Introduction

Amyotrophic lateral sclerosis (ALS) is a fatal, neurodegenerative disease causing the progressive loss of motor neurons, and death only 3 to 5 years after diagnosis [1]. There is no cure for ALS, and the only two drugs available to date are riluzole and, more recently, edaravone that present limited efficacy. The pathophysiology of ALS includes glutamatergic excitotoxicity, oxidative stress, mitochondrial dysfunctions, abnormal protein aggregations, and neuroinflammation [2,3,4,5]. Nevertheless, these mechanisms and their interconnectedness are still only partially understood. Thus, new “omics” strategies further integrating these patho-mechanisms and revealing undescribed targets remain of utmost importance for the development of ALS treatments. Metabolic alterations have been largely described in ALS through targeted and untargeted approaches in the body fluids and tissues of patients, as well as in animal and cellular models [6].

Patient fibroblasts, often used in the exploration of mitochondrial diseases, have proven to be of use to understand ALS etiology, but so far they have not been analyzed by metabolomics. In ALS, skin fibroblasts have shown to display numerous abnormalities found in motor neurons [7] suggesting that these apparently unaffected cells may share common pathogenic pathways with motor neurons. Moreover, human fibroblast cultures have demonstrated that they are an accessible model to study the function of mitochondrial metabolism in ALS [8]. Fibroblasts of sporadic ALS (sALS) patients display an increased, uncoupled mitochondrial respiration and glycolytic rate compared to same-age control fibroblasts [9]. Similar defects in mitochondrial respiration with increased glycolytic flux were observed in fibroblasts from patients carrying SOD1 pathogenic variants [10]. A higher mitochondrial membrane potential associated to decreased mitochondrial content was observed in ALS fibroblasts, and the correlation with age at diagnosis offers hope in discovering potential ALS biomarkers [11]. A recent study showed that fibroblasts from patients affected with sALS, primary lateral sclerosis, or familial ALS related to C9ORF72 expansion repeats, were hyper-metabolic in order to counteract the energetic impairment [7]. This highlights the relevance of mitochondrial membrane potential, respiration, glycolysis, and ATP content to the classification of patients. Mitochondrial function also depends on its structure, including membrane composition, and the cellular environment. According to the link of polyunsaturated fatty acids with inflammatory mediators, and with the breakdown of cell membranes during neuronal degeneration [12, 13], we hypothesize that specific lipid and metabolic profiles may characterize the ALS pathologic condition. Mitochondria and ER are physically closely related by the mitochondria-associated membranes (MAM) that constitute the sites of lipids, calcium, and proteins transferred from the ER towards mitochondria, with a crucial regulatory role in cell homeostasis. MAM display a key role in the mitochondrial fission, apoptosis, and ROS generation. Defects in lipid metabolism observed at the MAM contact sites might be involved in pathophysiology of various conditions such as neurodegenerative or metabolic diseases [14], especially in ALS [15]. Since endoplasmic reticulum (ER) stress has been observed in familial and sporadic ALS [16, 17], the association between mitochondrial and ER defects in ALS led us to specifically explore this subcellular fraction, using a quick extraction protocol for metabolomics and lipidomics analyses [18]. Indeed, recent studies reported mitochondrial lipidome [19,20,21,22], but the long procedures used to isolate mitochondria are incompatible with metabolomics analyses due to the risk of metabolite degradation.

Here, we report metabolomic and lipidomic explorations of fibroblasts from ALS patients, both on whole cells and their mitochondrial-endoplasmic reticulum (mito-ER) fraction, and correlate these data to the functional assessment of oxidative phosphorylation (OXPHOS) and complete mitochondrial DNA sequencing, in order to unravel the metabolic disturbances through a multi-omics approach and to specify the alterations of the mitochondria-ER fraction.

Materials and Methods

Patients

Skin fibroblasts were obtained from 13 sporadic ALS patients diagnosed and followed in two French ALS reference centers: Angers and Limoges University Hospitals. ALS patients satisfied the diagnostic criteria of “probable” or “definite” symptoms as defined by the El Escorial workshop [23]. Information on diagnosis, gender, current age, site of onset, diagnostic delay, and age at onset were obtained for each patient. We also collected the parameters of disease progression, such as the revised ALS functional rating scale (ALSFRS-r), the forced vital capacity (FVC), and the body mass index (BMI) at the time of diagnosis.

Control fibroblasts from 11 gender- and age-matched healthy subjects were selected in the Biological Resource Center (Centre de Ressource Biologique CRB) from the hospital of Angers. All the participants in this current study provided their informed consent for the use of their fibroblasts for research. The ethic committees of the Centre for Human Research of Angers and Limoges Hospitals approved the study and the consent process (no 2016/36, 2016-A01425-46).

Fibroblast Culture

Human fibroblasts were cultured in a medium consisting of two-third Dulbecco’s modified Eagle medium with nutrient mixture F12 (DMEM-F12, PAN Biotech), one-third AmnioMAX™ (Gibco, Invitrogen) supplemented with 10% fetal bovine serum (PAN Biotech) at 37 °C and 5% CO2. To avoid artifacts due to senescence, all experiments were conducted on fibroblast cultures between the 6th and the 15th passages [24].

A schematic flowchart of experiments is presented in Fig. 1.

Fig. 1
figure 1

Flowchart summarizing the global methodology

Full size image

OXPHOS Exploration

The mitochondrial respiration rates were measured using a two-channel, high-resolution oxygraph respirometer (Oroboros, Innsbruck, Austria) in a minimum of three replicates. Cells were grown to 70–90% confluency and approximatively three million cells in DMEM-F12 were used for analysis. Firstly, the basal respiration was measured (i.e., routine oxygen consumption rate (OCR)). Secondly, the non-phosphorylating respiration (proton leak) was evaluated by adding an ATP synthase inhibitor, oligomycin (4 μg/mL) (Oligomycin OCR). Then, maximal capacity of cellular respiration was assessed by adding successive doses of FCCP (p-trifluoromethoxy-carbonyl-cyanide-phenyl hydrazine). Finally, we used an inhibitor of mitochondrial complex III, antimycin (2 μg/mL) to evaluate the non-mitochondrial oxygen consumption. The respiration capacity rate (RCR) was calculated as the ratio between the maximal capacity of respiration and the non-phosphorylating respiration. The phosphorylating RCR was obtained by the subtraction of non-phosphorylating respiration to basal respiration divided by the maximal capacity of respiration. We measured the concentration of proteins of each sample to normalize oxygen uptake between samples.

Metabolomic and Lipidomic Analyses

Whole Cell Extraction

Samples were extracted as previously described [25]. After medium removal, the cellular monolayer was rinsed twice with an aqueous solution containing 0.22% NaCl before being quenched with cold methanol for metabolomic analysis and with a cold mixture of ammonium bicarbonate at 155 mM and methanol (67.7/32.3) for lipidomics. The cell suspension (4 million cells) obtained after mechanical scraping was then collected in aliquots of one million cells and stored at −80 °C until analysis.

Mitochondria and Endoplasmic Reticulum (Mito-ER) Extraction

The mitochondria plus endoplasmic reticulum (mito-ER) fraction was extracted as previously described [18]. Briefly, cell lysis of pellets containing 10 million cells was achieved in a Precellys24® homogenizer (Bertin Instruments) using 1 cycle of grinding 5 s at 2500 rpm at 4 °C. Then differential centrifugations were performed for mito-ER purification.

Metabolomics and Lipidomics

Metabolomics and lipidomics protocols and analytical systems have been detailed and recently published [18, 25]. HRMS system was based on UPLC Ultimate 3000 system (Dionex) coupled to a Q-Exactive Mass Spectrometer (ThermoFisher Scientific, Bremen, Germany) using electrospray ionization (positive and negative Electrospray Ionization (ESI+ and ESI-)). Quality controls (QC) were analyzed regularly on the plate (every 8 samples) to ensure the stability of the mass system over time. Accordingly, peaks that had high variability (coefficient of variation > 30%) were excluded from further analysis. Metabolites were identified at level 1 and lipids at level 2 according to the level of compound identification process [26].

Mitochondrial DNA Sequencing

Total DNA was extracted from patient fibroblasts using the Nucleospin® tissue Kit (Macherey-Nagel®) and the EZ1 DNA extraction automate (Qiagen) according to the manufacturer’s recommendations. Whole mitochondrial genome (mtDNA) was sequenced using Ion Proton high-throughput sequencing platform and analyzed as previously described [27]. Variants were filtered using the following criteria: at least three callers predicting the variant, strand bias between 0.25–0.75, and a minimum mtDNA heteroplasmy level of 2%. Putative variants were visualized on Integrative Genome Viewer [28] to avoid artifacts. Variant prioritization relies on databases, i.e., Mitomap [29], HmtDB [30], Mamit [31], and prediction tools, i.e., MitImpact [32] and MitoTip [33]. Variant heteroplasmy was considered as low under 30%, medium between 30% and 70% and high between 70% and 95%, and above 95% variants were considered as homoplasmic. Mitochondrial haplogroups were determined using Phy-Mer [34]. Structural mtDNA rearrangements were analyzed trough a home-made tool based on soft-clipping.

Statistical Analysis

We performed both multivariate and univariate analyses using Metaboanalyst, version 3.0 (http://www.metaboanalyst.ca) for metabolomics and lipidomics data. Classification method was based on Partial Least-Squares Discriminant Analysis (PLS-DA). Values of variable importance in projection (VIP) represent the importance of the compound (metabolite/lipid) for the PLS-DA models. The score plots show the classified samples and the loadings characterize the relation between the Y and X variables (metabolites, lipids). Metaboanalyst modeling includes leave-one-out cross-validation (LOOCV). We retained the 15 most discriminant compounds based on the VIP to explore pathways. To attest the relevance of these selected compounds, the quality of the model built with these selected discriminant metabolites was assessed by prediction accuracy and permutation test (set permutation n = 100). The performance measures of the permutated data usually form a normal distribution and if the performance score of the original data lies outside the distribution, then the results are considered significant.

The univariate analysis of metabolites and lipids levels was based on the fold-change values and the threshold of significance after non-parametric Wilcoxon test, using Benjamini and Hochberg correction for multiple tests.

Non-parametric Wilcoxon test was used to compare clinical data, oxygen consumption rate, and mtDNA sequencing data between ALS and controls.

Pathway Analysis

We used KEGG pathway database (http://www.genome.jp/kegg/pathway.html) on Metaboanalyst tool to discuss the highlighted metabolic pathways. To focus on the most relevant data, we highlighted only the pathways that had significant Holm p and a pathway impact > 0.05. The pathway impact value is the cumulative percentage from the matched metabolite nodes.

Results

Patients

No statistical difference was found regarding age and gender between the 13 ALS and 11 control subjects. The mean age at onset of ALS was 60.5 ± 9.9 years, and percentage of females was 38.5% (Table 1). About 77% of ALS patients had a spinal onset. None of these patients carried mutations or expansions in SOD1, TARDBP, FUS, or C9ORF72 genes. Characteristics of ALS patients at diagnosis defined by BMI, FVC, and ALSFRS-r score were as follows: 22.8 ± 3.2 kg/m2, 81.6 ± 18.9%, and 38.2 ± 4.5, respectively. There was no significant difference in mean passage number between ALS and control fibroblast cultures used for analyses (data not shown).

Table 1 Clinical characteristics of ALS patients and controls. SD = standard deviation
Full size table

Mitochondrial Exploration

OXPHOS Analysis

The OXPHOS analysis revealed no differences in basal and maximal respiration between ALS and control fibroblasts. We did not detect any difference in respiratory capacity rate (RCR) or in phosphorylating RCR (Table 2).

Table 2 Exploration of oxidative phosphorylation of ALS and control fibroblasts. SD = standard deviation
Full size table

mtDNA Sequencing

mtDNA from 12 ALS and 11 control fibroblasts was sequenced, while one ALS sample could not be processed. We did not identify any large-scale rearrangement in patients or controls. Considering all samples together, 505 punctual variants were found, among which 124 were unrelated to the mitochondrial haplogroups (combinations of polymorphisms retracing human evolution), including 66 heteroplasmic variants. Among the non-haplogroup variants, 32 (25.8%) occurred in the D-Loop, the non-coding region controlling replication and transcription of mtDNA, 69 (55.6%) were located in polypeptide-coding regions, 10 (8.1%) in tRNA genes, and 13 (10.5%) in rRNA genes. Figure 2 shows the repartitioning of these variants in ALS and control fibroblasts. Twenty-eight (42.4%) of the heteroplasmic variants were recognized as non-pathogenic variants, nine (13.6%) were associated with clinical phenotypes, and 29 (44%) were not reported in all mitochondrial databases. The level of heteroplasmy was considered to be low for 61 (63.5%) variants and medium for five of them (15.4%).

Fig. 2
figure 2

Circos representing mtDNA sequence (black: D-loop; gray: tRNA genes; purple: rRNA genes; green: complex I genes; yellow: complex III genes; blue: complex IV genes; orange: complex V genes). The pie chart in the center of each circos (ALS and controls) shows the repartition of heteroplasmic variants on the mtDNA

Full size image

Statistical comparison between the ALS and control groups did not disclose difference between the total number of variants, the number of non-haplogroup variants, or the number of variants in the different mtDNA genes (Table 2). Of note, the mean level of heteroplasmic variants per ALS patient was 3.4, while it was 2.3 per control, but this difference was not significant. Interestingly, only ALS patients had mtDNA variants with medium levels of heteroplasmy among the non-haplogroup variants (p = 0.0179) (Table 3). We did not observe any difference in the number of pathogenic variants between ALS and controls. Reported pathogenic variants found in the ALS and control fibroblasts are reported in supplementary Table S1.

Table 3 Summary of mtDNA variants found in ALS and control fibroblasts. *p value < 0.05 was considered as significant
Full size table

Metabolomic and Lipidomic Signatures of ALS Fibroblasts

We detected and identified 124 and 127 metabolites and 328 and 220 complex lipids in fibroblasts and mito-ER extracts, respectively (Supplementary Tables S2 and S3).

The PLS-DA model from metabolomic data on total fibroblasts revealed an excellent discrimination between ALS patients and controls (75% accuracy, p value of permutation test: 0.01) (Fig. 3a). The 15 most discriminant metabolites were attributed with VIP scores higher than 1 and are presented in Fig. 3b. Mito-ER metabolomics correctly distinguished ALS patients and controls (accuracy: 62.5%, p value of permutation test: 0.1) (Fig. 3d). VIP scores of the 15 most discriminant metabolites were higher than 1.5, and 3 of them were significant after univariate analysis: inosine, guanosine, and guanine (p < 0.03 and FC > 1.5) (Fig. 3e).

Fig. 3
figure 3

Metabolomic data analyses of whole cells and mito-ER fraction. A. Optimized PLS-DA 2D-score plot from metabolomic whole cells data (red triangle: ALS samples; green cross: control samples) showing discrimination between the two groups. B. List of the top 15 VIP (variable importance in projection) from non-optimized PLS-DA model from metabolomic whole cells data. C. Results of pathway analysis including the top 15 VIP metabolites of whole cells PLS-DA model (using MetaboAnalyst module). D. Optimized PLS-DA 2D-score plot from metabolomic mito-ER data (red triangle: ALS samples; green cross: control samples) showing the discrimination between the two groups. E. List of the top 15 VIP (variable importance in projection) from non-optimized PLS-DA model from metabolomic mito-ER data (red stars show metabolites significant in univariate analysis with p < 0.05 and fold change > 1.2). F. Results of pathway analysis including top 15 VIP metabolites of mito-ER PLS-DA model (using MetaboAnalyst module)

Full size image

Whole cell and mito-ER lipidomics also displayed correct discriminations between ALS and control subjects (64% and 50% accuracies, p values of permutation tests 0.17 and 0.75, respectively). In these two non-optimized models, the 15 most discriminant lipids had VIP scores higher than two (Fig. 4b and d). Optimized PLS-DA models using only the 15 top VIP metabolites improved the discriminations between ALS and controls in both whole cells and mito-ER lipidomic data, with 80% and 83% accuracy, and 0.03 and 0.08 for p values of permutation tests in whole cells and mito-ER extracts, respectively (Fig. 4a and c). Univariate analysis did not reveal significant variation in lipid concentrations.

Fig. 4
figure 4

Lipidomic data analyses of whole cells and mito-ER fraction. A. Optimized PLS-DA 2D-score plot from lipidomic whole cells data (red triangle: ALS samples; green cross: control samples) showing discrimination between the two groups. B. List of the top 15 VIP (variable importance in projection) from non-optimized PLS-DA model from lipidomic whole cells data. C. Optimized PLS-DA 2D-score plot from lipidomic mito-ER data (red triangle: ALS samples; green cross: control samples) showing discrimination between the two groups. D. List of the top 15 VIP (variable importance in projection) from non-optimized PLS-DA model from lipidomic mito-ER data

Full size image

Metabolomic Findings

Using the “pathway analysis” module of the MetaboAnalyst software, we explored the pathways involved in the metabolomic signatures of ALS whole fibroblasts and mito-ER extracts. The main pathways involving the 15 most discriminant metabolites identified with robustness in ALS fibroblasts were arginine and proline metabolism, pyrimidine, and purine metabolism (Fig. 3c). The pathway analysis of the metabolomic signature of ALS mito-ER notably underscored the involvement of purine and pyrimidine metabolism, as well (Fig. 3f).

The global analysis of metabolic disturbances based on fibroblasts and their mito-ER fraction disclosed modifications of purine (adenine and guanine) and pyrimidine bases (cytosine, uracil, and thymine), and various nucleosides and their derivatives (inosine, guanosine, cytidine, deoxyadenosine, deoxyguanosine, uridine-5′-monophosphate, and cytidine-5′-diphosphocholine). We also observed a decrease in xanthine in the mito-ER fraction.

Metabolomic signatures also highlighted several compounds involved in energy metabolism. Indeed, we noticed variations in glucose-6-phosphate levels in ALS whole fibroblasts and in lactate levels in the mito-ER fraction. This was associated with a modification of the phosphocreatine levels in both ALS fibroblasts and mito-ER fractions.

Lipidomic Findings

We found a reduction in several ceramide concentrations in the ALS mito-ER fraction and a gain in sphingolipid (SM) levels in ALS whole cells (Fig. 4). Lipidomic profiles of ALS fibroblasts highlighted modifications in phosphatidylcholine (PC) and phosphatidylethanolamine (PE) levels, both in whole cell and mito-ER extracts (Fig. 4). Interestingly, we observed several plasmalogen species among the VIP metabolites of both models—PE plasmalogen (PEp) in the whole cells with PE(18:1p/22:6) and PE(18:0p/22:6); and PE(18:0p/22:6) and PC(36:4p) in the mito-ER extract.

Discussion

This study provides the first report of a multi-omics profiling on sporadic ALS fibroblasts, both at the whole cell and mito-ER subcellular levels, complemented by OXPHOS measurement and mtDNA complete sequencing.

Multiple Implications of the Purine and Pyrimidine Metabolisms in Oxidative Stress and Inflammation in ALS

The alterations of purine and pyrimidine metabolisms recognized in this study are convergent with data from a previous study that we performed on rodent motor neurons and astrocytes [35], placing these pathways at the core of oxidative stress and inflammatory mechanisms observed in ALS [36]. Purine catabolism leads to the production of xanthine, which is converted to uric acid by xanthine oxidase, finally producing reactive oxygen species (ROS) [37]. This is consistent with the increase in uric acid and ROS production found in ALS patients; we also found that xanthine levels were lower in the mito-ER samples compared to the controls, possibly signifying an augmented depletion due to uric acid production. We did not highlight significant differences in uric acid levels in our study (p = 0.6635), but we had a small cohort reducing statistical power. Uric acid has numerous roles in cell homeostasis, and its function in oxidative stress, especially increased in ALS, may explain this finding. Indeed, uric acid concentrations increase in response to oxidative stress and act as a scavenger of singlet oxygen and hydroxyl radicals [38]. Therefore, uric acid acts as either an antioxidant (especially in plasma) or a pro-oxidant in cells [39], suggesting a role in oxidative stress as a primary phenomenon or as a neuroprotective effect [40]. Interestingly, multiple studies have displayed support for the latter. It has been shown that uric acid modulates neuroinflammation by decreasing the serum levels of IL-1β, IL-6, and TNF-α in Parkinson’s disease murine models [41]. Furthermore, primary astrocyte-conditioned medium provides a neuroprotective effect against both H2O2-induced and SOD1G93A astrocyte-induced oxidative stress in NSC-34 cells, when the normal or mutant astrocytes are first treated with uric acid [42].

Disturbances in Energy Metabolism in ALS

An imbalance in ATP production and consumption could explain an increased rate of purine catabolism with consequent modifications in purine end-product levels, as xanthine and inosine observed here. A recent study showed that multiple sclerosis is characterized by perturbations in energy metabolism, leading to significant changes in the serum profiles of compounds directly (purines) or indirectly (creatinine, lactate, and pyrimidines) connected to ATP homeostasis [43]. Our findings are consistent with the results of this study, since ALS also involves metabolic imbalances. In this respect, we also noticed modified amounts of phosphocreatine, which mediates ADP phosphorylation and could be a response to energetic impairment. It is also known to stabilize mitochondrial membranes by binding phospholipids [44]. In addition, phosphocreatine has neuroprotective effects against oxidative stress and mitochondrial dysfunction [45]. As in cancer cells, significant levels of aerobic glycolysis, the so-called Warburg Effect, have been documented in ALS due to the over-activation of the WNT/β-catenin pathway [46]. This is supported by the observation of modifications in the concentration of glucose-6-phosphate, a glycolytic intermediate, related with variations in lactate levels. All these findings support the alteration of the energy metabolism in ALS, associated with oxidative imbalance and inflammation. As explained by Lu et al., the reported concentrations of cellular ATP are notoriously inconsistent because of incomplete quenching and incomplete extraction for mass spectrometry method. Enzyme assays reveal that ATP concentration is more reflective of overall purine abundance than cellular energetics [47]. Thus, these results also lead us to program further explorations of energy metabolism through the measurement of glycolytic and mitochondrial production of ATP.

Involvement of Phospholipids and Sphingomyelins in Neuroprotective Mechanisms

Lipidomic results further suggest perturbations in the oxidative mechanism and inflammation. Indeed, higher concentrations of PCs observed in ALS fibroblasts may stimulate the metabolic activity of phospholipase A2 (PLA2), resulting in an increased production of lipid mediators, such as eicosanoids that promote inflammation and are considered to play a role in ALS pathophysiology [48, 49].

Three plasmalogen species, corresponding to phospholipids (PL) having an ether bond between fatty acids and glycerol, were discriminants between ALS and control fibroblasts: (PE(18:1p/22:6), PE(18:0p/22:6), PC(36:4p)). Plasmalogens are important components of myelin, subcellular membranes as the mitochondrial membranes, and the membranes of secretory granules. The main functions of plasmalogens are the generation of lipid second messengers and the conservation of membrane integrity. Plasmalogens, which terminate lipid oxidation and protect ROS-vulnerable myelin, are considered to act as endogenous antioxidants [50, 51], which, in the context of ALS, can prevent the oxidation of polyunsaturated fatty acids and other lipids.

Our results further disclosed higher levels of sphingomyelins (SM) in ALS fibroblasts, associated with lower ceramide (Cer) levels in mito-ER fractions. Classically, ceramides are synthesized using serine in the ER then are converted to SM. These results are consistent with the lipidomic signature of the cerebrospinal fluid (CSF) from ALS patients [52], with the higher levels of ceramides and glucosylceramides found in the spinal cord of ALS patients and in ALS mouse model [53], and with the fact that ceramides have been associated with apoptosis in response to cytotoxic humoral factors [54]. Indeed, the accumulation of ceramide-derived agents may be protective by reducing ceramide synthesis and increasing the entry of ceramides into the CerG1 pathway, thus limiting the direct toxic effect on motor neurons [53]. Higher ceramide and downstream glycosphingolipid levels have also been reported in the muscle of an ALS mouse model, as well as in other mice after muscle injury [55]. Taken together, these findings clearly show a perturbation of the glycosphingolipid metabolism in ALS fibroblasts. These results open perspectives of deeply studying ER function in ALS.

The identification of the fatty acid chain 22:6 in the discriminant phospholipids, i.e., (PE(18:0p/22:6), PE(18:1p/22:6), PG(18:1/22:6)), corresponds to docosahexaenoic acid (DHA), an omega-3 polyunsaturated fatty acid (n-3 PUFA) that is involved in the maintenance of neuronal membranes, signal transduction, neuronal differentiation, neurogenesis, and protection against synaptic loss and spinal cord injury [56]. Moreover, some authors have highlighted the significant increase in DHA in the frontal cortex of ALS patients [17]. This phenomenon may be considered as a potentially defensive response of the nervous tissue to brain ischemia and Alzheimer’s disease [57], or as an inflammatory process via the production of n-3-derived anti-inflammatory resolvins and docosatrienes by the brain cortex [58]. We also recently reported discriminant phospholipids containing DHA in the lipidomic signature of ALS cerebrospinal fluids [52].

Mitochondrial Instability as a Protective Mechanism?

The assessment of OXPHOS capacities revealed that there was no difference in cellular respiration between ALS and control fibroblasts. MtDNA sequences of ALS fibroblasts were similar to those of controls in terms of total number of variants per sample, frequency of mtDNA variants, type of mtDNA variants, and patterns of distribution. Nevertheless, we found that ALS cells exhibited higher levels of heteroplasmy, which may attest to a higher mtDNA instability. We could note that reported pathogenic variants found in ALS fibroblasts are associated with Parkinson’s disease, dystonia and myopathy, hypertrophic cardiomyopathy, and Leber Hereditary Optic Neuropathy (LHON) mitochondrial diseases. Despite the absence of obvious mtDNA alteration in ALS fibroblasts, mitochondrial dysfunction observed so far in ALS could be related to epigenetic mtDNA modifications, as already suggested in Alzheimer’s disease [59]. Intriguingly, it was demonstrated that mitochondrial heteroplasmy alters mtDNA methylation [60]. Moreover, a recent study in blood mtDNA of ALS patients, especially from SOD1 mutation carriers, showed a significant decrease in D-loop methylation accompanied by an increase in mtDNA copy number, the latter representing a mitochondrial compensation for metabolic deregulation caused by oxidative stress in ALS [61]. However, we did not evidence difference in the mtDNA copy number in ALS fibroblasts compared to controls (data not shown). It neither seems likely that the inverse relationship between mtDNA copy number and D-loop methylation in the blood specifically points to ALS; rather, it could point generally to a metabolic imbalance.

Conclusion

Provided that the discriminant metabolites found in this study are also available in the blood and/or CSF, a collective reading of purine, pyrimidine, arginine, proline metabolites, and DHA-containing lipids could reflects a specific ALS signature. These results in agreement with the ALS research literature, in different affected tissues and body fluids (motor neurons, CSF, blood etc.…), reinforce the interest of fibroblast cultures in the study of ALS. These findings reveal a robust methodology to apply omics, even in subcellular organelles, with data integration with promise to deeply understand ALS pathophysiology. Thus, this approach may be applied to other complex and multifactorial neurological diseases with high confidence.