Pulmonary arterial hypertension (PAH) is a rare disease characterised by remodelling of the pulmonary vasculature and subsequent increase in vascular resistance, leading to right ventricle overexertion and eventual heart failure. Responses to treatment and disease progression vastly differ between PAH patients, while standard clinical phenotypes are not sufficient for accurate patient stratification.
Molecular profiling through multi-omic approaches offers greater granularity for PAH patient characterisation and could improve initial risk stratification, treatment selection and monitoring, as well as providing insights into new biological pathways not yet targeted by current therapies.
Transcriptomic approaches, such as RNA-sequencing (RNAseq), permit comprehensive analyses of gene expression in tissue samples. Whole blood RNA analysis offers an alternative “liquid biopsy” to lung biopsy—which carries a high risk in PAH—and can be performed sequentially. Previous PAH whole blood transcriptomic studies have been limited by cohort numbers and methodologies used—such as the less sensitive and probe-dependent microarrays—, especially when compared with the comprehensive nature of RNAseq.
RNAseq followed by differential expression analysis were utilised to identify PAH-associated transcriptomic profiles in a cohort of 359 PAH patients and 72 age- and sex-matched healthy controls. A LASSO RNA diagnostic model of 25 differentially expressed genes best distinguishing between PAH patients and healthy controls derived from whole blood PAH RNA signature could effectively separate PAH patients in an independent group. RNA model scores were associated with disease severity (p= 0.008) and survival (p= 4.66x10-6) in patients. These results were validated externally in two different cohorts (including 58 and 156 PAH patients and 25 and 110 healthy controls respectively), which highlighted the potential role in PAH of adenosylmethionine decarboxylase 1 (AMD1) and polyamines. Mendelian randomisation analysis implicated SMAD5 in PAH pathogenesis (p= 0.028). A combined diagnostic model of 25 RNAs and 7 metabolites was developed which performed better at distinguishing PAH patients from healthy controls than the 7-metabolite model alone (p= 0.002).
A second analysis more sensitive to the specific characteristics of PAH was also conducted. A cohort of 147 PH patients and 45 disease controls was utilised to identify gene expression differences between PAH, other forms of PH and other symptomatic cardiovascular non-PH patients (“disease controls”). LASSO diagnostic models were developed from RNA profiles, and they were able to distinguish between disease controls and both PAH (p= 0.049) and all PH (p= 1.92x10-5) patients, but not between PAH and other PH (p= 0.719). Comparison of RNA profiles highlighted the potential role of SEC22B, ZNF254, PPA2, CAMKMT, FER and EDEM1 in PAH pathology.
An in-silico analysis for compound repurposing, the Connectivity Map (CMap) database, was utilised to predict small molecule compounds which reverse the identified whole blood PAH RNA signature. The protein synthesis inhibitor Homoharringtonine—approved for leukaemia treatment and one of the top six compounds identified by CMap—, was shown to effectively reverse PAH transcriptional changes in isolated PBMCs and hPAECs and induce positive functional changes in hPAECs in vitro.
These results emphasise the potential of transcriptomics—and general omics—for PAH research. PAH RNA profiles were identified that associated with disease progression and mortality and several transcripts were implicated in PAH pathology. RNA profiles separating PH patients from other symptomatic patients were also identified. Homoharringtonine showed therapeutic potential in in vitro assays using PBMCs and hPAECs and should be investigated further.