Reclustering of Pulmonary Hypertension Patients Along Molecular Pathophysiology Traits Better Predicts Long‐Term Outcome Than Currently Used Classification

Pulmonary hypertension (PH) is a common notification for a biologically heterogeneous group of diverse pathophysiological conditions that end up in a common clinical symptom: elevated pulmonary artery pressure. Our increasing understanding of the heterogeneity of its pathophysiology has fueled classification into subgroups of the disease, based on the presumed underlying cause together with clinical and hemodynamic parameters. The current World Symposium on Pulmonary Hypertension (WSPH) classification system divides PH into 5 groups. ¹ This classification system, although widely accepted in the field, has a weakness in that it is strongly focused on cause rather than prognosis and outcome.

During recent years, however, understanding the molecular and cellular biology that drives development and progression of diseases has become more and more important, as more specific therapeutics directed at halting the progression of specific disease mechanisms and aiming to improve long‐term patient outcome have been developed.

In some diseases, this process has led to the necessity of implementing a novel classification system for the disease. For example, chronic heart failure, long classified by purely clinical criteria into New York Heart Association classes I to IV, is now subdivided into heart failure with reduced/preserved/mildly reduced ejection fraction following the observation that some patients experience heart failure despite normal ejection fraction. ² Diastolic relaxation of the ventricle had long been ignored as an important contributor to normal left ventricular function in clinical routine. A diversification of therapeutic recommendations has been the consequence of this. ³

The field of oncology has long had a leading role in molecular diagnostics that has stipulated the development of personalized treatment strategies based not on the anatomical location of a tumor, but rather on the individual expression of receptors, growth factors, and cell cycle regulators in the tumor cells. ⁴ For example, trastuzumab, a therapeutic monoclonal antibody directed against the human epidermal growth factor receptor‐2 (HER2), has been approved for treatment not of breast cancer in general, but for HER2‐positive breast cancer according to biopsy results. The same drug is also approved for treatment of HER2‐positive gastric cancer. An even more diversified extreme of this concept called tumor‐agnostic therapy ⁵ is pembrolizumab, a therapeutic monoclonal antibody directed against the programmed death‐ligand 1 (PD‐L1) receptor that can be found on many different tumors, but only in subgroups of these tumors expressing the target antigen. ⁶ The principal effort to introduce personalized treatment, based on cellular and molecular biomarkers rather than on clinical or anatomical disease entity, is slowly but steadily entering other fields of medicine. The use of large patient databases from controlled clinical trials and carefully designed registries has helped in this, and new bioinformatics methodologies have paved the way to efficiently extract the hidden information that such databases hold.

In this issue of the Journal of the American Heart Association (JAHA), Farha and coworkers have now undertaken an effort to rethink classification of PH in a different way than currently practiced. ⁷ They started with the hypothesis that any classification system should focus on hard end points of patient outcome, such as death and transplantation, in the case of PH. Together with deep phenotyping for major pathophysiological processes driving the disease, this might, according to their hypothesis, help to reveal novel subgrouping with shared biological mechanisms and stipulate the development of individualized treatment in a disease notoriously known to go along with an extremely poor prognosis. ⁸ Next, the group defined major common pathomechanisms in patients with PH. Based on previous studies from their group and others, they identified dysfunction of the endothelium to maintain NO‐mediated vasodilation, elevated generation of D‐dimers secondary to endothelial activation and intravascular thrombosis, and preservation of the diffusion capacity of the lung for carbon monoxide as major pathomechanisms that vary widely across and within WSPH subgroups, suggesting a potential for different, novel clustering.

First, this group applied an unsupervised clustering approach that included a wide range of clinical parameters readily available for each patient using a large registry of 1529 patients with PH, the CC‐PH (Cleveland Clinic Pulmonary Hypertension) registry. Having shown that differentiation into 3 clusters was optimal for separation of transplantation‐free survival groups, they validated their model in the independent PVDOMICS (Pulmonary Vascular Diseases Phenomics) cohort that comprised 853 individuals. They showed that the 3 clusters outperformed conventional classification of PH with respect to predicting transplantation‐free survival.

Finally, the authors analyzed whether metabolites of the L‐arginine‐NO signaling pathway, D‐dimers, and lung diffusion capacity differed between the new clusters. They showed that the cluster with the worst prognosis had the lowest L‐arginine bioavailability and levels of nitrate (the oxidative metabolite of NO), the highest D‐dimer levels, and reduced diffusion capacity of the lung for carbon monoxide lung diffusion capacity.

Both novel ways of classification, either by multiple clinical variables or by pathophysiological markers of endothelial dysfunction, intravascular microthrombosis, and potential pulmonary vascular rarefication, resulted in clusters with clearly separated survival curves. By contrast, the authors demonstrated that in the same cohorts, the current WSPH classification generated largely overlapping survival curves for the 5 commonly used groups of PH. The authors' conclusion was that PH phenotypes defined by clusters of endothelial health and pulmonary vascular rarefication contribute more substantially to mortality across WSPH subgroups than the currently practiced focus on pulmonary artery pressure and right heart function.

When looking in detail at the alterations in L‐arginine, NO metabolism in the newly defined cluster 3, one finds elevated circulating concentrations of asymmetric dimethylarginine (ADMA) and symmetric dimethylarginine, 2 methylated derivatives of L‐arginine of which 1 (ADMA) directly inhibits NO synthesis and has been reported multiple times to be a marker of major cardiovascular event and mortality risk. ⁹ , ¹⁰ , ¹¹ Elevated ADMA has also been shown to be a marker of survival in WSPH group 1 PH. ¹² Symmetric dimethylarginine, by contrast, does not directly inhibit NO synthesis; nonetheless, it has been shown to be a signal for increased mortality risk in stroke patients ¹³ and to be associated with previously undetected atrial fibrillation. ¹⁴ It is a pity that the analytical method used in the study by Farha and colleagues ⁷ did not allow for discrimination between these 2 metabolites. In addition, the study revealed low L‐arginine plasma concentration along with elevated concentrations of L‐citrulline and L‐ornithine. L‐ornithine is a product of the conversion of L‐arginine by arginase; arginase activation has been reported previously to go along with L‐arginine deficiency and ensuing vascular disease. L‐citrulline concentration is more difficult to interpret, because L‐citrulline is a by‐product of NO synthase from L‐arginine and may thus indicate increased NO synthesis. However, increased NO synthesis is somewhat unexpected in the presence of elevated concentrations of the NO synthesis inhibitor ADMA. Rather, L‐citrulline is also a product of the enzymatic degradation of ADMA and in the presence of reduced concentrations of the oxidative metabolite of NO, nitrate. Thus, elevated L‐citrulline may also signify compensatory increased metabolic clearance of ADMA. Although these pathobiochemical details need to be clarified in future translational studies, it is clear from the present study that impaired NO production along with at least 2 of its common sequelae, intravascular microthrombosis (as signaled by elevated D‐dimer levels) and pulmonary vascular rarefication (possibly due to thrombotic occlusion of intrapulmonary arterioles) appears as a major driver of mortality risk in cluster 3 of this study.

When we think through this concept to the end, we may end up with a differentiated treatment strategy that is also guided by pathophysiological mechanisms in PH. It can be expected that a combination of drugs restituting the endothelial NO/cyclic guanosine monophosphate (cGMP) pathway with other compounds directed against intravascular microthrombosis will exert a maximized therapeutic benefit in cluster 3 patients, whereas other treatment concepts may be better in patients within clusters 1 and 2, independently of the conventional PH group. Although this hypothesis clearly needs to be tested in appropriately designed randomized clinical trials, the study presented by Farha and coworkers opens the perspective for individualized, pathophysiologically driven treatment strategies in PH.

Disclosures

None.