Clinical characteristics of the patients enrolled in this study

The clinical characteristics of PAH patients and controls are summarized in Table 1. In the iTRAQ assay, 9 PAH patients and 9 age- and sex-matched controls were involved. Among the 9 PAH patients, 6 were female. The mean age was 39.3 ± 16.9 years. Among the included patients, 1 patient was at low risk, 6 at intermediate risk, and 2 at high risk. The median mPAP was 54.0 (42.0, 90.5) mmHg, and the mean PVR was 1324.4 ± 721.2 dyn•s•cm−5. 2 patients received calcium channel blockers (CCBs) treatment, 2 patients received initial monotherapy, 4 patients received initial combination therapy and 1 patient who was diagnosed with pulmonary veno-occlusive disease (PVOD) did not receive any specific-PAH therapy (Table 1).

Table 1 Clinical characteristics of the patients enrolled in this study

In the subsequent ELISA verification, 28 PAH patients and 28 age- and sex-matched controls were included in the ELISA assay. 22 patients were female. The median age was 39.0 (29.3, 60.5) years. Most patients (19, 67.9%) were at intermediate risk. 5 (17.9%) patients were at low risk and 4 (14.3%) at high risk. The mean mPAP and PVR were 47.2 ± 15.4 mmHg and 746.4 ± 435.1 dyn•s•cm−5, respectively. Among the included patients, 2 patients (7.1%) received initial CCBs, 10 (35.7%) received initial monotherapy, 15 (53.6%) patients received initial combination therapy and 1 (3.6%) patient who was diagnosed with PVOD did not receive any specific-PAH therapy (Table 1).

Schematic workflow of screening PAH proteins

We collected 9 serum samples from each group and conducted proteomics (Fig. 1A). A total of 362 proteins were identified in proteomic analysis, and 256 of them contained quantitative information (Fig. 1B). 85 serum proteins showed differential expression, with 41 proteins upregulated and 44 downregulated in PAH patients compared to healthy controls (Fig. 1C, D).

Fig. 1

Proteomics analysis revealed differential protein expression in the serum of PAH patients. A Flow chart. B Total number of identified (green bar) and quantified (red bar) proteins in the iTRAQ experiment. C Volcano plots for the expression of differentially expressed proteins. D A total of 41 upregulated proteins and 44 downregulated proteins were identified

Functional enrichment analyses of DEPs

To exploit the potential functions of DEPs, we analyzed GO function and KEGG pathway enrichment. In GO Biological Process (BP) enrichment analysis, DEPs were centered on “platelet degranulation”, “humoral immune response”, “coagulation”, “hemostasis”, “protein activation cascade”, and “fibrin clot formation” (Fig. 2A). As for the Cellular Component (CC) category, the core DEPs were significantly enriched related to “blood microparticle”, “secretory granule lumen”, “cytoplasmic vesicle lumen”, “vesicle lumen”, “collagen-containing extracellular matrix”, and “platelet alpha granule” (Fig. 2B). In addition, DEPs were enriched in the Molecular Function (MF) category focused on “glycosaminoglycan binding”, “serine-type endopeptidase activity”, “heparin binding”, “serine-type peptidase activity”, “serine hydrolase activity”, and “sulfur compound binding” (Fig. 2C).

Fig. 2

Functional enrichment analyses of differentially expressed proteins. A The top 10 enrichment GO Biological Process (BP) pathways ranked by enrichment score. B The top 10 enrichment GO Cellular Component (CC) pathways ranked by enrichment score. C The top 10 enrichment GO Molecular Function (MF) pathways ranked by enrichment score. D The top 9 enrichment KEGG pathways ranked by enrichment score

Moreover, we were able to obtain a global perspective of the changes in protein expression patterns. The top enriched KEGG pathways of DEPs were involved in complement and coagulation cascades, platelet activation, neutrophil extracellular trap formation, hypertrophic cardiomyopathy, and dilated cardiomyopathy (Fig. 2D).

PPI network analysis of DEPs

To identify hub proteins that may serve as biomarkers or therapeutic targets for PAH, a protein–protein co-expression network was constructed for the DEPs. The PPI network contained a total of 80 nodes and 1196 edges (Fig. 3). Subsequently, the co-expression network was further analyzed to detect potential critical modules, and three significant modules (HPSE, HGFA, and GSN) were determined. HPSE interacted with SERPINF2. HGFA interacted with CPB2, FGB, FGG, F2, PROC, SERPINA5, KRT1, SPP1 and HABP2. GSN interacted with ALB, A2M, ACTN1, A1BG, ACTB, APCS, CP, HP, HPX, TPM3, TPM4, TLN1, TF, FN1, LTF, LYZ, GIG25, and SAA1. Four proteins (FGA, GC, APOA4, TTR) were noteworthy and interacted with GSN and HGFA at the same time.

Fig. 3

PPI network for differentially expressed proteins and key module analysis

Verification of significantly dysregulated proteins between the PAH and reference groups

According to the results of PPI analysis, potential functional and pathological significance, we selected some candidate proteins for in-depth research, including HPSE, GSN, HGFA, SAA1, and ECM1. To validate the expression level of these candidate proteins, serum derived from another 28 PAH patients and 28 healthy individuals were determined by ELISA. The ELISA results showed that the levels of GSN [PAH: 10,771.8 (4950.5, 14,198.0) vs. control: 14,482.5 (13,283.7, 15,494.1) ng/mL, P < 0.001] and HGFA [PAH: 8652.8 (6941.9, 10,931.5) vs. control: 21,850.9 (16,660.7, 26,080.6) ng/mL, P < 0.0001] were lower in PAH patients compared to normal controls according to ELISA, whereas HPSE was higher [PAH: 81.5 (52.9, 148.5) vs. control: 24.5 (12.9, 55.1) ng/mL, P < 0.001] (Fig. 4A–C). No significant differences were observed for SAA1 [PAH: 157.7 (85.9, 188.3) vs. control: 133.3 (88.4, 150.0) ng/mL, P > 0.05] and ECM1 [PAH: 9450.1 (9158.6, 10,236.9) vs. control: 8997.2 (8547.5, 9329.1) ng/mL, P > 0.05] (Fig. 4D, E).

Fig. 4

Verification of differentially expressed proteins by ELISA in validation cohort. A GSN, B HFGA, C HPSE, D SAA1, and E ECM1 in pulmonary arterial hypertension patients and healthy controls. F Receiver operating characteristic (ROC) results of different proteins between the PAHs and healthy controls. As the result of significance test, *P value < 0.05; **P value < 0.01; ***P value < 0.001; ****P value < 0.0001; ns P value > 0.05

Diagnostic performance of candidate serum biomarkers

Given the observed differences of serum HPSE, GSN, and HGFA concentrations between PAH patients and healthy controls, we intended to test the diagnosis performance of candidate serum biomarkers to help diagnose PAH. In Fig. 4F, the AUC of HGFA was 0.964 and the AUC of HPSE and GSN proteins were in the range of 0.7–0.8 in the PAH diagnosis. The sensitivity and the specificity of HPSE were 75.0% and 75.0%, respectively, at the cutoff value of 54.5 ng/mL; the sensitivity and the specificity of GSN were 67.9% and 78.6%, respectively, at the cutoff value of 13,145.1 ng/mL; the sensitivity and the specificity of HGFA were 89.3% and 89.3%, respectively, at the cutoff value of 14,675.0 ng/mL.

Correlation analysis between clinical data and candidate biomarkers

Spearman’s correlation test was employed to investigate the correlation of serum HPSE, GSN, and HGFA levels with a cluster of clinical parameters including hemodynamic parameters, 6-min walking distance (6MWD), echocardiographic parameters, WHO functional class (WHO-FC), and laboratory tests. In Fig. 5, serum HPSE concentrations were inversely correlated with the tricuspid annular plane systolic excursion (TAPSE). Furthermore, serum HPSE concentrations were positively correlated with pericardial effusion (PE) and main pulmonary artery internal diameter (MPA). Serum GSN concentrations were negatively correlated with 6MWD. HGFA was negatively correlated with right atrial transverse diameter (RA-t), WHO-FC, MPA, and N-terminal pro-brain natriuretic peptide (NT-proBNP). Serum HGFA concentrations were also significantly lower in patients with a high risk-stratification of the 2018 world symposium on PH (WSPH) [4].

Fig. 5

Correlation network of three biomarkers and clinical indicators in pulmonary arterial hypertension patients. Correlations are indicated in red for positive correlations and in blue for negative correlations. As the result of significance test, *P value < 0.05

MR verified the causal relationship of HGFA with PAH using plasma pQTL

We identified 34 independent SNPs as instrumental variables that exhibited significant associations with HGFA levels. As HGFA levels were genetically reduced, the risk of PAH increased using IVW (OR 0.70, 95% CI 0.55–0.89; P value = 0.003). These findings were consistent with the result of MR Egger analyses. Collectively, our data suggested a causal association of genetically reduced HGFA levels with increased risk of PAH (Fig. 6).

Fig. 6

Two-sample Mendelian randomization reveals causal evidence for HGFA on pulmonary arterial hypertension. A The forest plots illustrate the standardized beta (95% confidence interval) for each two-sample Mendelian randomization method; B Scatter plot to visualize the causal effect of HGFA on the risk of pulmonary arterial hypertension. The slope of the straight line indicates the magnitude of the causal association. IVW inverse-variance weighted

Validation of HGFA expression in PAH in GSE113439

HGFA expression in PAH tissues was verified. The results demonstrated that HGFA expression was downregulated in PAH compared to the controls in the GSE113439 datasets (P < 0.0001; Fig. 7A). Next, to assess the potential predictive value of key gene, ROC curves were generated. The AUC (0.903) suggested that HGFA had a high accuracy and good predictive value for PAH (Fig. 7B).

Fig. 7

Validation of HGFA expression. A Gene expression of HGFA in GSE113439 datasets; B receiver operating characteristic (ROC) results of HGFA between the PAHs and healthy controls in GSE113439 datasets; C Hgfa mRNA expression via RT-qPCR in lung tissues from SuHx-PH rat models or control; D Hgfa mRNA expression via RT-qPCR in lung tissues from MCT-PH rat models or control; E HGFA expression via ELISA in serum from SuHx-PH rat models or control; F HGFA expression via ELISA in serum from MCT-PH rat models or control; G relationship between HGFA expression and RVSP in SuHx-PH rat models or control; H relationship between HGFA expression and RVSP in MCT-PH rat models or control. **P value < 0.01. ***P value < 0.001. ****P value < 0.0001

The expression of HGFA in the lung tissues and serum from different PAH rat models

To investigate the role of HGFA in PAH, we determined HGFA expression levels in the lung tissues and serum from MCT-PH and SuHx-PH rat models using RT-qPCR and ELISA, respectively. The mRNA expression levels of Hgfa were significantly lower in both PAH rat models than in the control group (Fig. 7C, D). Additionally, HGFA was found to be downregulated in the serum of PAH rat models relative to controls (Fig. 7E, F). Serum HGFA concentrations negatively correlated with RVSP (Fig. 7G, H).