma sRAGE levels (in FHS), were uploaded to MR-base¹⁹ as the exposure, and pulmonary traits from prior GWAS (FEV₁, FVC, FEV₁/FVC, prevalent COPD, and %LAA-950) were selected as the outcomes.^16,20 MR inverse-variance weighted and Egger regression were conducted twice—the first time using all nonredundant sRAGE cis-pQTL variants (linkage disequilibrium r² < 0.1; N = 9) as instrumental variables and repeated using the same set of nonredundant sRAGE cis-pQTL variants excluding rs2070600 (Gly82Ser), a nonsynonymous variant at the ligand binding domain, with the Ser82 allele having nearly a twofold higher binding affinity (K_D for Ser82 = 77 nM; K_D for Gly82 = 122 nM).²¹ MR causal effect estimates were reported per standard error increment in inverse-rank normalized plasma sRAGE.

Study Population

The study sample for protein-trait association analysis (n = 6,276) consisted of FHS offspring (n = 2,512) and third generation (n = 3,764) cohort participants with plasma sRAGE measurements and pulmonary function measurements at the baseline examination (offspring cohort examination 7 [1999-2003] and third generation cohort examination 1 [2002-2005]; Table 1). For the longitudinal analyses, participants who did not attend the follow-up examination (offspring cohort examination 8 and third generation cohort examination 2) or had COPD, defined as FEV₁/FVC < 0.7 at the baseline examination²² (n = 1,161), were excluded. The final longitudinal study sample consisted of 4,136 FHS participants (e-Table 2). Secondary protein-spirometry trait analyses were conducted on FHS participants who were stratified by baseline smoking status (current, former, and never smokers; e-Table 3). Protein-trait association analyses were also conducted on 2,707 participants who had multi-detector CT scan measurement of emphysematous lung parenchyma (%LAA-950; e-Table 4).

Table 1.

Baseline Clinical Characteristics of FHS Participants With Spirometry Measures of Pulmonary Function (N = 6,276)

Characteristic	Mean	SD
Age, y	48.3	13.4
Height, in	66.7	3.72
BMI	27.4	5.42
Pack-years of smoking	9.27	16.71
Plasma sRAGE, pg/mL	3,585	1,150
FEV1, L	3.22	0.90
FVC, L	4.24	1.09
FEV1/FVC	0.76	0.074
	No.	%
Women	3,360	53.5
COPD	1,161	18.5
Current smoker	979	15.6
Never smoker	3,266	52.0
Former smoker	2,031	32.4

sRAGE = soluble receptor for advanced glycation end-products.

Clinical Measures

Continuous spirometry measures of pulmonary function included FEV₁, FVC, and FEV₁/FVC obtained at the same examination at which circulating sRAGE was measured. Chest CT was used to assess lung parenchymal density.²³ The demographic and clinical characteristics of the CT sample are shown in e-Table 4. Lung parenchymal density was analyzed as a continuous variable, defined as the percentage of low attenuation areas (below −950 Hounsfield units) as measured by CT (%LAA-950)²⁴; a high value is indicative of greater emphysematous lung parenchyma.

Protein Quantification

Baseline plasma sRAGE concentration was measured as part of the Systems Approach to Biomarker Research in Cardiovascular Disease Initiative. Briefly, fasting blood plasma samples were collected from FHS participants at the baseline examination (offspring cohort examination 7 [1999-2003] and third generation cohort examination 1 [2002-2005]; Table 1) and stored at −80 °C. Plasma sRAGE was quantified using a modified enzyme-linked immunosorbent assay, multiplexed on a Luminex xMAP platform (Sigma-Aldrich), using previously published methods.²⁵ To avoid cross-reactivity, all targets were initially developed as singleton assays before compatible targets were combined to form multiplex panels. For the sRAGE assay, the detection antibody was MAB11451 (R&D Systems), the capture antibody was product 1145-RG-050 (R&D Systems), and the reference protein was BAF1145 (R&D Systems). The mean inter- and intra-assay coefficients of variation for sRAGE were 5.6% and 14.5%, respectively.⁴

Statistical Methods

Statistical analyses were performed using R software version 3.1.1²⁶ and SAS software version 9.4. For all protein-trait association analyses, plasma sRAGE levels were log-transformed and standardized around a mean of 0 and SDof 1. For the cross-sectional analyses of pulmonary function, baseline plasma sRAGE levels were analyzed in relation to FEV₁, FVC, FEV₁/FVC, and COPD, after adjusting for age, sex, height, pack-years of smoking, study cohort, and current and former smoking status at the baseline examination. For the cross-sectional analyses of sRAGE in relation to CT measures of lung function, baseline plasma levels were analyzed in relation to %LAA-950 after adjusting for age, sex, BMI, study cohort, current and former smoking status, and smoking pack-years at the examination closest to the time of the CT scans. For the longitudinal analyses of changes in pulmonary function, baseline plasma sRAGE levels were analyzed in relation to follow-up minus baseline spirometry values: ΔFEV₁, ΔFVC, Δ(FEV₁/FVC, and incident COPD, after adjusting for baseline age, sex, height, pack-years of smoking, study cohort, smoking status, and baseline pulmonary function. A positive regression coefficient in the longitudinal protein-trait analysis was interpreted as higher baseline sRAGE being associated with a longitudinal increase in the respective spirometry measure (eg, a protective effect). Generalized estimating equations were applied to account for familial correlations among FHS participants. Secondary protein-trait analyses were stratified by smoking status.

Study Approval

Informed consent for genetic research was given by all participants in this investigation. The study protocol was approved by the Boston University Medical Center Institutional Review Board (protocol H-27984).

Results

Baseline Characteristics

The demographic and clinical characteristics of the FHS study participants in the cross-sectional spirometry analyses are shown in Table 1. The total sample for the cross-sectional analyses of sRAGE in relation to spirometry measures of pulmonary function included 6,276 FHS participants (mean age, 48 years; 54% women); 2,512 from the offspring and 3,764 from the third generation cohort. For the longitudinal spirometry analyses, participants who did not attend the follow-up examination and those with COPD at baseline were excluded, leaving a final study sample of 4,136 (mean baseline age, 46 years, 55% women; e-Table 2). Baseline clinical characteristics of the cross-sectional study sample, stratified by smoking status, are provided in e-Table 3. Characteristics of the study sample with CT assessment of emphysematous lung parenchyma (%LAA-950; n = 2,707) are provided in e-Table 4.

Colocalization Analyses

Colocalization of genetic signals for circulating sRAGE and lung traits at rs2070600 within the AGER locus was significant for FVC (PPH4 = 99.8%), FEV₁/FVC (PPH4 = 99.9%), and %LAA950 (PPH4 = 99.9%), but not for FEV₁ (PPH4 = 2.56E-16; e-Table 5).

MR Causal Inference Testing

MR results for causal inference of the relationship between sRAGE and pulmonary traits are summarized in Figure 2. Primary MR analyses using all nonredundant sRAGE cis-pQTL variants (n = 9 SNPs) demonstrated sRAGE to be causal for FVC (β = 0.081, P = .00000000089), FEV₁/FVC (β = −0.13, P = .000083), and %LAA950 (β = 0.30, P = .000000000099; Fig 2, top). MR analyses excluding missense variant rs2070600 (n = 8 SNPs) demonstrated similar results for FVC and %LAA950, whereas FEV₁/FVC was no longer statistically significant (Fig 2, bottom). MR analyses for COPD were not statistically significant whether including (OR = 1.08, P = .31) or excluding (OR = 0.91, P = .055) rs2070600. Forest plots showing individual effects for each SNP used as an instrumental variable in MR analyses for FEV₁ and FVC (e-Fig 1) demonstrated that rs2070600 significantly drove the MR association of sRAGE with FEV₁. MR Egger regression analyses (e-Table 6) including rs2070600 demonstrated statistically significant Egger intercepts for FEV₁ (P = .000833), FVC (P = .00397), and FEV₁/FVC (P = .00949), whereas analyses excluding rs2070600 resulted in nonsignificant Egger intercepts, further suggesting that rs2070600 significantly drove the MR results.

Figure 2.

Results from MR analyses conducted including rs2070600 (top) and excluding rs2070600 (bottom). SNPs (apart from rs2070600) used as IVs included: rs11796, rs1871665, rs204993, rs2073044, rs2170185, rs2523570, rs6923504, and rs9266529. SNP = single-nucleotide polymorphism. ∗SNP-outcome associations are from GWAS from Shrine et al.¹⁶ Regression coefficients for FEV₁ and FVC are reported as change in inverse rank-normalized lung volumes per SD increase in inverse rank-normalized plasma sRAGE. Regression coefficients for FEV₁/FVC are reported as change in FEV₁/FVC per SD increase in inverse rank-normalized plasma sRAGE. ∗∗SNP-outcome associations are from Cho et al¹⁴ %LAA-950 was defined as the percentage low attenuation areas (below −950 Hounsfield units) by chest CT. Regression coefficients for %LAA-950 are reported as change in log-transformed %LAA-950 per SD increase in inverse rank-normalized plasma sRAGE. Regression coefficients for FEV₁ and FVC are reported as change in inverse rank-normalized lung volumes per SD increase in inverse rank-normalized plasma sRAGE. Regression coefficients for FEV₁/FVC are reported as change in FEV₁/FVC per SD increase in inverse rank-normalized plasma sRAGE. Regression coefficients and SD are shown below each corresponding forest plot point estimate, formatted as beta (SD). All statistically significant P values (right column) are shown in bold.

Protein-Trait Association Analysis; Cross-Sectional and Longitudinal Associations of sRAGE With Pulmonary Traits

Plasma sRAGE was associated cross-sectionally with FEV₁ (β = 0.021, P = .00054), FVC (β = 0.038, P = .000000061), and FEV₁/FVC (β = −0.21, P = .017), but not prevalent COPD (OR = 0.95, P = .20). Results of the cross-sectional association analyses between sRAGE and pulmonary traits are shown in Table 2. Cross-sectional association between sRAGE and CT measures of lung parenchymal density showed higher sRAGE to be associated with lower %LAA-950 (β = −0.0017, P = .0012), consistent with a protective effect of sRAGE on the lung parenchyma (Table 2). Longitudinal analyses revealed that baseline plasma sRAGE was associated with greater ΔFEV₁ (β = 2.26, P = .00025), greater ΔFVC (β = 4.2, P = .0000000090), but lower Δ(FEV₁/FVC) (β = −0.026, P = .0018, Table 3). Cross-sectional protein-trait association analyses between sRAGE and pulmonary traits, when stratified by smoking status, remained directionally consistent with the primary analyses, but sRAGE had a larger (protective) effect size in current smokers than in former and never smokers on FEV₁ (β = 0.059, 0.031, and 0.036, respectively) and FVC (β = 0.087, 0.054, and 0.064; e-Table 7).

Table 2.

Observed Cross-sectional Associations Between sRAGE and Pulmonary Traits

Cross-sectional Trait	Cross-sectional Association With sRAGE
	β/OR	SE/95% CI	P
FEV1a	0.021	0.006	.000537
FVCa	0.038	0.007	.0000000614
FEV1/FVCb	−0.21	0.088	.0165
Prevalent COPD	OR, 0.95	95% CI, 0.88-1.02	.195
%LAA-950b	−0.002	0.000	.000101

Covariates for FEV₁, FVC, FEV₁/FVC, and COPD: age, sex, height (inches), pack-year of smoking, study cohort, current smoker (yes/no), former smoker (yes/no). COPD was defined as FEV₁/FVC < 0.7. ORs for COPD are reported per 1 SD increase in log-transformed sRAGE concentration. Statistically significant P values are shown in bold. %LAA-950 = percentage low attenuation area (below −950 Hounsfield units); sRAGE = soluble receptor for advanced glycation endproducts.

^aFEV₁ and FVC are expressed in L.

^bAnalyses for %LAA950 were additionally adjusted for BMI. A negative coefficient for %LAA-950 reflects less emphysema-like lung in relation to higher sRAGE levels.

Table 3.

Observed Longitudinal Associations Between sRAGE and Pulmonary Traits

Longitudinal Trait	Longitudinal Association With sRAGE
	β/OR	SE/95% CI	P
ΔFEV1a	2.256	0.616	.000251
ΔFVCa	4.168	0.725	.00000000895
Δ(FEV1/FVC)b	−0.026	0.008	.00176
Incident COPDc	OR, 1.02c	95% CI, 0.88-1.17	.830

Covariates in longitudinal analyses included age, sex, height (inches), pack-year of smoking, study cohort, current smoker (yes/no), former smoker (yes/no), baseline pulmonary function. Longitudinal regression coefficients reflect follow-up minus baseline differences in continuous spirometry values per SD increment in standardized long-transformed sRAGE. A positive regression coefficient for spirometry traits reflects greater preservation of a spirometry measure over time in relation to higher sRAGE levels. Statistically significant P values are shown in bold.

^aRegression coefficients for ΔFEV₁ and ΔFVC are reported as change in lung volume (mL; at follow-up minus baseline) per SD increase in standardized log-transformed plasma sRAGE.

^bRegression coefficients for FEV₁/FVC and %LAA-950 are reported as per SD increase in standardized log-transformed plasma sRAGE.

^cORs for COPD are reported per SD increase in standardized log-transformed plasma sRAGE. COPD was defined as FEV₁/FVC < 0.7.

Discussion

We provide evidence from an integrative genomics approach that sRAGE is a putatively causal and protective protein biomarker in relation to lung function. MR analyses revealed higher sRAGE levels to be causal and protective in relation to FVC, and protein-trait analyses in our large study sample demonstrated higher sRAGE levels to be associated with higher FVC, consistent with the protective effects seen in MR (Fig 2 and Table 2). Longitudinal analyses similarly revealed that higher baseline sRAGE levels were associated with preservation of FEV₁ and FVC over time (Table 3), suggesting a longitudinal protective effect of sRAGE on lung function. Our spirometry results for the association of sRAGE with FEV₁ and FVC remained significant in separate analyses of current, former, and never smokers (e-Table 7). The larger effect sizes observed in current smokers, however, suggested that the protective effect of sRAGE on pulmonary function is greatest among current smokers in whom the pro-inflammatory stimulus for lung parenchymal damage is greatest. Indeed, a prior mouse study demonstrated that the protective effects of knocking out AGER were only observed in mice exposed to cigarette smoke.²⁷ Of note, a recent RNA-seq analysis of AGER expression in bronchial biopsies demonstrated an up-regulation in sRAGE generation via alternative splicing in smokers as compared with never smokers.²⁸

Prior studies have shown that RAGE contributes to lung inflammation,²⁹ which is a contributing factor to lung disease, including COPD,³⁰ pulmonary emphysema,³¹ and pulmonary fibrosis.³² Membrane-associated RAGE (mRAGE) functions as a pro-inflammatory receptor stimulating nuclear factor kappa-B, mitogen-activated protein kinases, and oxidases.³³ Conversely, the soluble form of RAGE, sRAGE, has the opposite effect (ie, antiinflammatory), as it acts as a decoy receptor for RAGE ligands,³⁴ which is also supported by our findings. sRAGE is generated by dual processes: matrix metalloproteinase (MMP)-mediated ectodomain shedding of mRAGE and alternative splicing and removal of AGER exon 10, which encodes the transmembrane domain of the RAGE protein,³⁵ resulting in endogenously secreted RAGE (esRAGE).³⁶ Notably, rs1973612, an SNP in the KLKB1 gene locus, is a trans-pQTL both for sRAGE⁵ and MMP-2,³⁷ providing a potential genetic link between sRAGE, cleavage of membrane-bound RAGE by MMPs, and lung parenchymal damage in emphysema and pulmonary fibrosis.^32,38

RAGE is expressed at basal levels in diverse cell types and tissues, but it is highly expressed in lung tissue and particularly in type I pulmonary alveolar cells.³³ Addition of exogenous sRAGE was shown to ameliorate pulmonary inflammation in a rat model of lipopolysaccharide-induced inflammation,³⁹ a mouse model of acid-induced lung injury,⁴⁰ and multiple mouse models of allergic airways disease.^41,42 Therefore, factors that raise circulating levels of sRAGE may diminish the pulmonary inflammatory response to cigarette smoke and other pro-inflammatory stimuli. Considered as a whole, our results, in concert with prior animal experiments, indicate that sRAGE is protective against progressive loss of lung function, particularly in response to inflammation, such as that caused by cigarette smoke exposure.

We acknowledge several limitations of our study. First, our study participants were of European ancestry, and therefore, our results may not be generalizable to other ethnicities. Second, our study was not adequately powered to assess the prognostic utility of plasma sRAGE as a biomarker of dichotomous traits such as new-onset COPD or emphysema. Third, although sRAGE was positively associated with both FEV₁ and FVC—consistent with preservation of both measures of lung function—the direction of association between sRAGE and FEV₁/FVC was negative both in MR and protein-trait association analyses. Based on these findings, we hypothesize that sRAGE exerts a greater protective effect against restrictive than against obstructive lung disease, as reflected in the greater longitudinal preservation of FVC (β = 4.17) than of FEV₁ (β = 2.26; Table 3). Indeed, colocalization analyses also suggested shared genetic pathways between sRAGE and FVC (PPH4 = 99.9%, e-Table 5), but not with FEV₁ (PPH4 = .0000000000000256, e-Table 5). Of note, a prior study demonstrated that lungs from Ager^-/- mice had greater compliance and lower elastin mRNA and protein levels, but no significant changes in airway resistance.⁴³ Finally, analyses of CT emphysema-like lung density (%LAA-950) also did not provide evidence that sRAGE is protective against emphysema; rather, the body of evidence from our study suggests that sRAGE is a causal and protective biomarker in relation to restrictive rather than obstructive lung disease.

To our knowledge, this is the first study to use an integrative genomics strategy to establish sRAGE as a causal and protective protein biomarker of lung function. We have shown that sRAGE is protective against loss of lung function and provided evidence that the AGER/RAGE axis is a promising therapeutic target for the treatment and prevention of lung damage caused by inflammation, such as that caused by cigarette smoking. Overall, we have demonstrated the applicability of our integrative approach to elucidate putatively causal biomarkers for complex disease phenotypes that can be explored for therapeutic and prognostic utility.

Interpretation

sRAGE is produced by type I alveolar cells and acts as a decoy receptor to block the nuclear factor kappa-B inflammatory signaling cascade. Our findings provide mechanistic evidence from an integrative genomics approach that sRAGE has a causal and protective effect on lung function and that targeting the AGER/sRAGE axis may be therapeutically beneficial for the treatment and prevention of inflammation-related lung disease.

Take-home Points.

Study Question: Can an integrative genomic strategy using GWAS of plasma soluble receptor for advanced glycation end-products (sRAGE) levels in conjunction with GWAS of lung function traits identify putatively causal relations of sRAGE to lung function?

Results: Colocalization identified shared genetic signals for sRAGE with lung traits. Mendelian randomization analyses suggested protective causal relations of sRAGE to several pulmonary traits. Protein-trait association analyses demonstrated higher sRAGE levels to be cross-sectionally and longitudinally associated with preserved lung function.

Interpretation: sRAGE is produced by type I alveolar cells, and it acts as a decoy receptor to block the inflammatory cascade. Our integrative genomics approach provides evidence for sRAGE as a causal and protective biomarker of lung function, and the pattern of associations is suggestive of a protective role of sRAGE against restrictive lung physiology. We speculate that targeting the AGER/sRAGE axis may be therapeutically beneficial for the treatment and prevention of inflammation-related lung disease.