Prime Time for Proteomics in Pulmonary Arterial Hypertension Risk Assessment?

Precision-based approaches to pulmonary vascular disease have been a focus of research over the past decade (1, 2). The goal is to reclassify pulmonary hypertension (PH) in a way that more accurately aligns with pathobiology and precisely identifies patients at risk or those likely to respond to established and investigational therapeutics. One strategy to address these aims is to combine blood biomarkers and computational modeling to provide more robust and deeper phenotyping than can be achieved with clinical PH classification alone. Genomics, transcriptomics, proteomics, and metabolomics in isolation or combined in systems biology–based networks have, to date, uncovered novel insights into disease mechanisms and questioned how we clinically characterize patients (3–8). In this issue of the Journal, Rhodes and colleagues (pp. 1102–1111), who have been at the forefront of these efforts in pulmonary vascular disease (3, 4), build on a previous proteomics study (9) and present the largest unbiased analysis of plasma protein expression to date in idiopathic, heritable, and drug-induced pulmonary arterial hypertension (PAH) (10).

The burden that precision-based approaches in PAH bear is against established markers of disease progression and prognosis, such as functional class, 6-minute-walk distance, hemodynamics, and NT-proBNP (N-terminal pro–brain natriuretic peptide) levels. Newer but popular scores such as Registry to Evaluate Early and Long-Term PAH Disease Management (REVEAL) and the French risk equation (11, 12) incorporate these widely available clinical measures and are used to escalate treatment. The incorporation of omics into current risk assessment tools is a natural first step toward clinical integration. In the current study, the investigators performed a rigorous, unbiased analysis of over 4,000 plasma protein targets using an aptamer-based assay in derivation and replication cohorts from the UK National PAH Cohort to identify a weighted six-protein score predictive of a composite outcome of all-cause mortality or lung transplant, independent of NT-proBNP levels and 6-minute-walk distance. This six-protein score was further validated in the French EFORT (Evaluation of Prognostic Factors and Treatment Goals in PAH) cohort, which included incident patients. The score performed well at baseline (area under the curve [AUC], 0.73; 95% confidence interval, 0.63–0.85) and follow-up (AUC, 0.84; 95% confidence interval, 0.75–0.94). The six-protein score also outperformed the previously published nine-protein score (9). The six proteins identified (polydom, peroxidasin homolog, renin, neuropilin-1, thrombospondin-2, and peroxiredoxin-4) have biologic functions that are relevant to PAH pathobiology, including stabilization of collagen scaffolds in the extracellular matrix, promotion of angiogenesis, regulation of NF-κB, and compensatory inhibition of microvascular endothelial cell proliferation. There is a large unmet need for peripherally circulating biomarkers that capture pulmonary vascular remodeling, and the authors provide a nice discussion on the identified proteins and putative mechanisms in PAH. Strong biologic plausibility, as well as the ability to accurately predict ultimate clinical endpoints such as death or transplant, is a critical first step toward validation of such a score. Of the six proteins identified, commercially available ELISAs are available for four proteins (thrombospondin-2, renin, peroxiredoxin-4, and neuropilin-1). The authors performed a sensitivity analysis to narrow the model to these proteins and demonstrated sustained accuracy to predict survival in both United Kingdom and French cohorts.

If precision-based medicine in PAH is to reach its full potential, biomolecular risk stratification should offer improved discrimination and calibration over current clinical endpoints and risk calculators. If a blood signature reflective of ongoing pulmonary vascular remodeling reliably reclassified a patient from low to high risk, for example, it could have important implications for disease management and improve enrollment efficiency in clinical trials. More granular discrimination within intermediate-risk PAH is needed, as this group represents a significant proportion of prevalent patients; as much as 70% in some cohorts (13). The six-protein score was able to discern intermediate-risk subjects designated as low or high risk on the basis of the French risk equation. By our calculations, in low-risk subjects (those meeting at least two clinical criteria), adding the protein score reclassified 33% as high risk. This mirrors previous data demonstrating that the addition of newer endpoints (e.g., right ventricular function assessed by cardiac magnetic resonance imaging) can reclassify risk in patients when added to existing calculators (14). The addition of the six-protein score to NT-proBNP (a well-established blood marker in PAH but one that lacks specificity) alone modestly improved the AUC for transplant or death at 5 years from 0.72 (0.63–0.82) to 0.78 (0.71–0.86). The highest sensitivity and specificity cutoffs from the United Kingdom derivation cohort predicted survival with 89% sensitivity and 69% specificity in prevalent samples from the French cohort. Although this is encouraging, in high-stakes diseases such as PAH, excellent accuracy (AUC > 0.9) is the goal and remains elusive even with established clinical scores.

Whether this novel protein score can be applied outside of the three subtypes of PAH studied here remains to be seen. The discovery and replication cohorts included subjects with idiopathic and heritable PAH alone; the predictive ability of the six-protein score persisted in the validation cohort that included a small number (n = 12) of subjects with drug-associated PAH. The score also held up well in two distinct geographic locations. How this score will perform with other subtypes of PAH, including PAH associated with connective tissue disease, and outside of Europe remains unstudied. Two proteins in the proposed score do not yet have commercially available assays. The ability to scale these protein scores to the bedside depends on widely available cost-effective and automated platforms to facilitate their use in “prime time.” Finally, although biomolecular data may add more precise phenotypic discrimination to clinical risk assessment tools, clinical and molecular markers may have unidentified interactions, especially in this complex disease. When combined, interpretation should be done cautiously and with attention to weighting.

Rhodes and colleagues have made another important contribution to our understanding of how omics can improve our phenotyping and risk assessment for pulmonary vascular disease.

There remains a need for prospective clinical trials in which risk assessment tools are the intervention tested. Precision-based treatment regimens (with the exception of calcium channel blockers in vasoresponders) and tailored investigational drug trial eligibility for specific patients remain lofty aspirations. The six-protein score is an excellent step toward this goal for patients living with PAH.