All standard amines, hydroxylamine (NH₂OH, 50 wt. % in H₂O), 3-mercaptopropionic acid (MPA), 4-dimethylaminopyridine (DMAP), and N-hydroxysuccinimide (NHS) were purchased from Sigma-Aldrich. L-glutamic acid-¹³C5 (L-Glu-¹³C5), L-valine-¹³C5,¹⁵N (L-Val-¹³C5,¹⁵N), L-tyrosine-¹³C6 (L-Tyr-¹³C6), L-arginine-¹³C6 hydrochloride (L-Arg-¹³C6), L-glycine-¹³C2 (L-Gly-¹³C2) were purchased from Toronto Research Chemical. Methanol (MeOH, HPLC, MS), Acetonitrile (ACN, HPLC, MS), N,N-Dimethylformamide (DMF, HPLC), and Dichloromethane (DCM, HPLC) were purchased from Thermo Fischer. Polyacrylic acid (PAA), Ethylene glycol (EG), and Diethylene glycol (DEG) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. 1-(3-Dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC), and sodium acetate (NaAc) were purchased from Shanghai Macklin Biochemical Co., Ltd. FeCl₃.6H₂O was purchased from Sinopharm Chemical Reagent Co., Ltd.

FeCl₃.6H₂O (1.08 g), NaAc (4.00 g), and PAA (100 mg) were dissolved in a mixture of EG (14.0 ml) and DEG (24.0 ml) and stirred magnetically at room temperature for 1 h. Then the mixture was poured into the autoclave and heated at 200°C for 15 h to obtain Fe₃O₄/PAA magnetic beads. After the temperature of the autoclave was cooled to room temperature, Fe₃O₄/PAA particles were obtained by magnetic separation and rinsed alternately with water and ethanol. The cleaned Fe₃O₄/PAA beads were dried at 50°C.

1 mg Fe₃O₄/PAA magnetic particles were dispersed in acetonitrile (300 μL) and reacted with EDC (12.78 mg) and DMAP (1.22 mg) by shaking at room temperature for 6 h to obtain PMPs.

The PMPs were dispersed in 300 μL DCM and reacted with EDC (12.78 mg) and NHS (7.67 mg) by shaking for 12 h (Supplementary Figure S1A) at room temperature.

The treated biological samples were reacted with activated PMPs (3 mg) in a mixture of 1 ml DMF/MeOH (v/v, 9/1) and 1.5 μL TEA for 30 min at room temperature. After the reaction finished, the supernatant was discarded. The beads were washed six times with H₂O and DMF alternately, followed by the addition of 500 μL MeOH/H₂O/ACN (v/v/v, 1/1/3) and 6 μL NH₂OH (50 wt. % in H₂O) and shaking for 25 min. Then the PMPs were magnetic separated and the raffinate was centrifuge (4°C, 15,000 × g) for 15 min. 100 μL of the raffinate was added into 400 μL acetonitrile for dilution and MS detection.

Method validation is performed to demonstrate the reliability of various parameters and the developed method. The linearity, precision, accuracy, the limit of quantification, and the limit of detection were tested.

A mixed stock solution of amine standards was prepared (Supplementary Table S1). The gradient dilution of the mixed amines standard solution was added to the same concentration of internal standard (IS) solution for quantification. The absolute intensity ratio of amines standards to IS was ordinate values, and the concentration ratio was abscissa values. The calibration curve was plotted, and the equation of calibration curve and coefficient of determination (R ²) were obtained.

The mixture of amine standards was prepared by gradient dilution. The concentration at a signal-to-noise ratio (S/N) of 10 was used as the limit of quantification (LOQ) and the concentration at a signal-to-noise ratio (S/N) of 3 was used as the limit of detection (LOD) by MS detection.

Standard solutions with low, medium and high concentrations were mixed with ovalbumin and IS solution (Supplementary Table S2). The ovalbumin (50 mg/ml) was served as a substitute for the serum system. The concentration of the amine was obtained by calibration curve, and the ratio of the calculated concentration to the actual concentration was the recovery.

The precision of the method was assessed by analyzing the relative standard deviation (RSD) of intra-day and inter-day concentrations of amines in serum samples.

All mass spectra were acquired using ftmsControl 3.0.0 (Bruker Daltonics, Billerica, MA, United States) software. Molecular information (mass-to-charge ratio and absolute intensity) of amines in the mass spectra were set up and extracted using Data Analysis 4.0 software. Substance identification was performed by precise molecular weights from high-resolution mass spectrometry and library search matching identification through an open-source database (http://www.hmdb.ca/). GraphPad Prism 8 was used for scatter plotting and the Kruskal-Wallis test. Binary logistic regression analysis was performed by IBM SPSS Statistics 26 to screen potential biomarkers, and receiver operating characteristic (ROC) was used to evaluate the diagnostic ability of potential biomarkers.

The preparation procedure for PMPs is schematically illustrated in Figure 1. Briefly, Fe₃O₄/PAA particles were prepared by a modified solvothermal reaction. Then the Fe₃O₄/PAA particles were coated by polythioesters via polymerization of 3-mercaptopropionic acid to form core-shell magnetic particles (PMPs). The typical scanning electron microscope (SEM) images and transmission electron microscopy (TEM) of magnetic particles are shown in Figures 2A, B. The Fe₃O₄/PAA particles displayed uniform size with a mean diameter between 200 nm and 300 nm. After coated with polythioester, the size was increased and the shell thickness was about 10 nm. As shown in the thermogravimetric analysis of Fe₃O₄/PAA particles and PMPs (Figure 2C), the weight loss of Fe₃O₄/PAA particles was 8% with a step around 220°C, indicating the corresponding decomposition temperature of PAA. With the introduction of polythioester, the weight loss was increased to 47% with a pronounced step around 200°C. Figure 2D shows the X-ray diffraction patterns of magnetic particles. The similar profiling of diffraction patterns of Fe₃O₄/PAA particles and PMPs were observed with a series of characteristic peaks in good accordance with the inverse cubic spinel phase of Fe₃O₄ (magnetite, JCPDS card no. 85-1436), including (220), (311), (400), (422), (511), and (440). The results indicated that the crystal lattice structure of Fe₃O₄ was not changed by polymer functionalization. The infrared spectra of magnetic particles are shown in Figure 2E. The characteristic absorption peak at 1676 cm⁻¹ corresponds to the stretching vibration of C=O. After coated with polythioester, several new absorption peaks were added between 958 cm⁻¹ and 1059 cm⁻¹. These peaks correspond to the stretching vibration of C-S-C, indicating the thioester bond formation. As shown in Figure 2F, the superparamagnetic responsiveness of Fe₃O₄/PAA and PMPs allowed them to be easily separated from the solution. With polythioester modifications, the saturation magnetization value of PMPs was decreased from 65.39 to 24.68 emu/g. The structure of polythioester was identified by mass spectrometry (Supplementary Figure S1 and Supplementary Table S1), the major ions serie with a mass difference of 88 Da between peaks indicated that the repeating unit of the polythioester was -S-CH₂-CH₂-CO-.

The chemoselective capture of standards by PMPs was based on the condensation reaction between amino and NHS-activated carboxyl groups. The standard derivatives were released through the cleavage of the thioester bond by hydroxylamine (Figure 1) and detected by LC-MS/MS. The reaction conditions of polymerization, chemoselective capture, and release were optimized by comparing the peak area of derivatives of standards.

The addition of DMAP is crucial for the polymerization reaction and the capture efficiency of PMPs. With EDC as a condensation agent, the condensation reaction of MPA was carried out in the dichloromethane solution. The oligomers and by-products were generated in the transparent solution (Supplementary Figures S2A–F). With the addition of DMAP, polythioester precipitated in 6 h, indicating that self-polymerization of MPA was developed with the catalysis of DMAP (Supplementary Figures S2B, C, G–I). As shown in Figure 3A, the peak area of derivatives cleaved from PMPs was increased by 2 orders of magnitude compared with that of magnetic particles prepared without DMAP, indicating the dramatically enhanced capture efficiency of PMPs with the introduction of polythioester. Figures 3B–D shows that the peak area of standard derivatives increases first and then declines with increasing concentration of DMAP and MPA as well as the prolonged reaction time, which could lead to excessive polymerization and might be adverse for detection due to the interference of byproducts. Therefore, the concentrations of DMAP and MPA were optimized as 30 mM and 200 mM respectively and 6 h is the optimal reaction time for polymerization.

The conditions of chemoselective capture and release of standards were investigated (Figures 3E, F and Supplementary Figures S3A–E). The peak areas of standard derivatives increased with a prolonged reaction time of capture from 0 to 30 min and stayed broadly flat with proceeding on increased reaction time. Thus, 30 min was an appropriate duration for capture. The reaction time of carboxyl group activation was optimized as 10 h. Served as a proton receptor for condensation reaction between PMPs and standards, TEA was added and the concentration was optimized as 1.5 μL/ml in DMF/MeOH (v/v, 9/1). The excessive TEA might accelerate the deactivation process, which was adverse for capture. The standard derivatives were released from PMPs via cleavage of thioester bond by hydroxylamine. The peak areas of standard derivatives increased with the mounting concentration of hydroxylamine from 2 μL/ml to 12 μL/ml and stayed broadly flat with proceeding on increased concentration. Thus the concentration of hydroxylamine was optimized as 12 μL/ml in MeOH/H₂O/ACN (v/v/v, 1/1/3). There was no consistent trend for detection of all the standards with an increased reaction time of release from 0 to 55 min. Factoring in the results, the release time was optimized as 15 min.

The quantitative analysis of standards was based on direct infusion Electron Spray Ionization-Fourier Transform Ion Cyclotron Resonance Mass Spectrometer (ESI-FTICR MS). The theoretical m/z values, observed m/z values, and mass errors of the standard derivatives in negative ion mode are shown in Supplementary Table S2. The mass error was less than .0001 Da and the relative mass deviation was less than 1 ppm. The analysis method was of excellent linearity with the correlation coefficients higher than .99. The limits of quantification were between .0095 and .6075 μM and the limits of detection were between .0032 and .6075 μM. The intra-day relative standard deviations (RSD) of the method ranged from .34% to 4.42% and the inter-day RSDs ranged from 1.51% to 14.88%, indicating the excellent stability of the analysis method. The recoveries of the standards at the three levels ranged from 76.17% to 115.32%, suggesting that the method was accurate and reliable.

With excellent stability and accuracy, the method was further applied in the analysis of amines in biological samples. Intracellular amines from three lung adenocarcinoma cell lines, HCC827, NCI-H1650, and NCI-H2228 were analyzed. 26 amino acids and cystathionine (dipeptide) were quantified (Supplementary Table S3). As shown in Figure 4, the intracellular amines were involved in complicated and inconstant metabolic pathways, including several amino acid biosynthesis pathways, amino acid metabolism pathways, and the urea cycle. Kruskal-Wallis test for amines revealed that ornithine, N-acetylornithine, and arginine in the arginine biosynthesis pathway differed between lung adenocarcinoma cells (Figure 4ii, Supplementary Figure S4, and Supplementary Table S4), which was similar to previous reports (Lim et al., 2018).

The PMPs were applied in the analysis of amino metabolites from mouse tissues, including mouse heart tissue (n = 5), liver tissue (n = 5), spleen tissue (n = 5), lung tissue (n = 5), kidney tissue (n = 5), and brain tissue (n = 5). 25 amino acids, cystathionine (dipeptide), and alanylglutamine (dipeptide) were quantified (Supplementary Table S5). The amines are involved in several pathways, such as the TCA cycle, urea cycle, amino acid biosynthesis pathway, and amino acid metabolism pathway (Figure 4). The Kruskal-Wallis test showed that amines were significantly different between tissues (Supplementary Table S6). Among them, aromatic amino acids were less abundant in brain tissues than in other tissues (Supplementary Figure S5), which were involved in phenylalanine, tyrosine and tryptophan biosynthesis (Figure 4i). The results indicate that the analysis method demonstrates extensive application in the detailed study of the biochemical landscape.

The method was further applied in the analysis of amines in serum and biomarker discovery. As shown in Supplementary Table S7, the samples from 29 healthy controls (HCs), 37 patients with benign lung diseases (BLDs), and 37 patients with lung cancer (LCs) were processed and analyzed, and a QC sample was added to every 12 samples to verify the stability of the experimental process. 30 amino acids as well as cystathionine (dipeptide) and alanylglutamine (dipeptide) were quantified (Supplementary Table S8). Statistical analysis in this study showed that the profile of serum amines was significantly altered in patients with benign lung disease and lung cancer patients. Healthy controls, patients with benign lung disease, and patients with lung cancer were grouped according to sex and age (<49 and ≥49), respectively. The subjects were matched for age and sex by Wilcoxon rank-sum tests (Supplementary Figure S6). Kruskal-Wallis test was performed for each amine (Supplementary Table S9). Alanine, aspartic acid, glutamine, glutamate, phenylalanine, and 5-hydroxylysine were significantly different among the three groups (Figures 5A–F).

With PMPs treatment, a total of 103 amines were detected from biological samples (Supplementary Table S10). The amino acids were used to differentiate between healthy controls, patients with benign lung disease, and patients with lung cancer. Screening of biomarkers by the binary logistic regression model, the diagnostic ability of potential markers was evaluated by ROC analysis (Figures 5G, H). Panel 1 (alanine, glutamine, glutamic acid, 5-hydroxylysine) could distinguish well between healthy controls and lung diseases (including benign lung diseases and lung cancer) with an AUC of .964 (95%CI [.931-.997]), the sensitivity of 95.9% and the specificity of 82.8% (Table 1). Panel 2 (aspartic acid, glutamine, glutamic acid, phenylalanine, 5-hydroxylysine) distinguished BLDs from LCs with an AUC of .956 (95% CI [.910-1.000]), the sensitivity of 86.5%, and the specificity of 86.5% (Table 1). The results indicate that PMPs are promising nanoprobes for chemoselective capture of amines and biomarker discovery.