Supplementary Materials

Methods

LC-MS and LC-MS/MS

The sample analysis was performed on an Ultimate-3000 UPLC system coupled to a Q Exactive hybrid quadrupole-Orbitrap MS system (Thermo Scientific). The residues were re-suspended in platform specific solutions before injection. The untargeted metabolomic platform employed for this analysis was based on a combination of three conditions: (A) Hydrophilic interaction liquid chromatography (HILIC)-MS with ESI+/ESI- switching mode； (B) Reversed phase liquid chromatography (RPLC)-MS with ESI+ mode； and (C) RPLC-MS with ESI- mode. For HILIC-MS analysis, the obtained residue was reconstituted in 100 μL of acetonitrile/water (1:1, v/v), with 2-chloro-L-phenylalanine as internal standard. For RPLC-MS analysis, the residue was reconstituted in 100 μL of 80% methanol, with 2-chloro-L-phenylalanine, fatty acid 19:0, lysophosphatidylcholine 12:0, and phosphatidylcholine 22:0 as internal standards.

Gradient Profile for HILIC-MS Method An Acquity UPLC BEH Amide column (1.7 μm × 2.1 mm × 100 mm, Waters) with a VanGuard pre-column was used. The column temperature was set at 40°C. The mobile phases were 20 mM ammonium acetate and 20 mM ammonium hydroxide in water/acetonitrile (95:5, v:v) (A) and acetonitrile (B). The flow rate was 0.35 mL/min and the injection volume was 5 μL. The gradient was 85% B in 0-1.0 min, 85-65% B in 1.0- 12.0 min, 65-40% B in 12.0-12.1 min, 40% B in 12.1-15.0 min, 40-85% B in 15.0-15.1 min, and 85% B in 15.1-20.0 min. All the samples were kept at 4°C during the analysis.

Gradient Profiles for RPLC-MS Methods A Hypersil GOLD C8 column (1.9 μm × 2.1 mm × 100 mm, Thermo Scientific) with a guard filter was used. The column temperature was set at 40°C. For ESI+ mode, the mobile phases were 0.1% formic acid in methanol (A) and 5% methanol in water with 0.1% formic acid (B). For ESI- mode, the mobile phases were 6.5 mM ammonium bicarbonate in methanol with 5% water (A) and 6.5 mM ammonium bicarbonate in water with 5% methanol (B). For both positive and negative modes, the flow rate was 0.35 mL/min and the injection volume was 5 μL. The gradient was 100% B in 0-1.0 min, 100-0% B in 1.0-5.5 min, 0% B in 5.5-8.0 min, 0-100% B in 8.0-8.1 min, and 100% B in 8.1-12.0 min. All the samples were kept at 10°C during the analysis.

Electrospray ionization in both of the liquid condition was carried out using the following conditions. Full scan data was acquired in positive, negative, or positive/negative ion switching modes in profile mode at 70,000 resolution (at m/z range 70-1050), an automatic gain control target of 3e6, with spray voltages +3.0 kV and -2.5 kV. The capillary temperature was set at 350°C with the sheath gas flow rate at 45 arb and aux gas flow rate at 15 arb. The heater temperature was set at 350°C. The S-Lens RF level was set at 50.

The MS/MS was operated at a resolving power of 70,000 in full-scan mode (scan range: 70-1050 m/z)； automatic gain control target: 1e6 both in positive and negative ion mode) and of 17,500 in the Top10 data-dependent MS2 mode (stepped normalized collision energy: 15, 30 and 45 in positive mode and 30, 45, 60 in negative ion mode； Injection time: 100 ms； Isolation window: 4.0 m/z； automatic gain control target: 2e5) with dynamic exclusion setting of 6.0 s.

Table 1S. Model quality parameters for Figure 1 PLS-DA models

Table 2S. Abnormal metabolomic profiling in peripheral venous blood of HF patients.

Table 2S. Cont.

Table 2S. Cont.

Figure 1S. The validation graphs of Figure 1 PLS-DA score plots. (A), (B), and (C) are the corresponding validation models of Figure 1 PLS-DA models (A), (B), and (C). A default seven-round cross-validation in SIMCA was performed throughout to determine the optimal number of principal components and to avoid model overfitting. The PLS-DA models were also validated by a permutation analysis (200 times). The quality of the models is described by the R2X or R2Yand Q2 values. R2X or R2Y is defined as the proportion of variance in the data explained by the models and indicates goodness of fit. Q2 is defined as the proportion of variance in the data predicted by the model and indicates predictability, calculated by a cross-validation procedure. All the permuted R2 (cum) and Q2 (cum) values were shown on the figures, which indicating the relatively good predictive capability and the models are valid without overfitting. The detailed model quality parameters were presented in Table 1S.

Figure 2S. Correlations of the concentration level in peripheral venous blood with that in coronary sinus blood. (A): L-arginine； (B): Lysyl-gamma-glutamate.