Whole-Blood RNA Profiles Associated with Pulmonary Arterial Hypertension and Clinical Outcome

Abstract

Rationale: Idiopathic and heritable pulmonary arterial hypertension (PAH) are rare but comprise a genetically heterogeneous patient group. RNA sequencing linked to the underlying genetic architecture can be used to better understand the underlying pathology by identifying key signaling pathways and stratify patients more robustly according to clinical risk.Objectives: To use a three-stage design of RNA discovery, RNA validation and model construction, and model validation to define a set of PAH-associated RNAs and a single summarizing RNA model score. To define genes most likely to be involved in disease development, we performed Mendelian randomization (MR) analysis.Methods: RNA sequencing was performed on whole-blood samples from 359 patients with idiopathic, heritable, and drug-induced PAH and 72 age- and sex-matched healthy volunteers. The score was evaluated against disease severity markers including survival analysis using all-cause mortality from diagnosis. MR used known expression quantitative trait loci and summary statistics from a PAH genome-wide association study.Measurements and Main Results: We identified 507 genes with differential RNA expression in patients with PAH compared with control subjects. A model of 25 RNAs distinguished PAH with 87% accuracy (area under the curve 95% confidence interval: 0.791-0.945) in model validation. The RNA model score was associated with disease severity and long-term survival (P = 4.66 × 10^-6) in PAH. MR detected an association between SMAD5 levels and PAH disease susceptibility (odds ratio, 0.317; 95% confidence interval, 0.129-0.776; P = 0.012).Conclusions: A whole-blood RNA signature of PAH, which includes RNAs relevant to disease pathogenesis, associates with disease severity and identifies patients with poor clinical outcomes. Genetic variants associated with lower SMAD5 expression may increase susceptibility to PAH.

Keywords: RNAseq; pulmonary arterial hypertension; whole-blood RNA.

Figure 1.

Study design. Three hundred fifty-nine consecutively recruited patients with pulmonary arterial hypertension (PAH) from the UK National Cohort study of PAH were randomized into three groups. Age- and sex-matched control subjects for each group were then identified from the same study. The first two groups were analyzed separately for RNA discovery and RNA validation and then combined for modeling of the best combination of RNAs to distinguish between control subjects and PAH. This model was then tested in the final group of individuals. All patients were combined for subsequent clinical analyses. LASSO = least absolute shrinkage and selection operator.

Figure 2.

Identification of RNAs with different levels in patients with pulmonary arterial hypertension (PAH) and control subjects. (A and B) Volcano plot of RNA log fold differences between PAH and control subjects in discovery (A) and validation (B) analyses. Selected genes were identified in further analyses presented throughout the manuscript including modeling, external validation, and Mendelian randomization. (C and D) Box plots of SMAD5 (C) and TRPC1 (D) levels (log₁₀ reads) in control subjects and patients with PAH in the RNA discovery, validation, and model validation (validation 2) analysis groups. Horizontal dashed lines in C and D indicate the median level in healthy control subjects.

Figure 3.

Least absolute shrinkage and selection operator (LASSO) model performance in an independent validation group of age- and sex-matched subjects. (A) Box plot showing LASSO model scores for control subjects (n = 24) and patients with pulmonary arterial hypertension (PAH) (n = 119). (B) Receiver operating curve showing the performance of LASSO model scores for determining PAH status in the model validation group.

Figure 4.

Diagnostic RNA model and survival in pulmonary arterial hypertension (PAH). (A) Box plot of LASSO model score in control subjects and patients with PAH separated by survival status during follow-up. Dashed line shows the cutoff that identified 88.9% of nonsurviving patients with PAH. (B) Kaplan-Meier survival plot separating patients on the basis of the 88.9% sensitive cutoff. (C) Box plot of RNA model score in healthy volunteers and patients with PAH divided by WHO functional class. (D) Box plot of RNA model score in patients with PAH divided by presence of low, intermediate, or high levels of cardiac biomarkers BNP (brain natriuretic peptide) (<50, 50–300, or >300 pg/ml) or NT-proBNP (N-terminal pro-brain natriuretic peptide) (<300, 300–1,400, or >1,400 pg/ml), as per European guidelines for PAH assessment. LASSO = least absolute shrinkage and selection operator; WHO = World Health Organization.

Figure 5.

Whole-blood RNA levels of SMAD5 in control subjects and patients with pulmonary arterial hypertension stratified by genotype at the SMAD5 expression quantitative trait locus rs4146187.