Rationale: NT-proBNP (N-terminal pro–brain natriuretic peptide), a biomarker of cardiac origin, is used to risk stratify patients with pulmonary arterial hypertension (PAH). Its limitations include poor sensitivity to early vascular pathology. Other biomarkers of vascular or systemic origin may also be useful in the management of PAH.
Objectives: Identify prognostic proteins in PAH that complement NT-proBNP and clinical risk scores.
Methods: An aptamer-based assay (SomaScan version 4) targeting 4,152 proteins was used to measure plasma proteins in patients with idiopathic, heritable, or drug-induced PAH from the UK National Cohort of PAH (n = 357) and the French EFORT (Evaluation of Prognostic Factors and Therapeutic Targets in PAH) study (n = 79). Prognostic proteins were identified in discovery–replication analyses of UK samples. Proteins independent of 6-minute-walk distance and NT-proBNP entered least absolute shrinkage and selection operator modeling, and the best combination in a single score was evaluated against clinical targets in EFORT.
Measurements and Main Results: Thirty-one proteins robustly informed prognosis independent of NT-proBNP and 6-minute-walk distance in the UK cohort. A weighted combination score of six proteins was validated at baseline (5-yr mortality; area under the curve [AUC], 0.73; 95% confidence interval [CI], 0.63–0.85) and follow-up in EFORT (AUC, 0.84; 95% CI, 0.75–0.94; P = 9.96 × 10−6). The protein score risk stratified patients independent of established clinical targets and risk equations. The addition of the six-protein model score to NT-proBNP improved prediction of 5-year outcomes from AUC 0.762 (0.702–0.821) to 0.818 (0.767–0.869) by receiver operating characteristic analysis (P = 0.00426 for difference in AUC) in the UK replication and French samples combined.
Conclusions: The plasma proteome informs prognosis beyond established factors in PAH and may provide a more sensitive measure of therapeutic response.
Pulmonary arterial hypertension (PAH) is a highly morbid condition with high mortality and a heterogeneous response to established therapies. A number of studies have reported the association between plasma concentrations of candidate proteins and survival in PAH. Plasma proteomics is able to survey a range of proteins without bias and is well suited to identifying clinically useful and biologically relevant biomarkers in complex diseases. A comprehensive assessment of the plasma proteome in a prospectively designed multicenter study to define the most important proteins associated with PAH severity and prognosis is timely.
This study provides the most comprehensive coverage of plasma proteins measured in patients with PAH to date. By discovery–replication and model development in the multicenter UK PAH Cohort study followed by validation in the French EFORT cohort, we identify novel proteins strongly associated with prognosis and derive a six-protein model score as an easy-to-interpret single measurement. We directly demonstrate utility of this protein score in risk stratification, independent of established markers, notably specific measures of exercise capacity and cardiac blood biomarker NT-proBNP, as well as clinical risk equations.
Pulmonary arterial hypertension (PAH) is a rare condition associated with high morbidity and mortality (1). The current licensed treatments may improve symptoms and functional capacity, but the clinical response varies between patients, and the benefits can be short lived (2). Recent progress in the management of patients has been dependent on combining and escalating treatments based on a systematic assessment of clinical response (3). A number of multiparametric risk calculators, combining clinical, exercise, right ventricular function, and hemodynamic parameters, can be used to stratify patients into high, intermediate, and low risk of death before and at intervals after starting therapy (4–6). Treatment is adjusted to achieve and maintain patients at low risk (3).
A key component of these risk calculators is a measurement of circulating BNP (brain natriuretic peptide) or NT-proBNP (N-terminal proBNP). BNP and NT-proBNP are secreted by cardiac myocytes in response to increased ventricle wall strain (7) and raised plasma concentrations in patients with PAH predict increased mortality (8, 9). Although broadly useful, the threshold plasma concentrations used to classify patients as high, medium, and low risk have limitations. In a recent report, around 40% of patients in the low-risk category, with a NT-proBNP <300 ng/L, were in World Health Organization (WHO) functional class III, and 32% of patients in the high-risk category, with a NT-proBNP >1,400 ng/L, were in functional class II (9). Other circulating proteins arising from the pulmonary vascular pathology (10) or systemic involvement in pulmonary hypertension (11) may be useful in place of, or in addition to, BNP and NT-proBNP to inform risk stratification. To date, this hypothesis has been largely investigated on a candidate-by-candidate basis (12).
The plasma proteome, the totality of proteins in plasma secreted and leaked from tissues, including the vascular bed, offers a measure of health as circulating concentrations change with disease (13). We have previously shown plasma proteomics can be used to separate patients with PAH into low- and high-risk groups using an aptamer-based assay (14). Here we use the expanded version of the assay (SomaScan version 4; SomaLogic, Inc.), which allows quantification of more than 4,000 proteins, to develop and investigate the clinical utility of a small panel of plasma proteins identified from an unbiased screen. A panel of six proteins was derived that classified risk, both before and after initiation of PAH targeted therapy, independent of established clinical targets, including circulating NT-proBNP concentrations.
A prognostic protein panel was derived from patients with prevalent idiopathic, heritable, or drug-induced PAH aged 18–65 years (n = 357) recruited to the UK National Cohort Study of Idiopathic and Heritable PAH (clinicaltrials.gov NCT01907295) between February 19, 2014, and November 6, 2018. The diagnostic criteria for idiopathic and heritable PAH over the course of this study have been stable; a raised mean pulmonary arterial pressure ⩾ 25 mm Hg with mean pulmonary artery pressure ⩽ 15 and pulmonary vascular resistance > 3 Wood units at rest with exclusion of known associated diseases according to contemporary international consensus (15). Seventy age- and sex-matched (median age, 42; interquartile range, 33–49; 49 [70%] females) healthy control subjects without cardiovascular or respiratory diseases were recruited over the same period from the same centers. Survival status for patients with PAH was censored on March 14, 2020. After a median follow-up of 4.7 years from sampling, 65 deaths and 13 transplants had occurred; lung transplantation or death was used as a composite endpoint.
The panel was validated in the EFORT (Evaluation of Prognostic Factors and Therapeutic Targets in PAH) study in France (ClinicalTrials.gov: NCT01185730) (16). Here, 79 patients with PAH were sampled at diagnosis (incident cases), between January 11, 2011, and December 9, 2013, and resampled at follow-up visits averaging 7 months apart (range, 2.8–28.5 mo). Survival status for patients with PAH was censored on May 11, 2020. After a median follow-up of 7.1 years, 22 deaths and five transplants had occurred.
Peripheral venous plasma ethylenediaminetetraacetic acid samples were collected as previously described (14) and were obtained with informed consent and research ethics committee approval (13/EE/0203, 17/LO/0563, and 17/LO/0565). Patients were not fasting and were sampled at their routine clinical appointment visits. The plasma samples underwent one freeze–thaw cycle to aliquot 120 μl for proteomic analysis and provide other aliquots for NT-proBNP and targeted assays. Clinical and biochemical data were collected within 30 days and 7 days, respectively, of blood sampling.
Proteomic analysis was performed using the SomaScan version 4 (17), and technicians were blinded to patient status. The SomaScan version 4 assay lists 5,284 aptamers, and we included 4,349 (targeting 4,152 unique human proteins) for analysis after removal of 305 nonhuman/nonprotein targets and quality control to select only those with stable measurements, defined as <20% coefficient of variance in the repeated pooled plasma assay controls. Relative fluorescence units were log10 transformed to normalize protein concentrations and then corrected for the first two principal components (derived from all 4,349 aptamers) by linear regression to correct for population-stratification or sample-quality differences. Finally, protein concentrations were standardized to healthy control amounts (converted to z-scores using the mean and SD in controls as reference) for ease of interpretation of results and comparability of proteins. Four prognostic proteins with commercially available assays were selected for confirmation by ELISA from R&D Systems (DTSP20 1/5 dilution in standard buffer, DNRP10 1/200 dilution in standard buffer, DREN00 1:1 in calibrator diluent) and Abnova (KA2121 1/20 dilution) in plasma samples from patients with PAH selected for having a wide range of protein values from the proteomic measurements.
Patients and controls from the UK dataset were randomized into discovery and replication groups in a 2:1 ratio to adequately power discovery analysis of all proteins and replication of proteins meeting statistical significance (Table 1). All comparisons were corrected for multiple testing using Benjamini-Hochberg false discovery rate.
|UK Discovery PAH Patients (n = 238)||UK Replication PAH Patients (n = 119)||French EFORT PAH Patients (n = 79)||P Value|
|Idiopathic/heritable/drug-associated PAH, n (%)||216/22/0 (91/9/0)||106/13/0 (89/11/0)||53/14/12 (67/18/15)|
|Age at diagnosis/recruitment||39.1 (30.4–47.9)||41.9 (33.8–49.5)||52.1 (34.1–65.5)||<0.001|
|Sex, n (%)|
|F||177 (74)||86 (72)||56 (71)||0.81|
|M||61 (26)||33 (28)||23 (29)|
|WHO functional class, n (%)|
|I||3 (1)||5 (4)||4 (5)||0.054|
|II||45 (20)||23 (20)||26 (33)|
|III||146 (64)||69 (61)||43 (54)|
|IV||34 (15)||16 (14)||6 (8)|
|6-min-walk distance, m||353 (260–421)||352 (270–431)||360 (250–453)||0.84|
|Mean pulmonary artery pressure, mm Hg||55 (48–65)||54 (47–61)||50 (44–59)||0.003|
|Mean pulmonary artery wedge pressure, mm Hg||9 (7–12)||10 (7–11)||9 (7–11)||0.94|
|Mean right atrial pressure, mm Hg||9 (6–13)||8 (6–12)||8 (5–11)||0.24|
|Cardiac index, L/min/m2||2.0 (1.7–2.5)||2.0 (1.6–2.7)||2.39 (2.01–2.69)||0.005|
|Pulmonary vascular resistance, dyn · s · cm−5||1,006 (659–1,350)||817 (576–1,238)||792 (584–1,016)||0.005|
Prognostic proteins were identified by Cox regression analysis, correcting for age and sex, and all-cause mortality or lung transplant for severe PAH was included as an event (composite endpoint for primary analyses). To prioritize proteins independent of known prognostic factors, models including 6-minute-walk distance (6MWD) or NT-proBNP were constructed. To avoid analyzing tenascin and SVEP1 more than once, the most significant aptamer from the analysis including NT-proBNP was chosen for these two proteins. To identify the combination of markers that best predicted prognosis, a least absolute shrinkage and selection operator (LASSO) modeling approach using k-fold cross-validation (k = 10) was applied, with regularization parameter (lambda) determined by the lowest error plus 1 SE (to minimize overfitting) using the glmnet version 2.0–18 R package from CRAN. This produces a protein score from a Cox regression model of the proteins identified by the LASSO analysis. Receiver operating characteristic (ROC) analysis was performed using the survivalROC v1.0.3 R package.
This prognostic protein model score was then investigated in the samples from the French EFORT study and compared with risk indicators (clinical targets) as described in the international guidelines for PAH (15). The thresholds used were as follows: WHO functional class I or II, 6MWD > 440 m, cardiac index ⩾ 2.5 L/min/m2, mean right atrial pressure < 8 mm Hg, and BNP < 50 ng/L or NT-proBNP < 300 ng/L. Less than 5% of data were missing and were classed as not met. Kaplan-Meier survival estimates were calculated and plotted using the survival version 3.1–8 R package.
Data are presented as percentages, mean (±SD), 95% confidence interval (CI), or median and percentile range. Analysis was performed in R (version 3.6.3) and SPSS (version 26; IBM).
Table 1 details the baseline characteristics of the study groups, analyzed as summarized in Figure 1. UK patients were younger and had a higher mean pulmonary arterial pressure and pulmonary vascular resistance and a lower cardiac index.
Discovery and replication subgroups of patients from the UK National Cohort Study of Idiopathic and Heritable PAH were used to investigate the association of 4,152 plasma protein concentrations (measured by 4,349 aptamers) with survival. Cox regression models were adjusted for age and sex. A total of 606 proteins exceeded false discovery rate (q < 0.05) in the discovery cohort and 49 in the replication cohorts. 6MWD and plasma NT-proBNP concentrations are established noninvasive risk assessment tools in PAH. Thirty-one proteins predicted survival independent of 6MWD, age, and sex (P < 0.05) (Figure 2) and of NT-proBNP, age, and sex (see Table E1 in the online supplement).
LASSO analysis of combinations of proteins in a Cox survival regression model selected six proteins (SVEP1, PXDN [peroxidasin homolog], renin, NRP1 [neuropilin-1], TSP2 [thrombospondin-2], and PRDX4 [peroxiredoxin-4]) to create a single model score in the pooled discovery and replication patient subgroups (Table E2 and Figure E2). The Cox regression-based score accurately discriminated 5-year transplant-free survivors from nonsurvivors in the entire UK PAH cohort (area under the curve [AUC], 0.82; 95% CI, 0.766–0.875) (Figure E3). The score was not significantly different in heritable pulmonary arterial hypertension cases or between sexes; of the individual proteins, only NRP1 was associated with sex, but it was equally elevated in male or female patients who died or were transplanted before 5 years of follow-up (Figure E4).
We calculated the six-protein model score in the EFORT study and saw a very similar distribution (UK: mean 0.499 ± 0.499 SD; EFORT: 0.479 ± 0.525) (Figure E5 and Table E3). We confirmed that the six-protein model was able to predict 5-year outcomes (AUC, 0.73; 95% CI, 0.63–0.85) in this geographically separate group of patients (Figure E3). The score predicted survival at both diagnosis (Figure 3A) and follow-up sampling (Figure 3B). Combining the UK replication and EFORT samples, the addition of the six-protein model score to NT-proBNP improved prediction of 5-year outcomes from AUC 0.721 (0.627–0.815) to 0.783 (0.707–0.860) by ROC analysis (P = 0.0424 for difference in AUC) (Figure 4).
The best-performing cutoff (highest sensitivity + specificity) from the UK analysis, 0.56 (80% sensitivity, 74% specificity for 5-year transplant-free survival; AUC, 0.82 ± 0.028; P = 3.09 × 10−17), predicted survival with 89% sensitivity and 69% specificity in follow-up EFORT samples taken from patients after starting treatment (AUC, 0.84; 95% CI, 0.75–0.94). The PAH therapies of the EFORT patients are summarized in Table E4.
The six-protein score was able to further stratify risk in those patients already predicted to have a good outcome based on meeting two or more clinical targets at follow-up in the EFORT study (Figure 5 and Table E5). Those patients meeting at least two clinical targets but with higher protein scores met on average 2.6 (vs. 3.8 with low scores) clinical targets. This was also shown in subanalyses of both high-risk (meeting zero targets) and lower-risk patients (meeting at least two targets) in the UK PAH Cohort (Figure E6). The individual-level data for survivors and nonsurvivors in the EFORT study (Figure E7) show that scores change from diagnosis to follow-up post-therapy in survivors from 0.394 ± 0.531 to 0.367 ± 0.489 and in nonsurvivors from 0.768 ± 0.397 to 1.054 ± 0.519. In patients with low or high risk at baseline based on ROC-derived protein score cutoff, a change in the protein score was also associated with different outcomes despite no clear differences in clinical presentation (Figures 3C and 3D and Table E6).
Cox regression models confirmed the score was prognostic independent of achievement of clinical targets at follow-up and that change in the model was independent of risk at baseline (based on best prognostic ROC-derived protein score cutoff) (Table 2). Furthermore, the six-protein model independently predicted risk when evaluated against the established French risk equation in both the UK cohort and EFORT studies (Table 2) (4).
|Hazard Ratio (95% Confidence Intervals)||P Value|
|Model 1: clinical targets in EFORT|
|Six-protein model sample 2 ROC cutoff||3.72 (1.49–9.307)||0.0049|
|Clinical targets at follow-up (⩾2 or not)||0.232 (0.103–0.526)||0.000459|
|Model 2: change in model score in EFORT|
|Six-protein model sample 1 ROC cutoff||6.119 (2.102–17.819)||0.000894|
|Change in six-protein model||3.994 (1.711–9.324)||0.00137|
|Model 3: comparison with French risk equation in UK Cohort|
|Six-protein model score||5.486 (3.243–9.279)||2.20 × 10−10|
|French risk equation||1.652 (1.068–2.556)||0.0242|
|Model 4: comparison with French risk equation in EFORT baseline|
|Six-protein model score||2.138 (1.065–4.293)||0.033|
|French risk equation||4.818 (2.101–11.047)||0.0002|
|Model 5: comparison with French risk equation in EFORT follow-up|
|Six-protein model score||2.781 (1.315–5.88)||0.007|
|French risk equation||5.805 (1.297–25.986)||0.021|
The previously published nine-protein panel score (14) was prognostic in the UK samples (Figure E8) but was outperformed by the novel six-protein score in 5-year ROC analysis (Figure E8). We validated the SomaLogic proteomic measurements of four of the six proteins of interest, TSP-2, renin, PRDX4, and NRP-1, using commercially available targeted ELISA assays (Spearman’s rho = 0.3–0.9; P < 0.05 for all) (Figure E9). A sensitivity analysis, examining the power to predict survival using models constructed without one or more of the proteins, demonstrated that loss of one or two proteins was well tolerated in the UK dataset; a model combining the four proteins validated by ELISA achieved an AUC of 0.841 ± 0.026; P = 2.7 × 10−19 (Figure E10). This four-protein model also performed well in the follow-up EFORT samples (AUC, 0.791 ± 0.05; P = 0.000488).
The Human Peptide Atlas has, to date, cataloged more than 3,500 proteins in plasma, with additional evidence for another 1,300 proteins (13). We have used an aptamer-based assay to investigate how plasma concentrations of over 4,000 proteins relate to clinical outcome in a large well-phenotyped UK cohort of patients with idiopathic, heritable, or drug-induced PAH. We developed a prognostic score based on six proteins that predict survival independent of NT-proBNP and 6MWD and validated this score in a multicenter study of patients from France, where patients were sampled at diagnosis and again following initiation of targeted therapies.
The plasma proteome is stable in health (18). There is interest in how changes in the plasma protein profile can be used to track disease. This has been underexplored in PAH; despite significant numbers of observational studies indicating the potential utility of blood biomarkers (19), only plasma NT-proBNP and troponin concentrations are used to inform clinical decisions. These two proteins report on cardiac status, troponin indicating myocardial injury and necrosis, and BNP, a surrogate of myocardial wall stress. We reasoned that changes in plasma proteins that report directly upon the vascular remodeling process might inform risk in addition to proteins that report on right heart strain and cardiac function. By focusing on proteins that tracked survival independent of NT-proBNP and 6MWD, both measures of right heart function, a major predictor of mortality in PAH, we hypothesized we could build a score that was useful in addition to established clinical targets. In this context, the performance of the protein score in the EFORT study is clinically important. Specifically, the protein score added greater granularity to risk stratification by clinical targets based on functional class, 6MWD, cardiac index, mean right atrial pressure, and BNP/NT-proBNP and identified patients at risk even if two or more clinical targets were met at follow-up, suggesting it could be used in combination with established clinical targets.
The six proteins that comprise the score emerged from statistical modeling but have biological plausibility. Two of them have clear functional links to vascular remodeling and fibrosis. PXDN is induced by TGF-β (transforming growth factor β) in human pulmonary fibroblasts, and the secreted protein is incorporated into the extracellular matrix, colocalizing with fibronectin (20) and harnessing bromine to stabilize collagen IV scaffolds (21). It promotes angiogenesis through activation of Akt and FAK (focal adhesion kinase) (22). SVEP1, also known as Polydom, is a ligand for integrin α9β1 (23) and a breast cancer antigen (24), with genetic links to cardiovascular and specifically coronary disease (25, 26). It also has a crucial conserved role in lymphatic development (27).
Two other proteins have links to endothelial function. NRP1 is implicated in angiogenesis, cell survival and migration, acting as a coreceptor for VEGF (vascular endothelial growth factor) and semaphorins, and cardiac regeneration in zebrafish (28). Interestingly, NRP1 was also recently identified in a diagnostic protein panel for systemic sclerosis–associated PAH (29). NRP1 appears essential for the signaling of ANGPTL4 (angiopoietin-like 4). Treatment with sNRP1 (soluble fragment of NRP1) prevented ANGPTL4 from binding to NRP1, blocking ANGPTL4-induced activation of RhoA and endothelial permeability in vitro and retinal vascular leakage in vivo. NRP1 is hypoxia sensitive, and upregulation has been reported in studies of adaptation to high altitude (30).
TSP2 (thrombospondin-2) is a secreted matricellular protein, and increased concentrations are found in heart failure (31, 32), with cardiac fibroblasts a potential source (33, 34). However, gene expression in arterial tissue is high relative to other nondiseased human tissues (Genotype-Tissue Expression database, version 8 release) and changes in gene expression found in laser microdissected pulmonary vessels from patients with pulmonary hypertension associated with pulmonary fibrosis (35). TSP2 inhibits human microvascular endothelial cell proliferation (36), and TSP2 knockout mice show enhanced angiogenesis (37). The precise signaling pathways involved are unclear, as TSP2 is able to bind many different ligands (38), but TSP2 represses matrix metalloproteinases-2 and -9 and interactions with surface receptors (39). Impaired TSP2 activity would appear to be detrimental to vascular and cardiac homeostasis. Elevated circulating TSP2 would be consistent with a compensatory response in PAH, in an attempt to reduce pulmonary vascular damage.
PRDX4 is an antioxidant enzyme that regulates the activation of NF-κB in the cytosol by modulation of IκBα phosphorylation (40). PRDX4 concentrations are elevated in idiopathic pulmonary fibrosis, and overexpression worsens bleomycin-induced IPF in mice (41). Galectin-3 (which is elevated in heart failure) reduces PRDX4 concentrations, promoting cardiac fibrosis (42).
Finally, renin concentrations likely relate to the systemic consequences of PAH and the cross-talk between right heart function and the kidney (11). Reduced cardiac output affects efferent renal arteriolar blood flow, which stimulates release of renin from juxtaglomerular cells. The systemic activation of the renin–angiotensin–aldosterone system in patients with PAH is well documented (43, 44). Impaired renal perfusion may activate circulating factors, such as TNF-α and IL-1β and -6, that can aggravate pulmonary vascular disease (45).
The proteins that comprise the six-panel score emerged from robust statistical modeling of the largest plasma proteome study in PAH to date. It was generated as a practical tool for risk stratification, to be used in addition to NT-proBNP and clinical risk factors. The improvement in AUC in the replication analysis from 0.721 to 0.783 is significant but indicates that further improvements in risk stratification of PAH are needed, and this approach may be complemented by other noninvasive biomarkers, for example, cardiac magnetic resonance (46). The utility of the six-protein panel in routine patient management will need to be prospectively evaluated in a clinical study.
Such a study would be best conducted with specific targeted assays. The proteomic assay used here has the broadest protein coverage of the currently available platforms, but data are provided as relative concentrations rather than absolute units. We identified commercially available ELISAs to demonstrate reproducibility of four protein measurements (TSP2, renin, NRP1, and PRDX4) in this study. TSP2, renin, and NRP1 SomaScan measurements have also been validated by mass spectrometry (47), and SVEP1 and PXDN are supported by association with cis protein quantitative loci (47, 48). Reproducible immunoassays are not yet available for SVEP1 and PXDN, but our sensitivity analysis suggests the four ELISA-validated proteins could be used with marginal loss of information. Development of rapid, automated testing of these proteins on a widely available platform at costs comparable to other biomarker assays, such as BNP, would be required for the panel to be routinely useful.
Through an unbiased screen of the plasma proteome, we have identified and validated a minimal panel of six proteins, each a plausible candidate for a biological role in PAH, that complements the use of NT-proBNP and clinical risk factors to risk stratify patients with PAH.
The authors thank National Institute for Health Research (NIHR) BioResource volunteers for their participation and gratefully acknowledge NIHR BioResource centres, National Health Service Trusts, and staff for their contribution. The authors thank the NIHR Imperial Clinical Research Facility, NIHR Sheffield Biomedical Research Centre, and National Health Service Blood and Transplant.
|1.||National Audit of Pulmonary Hypertension, 10th Annual Report. National Health Service; 2019 [published 2019 Oct 24; accessed 2020 Feb 10]. Available from: https://digital.nhs.uk/data-and-information/publications/statistical/national-pulmonary-hypertension-audit/2019.|
|2.||Boucly A, Weatherald J, Savale L, Jaïs X, Cottin V, Prevot G, et al. Risk assessment, prognosis and guideline implementation in pulmonary arterial hypertension. Eur Respir J 2017;50:1700889.|
|3.||Galiè N, Channick RN, Frantz RP, Grünig E, Jing ZC, Moiseeva O, et al. Risk stratification and medical therapy of pulmonary arterial hypertension. Eur Respir J 2019;53:1801889.|
|4.||Sitbon O, Benza RL, Badesch DB, Barst RJ, Elliott CG, Gressin V, et al. Validation of two predictive models for survival in pulmonary arterial hypertension. Eur Respir J 2015;46:152–164.|
|5.||Benza RL, Gomberg-Maitland M, Elliott CG, Farber HW, Foreman AJ, Frost AE, et al. Predicting survival in patients with pulmonary arterial hypertension: the REVEAL risk score calculator 2.0 and comparison with ESC/ERS-based risk assessment strategies. Chest 2019;156:323–337.|
|6.||Sitbon O, Nikkho S, Benza R, Cq Deng C, W Farber H, Gomberg-Maitland M, et al. Novel composite clinical endpoints and risk scores used in clinical trials in pulmonary arterial hypertension. Pulm Circ 2020;10:2045894020962960.|
|7.||Goetze JP, Bruneau BG, Ramos HR, Ogawa T, de Bold MK, de Bold AJ. Cardiac natriuretic peptides. Nat Rev Cardiol 2020;17:698–717.|
|8.||Fijalkowska A, Kurzyna M, Torbicki A, Szewczyk G, Florczyk M, Pruszczyk P, et al. Serum N-terminal brain natriuretic peptide as a prognostic parameter in patients with pulmonary hypertension. Chest 2006;129:1313–1321.|
|9.||Chin KM, Rubin LJ, Channick R, Di Scala L, Gaine S, Galiè N, et al. Association of N-terminal pro brain natriuretic peptide and long-term outcome in patients with pulmonary arterial hypertension. Circulation 2019;139:2440–2450.|
|10.||Humbert M, Guignabert C, Bonnet S, Dorfmüller P, Klinger JR, Nicolls MR, et al. Pathology and pathobiology of pulmonary hypertension: state of the art and research perspectives. Eur Respir J 2019;53:1801887.|
|11.||Rosenkranz S, Howard LS, Gomberg-Maitland M, Hoeper MM. Systemic consequences of pulmonary hypertension and right-sided heart failure. Circulation 2020;141:678–693.|
|12.||Llucià-Valldeperas A, van Wezenbeek J, Goumans MJ, de Man FS. The battle of new biomarkers for right heart failure in pulmonary hypertension: is the queen of hearts NT-proBNP defeated at last? Eur Respir J 2021;57:2004227.|
|13.||Schwenk JM, Omenn GS, Sun Z, Campbell DS, Baker MS, Overall CM, et al. The human plasma proteome draft of 2017: building on the human plasma peptideatlas from mass spectrometry and complementary assays. J Proteome Res 2017;16:4299–4310.|
|14.||Rhodes CJ, Wharton J, Ghataorhe P, Watson G, Girerd B, Howard LS, et al. Plasma proteome analysis in patients with pulmonary arterial hypertension: an observational cohort study. Lancet Respir Med 2017;5:717–726.|
|15.||Galiè N, Humbert M, Vachiery JL, Gibbs S, Lang I, Torbicki A, et al. 2015 ESC/ERS guidelines for the diagnosis and treatment of pulmonary hypertension: The Joint Task Force for the Diagnosis and Treatment of Pulmonary Hypertension of the European Society of Cardiology (ESC) and the European Respiratory Society (ERS): Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC), International Society for Heart and Lung Transplantation (ISHLT). Eur Respir J 2015;46:903–975.|
|16.||Sitbon O, Clerson P, Cottin V, Degroote P, Prevot G, Savale L, et al. A dynamic prognostic model to predict survival and determine treatment goals in pulmonary arterial hypertension (PAH): the EFORT study. Eur Respir J 2019;54:OA497.|
|17.||Gold L, Ayers D, Bertino J, Bock C, Bock A, Brody EN, et al. Aptamer-based multiplexed proteomic technology for biomarker discovery. PLoS One 2010;5:e15004.|
|18.||Zhong W, Gummesson A, Tebani A, Karlsson MJ, Hong MG, Schwenk JM, et al. Whole-genome sequence association analysis of blood proteins in a longitudinal wellness cohort. Genome Med 2020;12:53.|
|19.||Hemnes A, Rothman AMK, Swift AJ, Zisman LS. Role of biomarkers in evaluation, treatment and clinical studies of pulmonary arterial hypertension. Pulm Circ 2020;10:2045894020957234.|
|20.||Péterfi Z, Donkó A, Orient A, Sum A, Prókai A, Molnár B, et al. Peroxidasin is secreted and incorporated into the extracellular matrix of myofibroblasts and fibrotic kidney. Am J Pathol 2009;175:725–735.|
|21.||McCall AS, Cummings CF, Bhave G, Vanacore R, Page-McCaw A, Hudson BG. Bromine is an essential trace element for assembly of collagen IV scaffolds in tissue development and architecture. Cell 2014;157:1380–1392.|
|22.||Medfai H, Khalil A, Rousseau A, Nuyens V, Paumann-Page M, Sevcnikar B, et al. Human peroxidasin 1 promotes angiogenesis through ERK1/2, Akt, and FAK pathways. Cardiovasc Res 2019;115:463–475.|
|23.||Sato-Nishiuchi R, Nakano I, Ozawa A, Sato Y, Takeichi M, Kiyozumi D, et al. Polydom/SVEP1 is a ligand for integrin α9β1. J Biol Chem 2012;287:25615–25630.|
|24.||Scanlan MJ, Gout I, Gordon CM, Williamson B, Stockert E, Gure AO, et al. Humoral immunity to human breast cancer: antigen definition and quantitative analysis of mRNA expression. Cancer Immun 2001;1:4.|
|25.||Kichaev G, Bhatia G, Loh PR, Gazal S, Burch K, Freund MK, et al. Leveraging polygenic functional enrichment to improve GWAS power. Am J Hum Genet 2019;104:65–75.|
|26.||Stitziel NO, Stirrups KE, Masca NG, Erdmann J, Ferrario PG, Konig IR; MyoCardial Infarction Genetics and CARDiGRAM Exome Consortia Investigators. Coding variation in ANGPTL4, LPL, and SVEP1 and the risk of coronary disease. N Engl J Med 2016;374:1134–1144.|
|27.||Karpanen T, Padberg Y, van de Pavert SA, Dierkes C, Morooka N, Peterson-Maduro J, et al. An evolutionarily conserved role for polydom/SVEP1 during lymphatic vessel formation. Circ Res 2017;120:1263–1275.|
|28.||Lowe V, Wisniewski L, Sayers J, Evans I, Frankel P, Mercader-Huber N, et al. Neuropilin 1 mediates epicardial activation and revascularization in the regenerating zebrafish heart. Development 2019;146:dev174482.|
|29.||Bauer Y, de Bernard S, Hickey P, Ballard K, Cruz J, Cornelisse P, et al. Identifying early pulmonary arterial hypertension biomarkers in systemic sclerosis: machine learning on proteomics from the DETECT cohort. Eur Respir J 2021;57:2002591.|
|30.||Gaur P, Saini S, Ray K, Asanbekovna KN, Akunov A, Maripov A, et al. Temporal transcriptome analysis suggest modulation of multiple pathways and gene network involved in cell-cell interaction during early phase of high altitude exposure. PLoS One 2020;15:e0238117.|
|31.||Kimura Y, Izumiya Y, Hanatani S, Yamamoto E, Kusaka H, Tokitsu T, et al. High serum levels of thrombospondin-2 correlate with poor prognosis of patients with heart failure with preserved ejection fraction. Heart Vessels 2016;31:52–59.|
|32.||Chan MY, Efthymios M, Tan SH, Pickering JW, Troughton R, Pemberton C. Prioritizing candidates of post-myocardial infarction heart failure using plasma proteomics and single-cell transcriptomics. Circulation [online ahead of print] 4 Sep 2020; DOI: 10.1161/CIRCULATIONAHA.119.045158. [Published erratum appears in Circulation 142:e234]|
|33.||Egerstedt A, Berntsson J, Smith ML, Gidlöf O, Nilsson R, Benson M, et al. Profiling of the plasma proteome across different stages of human heart failure. Nat Commun 2019;10:5830.|
|34.||Wells QS, Gupta DK, Smith JG, Collins SP, Storrow AB, Ferguson J, et al. Accelerating biomarker discovery through electronic health records, automated biobanking, and proteomics. J Am Coll Cardiol 2019;73:2195–2205.|
|35.||Hoffmann J, Wilhelm J, Marsh LM, Ghanim B, Klepetko W, Kovacs G, et al. Distinct differences in gene expression patterns in pulmonary arteries of patients with chronic obstructive pulmonary disease and idiopathic pulmonary fibrosis with pulmonary hypertension. Am J Respir Crit Care Med 2014;190:98–111.|
|36.||Tomii Y, Kamochi J, Yamazaki H, Sawa N, Tokunaga T, Ohnishi Y, et al. Human thrombospondin 2 inhibits proliferation of microvascular endothelial cells. Int J Oncol 2002;20:339–342.|
|37.||Kyriakides TR, Zhu YH, Yang Z, Huynh G, Bornstein P. Altered extracellular matrix remodeling and angiogenesis in sponge granulomas of thrombospondin 2-null mice. Am J Pathol 2001;159:1255–1262.|
|38.||Zhang K, Li M, Yin L, Fu G, Liu Z. Role of thrombospondin-1 and thrombospondin-2 in cardiovascular diseases (review). Int J Mol Med 2020;45:1275–1293.|
|39.||MacLauchlan SC, Calabro NE, Huang Y, Krishna M, Bancroft T, Sharma T, et al. HIF-1α represses the expression of the angiogenesis inhibitor thrombospondin-2. Matrix Biol 2018;65:45–58.|
|40.||Jin DY, Chae HZ, Rhee SG, Jeang KT. Regulatory role for a novel human thioredoxin peroxidase in NF-kappaB activation. J Biol Chem 1997;272:30952–30961.|
|41.||Hanaka T, Kido T, Noguchi S, Yamada S, Noguchi H, Guo X, et al. The overexpression of peroxiredoxin-4 affects the progression of idiopathic pulmonary fibrosis. BMC Pulm Med 2019;19:265.|
|42.||Ibarrola J, Arrieta V, Sádaba R, Martinez-Martinez E, Garcia-Peña A, Alvarez V, et al. Galectin-3 down-regulates antioxidant peroxiredoxin-4 in human cardiac fibroblasts: a new pathway to induce cardiac damage. Clin Sci (Lond) 2018;132:1471–1485.|
|43.||de Man FS, Tu L, Handoko ML, Rain S, Ruiter G, François C. Dysregulated renin–angiotensin–aldosterone system contributes to pulmonary arterial hypertension. Am J Respir Crit Care Med 2012;186:780–789.|
|44.||Wilkins MR, Aman J, Harbaum L, Ulrich A, Wharton J, Rhodes CJ. Recent advances in pulmonary arterial hypertension. F1000 Res 2018;7:F1000 Faculty Rev-1128.|
|45.||Husain-Syed F, McCullough PA, Birk HW, Renker M, Brocca A, Seeger W, et al. Cardio-pulmonary-renal interactions: a multidisciplinary approach. J Am Coll Cardiol 2015;65:2433–2448.|
|46.||Bello GA, Dawes TJW, Duan J, Biffi C, de Marvao A, Howard LSGE, et al. Deep learning cardiac motion analysis for human survival prediction. Nat Mach Intell 2019;1:95–104.|
|47.||Emilsson V, Ilkov M, Lamb JR, Finkel N, Gudmundsson EF, Pitts R, et al. Co-regulatory networks of human serum proteins link genetics to disease. Science 2018;361:769–773.|
|48.||Sun BB, Maranville JC, Peters JE, Stacey D, Staley JR, Blackshaw J, et al. Genomic atlas of the human plasma proteome. Nature 2018;558:73–79.|
Supported by the National Institute for Health Research (NIHR), and the UK Medical Research Council (MR/K020919/1), and in part by the Assistance Publique-Hopitaux de Paris, INSERM, University Paris-Sud, and Agence Nationale de la Recherche (Departement Hospitalo-Universitaire Thorax Innovation; LabEx LERMIT, ANR-10-LABX-0033; and RHU BIO-ART LUNG 2020, ANR-15-RHUS-0002); British Heart Foundation Centre for Research Excellence award RE/18/4/34215; and special project grant SP/18/10/33975. C.J.R. is supported by a BHF Intermediate Basic Science Research fellowship (FS/15/59/31839) and Academy of Medical Sciences Springboard fellowship (SBF004\1095). A.L. is a BHF Senior Investigator (FS/18/52/33808). N.W.M. is a BHF Professor and NIHR Senior Investigator. The views expressed are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health and Social Care.
Author Contributions: Conceptualization: C.J.R., J.W., L.H., O.S., M.H., and M.R.W. Data curation: C.J.R., J.W., E.M.S., B.G., A.B., and D.M.; Formal analysis and writing, original draft: C.J.R. and M.R.W. Data acquisition, data interpretation, and writing, review and editing: all authors. C.J.R., J.W., E.M.S., L.H., B.G., O.S., and M.R.W. have had access to, and verified, the data used in this article.
This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org.
Originally Published in Press as DOI: 10.1164/rccm.202105-1118OC on January 26, 2022