Plasma protein signatures of adult asthma

Abstract

Background: Adult asthma is complex and incompletely understood. Plasma proteomics is an evolving technique that can both generate biomarkers and provide insights into disease mechanisms. We aimed to identify plasma proteomic signatures of adult asthma.

Methods: Protein abundance in plasma was measured in individuals from the Agricultural Lung Health Study (ALHS) (761 asthma, 1095 non-case) and the Atherosclerosis Risk in Communities study (470 asthma, 10,669 non-case) using the SOMAScan 5K array. Associations with asthma were estimated using covariate adjusted logistic regression and meta-analyzed using inverse-variance weighting. Additionally, in ALHS, we examined phenotypes based on both asthma and seroatopy (asthma with atopy (n = 207), asthma without atopy (n = 554), atopy without asthma (n = 147), compared to neither (n = 948)).

Results: Meta-analysis of 4860 proteins identified 115 significantly (FDR<0.05) associated with asthma. Multiple signaling pathways related to airway inflammation and pulmonary injury were enriched (FDR<0.05) among these proteins. A proteomic score generated using machine learning provided predictive value for asthma (AUC = 0.77, 95% CI = 0.75-0.79 in training set; AUC = 0.72, 95% CI = 0.69-0.75 in validation set). Twenty proteins are targeted by approved or investigational drugs for asthma or other conditions, suggesting potential drug repurposing. The combined asthma-atopy phenotype showed significant associations with 20 proteins, including five not identified in the overall asthma analysis.

Conclusion: This first large-scale proteomics study identified over 100 plasma proteins associated with current asthma in adults. In addition to validating previous associations, we identified many novel proteins that could inform development of diagnostic biomarkers and therapeutic targets in asthma management.

Keywords: allergy; area under curve; biomarkers; precision medicine; proteomics.

Figure 1.

Overview of our proteome and asthma study. Each study examined associations between plasma protein levels and asthma. Participating studies were ALHS (Agricultural Lung Health Study) and ARIC (Atherosclerosis Risk in Communities). Our meta-analysis included three datasets from two ancestries: European ancestry (EA) and African ancestry (AA). Functional follow-up analyses included pathway enrichment analysis, Mendelian randomization, and a druggable targets search. We validated 19 of our significant findings in a proteome study (n=85). In our proteomic score analysis, we included proteins either FDR-significant or selected via LASSO feature selection. We derived the proteomic score in ALHS and validated in ARIC.

Figure 2.

Associations between plasma proteins and current asthma in adults. Meta-analyzed beta estimates for each protein are plotted against the negative log₁₀-transformed p-value of the estimate; each point represents one protein. Proteins either significant after Bonferroni correction (blue) or validated in a proteome study (n=85) (red) are labeled with the corresponding target protein name. Horizontal dashed line represents the significance thresholds after multiple-testing correction (FDR<0.05).

Figure 3.

Functionally enriched (FDR<0.05) pathways among proteins related (FDR<0.05) to adult asthma. A. Bar plot of -log₁₀(FDR) for enriched Ingenuity Pathway Analysis (IPA) canonical pathways, sorted by p-value. Dashed line represents the significance threshold (FDR<0.05). B. Interrelationships among enriched (FDR<0.05) pathways.

Figure 4.

Network of functional associations between proteins significant at FDR<0.05. This protein network had significantly more interactions than expected by chance when compared to a statistical background of all 4,860 measured proteins (P= 0.005). Colored nodes represent individual proteins and gray lines real or predicted interactions; line thickness denotes strength of evidence for association. Network made using STRING database version 12.0.

Figure 5.

Performance of proteomic score in derivation (ALHS) and validation (ARIC) cohorts. A. Receiver operating characteristics (ROC) curves comparing performance of the proteomic score with covariates (blue line) to the covariates alone (grey line) at classifying ALHS participants by asthma status. Area under the curve (AUC) of the proteomic score curve was significantly greater than the covariates-only curve. Covariates used: age (yrs.), sex, body mass index (kg/m²), smoking status (current vs. former, relative to never), pack-years of smoking, study center (IA or NC), season (fall vs. not), time in transit (hrs.). B. Displays the same comparison when the proteomic score is applied to classify ARIC EA participants. The proteomic score curve still performed significantly better than the covariates-only curve, though AUCs were lower than in the derivation cohort. Covariates used: age (yrs.), sex, body mass index (kg/m²), smoking status (current vs. former, relative to never), study center (NC, MS, MN, or MD).