Circulating Proteomic Biomarkers in Systemic Sclerosis Related Pulmonary Arterial Hypertension

Achieving the diagnosis of pulmonary hypertension (PH) is difficult as, given the non-specific nature of symptoms, this condition can masquerade as multiple other conditions, most of which are more common. PH is known to be a complication of certain other disease processes such as systemic sclerosis, and for this reason screening for PH in this disease population is now standard practice. The optimal screening process remains unclear, with multi-modality testing as part of the ERS and DETECT protocols being current practice, but for which improvements are needed.

Survival for patients with systemic sclerosis (SSc) related pulmonary arterial hypertension (PAH) is significantly poorer than in patients with idiopathic pulmonary arterial hypertension. The reasons for this are incompletely understood and may include differences in the nature of underlying pulmonary vasculopathy as well as differences in the ability of the myocardium to compensate for increased right ventricular afterload. A greater understanding of underlying pathophysiological mechanisms, and biomarkers with the ability to aid early diagnosis and guide therapy is therefore needed.

Using pre-treatment serum samples from a tightly phenotyped cohort of systemic sclerosis patients both with and without pulmonary arterial hypertension, through collaboration with industry partners, we have data from a large unbiased proteomic screen of 296 serum protein concentrations. We hypothesise that these data either for individual proteins, or a combined panel of proteins could be used to accurately classify for pulmonary arterial hypertension in these patients.

Several bio-informatic techniques were tested, with a final two-step process separating variable selection from classification modelling. Final modelling was done using logistic regression with backward step-AIC optimisation. A final panel of consisting of 3 serum proteins was derived including Tetranectin, Protein DJ-1 and Growth differentiation factor-15.

When combined in a predictive model, these three proteins classify PAH in SSc with an accuracy of 85%, which compares favourably when measured against the current ERS/ESC guideline screening at 86%, and the DETECT protocol at 74%. When tested in an external validation cohort this model performed well with an AU-ROC 0.79.

Neither Tetranectin nor Protein DJ-1 have previously been described in PAH. We have demonstrated the presence of these proteins in lung tissues of patients with PAH, and have presented cell culture results which go some way to support theoretical mechanisms of action for these proteins in the pathophysiology of this condition.