Dynamical systems analysis identifies aberrant alveolar epithelial cells as comprising a stable attractor basin
Year: 2026
Type: conference-abstract
Abstract: Background: Aberrant epithelial cells of alveolar and airway descent have been implicated as drivers of pulmonary fibrosis (PF). However, there has been little consensus as to these cells’ origins, fates, characteristics, or nomenclature, with claims including that these cells are stuck in transition between alveolar type 2 (AT2) and alveolar type 1 (AT1) cells, occupy a separate trajectory from normal differentiation, or are basal-like. Here, we consider cells as dynamical systems, allowing interrogation of these claims by quantifying cell identity based on stability. We hypothesize that these aberrant cells comprise stable “attractor basins” as opposed to unstable “saddle points” (two concepts rooted in dynamical systems theory [DST]). Viewed through this approach these aberrant cells would represent a distinct cell fate that is neither stuck in transition, nor basaloid, nor synonymous with normally differentiating AT2 cell intermediates. Methods: We sought to leverage dynamical systems theory to compare human aberrant alveolar epithelial cells across datasets and test whether or not they constitute a stable, unique identity. First, we used our published quantitative classification and visualization method, “single cell Type Order Parameters” (scTOP), to directly compare alveolar epithelial and basal cells from five human PF single cell RNAseq datasets against each other and against normal lung atlas cells. Next, we conducted differential expression analysis and pathway analysis (GSEApy prerank algorithm), comparing cell identities across diseased and control cell transcriptomes. Finally, we performed computational tests of stability, including: a) reclassification of holdout cells based on references composed of the training cells to determine if identities are separable from one another; b) in silico perturbations based on “spherical interpolation” of cells in the direction of other cell identities to find resistance to perturbation; and c) transcriptional distance between cell states projections to examine if expression can be explained by other states as intermediates. Results: Human AT2, AT1, basal, and aberrant epithelial cells consistently and exclusively projected (by scTOP) onto their annotated analogues across datasets, indicating they comprise separable cell fates with distinct molecular programs, whereas differentiating AT2 cells projected onto AT2 and AT1 cells, suggesting they lack transcriptional identity beyond that of AT2 and AT1 cells. Pathways such as extracellular matrix (ECM)-receptor interaction and focal adhesion were regularly overexpressed in disease cells compared to controls, and genes overexpressed in the aberrant population were central to the pathways (CLDN4, FN1, ITGB6, KRT17, ERBB2, etc.). In contrast, differentiating AT2 cells had few to no genes differentially expressed against AT1 and AT2 cells. Holdout aberrant cell populations reclassified as their original annotations 91% of the time, resisted directed in silico perturbations (with 0.29 mean scaled alpha value required to reclassify half the cells as other cell identities compared to 0.27 for AT2s and 0.32 for AT1 cells), and were distant from alveolar epithelial cell states (0.44 mean transcriptional distance compared to 0.41 for AT1 and 0.46 for AT2). Conversely, normally differentiating AT2 cells back-predicted 42% of the time and often caused misclassification of other lung cells, required mean perturbation alpha of only 0.15, and had mean distance to alveolar epithelial states of 0.08, suggesting the aberrant populations are much more stable and distinct than the normally differentiating populations. Conclusion: These findings suggest that human aberrant alveolar epithelial cells have a stable identity without overlap with basal, differentiating AT2, or other lung cells, and thus constitute a distinct attractor according to DST, whereas normally differentiating AT2 cells represent an unstable saddle point. Similar analysis would be instructive if performed on normal and injured murine alveolar epithelia. Applying insights into which cells represent attractors or saddles likely will reduce the differentiation landscapes feasible under DST and therefore assist discernment of aberrant cell trajectories driving fibrosis and clarify the varied nomenclature published to date for these cells.
Source: Figshare
Subfield: Pulmonary and Respiratory Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7202052859
Cites:
Cited by:
Related to: