The standards used in this study, including 32 AAs, 19 OAs, and 25 FAs, were purchased from Sigma-Aldrich (St. Louis, MO, USA) and Tokyo Chemical Industry (Kita-ku, Tokyo, Japan). Norvaline, 3,4-dimethoxybenzoic acid, and pentadecanoic acid (PDA) were purchased from Sigma-Aldrich (St. Louis, MO, USA) and were used as internal standards (ISs). HPLC-grade distilled water (DW) and acetonitrile (ACN) were purchased from J.T. Baker, Inc. (Phillipsburg, NJ, USA). Triethylamine (TEA) was purchased from Sigma Aldrich. Toluene, diethyl ether (DEE), ethyl acetate (EA), dichloromethane (DCM), and sodium chloride (NaCl) were purchased from Kanto Chemical Co. (Chuo-ku, Tokyo, Japan). Sodium hydroxide (NaOH) and sulfuric acid (H₂SO₄) were purchased from Daejung Reagent Chemicals (Siheung, South Korea). As derivatization reagents, methoxyamine hydrochloride, N-methyl-N-(tert-butyldimethylsilyl) trifluoroacetamide (MTBSTFA) + 1% tert-butyldimethylchlorosilane, and ethyl chloroformate (ECF) were obtained from Thermo Fisher Scientific (Bellefonte, PA, USA). All chemicals were analytical grade.

For standard and IS stock solutions, 32 AAs and norvaline (IS) were prepared at a concentration of 10 mg/mL in 0.1 M HCl. 19 OAs and 3,4-dimethoxybenzoic acid (IS) were prepared at a concentration of 10 mg/mL in methanol. 25 FAs and PDA (IS) were prepared at a concentration of 10 mg/mL in toluene. All standard stock solutions were stored at –20 ℃. For GC-MS/MS analysis, 19 OAs standard working solution mixtures were prepared at concentrations of 10 µg/mL and 100 µg/mL in methanol, and 3,4-dimethoxybenzoic acid (IS) was prepared at a concentration of 10.0 µg/mL in methanol. The 32 AAs standard working solution mixtures were prepared at concentrations of 10 µg/mL and 100 µg/mL in 0.1 M HCl, and norvaline was prepared at a concentration of 20 µg/mL in 0.1 M HCl. The 25 FAs standard working solution mixtures were prepared at concentrations of 10 µg/mL and 100 µg/mL in toluene, and PDA was prepared at a concentration of 10 µg/mL in toluene.

Eight male Sprague-Dawley rats (Osan, South Korea), aged 8 months (young) and 22 months (old), were used in the experiment. Rats were individually housed in polycarbonate cages with wood chip bedding, maintained in an air-conditioned animal room (temperature: 24 °C; relative humidity: 55 $\pm$ 5%) with 12 h light/dark cycle. The rats were provided tap water and ad libitum access to feed [22]. After young and old rats were euthanized with carbon dioxide, and all brain tissues (cortex, cerebellum, hypothalamus, and hippocampus) were carefully removed and collected and then stored at –80 °C until further analysis. The animal study was designed by the Aging Tissue Bank and approved by the Pusan National University Institutional Animal Care and Use Committee (approval No. PNU-2014-0601) [23].

Analyses of standards and samples were performed using a GCMS-TQ8040 (Shimadzu, Tokyo, Japan) interfaced with a triple quadrupole mass spectrometer (70 eV, electron impact ionization mode) in the selected reaction monitoring (SRM) mode for quantitative analysis. The injector, interface, and ion source temperatures were maintained at 260, 300, and 230 °C, respectively. An Ultra-2 (5% phenyl–95% methylpolysiloxane bonded phase; 25 m $\times$ 0.20 mm i.d; 0.11 µm film thickness) cross-linked capillary column (Agilent Technologies, Palo Alto, CA, USA) was used for analysis. Helium was used as the carrier gas at a flow rate of 0.5 mL/min in a constant flow mode. Samples (1.0 µL) were introduced using AOC-20i auto-injector and AOC-20s auto-sampler in the split-injection mode (10:1). The oven temperature was programmed as follows: AAs, 140 °C was maintained for 3 min, increased to 300 °C at a rate of 8 °C/min and maintained for 5 min; OAs and FAs, 100 °C was maintained for 2 min, increased to 300 °C at a rate of 10 °C/min and maintained for 8 min.

Pooled brain tissues including the cortex, cerebellum, hypothalamus, and hippocampus of rats were used for matrix validation of AAs, OAs and FAs. A mixed standard solution containing 18 AAs of 20–1000 ng and norvaline of 200 ng as IS was used for optimization of AA profiling analysis as EOC/TBDMS derivatives. For the method validation of simultaneous OA and FA profiling analyses as MO/TBDMS derivatives, mixed standard solution containing 13 OAs and 10 FAs of 20–1000 ng, and ISs [3,4-dimethoxybenzoic acid (100 ng) as IS of OAs and PDA (100 ng) as IS of FAs] were used in this study [24, 25]. Then, these were spiked to pooled brain tissue solution of 0.2 mg and was vortex-mixed with acetonitrile (150 µL) to precipitate proteins and centrifuged at 13,500 rpm for 3 min. Then, each method validation was performed according to the following method.

In method validation for AA profiling analysis, the supernatant was transferred to a vial containing ECF (40 µL) and DCM (2 mL), and then DW (820 µL) was added. The aqueous phase was then adjusted to pH $>$12 using 5.0 M NaOH and subjected to a sequential EOC reaction. The aqueous phase was adjusted to pH $\le$2 with 10% H₂SO₄, saturated with NaCl, and sequentially extracted using DEE (3 mL) and EA (2 mL). The extract was evaporated to dryness under a gentle stream of nitrogen at 40 °C. Toluene (15 µL), MTBSTFA (20 µL), and TEA (5 µL) were added to the residue, and then reacted at 60 °C for 60 min to obtain the TBDMS derivative. Finally, 1.0 µL was injected into the GC-MS/MS with SRM mode. Injector, interface, and ion source temperatures were maintained at 260, 300, and 230 °C, respectively.

In method validation for OA and FA profiling analyses, the supernatant was transferred to a vial and DW (820 µL) was added. Methoxyamine hydrochloride (1 mg) was added to the aqueous phase, adjusted to pH $\ge$12 using 5 M NaOH, and reacted at 60 °C for 60 min to form the MO derivative. The aqueous phase was adjusted to pH $\le$2 with 10% H₂SO₄, saturated with NaCl, and sequentially extracted using DEE (3 mL) and EA (2 mL). The extracts containing TEA (5 µL) were evaporated to dryness under a gentle stream of nitrogen at 40 °C. Toluene (10 µL) and MTBSTFA (20 µL) were added to the residue and reacted at 60 °C for 60 min for formation of TBDMS derivative. Finally, 1.0 µL was injected into the GC-MS/MS with SRM mode.

These methods were performed in triplicate under optimal conditions and were validated for analytical parameters including linearity, repeatability, accuracy, limits of detection (LOD) and limits of quantitation (LOQ). The values of slope, intercept and correlation coefficient were determined to linearity test using the least squares regression analysis on a calibration curve constructed based on the relative peak area ratios to IS. The LOD and LOQ value for each metabolite were calculated as three- and 10-times the standard deviation of the blank divided by the slope of calibration curve. Repeatability as relative standard deviation (% RSD) and accuracy as relative error (% RE) were performed in triplicate within the range that included metabolite levels detected.

For the metabolomics study, profiling analyses of AAs, OAs, and FAs were performed using GC-MS/MS as EOC/TBDMS and MO/TBDMS [24, 25]. Briefly, each brain tissue sample was homogenized in DW and centrifuged at 13,500 rpm for 3 min.

For AA profiling analysis, an aliquot of the homogenate equivalent to the weight of each brain tissue (4 mg cortex, 1 mg cerebellum, 1 mg hypothalamus, 0.5 mg hippocampus), and norvaline (0.2 µg) were added to 60 µL of ACN and centrifuged at 13,500 rpm for 3 min. The supernatant was transferred to a vial containing ECF (40 µL) and DCM (2 mL), and then DW (1 mL) was added. The aqueous phase was then adjusted to pH $>$12 using 5.0 M NaOH and subjected to a sequential EOC reaction. And then the aqueous phase was adjusted to pH $\le$2 with 10% H₂SO₄, saturated with NaCl, and sequentially extracted using DEE (3 mL) and EA (2 mL). The extract was evaporated to dryness under a gentle stream of nitrogen at 40 °C. Toluene (15 µL), MTBSTFA (20 µL), and TEA (5 µL) were added to the residue, and then reacted at 60 °C for 60 min to obtain the TBDMS derivative. Finally, the 1.0 µL was injected into the GC-MS/MS with SRM mode.

For OAs and FAs profiling analysis, an aliquot of the homogenate equivalent to the weight of each brain tissue (8 mg cortex, 2 mg cerebellum, 2 mg hypothalamus, 1 mg hippocampus) and 3,4-dimethoxybenzoic acid and PDA (0.1 µg) were added to 120 µL of ACN and centrifuged at 13,500 rpm for 3 min. The supernatant was transferred to a vial and DW (1 mL) was added. Methoxyamine hydrochloride (1 mg) was added to the aqueous phase, adjusted to pH $\ge$12 using 5 M NaOH, and reacted at 60 °C for 60 min to form the MO derivative. The aqueous phase was adjusted to pH $\le$2 with 10% H₂SO₄, saturated with NaCl, and sequentially extracted using DEE (3 mL) and EA (2 mL). The extracts containing TEA (5 µL) were evaporated to dryness under a gentle stream of nitrogen at 40 °C. Toluene (10 µL) and MTBSTFA (20 µL) were added to the residue and reacted at 60 °C for 60 min for formation of TBDMS derivative. Finally, 1.0 µL was injected into the GC-MS/MS with SRM mode. The steps of metabolomics analysis shown in Fig. 1.

Fig. 1.

Workflow the steps of metabolomics analysis. EOC, ethoxycarbonyl; AAs, amino acids; MO, methoximation; OAs, organic acids; FAs, fatty acids; TBDMS, tert-butyldimethylsilyl; GC-MS/MS, gas chromatography-tandem mass spectrometry.

The levels of AAs, OAs, and FAs in the brain tissues were calculated according to each calibration curve. For drawing a star graphic patterns, the mean levels of AAs, OAs, and FAs in the older group were normalized to the mean level in the young group, and then Microsoft Excel Office 365 (Microsoft, Redmond, WA, USA) to draw a star graphic pattern [20, 21, 24, 25]. Univariate statistical analysis was performed by a Student’s t-test to compare the average differences in all metabolites between the young and aging groups. The normality of the data was evaluated using the Shapiro-Wilk test, and the Wilcoxon rank sum test as a non-parametric test was used because the data did not follow a normal distribution. Due to the small sample size, multiple comparison tests were performed using the Benjamini-Hochberg procedure to control the false discovery rate (FDR) for enhancing the accuracy and reliability of the data. The p-values from the results of the Wilcoxon rank sum test were adjusted using FDR, and metabolites with a Q-value 0.05 were considered statistically significant. Multivariate analysis was used to evaluate differences between young and old groups. Principal component analysis (PCA) as unsupervised learning was used for trends and pattern analyses in data. Orthogonal partial least square discriminant analysis (OPLS-DA) as supervised learning was used for discriminating between groups to reveal biomarker candidates. To evaluate the predictability and quantification of the OPLS-DA model (permutation = 100), two parameters (Q2 and R2Y) were calculated. The high R2Y and Q2 indicated that the model was valid. The variable importance in projection (VIP) score was used to measure the importance of variables in the OPLS-DA model, and variables with a VIP score $>$1.0 were considered significant in the OPLS-DA model. For classification analysis and intuitive visualization in metabolite levels, a heatmap was applied to metabolites. The following analyses were performed with log10-transformed and auto-scaled data using MetaboAnalyst (version 6.0) based on the R project (http://www.metaboanalyst.ca) [20, 21, 24, 25].

Metabolites determined in each brain tissue were analyzed using the Wilcoxon rank-sum test by comparing the young and old groups.

In the cortex, the levels of 32 metabolites were determined in the young (n = 8) and old (n = 6) groups. In the young group, acetoacetic acid was the most abundant, followed by lactic acid and glutamic acid, whereas in the old group, lactic acid was the most abundant, followed by glutamic acid and acetoacetic acid (Table 4). In particular, 10 AAs (alanine, glycine, valine, leucine, isoleucine, threonine, serine, proline, phenylalanine, and tyrosine) and lactic acid were increased (p-value 0.05). After FDR correction, eight AAs (alanine, valine, leucine, isoleucine, threonine, serine, proline, and phenylalanine) except for glycine and tyrosine among 10 AAs were significantly increased in the old group (Q-value 0.05, Table 4).

In the cerebellum, the levels of 43 metabolites were measured in the young (n = 8) and old (n = 7) groups. In the young group, docosatetraenoic acid was the most abundant, followed by aspartic acid and acetoacetic acid, whereas in the old group, docosatetraenoic acid was the most abundant, followed by acetoacetic acid and aspartic acid (Table 5). Especially, 11 AAs (alanine, glycine, valine, leucine, isoleucine, threonine, serine, proline, methionine, phenylalanine, and cysteine) and 3-hydroxypropionic acid were increased, whereas two FAs (eicosadienoic acid and erucic acid) were decreased (p-value 0.05). After FDR correction, three metabolites (alanine, phenylalanine, and 3-hydroxypropionic acid) were significantly increased and eicosadienoic acid was significantly decreased in the old group compared to the young group (Q-value 0.05, Table 5).

In the hypothalamus, the levels of 45 metabolites were measured in the young (n = 8) and old (n = 7) groups. In the young group, docosatetraenoic acid was the most abundant, followed by aspartic acid and acetoacetic acid. In the old group, docosatetraenoic acid was the most abundant, followed by aspartic acid and glutamic acid (Table 6). Specifically, seven AAs (glycine, $\beta$-alanine, valine, proline, $\gamma$-aminobutyric acid, cysteine, and aspartic acid) were decreased in the old group compared to those in the young group (p-value 0.05, Table 6). After FDR correction, metabolites were not significantly altered in the old group compared to the young group.

In the hippocampus, the levels of 30 metabolites were measured in the young (n = 4) and old (n = 6) groups. In all groups, acetoacetic acid was the most abundant, followed by lactic acid and glutamic acid (Table 7). Acetoacetic acid was higher in the old group than that of the young group (p-value 0.05, Table 7). After FDR correction, metabolites were not significantly altered in the old group compared to the young group.

The AA levels in the old group were normalized to the corresponding mean levels of the young group. A star plots of the normalized AAs in each brain tissue were shown in Fig. 2. In the cortex, the normalized values of the 22 AAs ranged from 0.92 to 1.58 in the old group (Table 4). In particular, 13 AAs (alanine, valine, leucine, isoleucine, threonine, serine, proline, methionine, phenylalanine, cysteine, lysine, tyrosine, and tryptophan) were increased by 14–58% in the old group (Fig. 2a). In the cerebellum, normalized values of 21 AAs ranged from 0.88 to 1.80 in the old group (Table 5). Especially, 18 AAs (alanine, glycine, $\beta$-alanine, valine, leucine, isoleucine, threonine, serine, proline, $\gamma$-aminobutyric acid, pyroglutamic acid, methionine, phenylalanine, cysteine, aspartic acid, glutamic acid, asparagine, and ornithine) were increased by 10–80%, whereas glutamine was decreased by 12% in the old group (Fig. 2b). In the hypothalamus, normalized values of 22 AA ranged from 0.71 to 1.17 in the old group (Table 6). Specifically, valine was increased by 17%, whereas eight metabolites (glycine, $\beta$-alanine, proline, $\gamma$-aminobutyric acid, pyroglutamic acid, methionine, cysteine, and aspartic acid) were decreased by 10–29% in the old group (Fig. 2c). In the hippocampus, normalized values of 18 AAs ranged from 0.90 to 1.20 in the old group (Table 7). Additionally, five AAs (alanine, valine, leucine, threonine, and pyroglutamic acid) were increased by 11–20%, whereas glutamic acid was decreased by 10% in the old group (Fig. 2d).

Fig. 2.

Star symbol plots of AAs in the (a) in cortex, (b) cerebellum, (c) hypothalamus, and (d) hippocampus from young and old groups. Ray; 1 = Alanine, 2 = Glycine, 3 = $\beta$-Alanine, 4 = Valine, 5 = Leucine, 6 = Isoleucine, 7 = Proline, 8 = $\gamma$-Aminobutyric acid, 9 = Pyroglutamic acid, 10 = Methionine, 11 = Serine, 12 = Threonine, 13 = Phenylalanine, 14 = Cysteine, 15 = Aspartic acid, 16 = Glutamic acid, 17 = Asparagine, 18 = Ornithine, 19 = Glutamine, 20 = Lysine, 21 = Tyrosine, 22 = Tryptophan.

The OA and FA levels in the old group were normalized to the corresponding mean levels of the young group. A star plots of the normalized OAs and FAs of each brain tissue were shown in Fig. 3. In the cortex, the normalized values of nine OAs and one FA ranged from 0.46 to 1.23 in the old group (Table 4). In particular, two metabolites (lactic acid and succinic acid) were increased by 12–23%, whereas five metabolites (acetoacetic acid, glycolic acid, fumaric acid, oxaloacetic acid, and palmitic acid) were decreased by 10–54% in the old group (Fig. 3a). In the cerebellum, the normalized values of 10 OAs and 12 FAs ranged from 0.66 to 1.41 in the old group (Table 5). Especially, six metabolites (pyruvic acid, acetoacetic acid, glycolic acid, 3-hydroxypropionic acid, citric acid, and linolenic acid) were increased by 11–41%, whereas 10 metabolites (oxaloacetic acid, palmitic acid, oleic acid, stearic acid, arachidonic acid, eicosadienoic acid, gondoic acid, docosahexaenoic acid, docosatetraenoic acid, and erucic acid) were decreased by 10–34% in the old group (Fig. 3b). In the hypothalamus, the normalized values of 10 OAs and 13 FAs ranged from 0.61 to 1.18 in the old group (Table 6). Specifically, two metabolites (3-hydroxypropionic acid and linolenic acid) were increased by 10–18%, whereas six metabolites (acetoacetic acid, myristic acid, palmitoleic acid, arachidonic acid, docosahexaenoic acid, and docosatetraenoic acid) were decreased by 12–39% in the old group (Fig. 3c). In the hippocampus, the normalized values of nine OAs and three FAs ranged from 0.74 to 2.04 in the old group (Table 7). Additionally, seven metabolites (pyruvic acid, acetoacetic acid, glycolic acid, 3-hydroxypopionic acid, malic acid, oleic acid, and stearic acid) were increased by 12–104%, whereas oxaloacetic acid was decreased by 26% in the old group (Fig. 3d).

Fig. 3.

Star symbol plots of OAs and FAs in the (a) in cortex, (b) cerebellum, (c) hypothalamus, and (d) hippocampus from young and old groups. Ray; 23 = Pyruvic acid, 24 = Acetoacetic acid, 25 = Lactic acid, 26 = Glycolic acid, 27 = 3-Hydroxypropionic acid, 28 = Succinic acid, 29 = Fumaric acid, 30 = Oxaloacetic acid, 31 = Malic acid, 32 = Citric acid, 33 = Myristic acid, 34 = Palmitoleic acid, 35 = Palmitic acid, 36 = Linoleic acid, 37 = Oleic acid, 38 = Stearic acid, 39 = Arachidonic acid, 40 = Eicosadienoic acid, 41 = Gondoic acid, 42 = Docosahexaenoic acid, 43 = Docosatetraenoic acid, 44 = Erucic acid, 45 = Nervonic acid.

In the cortex, the PCA score plot explained 51.7% of total variance with PC1 and PC2 (Fig. 4a). To discriminate between the two groups, PCA loading scores were evaluated using loading 1 and loading 2. The variable that had the greatest influence on loading 1, the main component of PCA, was phenylalanine, and on loading 2 was pyroglutamic acid (Table 4). Although the young and old groups were not completely separated in the PCA score plots (Fig. 4a), they were clearly separated in OPLS-DA (Fig. 4b). The OPLS-DA presents R2Y and Q2 values equal to 0.927 and 0.736, respectively with p-values from permutation test less than or equal to 0.01. According to OPLS-DA analysis, 13 metabolites (leucine, phenylalanine, proline, isoleucine, valine, threonine, serine, alanine, tyrosine, aspartic acid, glycolic acid, $\beta$-alanine, and lactic acid) with a VIP score $>$1.0 were evaluated as major contributing metabolites for discrimination of the young and old groups (Table 4). A heatmap for the classification and visualization of 32 metabolites showed differences of metabolic changes in brain tissues of young and old rats with aging (Fig. 5a).

Fig. 4.

PCA (a,c,e,g) and OPLS-DA (b,d,f,h) score plots in the cortex, cerebellum, hypothalamus, and hippocampus from young and old groups.

Fig. 5.

Heatmap analysis in the in (a) cortex, (b) cerebellum, (c) hypothalamus, and (d) hippocampus from young and old groups.

In the cerebellum, the PCA score plot explained 53.6% of total variance with PC1 and PC2 (Fig. 4c). To discriminate between the two groups, PCA loading scores were evaluated using loading 1 and loading 2. The variable that had the greatest influence on loading 1 was eicosadienoic acid, and that of loading 2 was palmitoleic acid (Table 5). Although the young and old groups were not completely separated in the PCA score plots (Fig. 4c), they were clearly separated in OPLS-DA (Fig. 4d). The OPLS-DA presents R2Y and Q2 values equal to 0.872 and 0.595, respectively with p-values from permutation test equal to 0.02. According to OPLS-DA analysis, 20 metabolites (alanine, eicosadienoic acid, leucine, isoleucine, 3-hydroxypropionic acid, phenylalanine, cysteine, serine, glycine, docosatetraenoic acid, proline, erucic acid, methionine, arachidonic acid, threonine, valine, pyroglutamic acid, citric acid, oleic acid, and stearic acid) with a VIP score $>$1.0 were evaluated as major contributing metabolites for discrimination of the young and old groups (Table 5). A heatmap for the classification and visualization of 43 metabolites showed differences of metabolic changes in brain tissues of young and old rats with aging (Fig. 5b).

In the hypothalamus, the PCA score plot explained 39.8% of total variance with PC1 and PC2 (Fig. 4e). The variable that had the greatest influence on loading 1 was docosatetraenoic acid, and that of loading 2 was lactic acid (Table 6). Although the young and old groups were not completely separated in the PCA score plots (Fig. 4e), they were clearly separated in OPLS-DA (Fig. 4f). The OPLS-DA presents R2Y and Q2 values equal to 0.905 and 0.315, respectively with p-values from permutation test less than 0.03. According to OPLS-DA analysis, 13 metabolites (aspartic acid, $\gamma$-aminobutyric acid, glycine, $\beta$-alanine, cysteine, proline, valine, alanine, pyroglutamic acid, linolenic acid, arachidonic acid, docosahexaenoic acid, and myristic acid) with a VIP score $>$1.0 were evaluated as major contributing metabolites for discrimination of the young and old groups (Table 6). A heatmap for the classification and visualization of 45 metabolites showed differences of metabolic changes in brain tissues of young and old rats with aging (Fig. 5c).

In the hippocampus, the PCA score plot explained 68.8% of total variance with PC1 and PC2 (Fig. 4g). The variable that had the greatest influence on loading 1 was lactic acid, and that of loading 2 was acetoacetic acid (Table 7). Although the young and old groups were not completely separated in the PCA score plots (Fig. 4g), they were clearly separated in OPLS-DA (Fig. 4h). However, the OPLS-DA presents R2Y and Q2 values equal to 0.881 and 0.249, respectively with p-values from permutation test less than 0.16. According to OPLS-DA analysis, nine metabolites (acetoacetic acid, malic acid, pyroglutamic acid, aspartic acid, pyruvic acid, oleic acid, alanine, glutamic acid, and glycolic acid) with a VIP score $>$1.0 were evaluated as major contributing metabolites for discrimination of the young and old groups (Table 7). A heatmap for the classification and visualization of 30 metabolites showed differences of metabolic changes in brain tissues of young and old rats with aging (Fig. 5d).