Clustering Patients with Pulmonary Hypertension Using the Plasma Proteome

Background: Patients with pulmonary hypertension are classified according to clinical criteria to inform treatment decisions. Knowledge of the molecular drivers of pulmonary hypertension might better inform treatment choice. Objectives: To investigate plasma protein clusters in patients with a diagnosis of pulmonary hypertension. Methods: Between 2013 and 2021, 470 patients with pulmonary hypertension, 136 control subjects with symptomatic disease, and 59 healthy control subjects were enrolled as a discovery cohort. Plasma levels of 7,288 proteins were assayed (SomaScan 7K platform). Proteins that distinguished pulmonary hypertension from both control groups were selected for unsupervised clustering (k-means clustering of Uniform Manifold Approximation and Projection dimensions). Clinical characteristics and outcomes were compared across clusters. Separate cohorts of serially sampled patients from pulmonary hypertension centers in the United Kingdom (n = 229) and France (n = 79) provided independent validation. Measurements and Main Results: A total of 156 plasma proteins that distinguished pulmonary hypertension from control subjects with symptomatic disease and healthy control subjects formed four clusters with diverse 5-year survival rates: 78% (cluster 4), 62% (cluster 2), 44% (cluster 3), and 33% (cluster 1). The distinction and clinical relevance of the clusters were confirmed in validation cohorts by their association with survival. To further characterize the therapeutic relevance of the clusters, we investigated two experimental drug targets: the PDGF (Platelet-Derived Growth Factor) pathway was upregulated in cluster 3 compared with other clusters, and the TGF-β (Transforming Growth Factor-β) pathway was upregulated in cluster 1. Conclusions: Plasma proteomic profiling of patients with pulmonary hypertension distinguishes four clusters, independent of the clinical classification. These groups, based on differential plasma protein levels, could act as theragnostic biomarkers for new therapies targeting PDGF and TGF-β pathways.