Data quality assessment
After raw data processing and structure identification using our in-house metabolite database, 367 specific metabolites in plasma and 304 in tissue samples were identified and used for the following statistical analysis. To evaluate the stability and repeatability of the analytical method, quality control (QC) samples were constructed by pooling equal amounts of extracts from all plasma or tissue samples, and then aliquoted and analyzed every 10 sample runs. Univariate analysis of all annotated metabolites in QC samples showed that 99.73% and 94.08% of all identified metabolites had relative standard deviation (RSD) values less than 30% in plasma and tissues, respectively (Figure S2A, 2D). Moreover, the principal component analysis (PCA) scores plot showed that the QC samples clustered tightly, with the majority of samples falling within mean ± 2 SD in the first principal component direction (Figure S2B-C, E-F). These results demonstrate that the metabolomic data are reliable and robust.
The metabolic disturbance was more prominent in the symptomatic stage
To investigate the dynamic metabolic changes in the progression of ALS, SOD1G93A mice at 60 and 120 days of age and age-matched WT mice were studied by metabolomic assays. These time points correspond to the pre-symptomatic and symptomatic stages, respectively. At 60 days of age, there were no obvious motor symptoms, and only mild pathological changes were noted in SOD1G93A mice [31]. At 120 days of age, all SOD1G93A mice exhibited obvious motor symptoms and muscle denervation along with pathological changes [31]. Accordingly, the metabolic profiles of 60- and 120-day-old SOD1G93A mice could fully reveal metabolic alterations under different pathological conditions of ALS. As shown in the orthogonal partial least-squares discrimination analysis (OPLS-DA) scores plot, the metabolic signatures of both WT and SOD1G93A mice at the symptomatic stage differed from those at the pre-symptomatic stage. The metabolic discrepancy between WT and SOD1G93A mice was found to be more pronounced at the symptomatic stage (120d) in plasma, spinal cord, and cortex (Fig. 1A–C), suggesting that metabolic alterations in the symptomatic stage were more pronounced and complicated.
When comparing symptomatic SOD1G93A mice with WT mice, 77 differentially expressed metabolites (DEMs) (p < 0.05, t-test; variable importance in projection (VIP) > 1) were found in the plasma (Fig. 1D). These DEMs were primarily lipids (about 66%), including the free fatty acid (FFA), phosphatidylcholine (PC), lysophosphatidylcholine (LPC), phosphatidylethanolamine (PE), and phosphatidylethanolamine (LPE) families (Fig. 1E). The rest of the DEMs were mainly amino acids (Fig. 1E). Apart from FFAs, we found that most lipids were reduced in SOD1G93A mice. In addition, amino acids, including lysine, arginine, proline, several amino acid derivatives, and myo-inositol also decreased in SOD1G93A mice (Fig. 1D). To identify the relevant metabolic pathways that contributed to the pathophysiology of ALS, these DEMs were further subjected to pathway enrichment analysis. The results show that the most significantly perturbed pathways in the symptomatic stage were linoleic acid and α-linolenic acid metabolism, cysteine and methionine metabolism, arachidonic acid metabolism, and glycerophospholipid metabolism (Fig. 1F).
In the spinal cord, 47 metabolites (p < 0.05, t-test; VIP > 1) were found to be differentially expressed in SOD1G93A mice (Fig. 2A-B). A heat map was generated to better depict the DEMs levels in WT and SOD1G93A mice (Fig. 2C). In contrast with plasma, levels of differential lipids in the spinal cord were elevated, except glycerol 3-phosphate, PC 36:2, and PC 40:6. Meanwhile, intermediate metabolites in glycolysis (glucose 6-phosphate and fructose-1,6-bisphosphate) and cis-aconitic acid were found to be increased in SOD1G93A mice (Fig. 2C). Notably, we found a global reduction in purines and purine derivatives (adenine, adenosine 5’-monophosphate (AMP), guanosine 5’-diphosphate (GDP), guanosine 3’,5’-cyclic monophosphate (cGMP), 2’-Deoxyguanosine 5’-monophosphate (dGMP), and allantoic acid) in SOD1G93A mice (Fig. 2C). The pathway analysis also revealed that purine metabolism, glycerophospholipid metabolism, and sphingolipid metabolism were the most significantly disturbed pathways in the spinal cord of SOD1G93A mice during the symptomatic stage (Fig. 2D).
In contrast with spinal cord, only 28 DEMs were screened in the motor cortex during the symptomatic stage (p < 0.05, t-test; VIP > 1, Figure S3A–C). These DEMs were mainly involved in arachidonic acid metabolism, glutathione metabolism, starch and sucrose metabolism, and inositol phosphate metabolism, among others (Figure S3D).
Metabolic disturbance in the pre-symptomatic stage
Since pre-symptomatic ALS is the best time for early intervention and treatment of the disease, we investigated whether pre-symptomatic ALS had distinct metabolic profiles. Although the metabolic alterations in the pre-symptomatic stage (60d) were less prominent than those in the symptomatic stage (120d), the partial least-squares discrimination analysis (PLS-DA) scores plot showed that there was still clear discrimination in the metabolic profile of plasma between WT and SOD1G93A mice (Fig. 3A). In total, 23 DEMs were screened (Table 1). In addition to a decrease in most lipids, such as phospholipids, sphingolipids, and acyl-carnitines, a significant reduction in purine derivatives, such as AMP, adenosine 5’-diphosphate (ADP), and dGMP, were also noted in the plasma of SOD1G93A mice.
Table 1
Differential metabolites between WT and SOD1G93A mice in plasma at the pre-symptomatic stage.
Metabolites | WT_60 | SOD1G93A_60 | p value | Metabolites | WT_60 | SOD1G93A_60 | p value |
13-Cis-acitretin | 5.614 | 4.929 | 0.045 | LPC 17:0 sn-1 | 0.009 | 0.007 | 0.004 |
dGMP | 0.052 | 0.025 | 0.037 | LPC O-18:0 | 0.012 | 0.011 | 0.035 |
N-formyl-L-methionine | 0.007 | 0.004 | 0.012 | Sphingosine-1-phosphate | 0.121 | 0.109 | 0.037 |
AMP | 0.157 | 0.099 | 0.045 | 8[R]-HETE | 0.143 | 0.104 | 0.049 |
ADP | 0.008 | 0.006 | 0.036 | PC (33:1) | 0.007 | 0.005 | 0.009 |
Carnitine C12:0 | 0.006 | 0.004 | 0.034 | PC (33:2) | 0.052 | 0.040 | 0.007 |
Carnitine C14:2 | 0.005 | 0.003 | 0.010 | PC (31:0) | 0.003 | 0.003 | 0.013 |
Carnitine C18:0 | 0.003 | 0.002 | 0.042 | PE (38:3) | 0.027 | 0.023 | 0.022 |
Cer(d18:1/24:1)-9a | 0.001 | 0.001 | 0.039 | PE (38:4) | 0.024 | 0.020 | 0.005 |
Indolelactic acid | 3.477 | 4.212 | 0.030 | SM (36:3) | 0.002 | 0.001 | 0.037 |
Leucine | 3.109 | 3.745 | 0.035 | LPE 16:0 | 0.028 | 0.031 | 0.023 |
LPC 15:0 | 0.027 | 0.022 | 0.003 | AMP | 0.157 | 0.099 | 0.045 |
dGMP = 2’-Deoxyguanosine 5’-monophosphate; AMP = Adenosine 5’-monophosphate; ADP = Adenosine 5’-diphosphate; Cer = Ceramides; LPC = Lysophosphatidylcholine; PC = Phosphatidylcholine; PE = Phosphatidylethanolamine; SM = Sphingomyelin; LPE = Lysophosphatidylethanolamine; AMP = Adenosine 5’-monophosphate; 8[R]-HETE = 8(R)-Hydroxy-(5Z,9E,11Z,14Z)-eicosatetraenoic acid. |
For tissues, a visible regional difference was observed between the spinal cord and the motor cortex. In addition, the metabolic difference between SOD1G93A mice and WT mice was more pronounced in the spinal cord than in the motor cortex (Fig. 3B). We then focused on the metabolic alterations in the spinal cord, and 20 DEMs were found to be significantly altered in SOD1G93A mice (Table 2). Significant reductions in amino acids such as arginine, ornithine, and tyrosine were noted, with the exception of S-lactoylglutathione. In addition, we found a marked alteration in purine and its derivatives (inosine, hypoxanthine, oxypurinol, AMP, ADP, and dGMP) in the spinal cord of SOD1G93A mice. Although the trends of purine derivatives (AMP, ADP, and dGMP) in the plasma and spinal cord were inconsistent, disturbances in purine metabolism were found in both the central and peripheral circulation, suggesting that purine metabolism may have been disrupted in the early stage of ALS. More importantly, altered levels of a series of purine metabolites and disturbed purine metabolism were also found in the symptomatic stage, demonstrating the significance of purine metabolism in the development and progression of ALS.
Table 2
Differential metabolites between WT and SOD1G93A mice in spinal cord at the pre-symptomatic stage.
Metabolites | WT_60 | SOD1G93A_60 | SOD1G93A/WT | p value | Metabolites | WT_60 | SOD1G93A_60 | SOD1G93A/WT | p value | |
AMP | 1.083 | 1.547 | 1.428 | 0.002 | Cysteine-glutathione disulfide | 0.008 | 0.004 | 0.578 | 0.024 | |
FFA C20:3 | 0.002 | 0.002 | 0.801 | 0.003 | Tyrosine | 0.000 | 0.000 | 0.841 | 0.020 | |
Inosine | 0.050 | 0.030 | 0.594 | 0.003 | ADP | 0.207 | 0.245 | 1.180 | 0.020 | |
dGMP | 0.005 | 0.009 | 1.775 | 0.004 | Arginine | 0.014 | 0.010 | 0.738 | 0.026 | |
Hypoxanthine | 0.318 | 0.237 | 0.748 | 0.004 | Oxypurinol | 0.000 | 0.000 | 0.689 | 0.026 | |
UMP | 0.000 | 0.000 | 1.422 | 0.005 | N-acetyl-L-histidine | 0.001 | 0.001 | 0.848 | 0.049 | |
Methionine | 0.004 | 0.003 | 0.752 | 0.007 | Phenylalanine | 0.008 | 0.007 | 0.882 | 0.026 | |
Saccharopine | 0.003 | 0.002 | 0.745 | 0.010 | S-lactoyglutathione | 0.002 | 0.003 | 1.595 | 0.035 | |
Ornithine | 0.000 | 0.000 | 0.778 | 0.011 | FFA C20:2 | 0.003 | 0.003 | 0.844 | 0.043 | |
Indoline | 0.018 | 0.016 | 0.877 | 0.018 | | | | | | |
AMP = Adenosine 5’-monophosphate; FFA= Free fatty acids; dGMP = 2’-Deoxyguanosine 5’-monophosphate; UMP = Uridine monophosphate. |
Levels of the differentially expressed purine and purine derivatives in the spinal cord at different stages are presented in Fig. 3C–N. The results show that levels of adenine and its three derivatives (AMP, ADP, and ATP) increased at the pre-symptomatic stage and decreased at the symptomatic in SOD1G93A mice (Fig. 3C–F). The same trend was observed in the alteration of four guanosine derivatives (cGMP, GDP, dGMP, and 2’-deoxyguanosine) (Fig. 3G–J). In contrast, the trend of inosine and hypoxanthine, uric acid, and allantoic acid were inconsistent with these results. Inosine and hypoxanthine were significantly reduced at the pre-symptomatic stage; however, no significant alteration was found at the symptomatic stage of ALS (Fig. 3K–L). The levels of uric acid and allantoic acid were significantly altered at the symptomatic stage. As an end-product of purine metabolism, the level of uric acid was significantly increased, whereas the level of allantoic acid, the oxidative product of uric acid, was significantly decreased in SOD1G93A mice (Fig. 3M-N).
Proteomic analysis of the spinal cord in the symptomatic stage
To further understand the regulatory network of disturbed metabolic pathways and to unveil the underlying mechanism of metabolic alterations in ALS, the spinal cord samples from 120-day-old SOD1G93A mice and WT littermates were subjected to proteomic analysis. According to the PCA analysis, the proteomes of WT and SOD1G93A mice could be clearly discriminated (Fig. 4A). In total, 297 proteins were screened as differentially expressed proteins (DEPs) in WT and SOD1G93A mice (p < 0.05, FDR < 0.1, fold change > 1.2). Of these proteins, 193 DEPs were upregulated, while 104 DEPs were downregulated (Fig. 4B and Table S5).
To interpret the biological significance of these DEPs, we performed the Gene Ontology (GO) annotation analysis, which depicts protein functions in three categories: biological processes (BP), molecular functions (MF), and cellular components (CC) (Fig. 4C). In terms of CC, these DEPs were mostly found in the cytosol or on mitochondria, ribosomes, synapses, and myelin sheaths. The analysis results of MF show that differential proteins were primarily associated with protein binding, macromolecular complex binding, and structural constituent of ribosome. The most correlated BP of the DEPs were translation, mitochondrial respiratory chain complex I assembly, fatty acid beta-oxidation, fatty acid metabolic process, etc. To further investigate the functions and signaling pathways that the DEPs were involved in, KEGG enrichment analysis were performed. In total, 93 KEGG pathways were enriched by 297 DEPs (Table S6), and the top 20 enriched pathways related to metabolism are shown as a bubble plot in Fig. 4D. These DEPs were found to be mainly associated with metabolic pathways such as oxidative phosphorylation, fatty acid metabolism, fatty acid degradation, and carbon metabolism. Moreover, some of these DEPs in the spinal cord were involved in the regulation of purine metabolism, which was found to be significantly disrupted in metabolomic analysis.
Table 3
Cluster details of MCODE protein-protein interaction and involved pathways.
Cluster | Score | Nodes | Edges | KEGG Pathways |
1 | 24.56 | 26 | 614 | Ribosome, RNA transport |
2 | 12.5 | 13 | 150 | Oxidative phosphorylation, Thermogenesis, Retrograde endocannabinoid signaling, and etc. |
3 | 7.75 | 9 | 62 | Spliceosome |
4 | 4.889 | 28 | 132 | Fatty acid metabolism, Peroxisome, Biosynthesis of unsaturated fatty acids, and etc. |
To investigate the interactions among these DEPs, the STRING database was employed to construct a protein-protein interaction (PPI) network with a medium confidence level (0.4). The established PPI network consisted of 297 nodes and 1447 edges (Table S7). MCODE method was then applied with the default parameters (degree cutoff ≥ 2, node score cutoff ≥ 0.2, K-core ≥ 2, and max. depth = 100) to identify core modules in the established PPI network, following which seven clusters were found (scores > 3 and the number of nodes > 5) (Table S8). Proteins in the first four significant clusters were subjected to KEGG pathway enrichment analysis (Fig. 5A and Table 3). Cluster 1, including 26 nodes and 614 edges, was significantly enriched in the ribosome and RNA transport. Cluster 2, including 13 nodes and 150 edges, was markedly enriched in oxidative phosphorylation, thermogenesis, and retrograde endocannabinoid signaling. Proteins in Cluster 3 were involved in the spliceosome. The most correlated categories of Cluster 4 were fatty acid metabolism, peroxisome, and biosynthesis of unsaturated fatty acids (Table 3 and Table S9).
Oxidative phosphorylation is the process by which ATP is produced, coupled with the transport of electrons driven by five protein complexes (Complexes I–V) [32]. In our study, levels of eight nicotinamide adenine dinucleotide hydride (NADH) dehydrogenase (Complex I) subunits, two cytochrome C oxidase (Complex IV) subunits, and two ATP synthase (Complex V) subunits were found to be significantly reduced (Fig. 5B), suggesting impaired oxidative phosphorylation in SOD1G93A mice in the symptomatic stage. Meanwhile, DEPs involved in multiple FFA metabolism pathways were significantly elevated, indicating enhanced synthesis and utilization of FFAs (Fig. 5C).
Joint pathway analysis of proteomics and metabolomics
Since cellular metabolic activities could be regulated directly by proteins, and the concentration of metabolites could influence protein function and expression in turn, the joint pathway analysis of proteomics and metabolomics data was performed. The results show that DEPs and DEMs in the spinal cord of SOD1G93A mice in the symptomatic stage were simultaneously involved in metabolic pathways such as purine metabolism, tricarboxylic acid (TCA) cycle, arginine, proline metabolism, and fatty acid degradation (p < 0.05, Fig. 6 and Table S10). To visualize these changes at the metabolic and proteomic levels and their associations, we mapped the most important DEMs and DEPs into metabolic pathways (Fig. 7). Many purine metabolites were found to be altered in the spinal cord of SOD1G93A mice at both pre-symptomatic and symptomatic stages. Meanwhile, abnormal expressions of proteins involved in purine metabolism (PRPS2, ENTPD2, AMPD3, and NT5C) were also observed in the symptomatic stage. Moreover, expression levels of TCA cycle-related enzymes (ACLY, DLST, SUCLA2, and MDH2) were significantly reduced, while enzymes involved in fatty acid metabolism (HADHA, HADHB, ACADM, ACAA1, ACADSB, and ACOT1) were significantly increased in the spinal cord of SOD1G93A mice.