CAW was prepared from the dried aerial parts of Centella asiatica herb (Lot # 170300206; Oregon’s Wild Harvest, Redmond, OR) that was authenticated as described previously (Matthews et al., 2019). A voucher sample of the original plant material is deposited at the Oregon State University Herbarium (OSC-V-258629). Several batches of a crude water extract (CAW a-θ) were prepared as needed from the same lot of plant material using a standardized extraction method described previously (Matthews et al., 2019; Matthews et al., 2020); A voucher sample of each batch is stored at −20°C in our laboratory. Targeted and untargeted HPLC-HRMS/MS (Alcazar Magana et al., 2020) was performed on a representative CAW batch (CAW-iota). Results of targeted analysis of caffeoylquinic acids and triterpenes in CAW iota are shown in Table 1, while untargeted HPLC-HRMS/MS data has been archived at Oregon State University. Animals were provided with CAW (0, 2, 5, or 10 g/L) ad libitum in their water bottles for 5 weeks. We calculated dosing ranges based on the average daily decrease in water volume and denoted these treatments as 0, 200, 500, or 1,000 mg of CAW per kg of body weight per day.

All animal procedures were conducted in accordance with the NIH Guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of the Portland VA Medical Center (IACUC #: 3260-17). 5xFAD male mice were bred with a C57BL/6:SJL F1 female purchased from Jackson Laboratory (Bar Harbor, ME, United States). Identification of 5xFAD progeny and WT littermates was completed by polymerase chain reaction (PCR) of transgenic human amyloid precursor protein (hAPP) from DNA tail samples. Animals were housed in a climate-controlled environment with an alternating 12 h light–dark schedule. Male and female 5xFAD and WT littermates (7.6 ± 0.6 months) were treated with or without CAW (0, 200, 500, or 1,000 mg/kg/d) in the drinking water for 5 weeks. Animals were fed PicoLab Laboratory Rodent Diet 5L0D (LabDiet, St. Louis, MO, United States). Food and water were provided ad libitum. After 5 weeks of treatment, animals were sacrificed at 9 months of age. Brains were dissected and the left cortex was used for metabolite extraction.

Methanol and water (LC-MS-grade) were purchased from EMD Millipore (Burlington, MA, United States). Formic acid (certified ACS reagent) was from Fisher Chemicals (Suwanee, GA, United States). Labeled amino acid indole-3-acetic acid (D7) was used as internal standard (Cambridge Isotope Laboratories, Inc., MA, United States). For metabolomic analysis, mice cortical tissue was extracted with a protocol previously reported with some modifications (Kirkwood et al., 2013). Briefly, the weight of one cortex from one hemisphere was measured accurately and placed in a 2 ml homogenization tube prefilled with ceramic beads (1.4 mm). A proportional amount (20 µl/mg) of homogenization solvent [methanol: water solution (8:2, v/v)] was added to each sample. Samples were spiked with the deuterated amino acid (1 µM final concentration) to account for sample degradation and homogenized using a Precellys™ 24 bead ruptor homogenizer (Bertin Technologies, MD, United States) for three consecutive cycles of 20 s at 5,000 rpm with 30 s of cooldown in between. Samples were placed at −20°C for 1 hour and centrifuged (14,000 rpm, 10 min, 4°C) to collect the supernatant. The resultant supernatant from each sample was transferred to a HPLC vial (Microsolv, Leland, NC, United States) for HPLC–HRMS/MS analysis. A quality control sample was prepared by combining 5 µl of each sample in a single vial. Prior to analysis, 3 µl of a 50 ng/ml solution of CUDA [12-[[(cyclohexylamino)carbonyl]amino]-dodecanoic acid] was added to each sample to allow for platform stability and injection volume correction.

Untargeted HPLC–HRMS/MS analyses were performed using a previously published method (Kirkwood et al., 2013; Housley et al., 2018; Magana et al., 2020) with minor modifications. Briefly, data-dependent acquisition in the positive ion mode was conducted on an AB SCIEX TripleTOF^® 5,600 mass spectrometer (AB SCIEX, Concord, Canada) coupled to a Shimadzu Nexera HPLC system. Chromatographic separation was performed on an Inertsil Phenyl-3 column (4.6 × 150 mm, 100 Å, 5 μm; GL Sciences, Rolling Hills Estates, CA, United States) held at 50°C. A gradient with two mobile phases was used: (A) water (LC-MS grade) with 0.1% v/v formic acid; (B) methanol (LC-MS grade) with 0.1% v/v formic acid. After 1 min at 5% B, the linear elution gradient was as follows: 1 min, 5% B; 11 min, 30% B; 20 min, 100% B; 25 min, 100% B; 30 min, 5% B; and 35 min, 5% B. The injection volume was 6 μl with a flow rate of 0.4 ml/min. Samples were analyzed in a fully randomized batch. The quality control pooled sample was spiked with 5 µl of a 5-ppm solution in ethanol of the drugs verapamil and verapamil-D3 (Cambridge Isotope Laboratories, Inc., MA, United States), and analyzed every ten injections after a methanol blank sample. The isotopic pattern obtained for these internal standards was used to monitor for platform stability along the run. The IonSpray voltage was set at 4,500 V and the source temperature was 500°C. Period cycle time was 950 ms; accumulation time 100 ms; m/z scan range 100–1,400. The collision energy was set at 35 V with a collision energy spread setting of 15 V. Mass calibration of the time-of-flight (TOF) analyzer was performed automatically every five injections.

Raw data was loaded to PeakView™ with XIC Manager 1.2.0 (ABSciex, Framingham, MA) for peak-picking, retention time correction, and peak alignment. Metabolite identities were confirmed as previously described by matching with an in-house library comprising IROA standards (IROA Technology, Bolton, MA) and other commercially available standards (650 total) (García-Jaramillo et al., 2020). The list of identified peaks was exported to MultiQuant 3.0.2, and chromatograms integrated to obtain peak area values for all the assigned metabolites. To account for drift during the metabolomics run, and for potential differences in the injection volume, the annotated metabolites were normalized to the highest peak area measured for the internal standard CUDA. A complete list of annotated metabolites is provided in Supplementary Table S1.

To compare the cortical metabolomic profiles of untreated WT and 5xFAD transgenic mice, peak intensities of individual metabolites were log₂-transformed and mean-normalized by dividing each value by the within-mouse average (across all metabolites) and then scaling by the average of all values (across all metabolites and mice). Heatmaps of peak intensities across all metabolites were created to descriptively assess for differences in magnitude by sex and genotype. To identify changes in the metabolomic profiles, a linear regression model for the normalized log₂ intensities was implemented with indicator variables for sex, genotype, and the interaction between sex and genotype. Within-sex and within-genotype fold changes were back-transformed and reported, as well as the marginal effects for sex and genotype. For all comparisons, volcano plots were created with a horizontal line to indicate the p-value cutoff for the Benjamini-Hochberg control of false discovery rate, which was computed using an overall significance level of 0.05.

The same normalization and descriptive analysis described above were used for the CAW dose-response analysis. For each sex and genotype combination, the normalized log₂ intensities were plotted by dose then separate one-way ANOVAs were conducted with dose as the independent variable for each metabolite. The overall Benjamini-Hochberg adjusted p-values from the ANOVAs were reported. Pairwise t-tests for all possible comparisons were conducted with a Benjamini-Hochberg p-value adjustment. In addition, separate 3-way ANOVAs were conducted on the normalized log₂ intensities for each metabolite to determine if a significant dose-response curve exists and whether it differs by sex and genotype. The least squares slopes for the overall dose-response, the marginal dose-response by sex and genotype, and the within-sex-and-genotype dose-response were computed and reported. The fold changes from a dose of 0–1,000 were calculated using the slopes, and volcano plots of the fold changes with the Benjamini-Hochberg FDR p-value cutoff were generated. All analyses were conducted in StataSE 16.1 and R version 4.0.4. Partial least squares-discriminant analysis and plots (PLS-DA) were generated with MetaboAnalyst 5.0 (Pang et al., 2021).

Metabolite data was imported into MetaboAnalyst 5.0 (Pang et al., 2021) for pathway analysis to compare the metabolomic profiles of WT and 5xFAD mice and to investigate the effects of 500 mg/kg/d CAW on metabolomic profiles in WT and 5xFAD mice. A suitable match for the metabolite maleimide could not be found in MetaboAnalyst and so 119 of 120 annotated metabolites were included in the pathway analyses. A threshold of 0.1 was set for the impact value and a raw p value ≤0.05 was considered significant.

Cortical tissue from untreated (0 mg/kg/d CAW) male and female WT and 5xFAD mice was collected and a PLS-DA analysis was performed using 120 metabolites (Supplementary Table S1) identified through methods described in Section 2.5. As seen in Figure 1, there was separation between the WT and 5xFAD genotypes in both female (1A) and male (1B) mice, though the separation was more clearly defined in female mice than in male mice.

Statistically significant fold changes for metabolites were determined for each sex using a full factorial per metabolite model with a linear regression model estimator and Benjamini and Hochberg false discovery rate corrections (Supplementary Table S2). Relative changes in individual metabolites for male and female 5xFAD mice compared to WT littermates of the same sex are shown in Figure 2. While there were similarities between male and female mice in the direction of change for most individual metabolites resulting from transgenic status, there were clearly observable sex-related differences in either the magnitude of change or the direction of change for several metabolites.

Individual metabolites that were significantly altered in 5xFAD mice compared to WT mice can be seen in the volcano plots for female (Figure 3A) and male mice (Figure 3B). The experimental resolution used in the statistical analysis was such that fold changes between 0.7 and 1.4 could not be distinguished from measurement nonlinearities. Therefore, on the volcano plots, only metabolites outside of those thresholds (indicated by vertical lines) and above the indicated FDR cutoff (red horizontal line) can be considered statistically significant. Overall, 12 metabolites (6 upregulated and 6 downregulated) were significantly altered by the transgene in female 5xFAD mice, whereas there were five significantly altered metabolites (2 upregulated and 3 downregulated) in male 5xFAD (Figure 3). Of these, betaine and N-acetyl-D-mannosamine were the only metabolites that were significantly altered in both male and female 5xFAD mice compared to their sex-matched WT littermates.

A metabolomic pathway analysis was performed using MetaboAnalyst 5.0 to further compare WT and 5xFAD transgenic mice. Results are presented in Table 2. The total number of significantly altered pathways with an impact score ≥0.1 was eleven in 5xFAD females and two in 5xFAD males (Table 2). Amino sugar and nucleotide sugar metabolism was the only metabolic pathway with an impact score ≥0.1 that was significantly altered in both 5xFAD females (p < 0.001) and 5xFAD males (p = 0.027). Significantly altered pathways with the highest impact scores in 5xFAD females included nicotinate and nicotinamide metabolism (impact score = 0.639, p = 0.005), glycine, serine, and threonine metabolism (impact score = 0.591, p = 0.006), and pyrimidine metabolism (impact score = 0.493, p = 0.016). Pentose phosphate and glucuronate interconversion was also significantly impacted in 5xFAD males (impact score 0.250, p = 0.03). Taken together, the results of metabolomic analyses comparing untreated WT and 5xFAD mice indicate that there are significant metabolic alterations associated with the 5xFAD transgene but there is variability in the number, magnitude, and direction of these changes between female and male mice.

Male and female WT and 5xFAD mice were treated with CAW in their drinking water for 5 weeks at four different doses (0, 200, 500, and 1,000 mg/kg/d) and changes in cortical metabolomic profiles were assessed. A PLS-DA analysis was performed for each group using the same 120 identified metabolites (Supplementary Table S1). The PLS-DA plots for WT females (Figure 4A), 5xFAD females (Figure 4B), WT males (Figure 4C) and 5xFAD males (Figure 4D) showed separation between control and CAW-treated animals, particularly at the higher CAW concentrations.

A heat map of the dose-response effects of CAW on the cortical metabolomic profiles of male and female WT and 5xFAD mice compared to genotype- and sex-matched control animals is shown in Figure 5. While there were some similarities between sexes and/or genotypes in response to different concentrations of CAW, changes in individual metabolites were rarely consistent across all four groups of animals (WT females, 5xFAD females, WT males, 5xFAD males). Fold changes were calculated, and statistical significance determined using a linear regression model for each metabolite with log₂ peak intensity as the outcome and treatment dose as the independent variable. Pairwise t-tests for the normalized metabolite intensities at each concentration of CAW compared to genotype- and sex-matched control animals were conducted using a Benjamini Hochberg FDR correction (Benjamini and Hochberg, 1995) (Figure 6). Again, significantly altered metabolites were those with fold changes outside of the 0.7–1.4 range with p values lower than the FDR correction limit when compared to genotype- and sex-matched control animals (Figures 6A–D).

Metabolite changes varied in response to different doses of CAW and there were only a few examples of individual metabolites that were significantly altered at all three doses of CAW for any given genotype. In WT females, glycerol 2-phosphate was significantly downregulated and adenosine 5′-diphosphoribose was significantly upregulated at all three dose levels (Figure 6A). In 5xFAD females, inosine and hypoxanthine were significantly downregulated at all three dose levels (Figure 6B). In WT males, adenosine 3′,5′-cyclic monophosphate and spermidine were significantly upregulated at all three dose levels (Figure 6C). For 5xFAD males, there were several metabolites that were significantly altered at each dose level, but no single metabolite was significantly altered at all three dose levels (Figure 6D). Across all genotypes and sexes (WT females, 5xFAD females, WT males, 5xFAD males), we observed that there were more significantly changed metabolites at the 500 mg/kg/d CAW dose level (24 metabolites total) than at the 200 mg/kg/d (20 metabolites) or 1,000 mg/kg/d (9 metabolites) dose levels. In addition, there were four examples of metabolites that were significantly altered in 5xFAD mice compared to WT mice which were then significantly reversed by CAW treatment (erythritol, nicotinamide adenine dinucleotide (NAD+), nicotinamide hypoxanthine dinucleotide, spermine); these changes all occurred only in 5xFAD females and only at the 500 mg/kg/d dose level (Figures 3A, 6B).

Based on these results, we conducted a metabolomic pathway analysis for female and male 5xFAD and WT mice treated with 500 mg/kg/d CAW compared to genotype- and sex-matched control mice. All metabolic pathways that were significantly altered (raw p value <0.05) in at least one of the four groups are presented in Table 3 with their corresponding impact scores. Nine pathways were significantly altered in 5xFAD females, followed by seven pathways in 5xFAD males, six pathways in WT females, and four pathways in WT males. Three pathways with an impact score ≥0.1 were found to be significantly altered in at least three of the four groups (WT females, WT males, 5xFAD females, 5xFAD males): purine metabolism, nicotinate and nicotinamide metabolism, and glycerophospholipid metabolism.

Purine metabolism (impact score = 0.302; Table 3) was significantly altered by CAW treatment (500 mg/kg/d) in all four groups: 5xFAD females (p = 0.003), WT females (p = 0.011), 5xFAD male (p = 0.047) and WT males (p = 0.044). Fifteen of the 120 annotated metabolites in the dataset were involved in the purine metabolism pathway. Figure 7 shows the purine metabolism pathway with box-and-whisker plots for individual metabolites that significantly changed in at least one of the four groups. Inosine and hypoxanthine were significantly decreased by CAW treatment in both 5xFAD male (p = 0.026, p = 0.042, respectively) and female (p = 0.002, p = 0.006, respectively) mice. Three metabolites (guanosine 5′-monophosphate (GMP), urate, adenosine diphosphate (ADP)-ribose) were significantly increased in WT females only (p = 0.031, p = 0.042, and p = 0.001, respectively). Inosine 5′-monophosphate (IMP) was significantly decreased in both 5xFAD females (p = 0.001) and WT females (p = 0.013). Cyclic adenosine monophosphate (cAMP) was significantly increased in WT males (p < 0.001) and significantly decreased in WT females (p = 0.006).

Nicotinate and nicotinamide metabolism (impact score = 0.639; Table 3) was significantly altered by CAW treatment (500 mg/kg/d) in 5xFAD females (p = 0.010), 5xFAD males (p = 0.007) and WT males (p = 0.007), but not in WT females (p = 0.098). Of the 120 identified metabolites used in the analysis, four metabolites (NAD+, deamino-NAD+, nicotinamide, and aspartate) were involved in the nicotinate and nicotinamide metabolism pathway. As before, box-and-whisker plots in Figure 8 show metabolites that were significantly changed in at least one group. NAD+ was significantly increased in both 5xFAD males (p = 0.028) and 5xFAD females (p = 0.022). In addition, deamino-NAD+ was significantly increased in 5xFAD females (p = 0.026) and nicotinamide was significantly increased in WT females (p = 0.032).

Glycerophospholipid metabolism (impact score = 0.103; Table 3) was significantly altered by CAW treatment (500 mg/kg/d) in WT females (p = 0.050), 5xFAD females (p = 0.012) and 5xFAD males (p = 0.033), but not WT males (p = 0.099). Four of the 120 identified metabolites in our study (choline, cytidine 5′-diphospho (CDP)-choline, choline phosphate, and sn-glycero-3-phosphocholine) were involved in glycerophospholipid metabolism. Box-and-whisker plots in Figure 9 show metabolites involved in glycerophospholipid metabolism that were significantly changed in at least one of the four groups. The metabolite sn-Glycero-3-phosphocholine was significantly increased in 5xFAD males (p = 0.008) and 5xFAD females (p = 0.029), while CDP-choline was significantly increased in 5xFAD females only (p = 0.013). Choline phosphate was significantly decreased in WT males (p = 0.041).