Heart failure (HF) has multiple causes and often lead to poor prognosis and frequent hospital admission^{[1, 2]}. In 2015, HF was the underlying cause for 75 251 deaths (33 667 males and 41 584 females) worldwide, and its prevalence across Asia was estimated to range from 1.26% to 6.7%^[3]. An array of biomarkers have been explored for their values in the diagnosis of HF, including N terminal fragment of pro-B-type natriuretic peptide (NT-proBNP) ^{[4, 5]}, neutrophil gelatinase-associated lipocalin (NGAL)^[6], soluble ST2^[7], and troponin^[8], which were often used in combination to enhance the diagnostic power ^[9]. Up to now only NT-proBNP has been intensively studied for its diagnostic and prognostic value for HF^[10]. Nevertheless, a NT-proBNP-based diagnosis may fail in patients with the comorbidities of chronic kidney disease, chronic lung disease, and obesity and even in patients with an advanced age, as these conditions may substantially affect the level of NT-proBNP^[11].

Growth and differentiation factor-15 (GDF-15), also known as macrophage inhibitory cytokine-1 (MIC-1), placental transforming growth factor-β (TGF-β), prostatederived factor, or placental bone morphogenetic protein, is a divergent member of the TGF-β superfamily^{[12, 13]}. GDF-15, which is normally expressed weakly in most parenchymal tissues, is initially synthesized as an inactive precursor protein with a relative molecular mass of ~40 000, then proteolytically cleaved to release the active C-terminal fragment and finally assembled into a disulfide-linked homodimer (~28 000) to become biologically active GDF-15^[14]. GDF-15 was found to play a cardioprotective role by activating the PI3K-Akt signaling pathway in a mouse model^[15-17], and in humans, serum GDF-15 concentrations have been shown to be associated with the risk and prognosis of HF^{[18, 19]}. But so far, little is known about GDF-15 mRNA expression in circulating monocytes (MNCs) in patients with HF.

In China, a large number of patients with HF have underlying cardiac diseases of dilated cardiomyopathy (DCM) or coronary artery disease (CAD)^{[20, 21]}, but the association of plasma GDF-15 level and GDF-15 mRNA expression in circulating MNCs with the clinical outcomes have not been explored in these patients. In this study, we aimed to investigate plasma GDF-15 level and GDF-15 mRNA expression in the MNCs of patients with HF and different underlying cardiac conditions and assess their value in the estimation of NYHA functional class and the risk for cardiovascular disease (CVD) events.

This study was performed with approval by the Ethics Committee of the First Affiliated Hospital of Xi'an Jiaotong University and in accordance with the Declaration of Helsinki. Between January, 2018 and January, 2019, a total of 512 patients with a mean age of 59.0±10.7 years were recruited from the First Affiliated Hospital of Xi'an Jiaotong University. All the patients had an established diagnosis of HF in line with the Framingham criteria and were divided into 3 functional classes (Ⅱ, Ⅲ and Ⅳ) according to the New York Heart Association (NYHA) functional classification system. Among them, 261 patients had underlying DCM (including 86 NYHA class Ⅱ patients, 77 class Ⅲ patients, and 98 class Ⅳ patients) and 251 had CAD (including 118 class Ⅱ patients, 78 class Ⅲ patients, and 55 class Ⅳ patients). We excluded the patients with a previous diagnosis of acute myocardial infarction in the last 3 months; chronic renal failure or chronic hepatic failure; active infection or tumor; ongoing pregnancy; and chronic use of corticosteroids, as these co-morbidities and drugs interfered directly with keto acid metabolism^[22]. We also enrolled 132 gender- and age-matched healthy subjects as the control group. Written informed consent was obtained from all the participants.

Fasting blood sample (5 mL) was collected from all the patients on the next morning following the admission day, within 24 h after symptom onset. Peripheral blood MNCs were separated by Ficoll density gradient centrifugation for real-time quantitative PCR analyses. Plasma samples were was stored at -80 ℃ for GDF-15 measurement. The levels of NT-proBNP and other routine blood biochemical parameters were derived from blood test reports by the clinical laboratory of the hospital.

Enzyme-linked immunosorbent assay (ELISA) was used for detecting plasma levels of GDF-15 according to the manufacturer's instructions of the detection kit (R&D Systems). The minimal detectable concentrations (or the lowest concentration of the standard sample) were 4.4 pg/mL. The intra- and inter-assay variation coefficients for all ELISA were below 10%, and each sample was detected in triplicate.

Total RNA was isolated from the circulating MNCs by TRIzol extraction (Invitrogen) according to the manufacturer's instructions. The first-strand cDNA was synthesized using RevertAidTM First Strand cDNA Synthesis Kit (Fermentas), and then analyzed using realtime quantitative PCR with SYBR Ex TaqTM Ⅱ (TaKaRa). All the primers were purchased from TaKaRa Biosystems, including the sequences of the gene-specific primers. Real-time PCR was performed on the iQ5TM Multicolor Real-time PCR and Detection System (Bio-Rad), and the relative expression of GDF-15 was determined by calculating 2^-ΔΔCT, where ΔΔCT is ΔCt (different group)-ΔCt (control group), ΔCt is Ct (target gene)-Ct (GAPDH), and Ct is the cycle at which the threshold is crossed. The primers used are listed below:

GDF-15, 5'-GGTGAATGGCTCTCAGATG-3',

5'-CACTTCTGGCGTGAGTATC-3';

GAPDH, 5'-ACCACAGTCCATGCCATCAC-3',

5'-TCCACCACCCTGTTGCTGTA-3'.

Death and re-admission with HF were defined as CVD events. The patients were followed up during their outpatient visits or by telephone interviews. The end of the follow-up was the date when the first CVD event occurred, and all the patients were followed up until January 31, 2019.

The categorical variables are shown as the absolute number (n) and the relative frequency (%), which were compared among the groups using Pearson's Chi-square test. The Kolmogorov-Smirnov test was used to test the normal distribution of the continuous variables, and the standard deviation and independent groups were analyzed using Student's t-test. Non-parametric data are shown as the median with the interquartile range (25th-75th percentile). The continuous variables were compared by Mann-Whitney test. Comparisons between strata of the patients were performed by Kruskal-Wallis test or analysis of variance. Kruskal-Wallis test and Dunn test were used to analyze the continuous variables stratified by NYHA classification. Cox proportional hazards model analyses were performed to determine the independent factors associated with the cardiovascular events. The predictive accuracies of NT-proBNP alone and in combination with GDF-15 mRNA were evaluated by comparing the areas under the receiver-operating characteristic curve (ROC). The computations were performed with SPSS software 17.0 (SPSS Inc., Chicago, IL) and MedCalc (MedCalc Software, Mariakerke, Belgium). A statistically significant difference was defined by a P value < 0.05 (two-tailed).

As shown in Tab. 1, the disease course of DCM and CAD patients was 2.53±2.07 and 2.46±1.81, respectively. The healthy subjects were Metched for age, gender, smoking, diabetes, lipoprotein a (Lpa), apolipoprotein B (apoB), apoE, high-density lipoprotein cholesterol (HDL-C), white blood cell (WBC) count, and lymphocyte count. The percentage of hypertensive patients and the levels of C-reactive protein (CRP) and cardiac troponin I (cTnI) were significantly higher in CAD group than in the other two groups (P < 0.01). Furthermore, the left ventricular ejection fraction (EF%) was significantly lower in CAD and DCM groups than in the control group (P < 0.01), but the difference between the former two groups was not significant.

表 1(Tab.1)

Tab.1 Characteristics of the patients and the control subjects Variable Control DCM CAD n 132 261 251 Age (year) 57.0±9.3 58.0±10.4 61.4±10.4 Gender (male/female) 79/55 151/110 136/115 Hypertension [n(%)] 33 45 79**## Smoke [n(%)] 71 113 133 Diabetes [n(%)] 29 71 75 TC (mmol/L) 3.90±0.87 3.69±0.82* 3.76±0.91 Lpa (mg/L) 119±77.4 115±87 142±70 apo A (g/L) 0.96±0.23 0.90±0.31 0.81±0.19* apoB (g/L) 0.90±0.23 0.80±0.17 0.86±0.25 apo E (mg/L) 48.0±34.2 45.1±22.6 41.7±16.3 TG (mmol/L) 1.65±0.79 1.99±0.72* 1.77±0.85 HDL-C (mmol/L) 1.05±0.28 0.96±0.31 0.86±0.32 LDL-C (mmol/L) 2.42±0.83 2.06±0.75* 2.38±0.89 FBG (mmol/L) 6.01±1.56 6.96±2.38* 7.09±2.41* WBC (×109/L) 6.30±1.76 5.93±1.31 6.83±2.43 lymphocyte (×109/L) 1.73±0.55 1.81±0.58 1.61±0.60 hs-CRP (mg/L) 1.84±1.82 2.25±2.16 4.74±3.12**## NT-proBNP (pg/mL) 101±47 1210±931**## 886±749** EF (%) 65.3±7.71 36.5±9.50** 40.4±12.1** cTnI (ng/mL) 0.18±0.12 0.43±0.33* 0.76±0.89**## Disease course 0 2.53±2.07 2.46±1.81 Data are presented as Mean±SD unless indicated otherwise. DCM: Dilated cardiomyopathy. CAD: Coronary artery disease; BP: Blood pressure; TC: Serum total cholesterol; Lpa: Lipoprotein a; apoA: Apolipoprotein A; apoB: Apolipoprotein B; apoE: Apolipoprotein E; TG: Triglyceride; HDL-C: High-density lipoprotein cholesterol; LDL-C: Low density lipoprotein cholesterol; FBG: Fasting blood glucose; WBC: White blood cell count; hs-CRP: High sensitivity C-reactive protein; EF: Ejection fraction; cTnI: Cardiac troponin I. *P < 0.05, **P < 0.01 vs control group; #P < 0.05, ##P < 0.01 vs the other case group.

Tab.1 Characteristics of the patients and the control subjects

The plasma level of NT-proBNP was markedly higher in DCM group (1210±931 pg/mL, P < 0.001) and CAD group (886 ± 749 pg/mL, P < 0.001) than in the control group (101 ± 47 pg/mL), but showed no significant difference between DCM and CAD groups (P>0.05, Tab. 1). As shown in Fig. 1, in DCM patients, plasma NT-proBNP level was significantly higher in NYHA class Ⅳ patients than in class Ⅱ and Ⅲ patients (P < 0.01), but showed no significant difference between the latter two groups (P>0.05); in CAD patients, plasma NT-proBNP level did not differ significantly among class Ⅱ, Ⅲ and Ⅳ groups (P>0.05), suggesting that plasma NT-proBNP not only identified the patients with HF but also identified the most severe cases (class Ⅳ) of DCM.

图 1(Fig.1)

Fig.1 Plasma concentrations of NT-ProBNP and GDF-15 in the control group (n=132), dilated cardiomyopathy (DCM) group (n= 261, including 86 NYHA class Ⅱ, 77 class Ⅲ and 98 class Ⅳ patients) and coronary heart disease (CAD) group (n=251, including 118 NYHA class Ⅱ, 78 class Ⅲ and 55 class Ⅳ patients). *P < 0.05, **P < 0.01, ***P < 0.001 vs control group. NS: P>0.05.

Plasma GDF-15 level was markedly higher in DCM (1676±179 pg/mL, P < 0.001) and CAD patients (1426± 152 pg/mL, P < 0.001) than that in the control group (415± 56 pg/mL, P < 0.001), but was comparable between DCM and CAD groups (P>0.05, Fig. 1). In DCM patients, plasma GDF-15 level were significantly higher in NYHA class Ⅳ group (2829±460 pg/mL) than in class Ⅱ group (1222±140 pg/mL, P < 0.001) and class Ⅲ group (1162± 147 pg/mL, P < 0.001), but similar between the latter two groups (P>0.05). Similarly, in CAD patients, GDF-15 plasma level was the highest in NYHA class Ⅳ group (2489 ± 512 pg/mL, P < 0.001) and comparable between class Ⅱ and Ⅲ groups (1120±140 vs 1144±147 pg/mL, P>0.05, Fig. 1). These data demonstrate that plasma GDF-15 can identify HF patients and has better performance than NT-proBNP for distinguishing CAD patients with different NYHA classes and for identifying NYHA class Ⅳ cases among HF patients.

As shown in Fig. 2, the patients with DCM and CAD had comparable levels of GDF-15 mRNA in circulating MNCs, which were significantly elevated compared with levels in the control group (P < 0.001). In DCM group, the patients with different NYHA classes all had significantly higher GDF-15 mRNA levels in the MNCs than the control subjects (P < 0.01); the class Ⅳ patients and class Ⅲ patients had similar GDF-15 mRNA levels (P>0.05), which were much higher that in class Ⅱ patients (P < 0.01). In CAD patients, GDF-15 levels in MNCs across the 3 NYHA classes were all significantly higher than that in the control group (P < 0.01); GDF-15 mRNA level in the MNCs was significantly higher in class Ⅳ group than in both class Ⅱ group (P < 0.01) and class Ⅲ group (P < 0.05), and was also higher in class Ⅲ group than in class Ⅱ group (P < 0.05). These data suggest that GDF-15 mRNA in the MNCs can also identify patients with HF and helps to distinguish different NYHA classes in DCM and CAD patients, and is especially useful for identifying class Ⅱ patients.

图 2(Fig.2)

Fig.2 Relative mRNA expression of GDF-15 in peripheral blood monocytes (MNCs) in the 3 groups measured by real- time PCR. The data are presented as fold change relative to the control level. **P < 0.01, ***P < 0.001 vs control group.

After one year of follow-up, 3 CVD events occurred in the control group, 19 in DCM group and 24 in CAD group. We also analyzed the prediction value of GDF-15 mRNA in the MNCs and NT-proBNP for CVD events among the patients. The results of the Cox analysis of the factors for predicting CVD events are presented in Tab. 2. Both univariate and multivariate Cox analyses showed that GDF-15 mRNA combined with NT-proBNP independently predicted CVD events in the patients (HR:4.132, 95% CI: 3.377-5.096, P < 0.001 in univariate analysis; HR: 4.067, 95% CI: 3.184-5.355, P < 0.001 in multivariate analysis).

表 2(Tab.2)

Tab.2 Cox analysis of the factors predicting CVD events in patients with heart failure Variable Univariate Multivariate HR (95% CI) P HR (95% CI) P Age (per 1 year) 1.021 (1.017-1.031) 0.002 1.013 (1.004-1.021) 0.012 Diabetes 1.469 (1.074-1.939) < 0.001 1.253 (1.018-1.553) < 0.001 NT-proBNP 2.632 (2.998-4.400) < 0.001 2.191 (1.791-2.680) < 0.001 NT-proBNP+GDF15 plasma 3.102 (2.071-4.135) < 0.001 3.100 (2.088-4.112) < 0.001 NT-proBNP+GDF15 mRNA 4.132 (3.377-5.096) < 0.001 4.067 (3.184-5.355) < 0.001

Tab.2 Cox analysis of the factors predicting CVD events in patients with heart failure

ROC analysis of the CVD events showed that for predicting the CVD events at one year, the AUC for NT-proBNP alone was only 0.73 (95% CI: 0.69-0.77, P < 0.001), significantly lower than that of 0.83 (95% CI: 0.79-0.86, P < 0.001) for GDF-15 mRNA and NT-proBNP combined (Z=5.993, P < 0.001, Fig. 3). These results demonstrate that the combination of NT-proBNP and GDF-15 mRNA can better predict the long-term CVD events in patients with HF than either of them alone.

图 3(Fig.3)

Fig.3 Receiver-operating characteristic curve analysis of different indicators for prediction of CVD events at 1 year. The area under curve was 0.73 (95% CI: 0.69-0.77, P < 0.001) for NT-proBNP and 0.83 (95% CI: 0.79-0.86, P < 0.001) for NT-proBNP+GDF-15 mRNA.

We found in this study that NT-proBNP could serve as a diagnostic marker for HF, as is consistent with previous studies^[10]. But due to its susceptibility to the influences by such factors as gender, age, and obesity, NT-proBNP does not seem to be an ideal biomarker for a diagnostic purpose ^[23]. Previous studies have shown that GDF-15 may have a protective effect on the heart, and in mouse models, GDF-15 protects the heart from ischemia/ reperfusion injury^[18], and its over-expression attenuates ventricular dilation and HF. Conversely, GDF-15 knockout in mice results in enhanced cardiac hypertrophic growth^[19]. Such cardioprotective effects of GDF-15 have been shown to be mediated by the activation of the PI3K-Akt signaling pathway and the Smad protein^[20], but so far, it remains unclear whether GDF-15 can serve as a predictor of HF in humans.

In this study, we found that plasma GDF-15 level was sensitive for identifying patients with severe HF GDF15 mRNA NT-proBNP NT pro-BNP+GDF15 mRNA Reference line (NYHA class Ⅳ), while GDF-15 mRNA level in the MNCs was more reliable for identifying class Ⅱ patients with HF. Furthermore, the more pronounced difference in GDF-15 mRNA level in the MNCs than that in plasma GDF-15 level and NT-proBNP between HF patients and the control group suggests that GDF-15 mRNA in the MNCs is more effective for diagnosing HF than the other two indicators. The results of Cox analysis and ROC analysis showed that GDF-15 mRNA in the MNCs was an excellent independent predictor for CVD events at one year in HF patients, and its predictive effect is further enhanced when combined with NT-proBNP.

We show here that GDF-15 mRNA level is significantly elevated in the MNCs in HF patients. Our findings also suggest that GDF-15 mRNA level in the MNCs can potentially serve not only as a diagnostic indicator of HF but also as a predictor of long-term CVD events in HF patients. However, the relatively small sample size included and the short follow-up time in this study limit the confidence in this conclusion. In addition, a suitable kit for detecting GDF-15 mRNA in the MNCs has not been currently available. The value of GDF-15 mRNA in the MNCs in clinical diagnosis and treatment of HF still awaits further investigation by cohort studies with larger sample sizes and more reliable and efficient detection methods.

In conclusion, GDF-15 level is increased significantly in the plasma and MNCs in HF patients with underlying DCM or CAD, and the elevated GDF-15 mRNA expression in the MNCs is correlated with cardiac function classification. Both plasma and MNC mRNA levels of GDF-15 can serve as biomarkers for HF, and when combined with NT-proBNP, GDF-15 mRNA level in the MNCs can better predict the occurrence of long-term CVD events.