Stable isotope labeling combined with liquid chromatography-tandem mass spectrometry for comprehensive analysis of short-chain fatty acids

Analytica Chimica Acta 1070 (2019) 51e59

Contents lists available at ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Stable isotope labeling combined with liquid chromatography-tandem mass spectrometry for comprehensive analysis of short-chain fatty acids Jie Zheng, Shu-Jian Zheng, Wen-Jing Cai, Lei Yu, Bi-Feng Yuan, Yu-Qi Feng* Key Laboratory of Analytical Chemistry for Biology and Medicine (Ministry of Education), Department of Chemistry, Wuhan University, Wuhan 430072, PR China

h i g h l i g h t s

g r a p h i c a l a b s t r a c t


4-AMBA and AMBA-d5 were synthesized to label SCFAs.
The detection sensitivities of most SCFAs were enhanced by up to three orders of magnitude after 4-AMBA labeling.
The developed method was applied in the analysis of SCFAs from the feces of mice.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 23 February 2019 Received in revised form 6 April 2019 Accepted 9 April 2019 Available online 11 April 2019

Short-chain fatty acids (SCFAs) are one class of bacterial metabolites mainly formed by gut microbiota from undigested fibers and proteins. These molecules are able to mediate signal conduction processes of cells, acting as G protein-coupled receptors (GPR) activators and histone deacetylases (HDAC) inhibitors. It was reported that SCFAs were closely associated with various human diseases. However, it is still challenging to analyze SCFAs because of their diverse structures and broad range of concentrations. In this study, we developed a highly sensitive method for simultaneous detection of 34 SCFAs by stable isotope labeling coupled with ultra-high performance liquid chromatography-electrospray ionizationmass spectrometry (UHPLC-ESI-MS/MS) analysis. In this respect, a pair of isotope labeling reagents, N-(4(aminomethyl)benzyl)aniline (4-AMBA) and N-(4-(aminomethyl)benzyl)aniline-d5 (4-AMBA-d5), were synthesized to label SCFAs from the feces of mice and SCFA standards, respectively. The 4-AMBA-d5 labeled SCFAs were used as internal standards to compensate the ionization variances resulting from matrix effect and thus minimize quantitation deviation in MS detection. After 4-AMBA labeling, the retention of SCFAs on the reversed-phase column increased and the separation resolution of isomers were improved. In addition, the MS responses of most SCFAs were enhanced by up to three orders of magnitude compared to unlabeled SCFAs. The limits of detection (LODs) of SCFAs were as low as 0.005 ng/mL. Moreover, good linearity for 34 SCFAs was obtained with the coefficient of determination (R2) ranging from 0.9846 to 0.9999 and the intra- and inter-day relative standard deviations (RSDs) were <17.8% and 15.4%, respectively, indicating the acceptable reproducibility of the developed method. Using the developed method, we successfully quantified 21 SCFAs from the feces of mice. Partial least squares discriminant analysis (PLS-DA) and t-test analysis showed that the contents of 9 SCFAs were significantly different between Alzheimer's disease (AD) and wide type (WT) mice fecal samples. Compared to WT mice, the contents of propionic acid, isobutyric acid, 3-hydroxybutyric acid, and 3-hydroxyisocaleric acid were decreased in AD mice, while lactic acid, 2-hydroxybutyric acid, 2-hydroxyisobutyric acid, levulinic acid, and valpronic acid were increased in AD mice. These significantly changed SCFAs in the feces of AD mice may afford to a better understanding of the pathogenesis of AD. Taken together, the developed

Keywords: Short-chain fatty acids Stable isotope labeling Liquid chromatography-tandem mass spectrometry Quantification Alzheimer's disease

* Corresponding author. E-mail address: [email protected] (Y.-Q. Feng). https://doi.org/10.1016/j.aca.2019.04.021 0003-2670/© 2019 Elsevier B.V. All rights reserved.

52

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

UHPLC-ESI-MS/MS method could be applied for the sensitive and comprehensive determination of SCFAs from complex biological samples. © 2019 Elsevier B.V. All rights reserved.

1. Introduction Short chain fatty acids (SCFAs) usually refer to carboxylic acids that contain less than 6 carbons [1]. They belong to one important class of intestinal bacterial metabolites and usually fermented from undigested dietary fibers by gut microbiota in the cecum and colon [2]. Acetate, propionate, and butyrate make up for the 90% of the whole SCFAs [3]. Protein fermentation can also contribute to the SCFA pool when fermentable fibers are in short supply, but mostly gives rise to branched-chain fatty acids such as isobutyric acid, 2methylbutyric acid, 3-hydroxyisovaleric acid, and isovaleric acid, exclusively originating from branched-chain amino acids, valine, isoleucine and leucine [4]. Once SCFAs are produced, some of them are eliminated from the body and the other enter into the circulatory system. It has been reported that SCFAs, especially propionic acid, butyric acid, and 3-hydroxybutyric acid, were important signaling molecules which could act as G protein-coupled receptors (GPR) activators and histone deacetylases (HDAC) inhibitors [5,6]. SCFAs take an important part in almost all circulation systems and regulate human physiological functions [1]. Additionally, there is a strong relationship between SCFAs and Type 2 diabetes mellitus (T2DM) [7], colorectal cancer (CRC) [8], inflammatory bowel diseases (IBD) [9], and Alzheimer's disease (AD) [10]. Thus, accurate and comprehensive quantification of SCFAs is vital for the diagnosis and treatment of associated diseases. Although the chemical structures of SCFAs are simple, comprehensive analysis of SCFAs is still challenging due to their structural diversities and broad range of concentrations. Nuclear magnetic resonance (NMR), gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LCMS) have been used to detect and quantify fecal SCFAs [11e13]. However, the low detection sensitivity of NMR made the detection of low-abundant SCFAs challenging [11]; while GC-MS was suitable for analytes that were thermally stable and easily volatile [12]. SCFAs were also possible to be directly determined by LC-MS/MS under negative-ion mode. However, the separation of the isomers of native SCFAs was generally not satisfying. Similarly, the negativeion mode of ESI-MS/MS also did not achieve the required sensitivity for the analysis of low-abundant SCFAs [13]. In the past two decades, stable isotope labeling strategies were developed and employed in the LC-MS analysis to increase the detection sensitivities of analytes [14e19]. This analytical strategy also has been used to improve the LC separation and enhance the detection sensitivities of the SCFAs derivatives in LC-ESI-MS/MS. In addition, the collaborative use of a pair of isotope reagents could reduce the co-eluted interferences [20e23]. For example, Han et al. used 3-nitrophenylhydrazine to analyze 10 SCFAs in the feces of healthy people and T2DM patients and the labeled products were analyzed by LC-MS/MS under negative ion mode [24]. Recently, Chan et al. used aniline and its isotope counterpart to label 15 SCFAs in infant fecal samples at different time points [25]. This analytical strategy provided the potential for the sensitive detection of SCFAs. In general, the previous studies mainly focused on alkyl SCFAs. Many other important bioactive fecal SCFAs with carbonyl, hydroxy, and alkenyl groups have not been studied. In addition, these studies mainly focused on the SCFAs with high contents and excluded the relatively low-abundant SCFAs probably due to the limit of

detectability. In this work, we developed a sensitive method for simultaneous quantification of 34 SCFAs that carried alkyl, carbonyl, hydroxy, and alkenyl groups in fecal samples. We designed and synthesized a pair of hydrophobic reagents containing both tertiary and primary amino groups, N-(4-(aminomethyl)benzyl)aniline (4-AMBA) and N-(4-(aminomethyl)benzyl)aniline-d5 (4-AMBA-d5). After labeling, the limits of detection (LODs) of SCFAs were as low as 0.005 ng/mL. The detection sensitivities of most analytes were improved by up to three orders of magnitude after 4-AMBA labeling. In addition, 4AMBA-d5 labeled SCFAs were used as internal standards to obtain accurate quantification results through compensating the possible ionization variances due to matrix effect. The developed method was applied in the analysis of SCFAs in the feces of mice and 21 SCFAs were successfully detected in the feces of mice. PLS-DA and ttest analysis demonstrated that some SCFAs exhibited significant difference between AD and WT mice fecal samples. 2. Experimental 2.1. Chemicals and reagents All SCFA standards were purchased from J&K Chemical (Beijing, China), Aladdin (Shanghai, China), Bide Pharmatech (Shanghai, China), and Innochem (Beijing, China). The detailed information of 34 SCFA standards was listed in Table S1 in Supporting Information. Tert-butyl-4-formylbenzylcarbamate, aniline, dimethyl sulfoxided6 (DMSO‑d6), and sodium cyanoborohydride (NaBH3CN) were purchased from J&K Chemical (Beijing, China). Aniline-d5 and triphenylphosphine (TPP) were purchased from Aladdin (Shanghai, China). 2,20 -dithiodipyridine (DPDS), trifluoroacetic acid (TFA), and concentrated hydrochloric acid were purchased from Bide Pharmatech (Shanghai, China). Analytical grade petroleum ether (PE), ethyl acetate (EA), dichloromethane (DCM), ether (DEE), formic acid (FA), and methanol (MeOH) were supplied by Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). HPLC-grade acetonitrile (ACN) and MeOH were obtained from Tedia Co. (Fairfield, OH, USA). Ultrapure water was purified by a Milli-Q apparatus (Millipore, Bedford, MA). The stock solutions of SCFA standards were dissolved in HPLCgrade ACN and ultrapure water at a concentration of 1 mg/mL. The stock solutions of 4-AMBA (100 mmol/mL) and 4-AMBA-d5 (100 mmol/mL) were dissolved in HPLC-grade ACN. All stock solutions were stored at 20  C. 2.2. Synthesis and characterization of 4-AMBA and 4-AMBA-d5 The first step of the synthesis of 4-AMBA was based on previous report [26]. Briefly, tert-butyl-4-formylbenzylcarbamate (0.5 g, 2.13 mmol) mixed with NaBH3CN (0.4 g, 6.38 mmol) were dissolved in MeOH (25 mL) in a 50-mL round-bottom flask and stirred at room temperature for 5 min. Then, aniline (0.22 mL, 2.34 mmol) was dropwise added to the mixture and stirred for 2 h. The reaction process was monitored by TLC. After the solvent was removed under vacuum, the residue was diluted with water (30 mL) and extracted with EA (3  30 mL). The obtained organic layer was washed with brine (3  30 mL) and dried over Na2SO4. Finally, the

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

residue was purified by flash column chromatography on silica gel eluted with EA/PE (1:6, v/v) to give the pure intermediate (0.46 g, 1.47 mmol, yield 69%). The obtained intermediate was then dissolved in DCM (15 mL) under ice-bath, and TFA (5 mL) was added to the mixture. After complete consumption of the starting material (about 45 min, monitored by TLC), the mixture was evaporated and neutralized with aqueous NaHCO3 solution. The aqueous layer was extracted twice with DCM (15 mL), and the organic layer was washed with brine and dried over anhydrous Na2SO4. The solvent was removed under vacuum to afford the final product (0.18 g, 0.85 mmol, yield 58%). The isotope labeling reagent of 4-AMBA was synthesized according to the same procedure. The detailed procedures for the synthesis of 4-AMBA and 4-AMBA-d5 were shown in Figure S1 in Supporting Information. HPLC-MS was carried out to assess these two reagents (Figure S2 in Supporting Information). 1H NMR (400 MHz, DMSO‑d6): 7.30 (s, 4H), 7.02 (d, J ¼ 8.6, 7.3 Hz, 2H), 6.55 (d, J ¼ 8.6, 1.0 Hz, 2H), 6.52e6.47 (m, 1H), 6.24 (t, J ¼ 5.9 Hz, 1H), 4.23 (d, J ¼ 5.8 Hz, 2H), 3.74 (s, 2H) ppm (Figure S3 in Supporting Information). High-resolution characterization: [MþH]þ: 213.1386 Da (Calc. 213.1409 Da). 2.3. Sample preparation and derivatization Mice fecal samples were collected from WT and APP/PS1 mice. It was reported that the pathological changes in APP/PS1 mice showed similarities with AD [27]. The APP/PS1 double transgenic mice C57BL/6J (C57) expressing both human mutant APPK670 N/ M671 N and mutant PS1M146L were utilized as the AD mice model. The phenotypes of these AD mice were verified by genotyping using polymerase chain reaction (PCR) with APP and PS1 primers according to our previous work [23]. 28 WT C57BL/6J mice (Hunan SJA Laboratory Animal Co., Ltd., P.R. China) and 28 AD mice were raised and kept separately under 23  C in lighting-controlled house (12 h of light and dark) with free access to a standard chow and water. At approximate three months of age (body weight 16e20 g), the fecal samples from the above 56 mice were collected. Samples collection time was from 7 a.m. to 9 a.m. The fresh fecal samples were kept at room temperature for less than 2 h after collection and then stored at 80  C for further analysis. Sample preparation was performed according to a reported method [28,29]. Briefly, feces were grinded into powers under liquid nitrogen. 20 mg feces were precisely weighed and 0.5 mL of ice water was added. The mixture was homogenized for 2 min. Next, hydrochloric acid solution was added to approximately adjust pH1.0, followed by vortexing for 3 min. Then, the mixture was extracted with ice-cold DEE (3  0.5 mL) and the liquid supernatant was collected followed by dried under gentle stream of nitrogen. The residue was redissolved with 1.5 mL ACN and an aliquot of 50 mL was subsequently mixed with 20 mL of 40 mmol/mL TPP/DPDS and 20 mL of 80 mmol/mL 4-AMBA. The mixture was vibrated at 40  C for 10 min. After the reaction was completed, 4-AMBA-d5 labeled SCFA standards were added to the mixture before injection to ensure accurate quantification (Fig. 1). 2.4. Optimization of 4-AMBA labeling conditions To obtain an efficient derivatization, 10 ng/mL of 7 SCFAs as representative analytes (alkenyl SCFAs: crotonic acid, 4-pentenoic acid, carbonyl SCFAs: 5-oxohexanoic acid, 2-oxopentanoic acid, hydroxy SCFAs: 2-hydroxyhexanoic acid, 3-hydroxybutyric acid, and alkyl SCFAs: 2-methylbutyric acid) were at first selected as the substrates in 400 mL ACN to optimize the reaction conditions. TPP/ DPDS were used as the catalysts according to the reported work [30]. We evaluated the effects of different concentrations of TPP/

53

DPDS and 4-AMBA on the reaction. Meanwhile, reaction temperature and reaction time were also optimized to achieve good derivatization efficiency. After the reaction was completed, the resulting solvent was evaporated under nitrogen stream. Finally, each sample was dissolved with 400 mL 30% ACN and then subjected to UHPLC-ESI-MS/MS analysis. 2.5. UHPLC-ESI-MS/MS The mass spectrometry analysis was performed on a LC-MS system consisting of a Shimadzu MS-8045 mass spectrometer (Tokyo, Japan) with an electrospray ionization source (ESI) (Turbo Ionspray) and a Shimadzu LC-20AD HPLC system (Tokyo, Japan). The chromatographic separation of 4-AMBA labeled SCFAs was performed on a Waters Acquity UPLC BEH C18 (50  2.1 mm, 1.7 mm) with a flow rate of 0.4 mL/min at 40  C. FA in water (0.1%, v/ v, solvent A) and ACN (solvent B) were employed as the mobile phases. A post-gradient for column equilibration of 0e3 min 5% B, 3e11 min 5%e22% B, 11e26 min 22% B, 26e41 min 22%e55% B, 41e43 min 55% B, and 43e45 min 55%e5% B was used. Autosampler was kept at 4  C during the analysis. Triplicate measurements were performed. All 4-AMBA labeled SCFAs were detected by multiple reaction monitoring (MRM) under the positive ion mode. The LC-MS system provided an automatic optimization for the collision energy (CE). The two most sensitive MRM transitions per analyte were monitored in subsequent analysis, one for quantitation and the other as a qualifier for the verification of analytes. As for metabolites that could be fully separated, the peak areas were used to represent their intensity; while the peak heights were used for the quantification of metabolites that could not be fully separated. The optimal electrospray ionization (ESI) source conditions were as follows: DL temperature 250  C, heat block temperature 400  C, nebulizing gas 3 L/min and drying gas 15 L/min, heating gas 10 L/min, interface temperature 300  C. The MRM mass spectrometric parameters and chromatographic retention times of 4-AMBA labeled SCFAs were summarized in Table S2 in Supporting Information. Labsolution software (version 5.53 sp2, Shimadzu, Tokyo, Japan) was used for the system control and data processing. 2.6. Method validation Under the optimized labeling conditions, 4-AMBA labeled SCFA standards at concentrations of 0.020, 0.050, 0.10, 0.20, 0.50, 1.0, 5.0, 20.0, 50.0, 100.0, 200.0, 500.0, and 1000.0 ng/mL and fixed amounts of internal standards (4-AMBA-d5 labeled SCFAs) were used to construct calibration curves by plotting the peak areas/heights ratio (analyte/IS). The ISs were added before UHPLC-ESI-MS/MS analysis. The LODs and limits of quantitation (LOQs) were determined at a concentration where the S/N ratios were 3 and 10, respectively. We also determined the LOQs of unlabeled SCFAs in MRM () to evaluate the change of detection sensitivities before and after 4AMBA labeling. The accuracy and precision of the developed method were evaluated by the recoveries and intra- and inter-day RSDs. All the recoveries and intra- and inter-day RSDs were calculated with SCFA standards spiked in the fecal extracts of mice at three different concentrations. The intra-day variations were evaluated by repeating the process for five times within one day and the interday variations were assessed by the RSDs of the recoveries on three successive days. 3. Statistical analysis Model building and identification of the characteristic SCFAs

54

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

Fig. 1. (a) Labeling reaction for 4-AMBA/4-AMBA-d5; (b) Work flow of quantification of SCFAs. LLE, liquid-liquid extraction.

with significant difference of contents between WT and AD mice were performed using the supervised partial least squares discriminant analysis (PLS-DA). In PLS-DA, the R2X, R2Y and Q2 (cum) were used for the model evaluation. R2X is the percentage of all response variables explained by the model. R2Y is the percentage of all observation or sample variables explained by the model. Q2 is the percentage of all observation or sample variables predicted by the previously established model [31]. The contribution of each SCFA in the PLS-DA was evaluated by the values of variable importance for the projection (VIP). The values of VIP positively reflected the influence of certain SCFAs on the classification. Generally, metabolites with VIP values > 1 were considered to be statistically significant. The statistical data of independent t-test was processed with IBM SPSS 22.0 software to evaluate the differences of the concentrations of SCFAs between WT and AD mice fecal samples. All p values were two sided and generally p values < 0.05 were considered statistically significant. 4. Results and discussion 4.1. Mass spectrometry characteristics of 4-AMBA labeled SCFAs First, we explored the MS fragmentation pathways of 4-AMBA labeled SCFAs under collision-induced dissociation (CID). Matched precursor ions revealed that 34 SCFAs could be effectively labeled by 4-AMBA. Interestingly, the neutral loss (aniline, molecular weight, 93.1) from the 4-AMBA occurred in all derivatives (Fig. 2). However, the fragmentation patterns of alkyl, carbonyl, hydroxy, and alkenyl SCFAs were not the same. For 4-AMBA labeled ketonic acids, the product ions of m/z 120.1 and 196.1 occurred under CID (Figure S4 (a), S4 (b), S4 (c), and S4 (d) in Supporting Information).

These two product ions were also found in the MS/MS spectra of 4AMBA labeled alkyl SCFAs, but the intensity of m/z 196.1 gradually diminished with the increase of carbon numbers (Figure S4 (e), S4 (f), S4 (g), and S4 (h) in Supporting Information). In addition, there was a loss of H2O from 4-AMBA labeled SCFAs that carried hydroxy group (Figure S4 (i) and S4 (j) in Supporting Information). The fragmentation behavior of alkenyl SCFAs was variable with unknown reasons (Figure S4 (k), S4 (l), and Table S2 in Supporting Information). Taken together, 4-AMBA labeled SCFAs could produce the neutral loss of 93.1, fragments of m/z 120.1 and m/z 196.1, which met both qualitative and quantitative analysis.

4.2. Optimization of 4-AMBA labeling conditions To obtain relatively high reaction efficiency, we optimized four important labeling conditions with 10 ng/mL of 7 SCFA standards in 400 mL ACN. As shown in Fig. 3a, the labeling reaction reached to a plateau when the concentration of TPP/DPDS was 6.5 mmol/mL, so 20 mL of 6.5 mmol/mL TPP/DPDS was applied to the further experiments. Similarly, we obtained the best labeling efficiency when 20 mL of 3.9 mmol/mL of 4-AMBA was used (Fig. 3b). Temperature slightly influenced the labeling reaction ranging from 20  C to 70  C, which was mainly due to the high reaction activity. Finally, 40  C was used for the following analysis (Fig. 3c). Fig. 3d showed that 4-ABMB labeling reaction was very fast and 10 min was enough for the efficient reaction. Collectively, we chose the optimal 4AMBA derivatization conditions as follow: TPP/DPDS, 20 mL of 6.5 mmol/mL; 4-AMBA, 20 mL of 3.9 mmol/mL; Labeling time, 10 min; Temperature, 40  C. In addition, we also optimized the reaction conditions of SCFAs in the feces of mice and the results were shown in Figure S5 in Supporting Information.

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

55

Fig. 2. General fragmentation behavior of 4-AMBA labeled SCFAs.

Under the optimal labeling conditions, we investigated the efficiency of labeling reaction by examining the remaining native SCFAs after 4-AMBA labeling. The results showed that almost all native SCFAs could not be detected after 4-AMBA labeling (>90%) (Figure S6 in Supporting Information). At the same time, we also investigated the stability of 4-AMBA labeled SCFAs. It was found that these derivatives were stable at least for 20 h, indicating that this labeling reaction could be applied to analyze SCFAs (Figure S7 in Supporting Information).

showed that 4-AMBA-d5 labeled SCFAs were eluted 0.1e0.4 min earlier than 4-AMBA labeled SCFAs, suggesting a slight isotope effect. However, it did not influence the quantification (Supplement for method validation, Table S6 and Figure S8 in Supporting Information). Good linear correlations were obtained with the coefficient of determination (R2) ranging from 0.9846 to 0.9999. The majority of recoveries of SCFAs were between 80% and 120%, except for a few SCFAs that contain hydroxy group with high hydrophilicity. The intra- and inter-day RSDs were <17.8% and 15.4%, respectively, suggesting acceptable precision and accuracy of the developed method (Table S7 in Supporting Information).

4.3. Method validation A series of experiments including LODs, LOQs, linearities, recoveries, and precisions were performed to evaluate the proposed method. As listed in Table S3 in Supporting Information, after 4AMBA labeling, the LODs and LOQs were in the range of 0.005e0.4 ng/mL and 0.015e0.8 ng/mL, respectively. Take several SCFAs as an example, the LOQs of 4-ABMA labeled SCFAs in the current method were comparable to or better than those reported before (Table S4 in Supporting Information) [24,25,32,33]. The detection sensitivities of most SCFAs were enhanced by up to three orders of magnitude (Table S5 in Supporting Information), which endowed the detection of low-abundant SCFAs and also reduced the consumption of biological samples. The isotope effect was also investigated in the separation of 4AMBA-d5 and 4-AMBA labeled SCFAs on C18 column. The results

4.4. Content difference of SCFAs between AD and WT mice fecal samples The feces of 56 mice including 28 WT and 28 AD mice were analyzed by stable isotope labeling coupled with UHPLC-ESI-MS/ MS. Totally, 21 SCFAs were detected and quantified in mice fecal samples. The extracted-ion chromatograms of the detected SCFAs from the feces of mice were shown in Fig. 4. The measured contents of propionic acid, butyric acid, and lactic acid were 530.4 ng/mg, 560.2 ng/mg, and 179.7 ng/mg, respectively, which were comparable to the results reported by previous studies [34,35]. The measured contents of SCFAs in the feces of WT and AD mice were listed in Table S8 in Supporting Information. PLS-DA was performed to assess the classification performance

56

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

Fig. 3. Optimization of labeling reaction by 4-AMBA for 7 SCFAs. (a) Concentration of TPP/DPDS. (b) Concentration of 4-AMBA. (c) Reaction temperature. (d) Reaction time.

of the contents of SCFAs between WT and AD mice fecal samples. As shown in Fig. 5, the PLS-DA score plot of the feces from 28 WT mice (black circle) and 28 AD mice (red circle) indicated that the feces between these two groups could be distinctly separated. The prediction model yielded R2X ¼ 47.6%, R2Y ¼ 89.1%, and a high predictive parameter Q2 (cum) of 85.8%. The VIP values were listed in Table S9 in Supporting Information. Finally, 9 SCFAs as shown in Fig. 6 were found to have significant difference between 28 AD and 28 WT mice fecal samples based on both the VIP value > 1 in PLSDA analysis and p < 0.05 in t-test analysis. AD is a multifactorial central nervous system disease with a high incidence in old people, but the pathogenesis is still not fully understood [36]. It was demonstrated that propionic acid acted as GPR41 and GPR43 activators and 3-hydroxybutyric acid promoted the activity of GPR109A [4]. So, the decreased contents of propionic acid and 3-hydroxybutyric acid in the feces of AD mice (p ¼ 2.2  1010 for propionic acid; p ¼ 5.7  1010 for 3hydroxybutyric acid in Fig. 6) were probably due to the gut microbiota alteration, which weakened the function of GPR activators and usually caused central inflammation [37]. In recent years, a model of gut-brain axis involvement in AD physiopathology was proposed [38]. Pathogenic factors could induce gut inflammation and cause the impairment of the barrier function, and subsequently increase the translocation of SCFAs from intestinal

tract to circulation system; therefore, the contents of certain SCFAs showed lower levels in feces of AD mice [37]. AD and T2DM patients shared several anomalies such as impaired glucose metabolism and insulin signaling, low grade chronic inflammation, and oxidative stress [39], as well as insulin resistance (IR), mainly resulting from age [40]. Interestingly, almost 81% of AD patients have been found to have T2DM or alterations in glucose metabolism [41]. Increased 2-hydroxybutyric acid have also been identified as a potential biomarker of T2DM [42]. In our study, the measured mean concentrations of 2-hydroxybutyric acid were 55.5 ng/mg and 112.8 ng/mg in the feces of WT and AD mice, respectively, which were consistent with the results of reported work [43]. Furthermore, high incidence rate of diabetic acidosis possibly explained the excess content of lactic acid in the feces of AD mice [44]. It has been reported that decreased concentrations of valine, leucine and isoleucine were found in T2DM [7,43,45]. Here we also observed the lowered contents of isobutyric acid and 3hydroxyisovaleric acid fermented by gut microbiota from valine and leucine in AD mice. Further investigation is still necessary to fully elucidate the roles of SCFAs in the regulation of AD. 5. Conclusion In conclusion, we developed a highly sensitive method for

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

57

Fig. 4. Extracted ion chromatograms of 4-AMBA labeled SCFAs from feces of mice.

analyzing 34 SCFAs using stable isotope labeling combined with UHPLC-ESI-MS/MS. SCFAs labeled by 4-AMBA could improve the separation of isomers through increasing the retention of SCFAs on the reversed-phase column and enhance the ionization efficiencies

of SCFAs. In addition, 4-AMBA-d5 labeled SCFA standards acted as internal standards to ensure accurate quantification in complex matrix. With the proposed method, 21 SCFAs that contain different functional groups were successfully detected in the feces of mice.

58

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59

Fig. 5. PLS-DA score plot of the detected SCFAs in mice feces (R2X ¼ 47.6%, R2Y ¼ 89.1%, Q2 (cum) ¼ 85.8%). Black boxes represent WT mice and red dots represent AD mice. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6. SCFAs that differed significantly between WT and AD mice (WT, black; AD, red). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

PLS-DA analysis could effectively distinguish WT from AD mice and 9 SCFAs were considered as differential metabolites. Compared to the WT mice, the contents of propionic acid, isobutyric acid, 3hydroxybutyric acid, and 3-hydroxyisocaleric acid were decreased in AD mice; while lactic acid, 2-hydroxybutyric acid, 2hydroxyisobutyric acid, levulinic acid, and valpronic acid were increased. These results may facilitate the further elucidation of the pathogenesis of AD. Taken together, the developed method provided an accurate view of global SCFAs and could be applied to targeted metabolomics analysis.

thank Professor Fuqiang Xu and Dr. Jie Wang for provide model mice and Jing Xu, Quan-Fei Zhu, Ning Guo, Hua-Ming Xiao, Ya-Lan Wang, Ke He, Shi-Jie Liu, and Jiang-Hui Ding for fecal sample collection. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.aca.2019.04.021. References

Conflict of interest The author declares that there are no conflicts of interest. Acknowledgements The work is supported by the National Key R&D Program of China (2017YFC0906800), and the National Natural Science Foundation of China (21635006, 31670373, 21721005). We would like to

[1] T.S. Postler, S. Ghosh, Understanding the holobiont: how microbial metabolites affect human health and shape the immune system, Cell Metabol. 26 (2017) 110e130. [2] E.S. Chambers, A. Viardot, A. Psichas, D.J. Morrison, K.G. Murphy, S.E. ZacVarghese, K. MacDougall, T. Preston, C. Tedford, G.S. Finlayson, J.E. Blundell, J.D. Bell, E.L. Thomas, S. Mt-Isa, D. Ashby, G.R. Gibson, S. Kolida, W.S. Dhillo, S.R. Bloom, W. Morley, S. Clegg, G. Frost, Effects of targeted delivery of propionate to the human colon on appetite regulation, body weight maintenance and adiposity in overweight adults, Gut 64 (2015) 1744e1754. [3] M.G. Rooks, W.S. Garrett, Gut microbiota, metabolites and host immunity, Nat. Rev. Immunol. 16 (2016) 341e352.

J. Zheng et al. / Analytica Chimica Acta 1070 (2019) 51e59 €ckhed, From dietary fiber to [4] A. Koh, F. De Vadder, P. Kovatcheva-Datchary, F. Ba host physiology: short-chain fatty acids as key bacterial metabolites, Cell 165 (2016) 1332e1345. [5] Y. Xiong, N. Miyamoto, K. Shibata, M.A. Valasek, T. Motoike, R.M. Kedzierski, M. Yanagisawa, Short-chain fatty acids stimulate leptin production in adipocytes through the G protein-coupled receptor GPR41, Proc. Natl. Acad. Sci. U. S. A 101 (2004) 1045e1050. [6] R.W. Johnstone, Histone-deacetylase inhibitors: novel drugs for the treatment of cancer, Nat. Rev. Drug Discov. 1 (2002) 287e299. [7] A. Puddu, R. Sanguineti, F. Montecucco, G.L. Viviani, Evidence for the gut microbiota short-chain fatty acids as key pathophysiological molecules improving diabetes, Mediat. Inflamm. 2014 (2014) 1e9. [8] P. Louis, G.L. Hold, H.J. Flint, The gut microbiota, bacterial metabolites and colorectal cancer, Nat. Rev. Microbiol. 12 (2014) 661e672. [9] A. Trompette, E.S. Gollwitzer, K. Yadava, A.K. Sichelstiel, N. Sprenger, C. NgomBru, C. Blanchard, T. Junt, L.P. Nicod, N.L. Harris, B.J. Marsland, Gut microbiota metabolism of dietary fiber influences allergic airway disease and hematopoiesis, Nat. Med. 20 (2014) 159e166. [10] I.J. Martins, W.M.A.D.B. Fernando, High fibre diets and Alzheimer's disease, Food Nutr. Sci. 5 (2014) 410e424. [11] A. Smolinska, L. Blanchet, L.M. Buydens, S.S. Wijmenga, NMR and pattern recognition methods in metabolomics: from data acquisition to biomarker discovery: a review, Anal. Chim. Acta 750 (2012) 82e97. [12] K. Bielawska, I. Dziakowska, W. Roszkowska-Jakimiec, Chromatographic determination of fatty acids in biological material, Toxicol. Mech. Methods 20 (2010) 526e537. [13] K. Shimizu, H. Ogura, T. Asahara, K. Nomoto, M. Morotomi, Y. Nakahori, A. Osuka, S. Yamano, M. Goto, A. Matsushima, O. Tasaki, Y. Kuwagata, H. Sugimoto, Gastrointestinal dysmotility is associated with altered gut flora and septic mortality in patients with severe systemic inflammatory response syndrome: a preliminary study, Neuro Gastroenterol. Motil. 23 (2011) 330e335. [14] B.L. Qi, P. Liu, Q.Y. Wang, W.J. Cai, B.F. Yuan, Y.Q. Feng, Derivatization for liquid chromatography-mass spectrometry, Trac. Trends Anal. Chem. 59 (2014) 121e132. [15] W. Huang, M.D. Lan, C.B. Qi, S.J. Zheng, S.Z. Wei, B.F. Yuan, Y.Q. Feng, Formation and determination of the oxidation products of 5-methylcytosine in RNA, Chem. Sci. 7 (2016) 5495e5502. [16] H.P. Jiang, T. Liu, N. Guo, L. Yu, B.F. Yuan, Y.Q. Feng, Determination of formylated DNA and RNA by chemical labeling combined with mass spectrometry analysis, Anal. Chim. Acta 981 (2017) 1e10. [17] H. Zeng, C.B. Qi, T. Liu, H.M. Xiao, Q.Y. Cheng, H.P. Jiang, B.F. Yuan, Y.Q. Feng, Formation and determination of endogenous methylated nucleotides in mammals by chemical labeling coupled with mass spectrometry analysis, Anal. Chem. 89 (2017) 4153e4160. [18] H.P. Jiang, J. Xiong, F.L. Liu, C.J. Ma, X.L. Tang, B.F. Yuan, Y.Q. Feng, Modified nucleoside triphosphates exist in mammals, Chem. Sci. 9 (2018) 4160e4167. [19] S.J. Zheng, S.J. Liu, Q.F. Zhu, N. Guo, Y.L. Wang, B.F. Yuan, Y.Q. Feng, Establishment of liquid chromatography retention index based on chemical labeling for metabolomic analysis, Anal. Chem. 90 (2018) 8412e8420. [20] J.M. Chu, C.B. Qi, Y.Q. Huang, H.P. Jiang, Y.H. Hao, B.F. Yuan, Y.Q. Feng, Metal oxide-based selective enrichment combined with stable isotope labelingmass spectrometry analysis for profiling of ribose conjugates, Anal. Chem. 87 (2015) 7364e7372. [21] S. Zhao, M. Dawe, K. Guo, L. Li, Development of high-performance chemical isotope labeling LC-MS for profiling the carbonyl submetabolome, Anal. Chem. 89 (2017) 6758e6765. [22] P. Liu, Y.Q. Huang, W.J. Cai, B.F. Yuan, Y.Q. Feng, Profiling of thiol-containing compounds by stable isotope labeling double precursor ion scan mass spectrometry, Anal. Chem. 86 (2014) 9765e9773. [23] B.F. Yuan, Q.F. Zhu, N. Guo, S.J. Zheng, Y.L. Wang, J. Wang, J. Xu, S.J. Liu, K. He, T. Hu, Y.W. Zheng, F.Q. Xu, Y.Q. Feng, Comprehensive profiling of fecal metabolome of mice by integrated chemical isotope labeling-mass spectrometry analysis, Anal. Chem. 90 (2018) 3512e3520. [24] J. Han, K. Lin, C. Sequeira, C.H. Borchers, An isotope-labeled chemical derivatization method for the quantitation of short-chain fatty acids in human feces by liquid chromatography-tandem mass spectrometry, Anal. Chim. Acta 854 (2015) 86e94. [25] J.C. Chan, D.Y. Kioh, G.C. Yap, B.W. Lee, E.C. Chan, A novel LC-MS/MS method for quantitative measurement of short-chain fatty acids in human stool derivatized with 12C- and 13C-labelled aniline, J. Pharm. Biomed. Anal. 138

59

(2017) 43e53. [26] L. Yu, J. Ding, Y.L. Wang, P. Liu, Y.Q. Feng, 4-phenylaminomethyl-benzeneboric acid modified tip extraction for determination of brassinosteroids in plant tissues by stable isotope labeling-liquid chromatography-mass spectrometry, Anal. Chem. 88 (2016) 1286e1293. [27] A. Bilkei-Gorzo, Genetic mouse models of brain ageing and Alzheimer's disease, Pharmacol. Ther. 142 (2014) 244e257. [28] A.J. Wallace, S.L. Eady, D.C. Hunter, M.A. Skinner, L. Huffman, J. Ansell, P. Blatchford, M. Wohlers, T.D. Herath, D. Hedderley, D. Rosendale, H. Stoklosinski, T. McGhie, D. Sun-Waterhouse, C. Redman, No difference in fecal levels of bacteria or short chain fatty acids in humans, when consuming fruit juice beverages containing fruit fiber, fruit polyphenols, and their combination, Nutr. Res. 35 (2015) 23e34. [29] M. Primec, D. Micetic-Turk, T. Langerholc, Analysis of short-chain fatty acids in human feces: a scoping review, Anal. Biochem. 526 (2017) 9e21. [30] T.M. Li, J. Chen, X. Li, X.J. Ding, Y. Wu, L.F. Zhao, S. Chen, X. Lei, M.Q. Dong, Absolute quantification of a steroid hormone that regulates development in caenorhabditis elegans, Anal. Chem. 85 (2013) 9281e9287. [31] W. Zhang, S. Likhodii, Y. Zhang, E. Arefeshghi, P.E. Harper, E. Randell, R. Green, G. Martin, A. Furey, G. Sun, Classification of osteoarthritis phenotypes by metabolomics analysis, Bmj, Open 4 (2014) e006286. [32] N.H. Park, M.S. Kim, W. Lee, M.E. Lee, J. Hong, An in situ extraction and derivatization method for rapid analysis of short-chain fatty acids in rat fecal samples by gas chromatography tandem mass spectrometry, Anal. Methods. 9 (2017) 2351e2356. [33] M. Zeng, H. Cao, Fast quantification of short chain fatty acids and ketone bodies by liquid chromatography-tandem mass spectrometry after facile derivatization coupled with liquid-liquid extraction, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 1083 (2018) 137e145. [34] A. Kotani, Y. Miyaguchi, M. Kohama, T. Ohtsuka, T. Shiratori, F. Kusu, Determination of short-chain fatty acids in rat and human feces by highperformance liquid chromatography with electrochemical detection, Anal. Sci. 25 (2009) 1007e1011. [35] T. Torii, K. Kanemitsu, T. Wada, S. Itoh, K. Kinugawa, A. Hagiwara, Measurement of short-chain fatty acids in human faeces using high-performance liquid chromatography: specimen stability, Ann. Clin. Biochem. 47 (2010) 447e452. [36] C. Jiang, G. Li, P. Huang, Z. Liu, B. Zhao, The gut microbiota and Alzheimer's disease, J. Alzheimer's Dis. 58 (2017) 1e15. €hler, M. Maes, A. Slyepchenko, M. Berk, M. Solmi, K. Lancto ^t, A. Carvalho, [37] C. Ko The gut-brain axis, including the microbiome, leaky gut and bacterial translocation: mechanisms and pathophysiological role in Alzheimer's disease, Curr. Pharmaceut. Des. 22 (2016) 6152e6166. [38] C. Clemmensen, T.D. Muller, S.C. Woods, H.R. Berthoud, R.J. Seeley, M.H. Tschop, Gut-brain cross-talk in metabolic control, Cell 168 (2017) 758e774. [39] M.Z. Alam, Q. Alam, M.A. Kamal, A.M. Abuzenadah, A. Haque, A possible link of gut microbiota alteration in Type 2 diabetes and Alzheimer's disease pathogenicity: an update, CNS Neurol. Disord. - Drug Targets 13 (2014) 383e390. [40] W.Q. Zhao, M. Townsend, Insulin resistance and amyloidogenesis as common molecular foundation for Type 2 diabetes and Alzheimer's disease, Biochim. Biophys. Acta 1792 (2009) 482e496. [41] J. Janson, T. Laedtke, J.E. Parisi, P. O'Brien, R.C. Petersen, P.C. Butler, Increased risk of Type 2 diabetes in Alzheimer disease, Diabetes 53 (2004) 474e481. [42] S.A. Varvel, J.V. Pottala, D.L. Thiselton, R. Caffrey, T. Dall, M. Sasinowski, J.P. Mcconnell, G.R. Warnick, S. Voros, T.E. Graham, Serum a-hydroxybutyrate (a-HB) predicts elevated 1 h glucose levels and early-phase b-cell dysfunction during OGTT, BMJ Open. 2 (2014) e000038. [43] D. Trico, H. Prinsen, C. Giannini, R. de Graaf, C. Juchem, F. Li, S. Caprio, N. Santoro, R.I. Herzog, Elevated a-hydroxybutyrate and branched-chain amino acid levels predict deterioration of glycemic control in adolescents, J. Clin. Endocrinol. Metab. 102 (2017) 2473e2481. [44] C. Moroti, L.F.S. Magri, M.D. Costa, D.C.U. Cavallini, K. Sivieri, Effect of the consumption of a new symbiotic shake on glycemia and cholesterol levels in elderly people with Type 2 diabetes mellitus, Lipids Health Dis. 11 (2012) 29. [45] L. Zhao, F. Zhang, X. Ding, G. Wu, Y.Y. Lam, X. Wang, H. Fu, X. Xue, C. Lu, J. Ma, L. Yu, C. Xu, Z. Ren, Y. Xu, S. Xu, H. Shen, X. Zhu, Y. Shi, Q. Shen, W. Dong, R. Liu, Y. Ling, Y. Zeng, X. Wang, Q. Zhang, J. Wang, L. Wang, Y. Wu, B. Zeng, H. Wei, M. Zhang, Y. Peng, C. Zhang, Gut bacteria selectively promoted by dietary fibers alleviate Type 2 diabetes, Science 359 (2018) 1151e1156.