Comparative metabolic profiling of posterior parietal cortex, amygdala, and hippocampus in conditioned fear memory

We first used a fear-conditioning paradigm to examine metabolic changes in brain regions induced by fear memory retrieval. Mice were subjected to a fear-conditioning task in which they learned to associate a cue (tone) with a US (foot shock) in a context. Mice in the fear-conditioning (FC) group were first exposed to six pairings of the tone and foot shock. Mice were then placed in the same context and received the tone twice without the foot shock during the next day. Control mice were exposed to the same task as the FC group but did not receive any foot shocks (Fig. 1a). During the task, conditioned fear responses were measured on the basis of the duration of freezing behavior (Fig. 1b, c).

Significant differences in freezing behavior were found between the control and FC groups during fear conditioning on day 1 [time × task interaction, F (6, 186) = 51.47, p < 0.0001; time, F (6, 186) = 49.30, p < 0.0001; task, F (1, 31) = 107.9, p < 0.0001, two-way repeated-measures analysis of variance (ANOVA) (Fig. 1b)]. On day 2, significant differences in fear responses to the tone and context were found between the control and FC groups during the retrieval test [time × task interaction, F (2, 62) = 8.67, p = 0.0005; time, F (2, 62) = 9.79, p = 0.0002; task, F (1, 31) = 135.6, p < 0.0001, two-way repeated-measures ANOVA (Fig. 1b)].

In the retrieval test, the levels of context-dependent freezing behavior were measured for the first 300 s (context test). To determine further changes in freezing by tone in the retrieval test, mice were then exposed to two tones in the same context. The level of freezing behavior was significantly higher in the FC group than the control group in the context only, and context and tone stimuli. After exposure to the tone with the same context, the levels of freezing behavior increased significantly in the FC group in comparison with the context test only (Fig. 1c).

To identify metabolites specific to conditioned fear memory in each brain region, we performed untargeted metabolite profiling of PPC, AMG, and HPC samples and focused on differences between the FC and control groups. In the three selected brain regions, partial least squares-discriminant analysis (PLS-DA) showed the variance in the data among five components (Fig. 2). In PLS-DA of PPC, 36.2% of the variance was explained by Component 1, 10% by Component 2, 10.1% by Component 3, 9.1% by Component 4, and 5.9% by Component 5 (Fig. 2a). In the AMG dataset, 27.2% of the variance was explained by Component 1, 15.8% by Component 2, 6.7% by Component 3, 6.2% by Component 4, and 3.8% by Component 5 (Fig. 2b). In HPC, 37.4% of the variance was explained by Component 1, 10.9% by Component 2, 5.2% by Component 3, 4.5% by Component 4, and 3.3% by Component 5 (Fig. 2c). The score plot for each brain region between Components 1 and 2 is shown in Fig. 2. PLS-DA indicated nearly perfect separations of FC and control groups in all brain regions. 3D score plots of the PLS-DA of PPC, AMG, and HPC are shown in Additional file 1: Fig. S1.

To identify metabolites driving the separation between the FC and control groups, metabolites from each brain tissue were ranked by variable importance in projection (VIP) scores generated from the PLS-DA. Metabolites with high VIP scores (≥ 1.0) were considered contributing to observed separation; 18 fear memory–relevant metabolites with important variations were identified by ascribing VIP scores in PPC, 25 in AMG, and 29 in HPC. The VIP scores of metabolites in each brain region and the patterns of their changes are shown in Fig. 3a–c. On the basis of the VIP score ≥ 1.0, some overlapping metabolites tended to be increased and some tended to be decreased (p-value < 0.1, fold change ≥|1.1|) after fear retrieval in PPC, AMG, and HPC, as revealed by intersection analysis. Twenty-two metabolites were identified in more than one brain region. Fifteen metabolites differed between FC and control-group mice in both PPC and AMG, 8 in PPC and HPC, and 7 in AMG and HPC. Four metabolites overlapped among all three selected brain regions (Fig. 3d). A heat map in Fig. 3e illustrates the trends of changes in the relative levels of the 65 metabolites compared between FC and control groups in PPC, AMG, and HPC. Of note, 13 of the 15 metabolites overlapping between PPC and AMG had similar patterns of increases and decreases (Fig. 3e). The detailed data of metabolites such as fold changes and overlaps are provided in Additional file 2: Dataset S1.

To further determine the biological significance of the fear memory retrieval–relevant metabolites in each brain region, we performed a metabolite set enrichment analysis. Metabolites that were significantly changed were selected and the metabolic pathways that involve these metabolites were examined by enrichment analysis. Metabolites with p-values < 0.05 were selected for the enrichment analysis (26 in PPC, 35 in AMG, and 43 in HPC). Additionally, metabolites with p-values < 0.1 (10 in PPC, 14 in AMG, and 14 in HPC) were included in the list for the enrichment analysis. Thus, the datasets included 36 metabolites for PPC, 49 for AMG, and 57 for HPC. Enrichment analyses using these datasets identified the top 25 metabolic pathways contributing to the separation between FC and control groups (Fig. 4a–c). Details of these pathways are provided in Fig. 4 and Additional file 3: Dataset S2. On the basis of the overlap analysis, sixteen pathways overlapped between PPC and AMG, 9 between PPC and HPC, and 13 between AMG and HPC (Fig. 4d). Thus, more metabolic pathways, including metabolism of amino acids, purine metabolism, and oxidation of fatty acids, overlapped between PPC and AMG than in the other comparisons. The sets of metabolites that differed between FC and control groups in all three brain regions exhibited enrichment in 8 pathways: arginine and proline metabolism, citric acid cycle, glutamate metabolism, glycine and serine metabolism, purine metabolism, pyruvate metabolism, tryptophan metabolism, and Warburg effect (Fig. 4d and Additional file 3: Dataset S2). Among 22 overlapping metabolic pathways on comparing each brain tissue with all other brain tissues, 17 pathways with high ranks in PPC were identified (Additional file 3: Dataset S2). Therefore, we performed pathway analysis that integrated metabolite set enrichment analysis and pathway topology analysis using the datasets of PPC metabolites and KEGG database in MetaboAnalyst software to identify the impact value of each pathway and the importance value of each metabolite (Additional file 4: Fig. S2). The pathways with high significance (p-value < 0.01) and impact value, and the metabolites with p-values < 0.1 and importance value in these pathways are listed in Table 1.

In PPC, pantothenate and CoA biosynthesis had the highest significance and impact value. 4′-Phosphopantetheine had the highest significance and importance value in this pathway. Purine metabolism had a high significance level in pathway analysis. Various metabolites were significantly changed in purine metabolism, although each metabolite had a relatively low importance value. Glutathione metabolism had comparatively high significance and impact value. Glutathione had a high importance value in this pathway (Table 1). ADP-ribose (ADPR), a major metabolite in NAD⁺-dependent signaling, had the highest significance among all metabolites in PPC. Other metabolites with significance (p-value < 0.05), such as ADP-ribose 2′-phosphate (ADPRP; p-value < 0.0001), cyclic ADP-ribose (cADPR; p-value = 0.0275), and NADH (p-value = 0.0199) are also involved in NAD⁺-dependent signaling. Therefore, NAD⁺-dependent signaling seems to be one of the pathways perturbed by fear retrieval in PPC.

Analysis of significantly altered metabolites and pathways prompted us to depict representative pathways (Fig. 5). Pantothenate and CoA biosynthesis, purine metabolism, glutathione metabolism, and NAD⁺-dependent signaling were the pathways most perturbed in PPC by fear memory retrieval. Major metabolites of these pathways were significantly regulated in PPC with p-values < 0.05 and fold change ≥|1. 2|. Fear memory retrieval increased the level of 4′-phosphopantetheine, xanthine, and glutathione and decreased the level of ADP in PPC (Fig. 6a-d). In AMG, the level of xanthine tended to increase and ADP level decreased significantly, similar to PPC (Fig. 6b and c). In contrast, glutathione level was significantly reduced in HPC (Fig. 6d). In PPC, ADPR and ADPRP were increased, whereas cADPR and NADH were decreased by fear memory retrieval (Fig. 6e–h). Fear retrieval also increased the level of ADPR and ADPRP but decreased the level of cADPR in AMG (Fig. 6e–g). Notably, the trends of changes in NADH in both AMG and HPC were similar to that in PPC (Fig. 6h). Those metabolites were significantly regulated and could be used as biomarkers specific to conditioned fear memory in PPC.

Male C57BL/6 mice (8 weeks old; purchased from The Koatech) were maintained in a controlled-light environment (12 h light and 12 h dark cycle; lights on from 7 am to 7 pm) with ad libitum access to standard laboratory chow (Teklad 2018) and water. Mice were handled daily for 1 week prior to behavioral procedures for fear conditioning and retrieval tests. All behavioral procedures were performed in accordance with the guidelines on care and use of laboratory animals as approved by the DGIST Institutional Animal Care and Use Committee to minimize animal suffering.

The fear conditioning procedure was performed in accordance with a previous protocol [9]. For the retrieval test, the context and day conditions of the procedure were modified because fear extinction training was not required.

The mouse test cage (17.78 × 17.78 × 30.48 cm; Coulbourn Instruments) used for fear conditioning consisted of gray walls and a stainless-steel shock grid floor. The grid floor was composed of bars that allowed the delivery of electric foot shocks. Before fear conditioning and the retrieval test, the wall and floor were cleaned with 75% ethanol.

On the day of fear conditioning, mice in the FC group were acclimated in the test cage for 300 s, and a tone (CS; 2.8 kHz, 85 dB, 30 s) was then co-terminated with an aversive foot shock (US; 0.2 mA, 0.5 s), which was presented 6 times at 300, 390, 500, 630, 740, and 870 s [62]. The variable time interval between trials which are tones co-terminated with a foot shock was used to minimize the possibility of the animals expecting the next trial. The retrieval test was performed on the next day. After exploration for 300 s, the same tone, without an electric foot shock, was played twice, at 300 and 390 s. The mice in the control group were exposed to the same context and tones at the same times as in the FC group without receiving foot shocks. Freezing is defined as the suppression of movement with the exception of respiration. FreezeFrame software (Coulbourn Instruments) was used to score freezing behavior, defined as mouse immobility lasting more than 1 s in accordance with a reference [9].

After behavioral procedures, the mice were anesthetized with isoflurane, and brains were collected. Brains were sliced coronally at 1 mm thickness by using a brain matrix. The PPC and AMG punches (2.0 mm diameter for PPC and 1.5 mm diameter for AMG) were collected bilaterally from a slice section (1.0 mm; from − 1.0 to − 2.0 mm relative to bregma). The HPC tissues were manually dissected and cut bilaterally in two slice section (1.0 mm each; from − 1.0 to − 3.0 mm relative to bregma). The tissues were flash frozen in liquid nitrogen and stored at − 80 °C until metabolite extraction. In some brain regions, tissue was not assayed because of inaccurate dissections.

PPC and AMG tissues were extracted with 500 µl of Chilled extraction solvent (acetonitrile:methanol:water, 2:2:1, v/v) containing 5 µM ATP-¹³C₁₀ (Sigma-Aldrich), which was used as an internal standard. HPC tissue was extracted with 800 µl of the same extraction solvent with the internal standard. The mixtures were sonicated for about 7 min (10 cycles), vortexed for 10 min, and then incubated at − 20 °C for 1 h. The samples were centrifuged at 18,000×g for 10 min at 4 °C. The supernatants were collected and stored at − 20 °C. The pellets were re-extracted using 200 µl aliquots of the extraction solvent and the procedure was repeated. The supernatants were combined and dried under nitrogen flow in a nitrogen evaporator. The dry extracts were reconstituted with 100 µl of 60% acetonitrile (ACN) and filtered through a 0.22 µm PTFE filter before analysis.

All samples were analyzed using an Agilent 1290 UPLC system coupled to an Agilent 6530 Q-TOF MS (Agilent Technologies) in positive or negative ESI mode. The instrument parameters were as follows: gas temperature 325 °C, drying gas 11 l/min, nebulizer 30 psig, fragment voltage 170 V, skimmer 60 V, and capillary voltage ± 3,500 V (nozzle voltage 500 V; only negative mode). Internal mass calibration was performed using reference masses during the runs.

The sample (10 µl) was injected into a Merck ZIC-cHILIC column (100 × 2.1 mm, 3 µm) heated to 40 °C in a column oven. The mobile phase A consisted of 9/1 (v/v) water/ACN containing 10 mM ammonium acetate (pH 6.9) and the mobile phase B consisted of 1/9 (v/v) water/ACN containing 10 mM ammonium acetate and 0.1% acetic acid (pH 7.3). Flow rate was 0.3 ml/min and the gradient program was as follows: 0–20 min, 87.5–37.5% B; 20–21 min, 37.5%; column wash: 21–21.5 min, 37.5–0% B; 21.5–23.5 min, 0% B; 23.5–24 min 0–87.5% B; the initial condition (87.5% B) was maintained for 7 min for equilibration. Untargeted data acquisition on the UPLC-MS was performed using Agilent MassHunter Workstation Data Acquisition software (Agilent Technologies).

The UPLC-MS data were deconvoluted into individual chemical peaks and used to find a “feature” that describes a chemical entity in a chromatogram via the algorithms “Molecular Feature Extractor” and “Find by Formula” in Agilent MassHunter Qualitative Analysis B.07.00 software (Agilent Technologies). The results were used for subsequent statistical analysis and data visualization in Mass Profiler Professional (MPP) software (Agilent Technologies). To create feature lists, the data were filtered on the basis of the number of entities detected in at least one condition (FC or control) in each brain region. Features appearing in more than half of the samples in at least one group were used for subsequent analyses.

PLS-DA was performed to display the variance in samples within the groups. To generate the PLS-DA data, the MetaboAnalyst web tool (www.​metaboanalyst.​ca) was used. All data were normalized by using an auto-scaling method that attempts to scale and center the data by dividing each value by standard deviation. By extracting the PLS-DA score values, the visualization aid was generated with a 95% confidence interval. VIP score plots were also generated from the PLS-DA function of MetaboAnalyst. After PLS-DA, the fold change data of metabolites with high VIP scores were log 2 transformed and then displayed as a heat map in GraphPad Prism 8 software. For visual comparison, the concentration changes in the heat map were shown using a color scale ranging from + 3 (red) to − 3 (blue).

The metabolic features that differed between FC and control were identified by accurate mass and isotopic abundance pattern matching to the METLIN database in MPP. To confirm the identity of these metabolites, UPLC-MS/MS analysis was subsequently performed. The lists of compounds created in MPP, which included m/z and retention time values, were used as targets in MS/MS analysis. Several brain samples were re-analyzed by targeted MS/MS analysis on the Q-TOF MS. The sample sizes of PPC, AMG, and HPC for UPLC-MS/MS are 25, 26, and 29, respectively. MS/MS spectra were generated at a given collision energy and the sample spectra were matched to a MS/MS spectral library of standards in METLIN database by using Agilent MassHunter Molecular Structure Correlator B.05.00 software (Agilent Technologies).

Metabolite set enrichment analyses in the three brain regions were performed using MetaboAnalyst. The analyses were based on SMPDB libraries containing 99 groups of metabolite sets that are linked to KEGG database and HMDB. To determine the significance of changes in metabolites and related pathways revealed by enrichment analysis, pathway analysis of PPC was performed using KEGG database in MetaboAnalyst. The pathway analysis module combines results from enrichment analysis and pathway topology analysis.

Total RNA from brain tissue was isolated with TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. The concentration of RNA was determined using a NanoDrop spectrophotometer (Thermo Scientific). cDNA was synthesized from 1.6 µg RNA using the GoScript™ Reverse Transcription System (Promega) as described previously [63]. In a CFX 96 Real-Time system (Bio-Rad), TB Green (TaKaRa Biotechnology) was used for quantitative PCR estimation of the expression of D1r and Gapdh genes. The following primer pairs were used: D1r, forward primer, 5′-CAGCCTTCATCCTGATTAGCGTAG-3′, reverse primer, 5′-CTTATGAGGGAGGATGAAATGGCG-3′. Gapdh, forward primer, 5′-ATCACTGCCACCCAGAAGAC-3′, reverse primer, 5′-ACACATTGGGGGTAGGAACA-3′. The expression level was normalized to that of Gapdh as an endogenous control.

Data were analyzed using univariate and multivariate analyses, including unpaired t-test for independent pairs of groups. The effects of fear retrieval on classification of the samples of each brain tissue were compared. A cut-off value of asymptotic p < 0.05 and threshold > 1.1 were considered significant in unpaired t-test using MPP software. The pathway analyses were performed and the p-value and impact value were assessed using MetaboAnalyst software. All data are presented as mean ± standard error of the mean (SEM). For comparisons of data between two groups, the two-tailed unpaired t-test was used. For comparing freezing percentage data of more than three groups, one-way ANOVA followed by Tukey’s multiple comparisons or two-way repeated-measures ANOVA was performed using GraphPad Prism 8 Software. A p-value of < 0.05 was considered statistically significant.