*2.3. 1H-NMR Mediated Metabolomics Analysis of SpD-Treated AC16 Cells*

To identify metabolite alterations that are induced by SpD, we used 1H-NMR spectroscopy to characterize 30 mg of AC16 cells incubated with SpD (10 μM) for 24 h. Metabolic profiling (Chenomx, Edmonton, AB, Canada) was used to identify 32 metabolites in SpD-treated AC16 cells and untreated control cells (Table 1).

**Table 1.** Identified metabolites and their corresponding concentrations (mM; mean, standard deviation), as determined by Chenomx NMR Suit 7.1® peak fitting of individual 1H-NMR spectra (600 MHz) for SpD (10 μM) treated AC16 cells (30 mg, *n* = 3).


After normalization of the data (Figure 3), univariate and multivariate statistical analyses were used to comprehensively evaluate the effects of SpD on AC16 cells.

**Figure 3.** Normalization of 1H-NMR aquired metabolite concentrations. The concentrations of metabolites were normalized by log-transformation followed by Pareto scaling (mean-centered and divided by the square root of the standard deviation of each variable). Changes of metabolites are represented as ratios of control metabolites.

Univariate volcano plots of log2(FC) > 1.2 (*p* < 0.05) metabolites showed that the levels of sn-glycero-3-phosphocholine (GPC), glutathione, myo-inositol, taurine, and O-phosphocholine were increased, while the levels of acetate and glutamine were decreased, by SpD (Figure 4).

**Figure 4.** Volcano plots for SpD-induced metabolic changes compared with controls (*n* = 3). Metabolites are considered significant if log2(fold change) > 1.2. The *p*-value threshold was 0.05. The significantly changed metabolites included acetate, glutamine, myo-inositol, glutathione, taurine, O-phosphocholine, and sn-Glycero-3-phosphocholine (GPC).

Multivariate analysis is used to determine the relative differences in two or more systems that are large and complex. Therefore, as shown in Figure 5A, we performed principal component analysis (PCA) of metabolites from SpD-treated AC16 cells. The aim of PCA is to reduce the dimensionality of original data within the preservation of the variance.

To calculate variable importance in projection (VIP) scores of metabolites, we performed partial least-squares projections for latent structures-discriminant analysis (PLS-DA). Metabolites with VIP scores larger than 1.0 were considered as important (Figure S2). To confirm the "goodness" of the model and the predictive quality, we tested orthogonal partial least-squares projections to latent structures-discriminant analyses (OPLS-DA) on data from SpD-treated AC16 cells and control cells (Figure S3). In PCA, the SpD-treated group and control group revealed class differences showing 95% confidence regions separating each other. We extended the supervised PLS regression using orthogonal signal correction filters after selecting VIP > 1.0 metabolites. The metabolites from the SpD-treated AC16 cells significantly differed from the control cell group in the OPLS-DA model. The R2Y model quality parameter was 0.937, demonstrating that the OPLS-DA model was robust (R2Y value near 1.0), and the Q<sup>2</sup> parameter was 0.597, showing that the model was predictive (Q<sup>2</sup> > 0.5) (Figure S3). The loading plot of OPLS-DA is shown in Figure 5C. The heat-map analysis of VIP > 1.0 metabolites was represented with logarithmic fold changes (Figure 5B). In comparison with the control group, the most increased and decreased metabolites with SpD treatment were GPC and acetate, respectively.

**Figure 5.** 1H-NMR metabolomics for SpD-treated AC16 cells. (**A**) Principal component analysis (PCA) indicated that metabolites from the SpD-treated (10 μM, 24 h) group were significantly different from those in the control group; (**B**) Heat-map analysis of metabolites with variable importance in projection (VIP) score > 1.0. The logarithmic fold changes are shown below. GPC, sn-glycero-3-phosphocholine; (**C**) The loading plots from orthogonal partial least-squares discriminant analysis (OPLS-DA) for SpD metabolites compared with the control group.

To interpret metabolic changes, correlation networks were generated according to Pearson's correlation coefficients (|r| > 0.9) between metabolites, in a pair-wise fashion. In untreated controls, acetate shared 28 correlations and GPC shared one correlation with other metabolites. Upon SpD treatment, the number of correlations with acetate decreased to eight metabolites, while the number of metabolites correlating with GPC increased to 11 (Figure 6A,B). Pathway enrichment analyses showed that various metabolic processes, including inositol phosphate metabolism, glycerolipid metabolism, and glutathione metabolism, were involved in the SpD treatment effects (Figure 6C and Figure S4).

**Figure 6.** Network analysis of metabolites altered by SpD (10 μM, 24 h) treatment of AC16 cells. Networks of metabolites according to their Pearson's correlation coefficients were drawn using Cytoscape program. The networks with significantly increased GPC and decreased acetate are marked as red lines. (**A**) control and (**B**) SpD-treated cell metabolites; and, (**C**) Pathway impact analysis shows the most affected metabolic pathways affected by SpD. Varying colors from yellow to red represent metabolites' significance in the data.

## *2.4. Glutathione Metabolism in AC16 Cells Was Significantly Influenced by SpD Treatment*

The integration of the metabolomic and proteomic data was carried out by Integrated Molecular Pathway Level Analysis (IMPaLa) to identify significantly influenced pathways from SpD-treated AC16 cells. The KEGG database showed that VIP > 1.0 metabolites were related to nine over-represented pathways, including glutathione metabolism, protein digestion and absorption, gap junction, and sulfur metabolism (Table 2). The related genes and metabolites are presented in Table S1. The directions of changes in the metabolite-specified proteins are listed in Table 3.

**Table 2.** Pathways determined from integration of metabolomic and proteomic data of SpD treated AC16 cells. Identified proteins and metabolites were analyzed using Integrated Molecular Pathway Level Analysis (IMPaLA) for pathway enrichment.


**Table 3.** Direction of log (FC) of metabolite-specified genes/proteins in SpD/Dox treated AC16.

