Aberrant mucin expression and keratinization distinguishing severe from mild asthma revealed by interpretable machine learning
Sagar L. Kale, Augusta Vincent, Mark A. Ross, Isha Mehta, Michael J. Calderon, Richard P. Ramonell, Himanshu Setya, Jessica C. McCreary-Partyka, Huijuan Yuan, Stephanie A. Christenson, Prescott G. Woodruff, Mario Castro, Kaharu Sumino, Nizar N. Jarjour, Loren C. Denlinger, Benjamin Gaston, Eugene R. Bleecker, Deborah A. Meyers, Wendy C. Moore, Elliot Israel, Bruce D. Levy, David Mauger, Serpil Erzurum, Anthony Newbrough, Taylor J. Nee, Prabir Ray, Claudette M. St. Croix, Sally E. Wenzel, Jishnu Das, Anuradha Ray, Marc C. Gauthier
Sagar L. Kale, Augusta Vincent, Mark A. Ross, Isha Mehta, Michael J. Calderon, Richard P. Ramonell, Himanshu Setya, Jessica C. McCreary-Partyka, Huijuan Yuan, Stephanie A. Christenson, Prescott G. Woodruff, Mario Castro, Kaharu Sumino, Nizar N. Jarjour, Loren C. Denlinger, Benjamin Gaston, Eugene R. Bleecker, Deborah A. Meyers, Wendy C. Moore, Elliot Israel, Bruce D. Levy, David Mauger, Serpil Erzurum, Anthony Newbrough, Taylor J. Nee, Prabir Ray, Claudette M. St. Croix, Sally E. Wenzel, Jishnu Das, Anuradha Ray, Marc C. Gauthier
View: Text | PDF
Research Article Immunology Pulmonology

Aberrant mucin expression and keratinization distinguishing severe from mild asthma revealed by interpretable machine learning

Abstract

Type 2 (T2) immune cells dominate the airways of patients with mild-moderate asthma (MMA) with a more complex type 1 (T1)-T2 mixed immune response evident in treatment-refractory severe asthma (SA). We hypothesized that comparing the transcriptomes of the airway epithelium of patients with SA and MMA would reveal molecular signatures associated with more severe disease in the context of a complex immune response. Using our interpretable machine learning tool, SLIDE, meaningful latent factors (context-specific gene co-expression networks) were revealed that distinguished SA from MMA. Unexpectedly, an aberrant high expression of normally host-protective, membrane-tethered, and IFN-inducible mucins, MUC1 and MUC4, was identified in SA. Gene networks in the significant latent factors discriminating SA from MMA corresponded to enrichment of a keratinization program in SA airways. Keratinization was marked by increased expression of the stress keratin KRT16, signifying squamous metaplasia suggesting adaptive reprogramming of the airway epithelium in response to chronic stress. These mucins and KRT16 were inversely associated with lung function in 2 separate asthma cohorts. Imaging of endobronchial biopsies revealed significantly higher KRT16 protein expression in SA compared with MMA that strongly correlated with MUC1 protein expression. Our study identifies dysregulated host-protective and maladaptive repair responses in SA distinguishing from MMA.

Authors

Sagar L. Kale, Augusta Vincent, Mark A. Ross, Isha Mehta, Michael J. Calderon, Richard P. Ramonell, Himanshu Setya, Jessica C. McCreary-Partyka, Huijuan Yuan, Stephanie A. Christenson, Prescott G. Woodruff, Mario Castro, Kaharu Sumino, Nizar N. Jarjour, Loren C. Denlinger, Benjamin Gaston, Eugene R. Bleecker, Deborah A. Meyers, Wendy C. Moore, Elliot Israel, Bruce D. Levy, David Mauger, Serpil Erzurum, Anthony Newbrough, Taylor J. Nee, Prabir Ray, Claudette M. St. Croix, Sally E. Wenzel, Jishnu Das, Anuradha Ray, Marc C. Gauthier

×

Figure 1

SLIDE analysis of airway brushing RNA-seq data in HCs and participants with MMA or SA.

Options: View larger image (or click on image) Download as PowerPoint
SLIDE analysis of airway brushing RNA-seq data in HCs and participants w...
(A) Distributions shown correspond to performance of SLIDE models fitted on actual data versus those fitted using permuted labels (negative controls) across replicates of k-fold cross-validation. Bounds of the box represent the first and third quartile, respectively. The line inside the box corresponds to the median. Whiskers correspond to 1.5 times the interquartile range, and dots represent values beyond that. ****P < 0.0001. P values calculated from a permutation test. (B) Correlation network of features in latent factors (LFs) 6 and 42, which are 2 of 3 latent factors discovered by SLIDE differentiating SA from HC. (C) Scatter plot with SA and HC z scores, a weighted average of each patient’s transcriptomic expression in terms of the features present within each LF. The x and y axes represent each participant’s respective LF score. Dots are color coded by participant condition. (D–F) The corresponding data for SA versus MMA comparison, with correlation network of features in LFs 63 and 30 depicted, which are 2 out of 5 LFs discovered by SLIDE differentiating SA from MMA. (G–I) For MMA versus HC comparison, correlation network of features in LFs 15 and 33 shown are 2 of 4 LFs discovered by SLIDE differentiating MMA from HC. For correlation networks, triangle nodes (Δ) indicate greater expression in HCs, whereas squares (□) and circles (○) represent the same for patients with MMA and SA, respectively. Purple edges indicate positive correlation, and green edges show negative correlation. Edges are only present if they have a strength of |R| > 0.4. Edge thickness indicates strength of correlation. Only nodes that passed a loading and AUC threshold were included.