Introduction

Dimethylguanidino valeric acid (DMGV) has been reported to be a potentially important biomarker associated with several metabolic disorders, such as obesity1, non-alcoholic fatty liver disease (NAFLD)2, type 2 diabetes2 and cardiovascular disease3. O’Sullivan et al. (2017) first discovered the importance of DMGV as a biochemical compound in metabolic diseases. In this work, DMGV levels were found to be an independent biomarker of CT-defined NAFLD in the Swedish cohort of the Framingham Heart Study, and the baseline levels of DMGV independently predicted future diabetes up to 12 years before disease onset in three distinct human cohorts2. Furthermore, elevated DMGV levels in plasma were associated with the incidence of coronary heart disease and cardiovascular mortality in three different Swedish cohorts involving 5768 participants4. A multicenter prospective cohort study of patients with suspected coronary artery disease also showed that DMGV was associated with the occurrence and severity of coronary artery disease3. Given that dietary habits and physical activity are the primary means of preventing metabolic disorders, studies have also shown that DMGV levels in the blood were highly associated with lifestyles5. Ottosson et al. (2019) revealed that individuals with higher plasma DMGV levels tended to consume more sweetened beverages while taking less vegetables4. It was further demonstrated that macronutrients affected DMGV-related metabolites in mouse models5,6. In addition, high levels of DMGV were associated with poor metabolic responses to long-term cardiovascular prognosis7, and both long-term exercise training intervention8 and acute exercise9 lowered plasma DMGV levels.

According to the latest published research10, the endogenous metabolites DMGV has two stereoisomers, including asymmetric dimethylguanidino valeric acid (ADGV) and symmetric dimethylguanidino valeric acid (SDGV). ADGV and SDGV are produced from asymmetric dimethylarginine (ADMA) and symmetric dimethylarginine (SDMA), respectively, by alanine-glyoxylate aminotransferase 2 (AGXT2)11,12, a mitochondrial enzyme mainly expressed in kidney and liver. The published results on DMGV mainly refer to ADGV, and only one article reported SDGV and showed that SDGV is also one of the determinants in NAFLD progression13.

Although endogenous DMGV, including ADGV and SDGV, are closely related to metabolic diseases, the lack of commercial standards has hindered their mechanistic research and potential clinical application. Strictly speaking, both external and internal standards are required for absolute quantitation for setting up a multiple reaction monitoring (MRM) method using HPLC-QQQ-MS/MS which is the most sensitive and quantitative MS-based approach. However, to the best of our knowledge, only some published articles on DMGV were measured using in-house synthesized DMGV standards1,10,14, and some were based on untargeted metabolomics without DMGV standards7,15,16,17,18. According to the literature, ADGV was first synthesized and detected using a derivatization procedure19. Besides, two other articles mentioned the process of chemical synthesis1,10. However, after numerous attempts in our own lab, we found that the chemical synthesis of DMGV, especially the process for ADMA synthesis, is very time-consuming and has relatively high technical requirements in chemical operation. The synthesized DMGV from ADMA had low yield (5.6%) and low purity (78.4%). Therefore, the establishment of alternative synthetic approaches is necessary.

Biological catalysis is currently a leading method for small-molecule synthesis, with the advantages of being environmentally friendly, efficient, easy to operate, and having low production costs20,21. Studies22,23 reported the biocatalytic production of keto-acid 6-guanidino-2-oxocaproic acid from homoarginine in HepG2 cells with the help of the AGXT2 enzyme overexpressed by an adenoviral vector, Therefore, it is promising to investigate the biocatalytic process of producing ADGV and SDGV from ADMA and SDMA, respectively, under the action of AGXT2 enzyme.

Thus, this study aimed to develop a method for the biosynthesis of DMGV with high yield, purity, and efficiency. Moreover, DMGV-related metabolites play a vital role in the occurrence and progression of metabolic diseases, especially ADMA and SDMA, which are important biomarkers for predicting cardiovascular diseases6,24. Consequently, an HPLC-MS/MS-based method was also established and comprehensively validated for the quantification of a panel of metabolites in the DMGV metabolic pathway, including ADGV, ADMA, β-aminoisobutyric acid (BAIBA), citrulline, homoarginine, NG-monomethyl-L-arginine (NMMA), SDGV, SDMA, and thymine with good analytical performance. The analytical method was further applied to determine the concentration profiles of these metabolites in human plasma samples (Fig. 1).

Fig. 1
figure 1

A general schematic diagram of the study.

Full size image

Materials and methods

DMGV synthesis

The chemical synthesis of ADGV mainly consists of two parts: (A) synthesis of ADMA from ornithine1; (B) synthesis of ADGV from ADMA, according to the references1,10,19. The detailed synthesis process is shown in the supplementary.

For biosynthesis, the gene of truncated human AGXT2Δ1-41 with a fused C-terminal His-tag was codon-optimized and constructed into the expression vector pET28b by Genewiz (Nanjing, China). The pET28b-AGXT2Δ1-41 plasmids were then transformed into E. Coli BL21 (DE3) competent cells by heat shock. A single colony of the AGXT2Δ1-41-His6 expression strain was inoculated into LB broth (100 mL) containing 50 µg/mL kanamycin and shaken at 37 °C overnight. When the culture reached an OD600 of 0.8, protein expression was stimulated by adding 0.4 mM isopropyl-β-D-1-thiogalactopyranoside and incubated at 20 °C, 200 g for 18 h.

Cells were harvested by centrifugation at 8000 g, 4 °C for 25 min and resuspended in phosphate buffer A (20 mM sodium phosphate, 300 mM NaCl, 10 mM imidazole, and 100 µM pyridoxal phosphate; pH = 7.4) supplemented with 1 mM phenylmethanesulfonyl fluoride. Then, cells were lysed under sonication (Misonix) in an ice bath for 10 min at 50% amplitude with a processing time pulse of 10s and an interval of 30s. The lysates were centrifugated at 18,000 g, 4 °C for 1 h. The soluble supernatants were filtered with a 0.45 µM filter and loaded onto a pre-equilibrated HisTrap FF column (Cytavia) with the same phosphate buffer. The enzyme activities were measured by an ultraviolet spectrophotometer (Agilent, USA). The enzyme product was aliquoted and stored at -80 °C before use.

The enzymatic catalysis of ADGV was conducted overnight at room temperature using the above-mentioned purified human AGXT2 enzyme (2.04 mg/mL) and a reaction solution containing 30 mM ADMA, 150 mM glyoxylate, and 10 mM potassium phosphate buffer (pH = 8). After lyophilization, compounds were extracted by protein precipitation using an organic solvent (ACN: MeOH, 9:1, v: v). After centrifugation at 18,000 g, 4 °C for 10 min, the supernatant was collected for further product purification and identification. Other DMGV standards, including d6-ADGV, SDGV, and d6-SDGV were successfully synthesized using the same protocol.

After chemical and biological synthesis, the synthesized standards were purified using preparative HPLC (1260 HPLC, Agilent) and qualified by HPLC-QToF MS/MS (1290II HPLC-6530 MS/MS, Agilent) and proton nuclear magnetic resonance 1H NMR; 400 MHz, Avance NEO, Bruker). The details are shown in the supplementary.

UPLC-QQQ MS/MS-based quantitative method for DMGV metabolite panel

A measurement method for a set of DMGV-related metabolites was established by using UPLC (Vanquish core, Thermo Scientific) coupled to a triple quadrupole (QQQ) MS/MS (TSQ Altis MS, Thermo Scientific), with d6-ADGV, d7-ADMA, d7-arginine, d3-BAIBA and d2-citrulline as internal standards. Moreover, method performance validation was carried out according to the “Guidance for Industry: M10 Bioanalytical Methods Validation and Study Sample Analysis”25 and published articles26,27. This section is described in detail in the supplementary materials.

Analysis of human plasma samples

Human plasma samples from the hospitals were prepared and analyzed using the same protocol as the standard samples in the AB Sciex 6500 QTrap MS/MS. In addition, to determine the levels of SDGV in plasma, another 10 human plasma samples (n = 5 for each gender) were applied. Wilcoxon-Mann–Whitney test was conducted using R (version 4.3.0) to determine the differences in metabolites between sexes. A p-value < 0.05 was considered statistically significant. Detailed information can be found in the supplementary materials.

The study involving human samples was ethically approved, and all experiments were performed in accordance with the Declaration of Helsinki.

Results and discussion

Synthesis of DMGV standards

Generally, the chemical synthesis of ADMA went through a six-step reaction, starting from Boc-Orn(Z)-OH (Suppl. Fig. S1A). During this process, silica gel-based purification was required after each reaction step of reaction to improve the product purity. Thus, the first step of the chemical synthesis from Boc-Orn(Z)-OH to ADMA was very time-consuming and required relatively high chemical experimental operation skills.

The second part of the chemical synthesis was the synthesis of ADGV from ADMA, which was conducted via a two-step reaction (Suppl. Fig. S1B). In total, this two-step reaction took about 6 h, whereas based on the ratio of the amount of ADMA converted after the two-step reaction to the initial amount of ADMA, the conversion rate was about 98.8%. By calculating the ratio of the actual amount of ADGV obtained to the theoretical amount of ADGV, the chemical synthesis yield was about 5.6%. After purification using preparative HPLC, the product purity reached 78.4% (Fig. 2A).

In contrast, the biocatalytic synthesis of ADGV from ADMA with the help of AGXT2 enzyme took about 12 h. The AGXT2 enzyme was expressed and purified from engineered E. coli, which is highly efficient and easy to operate, making it particularly suitable for biological laboratories. According to the ADMA content before and after the reaction, the conversion rate of ADGV was about 99.6%. The product purity easily reached 97.6% after purification by preparative HPLC (Fig. 2B). Therefore, our experimental results showed that the established biocatalytic process of ADGV had a relatively high yield (~ 16.2%) and purity (~ 97.6%).

For compound structure confirmation, QToF MS/MS (m/z 71.1, 70.1, and 46.2 are the characteristic fragments of ADGV with [M + H]+ ion m/z 202.1; Fig. 2C) and 1H NMR (400 MHz, D2O; δ ppm 3.65 (2H, t, CH2CH2CH2), 2.19 (6H, s, N-CH3 × 2), 3.16 (2H, t, CH2CH2CH2), 1.56 (2H, t, CH2CH2CH2; Fig. 2D) were used. The other DMGV standards, d6-ADGV (conversion rate = 99.7%, purity = 76.2%), SDGV (conversion rate = 99.5%, purity = 94.6%), and d6-SDGV (conversion rate = 99.5%, purity = 17.6%), were also catalytically synthesized in the study using the same protocol (Suppl. Fig. S2). To our knowledge, this is the first study on the biosynthesis of these endogenous compounds.

Fig. 2
figure 2

HPLC chromatogram of (A) chemical synthesized ADGV and (B) biocatalytic synthesized ADGV after purification, (C) QTOF MS/MS spectrum, and (D) 1H NMR spectrum of biocatalytic synthesized ADGV.

Full size image

UPLC-MS/MS-based quantitative method for DMGV metabolite panel

According to published articles5,10, a quantitative method for DMGV-related metabolites was established using UPLC-QQQ MS/MS in MRM acquisition mode, while a protein precipitation method using organic solvents, which was commonly used for sample preparation was adopted. For each analyte, the product ion with the highest response was usually chosen as the quantifier ion, and the next highest product ion as the qualifier ion. Representative extracted chromatograms (EIC) and MRM parameters of these DMGV metabolites from a human plasma sample are shown in Fig. 3.

The DMGV metabolite panel has three pairs of isomers: ADGV and SDGV, ADMA and SDMA, and homoarginine and NMMA. They have the same MRM transitions and the same highest product ions, such as ADGV and SDGV sharing 202.1→71.1 and 70.0, ADMA and SDMA sharing 203.1→70.1, and homoarginine and NMMA sharing 189.1→144.1. Therefore, unique ion transitions were selected for these isomers to ensure quantitative accuracy when creating the panel analysis. Still, the unique product ion intensities were sometimes relatively low, largely affecting their detection sensitivity, especially for SDGV. Moreover, arginine was removed from the panel due to severe carryover after sample runs (data not shown).

Furthermore, to mimic the matrix effects of plasma, different ratios of BSA (0.1%, 1%, 5%) were tested as matrix samples by comparing the intensity of the internal standards in these matrix samples with that in human plasma. Finally, 5% BSA was chosen as a surrogate matrix. These results are shown in Suppl. Table S1.

In addition, chromatographic columns commonly used for metabolite analysis were tested. The C18 column (100 × 2.1 mm, 2.2 μm, Acclaim RSLC column, Thermo Fisher) had limited ability to retain these water-soluble metabolites, which were washed out within the first two minutes with poor selectivity. Therefore, considering the analytical performance of RT, selectivity, peak intensity, and peak shape, the HILIC column (150 × 2.1 mm, 1.7 μm, Atlantis, Waters) was finally selected for analysis (Suppl. Fig. S3).

Fig. 3
figure 3

(A) Representative extracted ion chromatograms (EIC) of DMGV-related analytes in a plasma sample with spiked standards. The metabolites from left to right are: (a) thymine; (b) ADGV; (c) d6-ADGV; (d) BAIBA; (e) d3-BAIBA; (f) citrulline; (g) d2-citrulline; (h) d7-arginine; (i) homoarginine; (j) NMMA; (k) SDMA; (l) ADMA; (m) d7-ADMA, and (B) MRM parameters of DMGV analytes in detail.

Full size image

Validation of HPLC-MS/MS detection method

Linearity, accuracy, precision, and sensitivity of the calibration curves

The calibration curves were established based on eight-point concentrations using three batches of standard samples dissolved in 5% BSA. The linear regression models of all analytes and the Limits of Detection (LOD) are shown in Table 1. Overall, the standard curves of all analytes satisfied the method requirements in accuracy (85 ‒ 115%) and precision (< 15%; Suppl. Table S2).

Table 1 Regression curves for linearity in 5% BSA in PBS.
Full size table

The linear ranges of ADGV and SDGV calibration curves were 0.006–0.8 µM (R2 > 0.99). The average accuracy of ADGV at all 8-point concentrations was 98.9% (94.8–107.2%), and their average precision was 2.9% (0.6–6.3%) (Suppl. Table S2a). The precision and accuracy results of other metabolites were listed in Suppl. Table S2. Regarding all analytes, most metabolites passed the acceptance criteria for precision and accuracy, except for NMMA at low levels of concentrations (0.003 and 0.006 µM).

Selectivity, specificity, and carryover of the analytes

The selectivity and specificity of the analytes and internal standards were carefully assessed by using 5% BSA as the matrix samples (n = 6) and human plasma samples (n = 6), respectively. There were no significant interference peaks at adjacent RTs of the analytes in the 5% BSA matrix samples (< 10.6% of the corresponding LLOQ level). The peak areas at the RTs adjacent to the internal standards were also negligible in the 5% BSA matrix samples (< 3.6% of the corresponding internal standards peak area; Suppl. Table S3) and in the human plasma samples (< 4.5% of the corresponding internal standards peak area; Suppl. Table S4).

Carryover of all analytes was evaluated by running a blank sample (50% ACN) after running the standard mixture with the highest concentration in the calibration curve. The carryover amounts in the blank samples were < 4.8% of the corresponding LLOQ for all analytes (Suppl. Table S5). Thus, the method was highly selective for the analytes, with no significant carryover observed at the highest concentration.

Accuracy and precision of the matrix samples with spiked standards

The standard mixtures at low, medium, and high concentrations were added into the 5% BSA matrix, denoted as Low-matrix, Medium-matrix, and High-matrix samples. Intra- and inter-assay precision and accuracy were calculated for these samples (n = 6) and shown in Suppl. Table S6. Taking ADGV as an example (Suppl. Table S6a), the average inter-assay accuracy of ADGV was 98.6% (94.6–102.7%), and the precision was 4.7% (3.6–6.1%). These results are within the acceptance criteria for accuracy (100 ± 15%) and precision (< 15%), demonstrating the reliability and reproducibility of the assay. For the other analytes, most metabolites had high accuracy and precision at all 3 spiking levels, except for low levels of NMMA (Suppl. Table S6f).

Recovery of standards during extraction

The extraction process may result in loss of analytes or internal standards, so recovery was evaluated by comparing the peak areas of the analytes between prior-to-extraction plasma samples spiked with standards and post-to-extraction plasma samples spiked with standards at different concentrations. They are denoted as Low-plasma, Medium-plasma, and High-plasma samples. The recovery results for these samples are shown in Table 2 and 3. The average recoveries of all analytes ranged from 92.7 to 128.6%, which means the extraction procedure had relatively small impacts on the analyte concentration.

Table 2 Recoveries (%)* of plasma samples spiked with standards at low, medium, and high concentrations (n = 6).
Full size table
Table 3 Recoveries (%)* of plasma samples spiked with internal standards at low, medium, and high concentrations (n = 6).
Full size table

Matrix effects of ADGV in plasma samples

The matrix effect was measured by calculating the ratio of the intensity of ADGV or internal standard d6-ADGV added in plasma samples to the intensity in 5% BSA matrix samples and the corresponding IS-normalized MF (Suppl. Table S7).

The average matrix correction factor (MF) was 42.4% for ADGV and 43.9% for d6-ADGV at low, medium, and high spiking concentrations. The internal standard normalized matrix correction factor (IS-normalized MF) which is defined as the ratio of MFanalyte to MFIS, was further used to evaluate the matrix effect. The IS-normalized MFs of ADGV was 96.6% with an RSD of 8.7%, which indicates the internal standard compensated for the matrix effect of the analyte and eliminated interference with the detection of ADGV. In summary, the matrix effect on ADGV was negligible, and potential effects on quantitation were eliminated by using a stable-isotope internal standard (Suppl. Table S7).

Reinjection reproducibility and Dilution integrity

The reinjection reproducibility of analytes at different concentrations is satisfactory (precision < 10%, 85%< accuracy < 115%) based on 5 replicates of standard samples, except for the low concentration of NMMA (0.006 µM; Suppl. Table S8).

Dilution integrity assesses the effect of sample dilution on the accuracy and precision of the measured analyte concentrations. According to our results, the recalculated analyte concentrations were close to the original levels, except for NMMA in the 128-time diluted sample (0.66µM; Suppl. Table S9).

The results of reinjection reproducibility and dilution integrity showed that the analytical performance of NMMA at a low concentration (0.006 µM) was not good enough.

Stability studies of processed samples

Because large sample preparation should be avoided for long sequence analysis to maintain batch-to-batch consistency, cohort samples are typically prepared in batches of 80–120 samples per batch. A single sample run time was approximately 17 min. Therefore, we evaluated the stability of the processed samples after 36 h of storage at the autosampler temperature of 4ºC, which is equivalent to the run time of one batch of samples. The stabilities of the analytes in the processed samples were good after storage, except for high concentrations of thymine (RSD > 15%; Suppl. Table S10).

Impact of internal standards and instruments on ADGV quantification

Using stable isotope-labeled internal standards is considered the most effective way to compensate for the matrix effect. However, considering the cost and availability of isotope-labeled internal standards, internal standards can sometimes be replaced by alternatives, but their impacts on ADGV quantification have not been evaluated. In this study, the impacts of different internal standards (d6-ADGV, d7-arginine, d7-ADMA, d3-BAIBA, and d2-citrulline) on the ADGV quantification were evaluated by using plasma samples spiked with different levels of ADGV (0.045 and 0.650 µM).

The concentrations of ADGV were calculated using the standard curves generated with different internal standards. The results showed that d6-ADGV had the highest quantitative accuracy, around 92.2%, compared with other isotope-labeled internal standards (Fig. 4). When isotopic ADGV is not available, we would recommend choosing d3-BAIBA as the internal standard.

Fig. 4
figure 4

(A) ADGV standard curves using different internal standards (d6-ADGV, d7-arginine, d7-ADMA, d3-BAIBA, and d2-citrulline) on Vanquish HPLC (Thermo Scientific) coupled to TSQ Altis MS/MS; (B) ADGV quantitative accuracy (%) by using different internal standards.

Full size image

In addition, there were no distinguished differences in ADGV quantification between analytical instruments (Suppl. Fig. S4).

Analysis of human plasma samples

The DMGV metabolites in human plasma samples were measured using the above-established quantitative method (Table 4). The levels of these analytes from our population were commensurate with previous literatures10,13,14. We determined the concentration distribution of ADGV in healthy plasma (n = 488; [2.9–16.1 nM]) which is within the reported range by measuring the plasma ADGV levels in 240 human subjects from outpatient clinics, preadmission clinics and the emergency department [1.8 to 74.2 nM]10. However, SDGV was reported to be less than 150 nM in recent research13, but it was not detected in our study. It is mainly due to the low intensity of its unique product ion for quantification (m/z 171.1) in the instrument and low content of SDGV in healthy human plasma. The levels of the other DMGV-related metabolites in our study agree with the reported results from others’ studies28,29. Thus, it is feasible to incorporate these DMGV-related metabolites in one analytical panel.

The content of these metabolites was further compared between males and females. Several metabolites (ADGV, citrulline, NMMA, homoarginine, SDMA, thymine) were significantly different between genders (Table 4, and Suppl. Fig. S5). For instance, ADGV in males (9.4 nM [4.1–16.1]) was much higher than in females (6.2 nM [2.9–11.2]; P < 0.001), and SDMA in males (269.2 nM [238.5-299.2]) is also higher than female (236.5 nM [209.1-266.3]; P < 0.001). However, obvious sexual dimorphism in the distribution of ADMA and BAIBA was not found.

Table 4 The median [Q1, Q3] concentrations of the DMGV-related metabolites in male and female plasma samples (male n = 247 and female n = 241).
Full size table

Conclusion

The study successfully synthesized the endogenous metabolites DMGV, including ADGV, SDGV, d6-ADGV, and d6-SDGV, using a biocatalytic synthetic approach for the first time. Compared to the chemical synthetic method, the biocatalytic method is superior from the perspectives of product conversion rate, yield, product purity, and ease of operation. Furthermore, a targeted MS/MS-based method with high sensitivity, precision, accuracy, selectivity, and high throughput was established to quantify DMGV-related metabolites. The developed method has been applied to determine the concentration distribution of these metabolites in human plasma samples by measuring samples from 488 healthy participants, especially for DMGV. These findings provide an alternative way for DMGV synthesis and MS-based quantification, which will facilitate future mechanisms and biomarker studies on DMGV in different scenarios.