A blood and bronchoalveolar lavage protein signature of rapid FEV1 decline in smoking-associated COPD

Abstract

Accelerated progression of chronic obstructive pulmonary disease (COPD) is associated with increased risks of hospitalization and death. Prognostic insights into mechanisms and markers of progression could facilitate development of disease-modifying therapies. Although individual biomarkers exhibit some predictive value, performance is modest and their univariate nature limits network-level insights. To overcome these limitations and gain insights into early pathways associated with rapid progression, we measured 1305 peripheral blood and 48 bronchoalveolar lavage proteins in individuals with COPD [n = 45, mean initial forced expiratory volume in one second (FEV1) 75.6 ± 17.4% predicted]. We applied a data-driven analysis pipeline, which enabled identification of protein signatures that predicted individuals at-risk for accelerated lung function decline (FEV1 decline ≥ 70 mL/year) ~ 6 years later, with high accuracy. Progression signatures suggested that early dysregulation in elements of the complement cascade is associated with accelerated decline. Our results propose potential biomarkers and early aberrant signaling mechanisms driving rapid progression in COPD.

Conflict of interest statement

KMD, KCN, CMF, SAC, NEA, NEA, WHA, IZB, RGB, APC, EB, RCB, CBC, DJC, GJC, CMD, JMW, EAH, NNH, ATH, RJK, JAK, DAM, WKO, VEO, RP, SPP, PGW, RPB, JLC, KBA have no competing interests to declare. WWL reports personal fees from Konica Minolta and Continuing Education Alliance. MKH reports personal fees from GlaxoSmithKline, AstraZeneca, Boehringer Ingelheim, Cipla, Chiesi, Novartis, Pulmonx, Teva, Verona, Merck, Mylan, Sanofi, DevPro, Aerogen, Polarian, Regeneron, United Therapeutics, UpToDate, Medscape and Integrity. She has received either in kind research support or funds paid to the institution from the NIH, Novartis, Sunovion, Nuvaira, Sanofi, Astrazeneca, Boehringer Ingelheim, Gala Therapeutics, Biodesix, the COPD Foundation and the American Lung Association. She has participated in Data Safety Monitoring Boards for Novartis and Medtronic with funds paid to the institution. She has received stock options from Meissa Vaccines. FJM. has received personal fees from Forest, Janssen, GlaxoSmithKline, Nycomed/Takeda, Amgen, AstraZeneca, Boehringer Ingelheim, Ikaria/ Bellerophon, Genentech, Novartis, Pearl, Pfizer, Roche, Sunovion, Theravance, Axon, CME Incite, California Society for Allergy and Immunology, Annenberg, Integritas, InThough, Miller Medical, National Association for Continuing Education, Paradigm, Peer Voice, UpToDate, Haymarket Communications, Western Society of Allergy and Immunology, Informa, Bioscale, Unity Biotechnology, ConCert, Lucid, Methodist Hospital, Prime, WebMD, Bayer, Ikaria, Kadmon, Vercyte, American Thoracic Society, Academic CME, Falco, Axon Communication, Johnson & Johnson, Clarion, Continuing Education, Potomac, Afferent, and Adept; and has collected nonfinancial support from Boehringer Ingelheim, Centocor, Gilead, and Biogen/Stromedix; and declares other interests with Mereo, Boehringering Ingelheim, and Centocor. All are outside of this project.

© 2023. The Author(s).

Figures

Figure 1

Individual blood and BAL proteins cannot discriminate between annualized greater versus lesser rates of FEV1 decline in COPD. (a) Comparison of annualized post-bronchodilator FEV1 decline from V1 to V5. Decline was calculated as (V5 FEV1 – V1 FEV1)/ time, where time is the duration in years between V1 and V5 for each participant. (b) Volcano plot of blood and BAL proteins. Light and dark blue protein markers have a p-value < 0.05 and < 0.01, respectively, after a two-sampled two-tailed t-test. All depicted p-values are before correction for multiple comparisons. No proteins remained significant after applying the Benjamini-Hochberg false discovery rate (FDR) correction for multiple comparisons (α = 0.05).

Figure 2

A 52-feature Elastic Net (EN) signature identified individuals at-risk for FEV1 decline ≥70 mL/year with high accuracy. (a) PLSDA scores plot highlighting strong differentiation between greater decliners (magenta) and lesser decliners (yellow), separating the two groups with 98.4% cross-validation (CV) and calibration accuracy. (b) Loadings on latent variable 1 (LV1) (with negatively loaded proteins being comparatively increased in greater decliners and positively loaded proteins being comparatively reduced) captured 11.9% of the total variance in the data. (c) ROC curve of 52-feature signature suggests greater decliners classification with 100% sensitivity and 96.8% specificity in the cross-validated model. (d) LV1 scores were associated with annualized FEV1 decline (mL/yr). P-values and fit line shown for linear models adjusted for age, race, height, sex, baseline FEV1% predicted, smoking status, pack-years, and inhaled corticosteroids (ICS) use within three months of baseline visit. (e,f) Comparison of (e) 6-fold CV accuracies, (f) sensitivities, and specificities between the 52-feature EN signature, a collection of the six top proteins identified in Fig. 1, and literature-based models. All reported values are from cross-validated PLSDA models, unless otherwise noted. One-way ANOVA with Dunnett’s post hoc test; **p<0.01, ****p < 0.0001.

Figure 3

Clustering of COPD subjects by the EN-identified signature highlights distinct regulation of immune-associated processes. (a) Hierarchical clustering of the 52-feature signature highlights distinct clustering of greater decliners (magenta) and lesser decliners (yellow). Only 5 out of the 45 subjects were misclassified (Sensitivity: 85.7%, Specificity: 90.3%). BAL proteins denoted by blue text. (b) Significantly enriched ontology clusters by Metascape analysis. (c, d) Pathways encompassed in the (c) complement system cluster and in the (d) positive regulation of cytokine production cluster are listed in the table. Hatched squares indicate protein involvement in a particular pathway, colorations of magenta or yellow represent a relative elevation of the protein concentration in greater decliners or lesser decliners, respectively.

Figure 4

Complement profiles in COPD lesser decliners behave more similarly to TEPPS than COPD greater decliners. (a) Scores plot from PCA completed using all complement proteins measured in plasma (C1q, C1qBP, C1r, C2, C3d, C3b, C3, C3a, iC3b, C3a des Arg, C4, C4b, C5, C5a, C5-6, C6, C7, C8, C9, Factor B, Factor D, Properdin) of greater decliners (circles), lesser decliners (squares), and a reference group of tobacco-exposed people with preserved spirometry (TEPPS) (diamonds). First two principal components (PCs) capture 33.9% of the variance in the dataset. (b) Comparison of scores on PC1 (one-way ANOVA with Tukey’s post-hoc test; **p<0.01, ***p<0.001).

Figure 5

Select subgroup of signature proteins retain high predictive value for accelerated FEV1 decline. (a) Table denote AUCs for both ROC curves generated from calibration and cross-validation PLSDA models. AUCs of all calibration models were compared to that of the optimal (52-feature model) using the Hanley and McNeil method. (b) Comparisons of 6-fold CV accuracies of multi-compartment and (c) blood-only biomarker models to a collection of the 6 top proteins identified in Fig. 1, and literature models, as determined by ANOVA with Bonferroni’s post hoc test (*p<0.05, **p < 0.01, ***p<0.001). (d) Sensitivity and (e) specificity of signatures. All reported values are from cross-validated PLSDA models, unless otherwise noted. (†: multi-compartment model, #: blood-only model). CV: cross-validated; AUC: area under curve.