Abstract
X-linked adrenomyeloneuropathy (AMN) is an inherited neurometabolic disorder caused by malfunction of the ABCD1 gene, characterized by slowly progressing spastic paraplegia affecting corticospinal tracts, and adrenal insufficiency. AMN is the most common phenotypic manifestation of adrenoleukodystrophy (X-ALD). In some cases, an inflammatory cerebral demyelination occurs associated to poor prognosis in cerebral AMN (cAMN). Though ABCD1 codes for a peroxisomal transporter of very long-chain fatty acids, the molecular mechanisms that govern disease onset and progression, or its transformation to a cerebral, inflammatory demyelinating form, remain largely unknown. Here we used an integrated -omics approach to identify novel biomarkers and altered network dynamic characteristic of, and possibly driving, the disease. We combined an untargeted metabolome assay of plasma and peripheral blood mononuclear cells (PBMC) of AMN patients, which used liquid chromatography coupled to quadrupole-time-of-flight mass spectrometry (LC-Q-TOF), with a functional genomics analysis of spinal cords of Abcd1− mouse. The results uncovered altered nodes in lipid-driven proinflammatory cascades, such as glycosphingolipid and glycerophospholipid synthesis, governed by the β-1,4-galactosyltransferase (B4GALT6), the phospholipase 2γ (PLA2G4C) and the choline/ethanolamine phosphotransferase (CEPT1) enzymes. Confirmatory investigations revealed a non-classic, inflammatory profile, consisting on the one hand of raised plasma levels of several eicosanoids derived from arachidonic acid through PLA2G4C activity, together with also the proinflammatory cytokines IL6, IL8, MCP-1 and tumor necrosis factor-α. In contrast, we detected a more protective, Th2-shifted response in PBMC. Thus, our findings illustrate a previously unreported connection between ABCD1 dysfunction, glyco- and glycerolipid-driven inflammatory signaling and a fine-tuned inflammatory response underlying a disease considered non-inflammatory.
Introduction
With an incidence of one in 17 000 in newborns, X-linked adrenoleukodystrophy (X-ALD, OMIM number 300100) is the most common monogenic leukodystrophy and peroxisomal disorder. X-ALD is characterized by central inflammatory demyelination in the brain and/or slowly progressing spastic paraparesis resulting in axonal degeneration in the spinal cord (1–4). X-ALD is caused by mutations in the ABCD1 gene (Xq28), which encodes the ATP-binding cassette transporter, an integral peroxisomal membrane protein involved in the import of very long-chain fatty acids (C≥22:0) and very long-chain fatty acids-CoA esters into the peroxisome for degradation (5,6). The defective function of the ABCD1 transporter leads to very long-chain fatty acids accumulation and impaired β-oxidation of very long-chain fatty acids in organs and tissues (6–8), particularly hexacosanoic acid (C26:0), the pathognomonic disease marker.
Three major disease variants have been described. One is a late-onset form affecting adults, which is known as adrenomyeloneuropathy (AMN). Patients present with peripheral neuropathy and distal axonopathy involving corticospinal tracts of spinal cord—but without signs of overt brain neuroinflammation or major demyelination—and spastic paraparesis as main symptoms. This form can evolve ultimately into a lethal form, with cerebral demyelination and neuroinflammation in adults, cAMN. The childhood cerebral form cALD has a similar outcome (1–3).
For cALD patients, the only treatment to date is allogeneic bone marrow transplantation, which is associated with a high morbidity and mortality and is available only to nearly asymptomatic X-ALD children (9). Very recently, a gene therapy approach correcting CD34+ cells with the ABCD1 cDNA, using a lentiviral vector, has proved to be successful and, being less invasive, is a good alternative to transplantation (10,11).
In contrast, for AMN patients, there is no satisfactory treatment to date (12,13). Studies on pathomechanisms underlying disease posit that the excess of C26:0 disturbs mitochondrial oxidative phosphorylation (OXPHOS) function and elicits mitochondrial ROS. This interferes with mitochondrial biogenesis and mitochondrial calcium signaling (14–17), which underline a cross-talk between mitochondria and peroxisomes and pinpoint a secondary mitochondrial involvement as culprit in this disease (18). The identification of targets in these pathways, such as PGC-1α, Sirt1, or mTOR, which can be targeted with drugs, has led to successful preclinical tests (17,19–22) in the Abcd1− mouse model (17,19–22) [a model for AMN (23)], which warrant clinical trials.
Although the entire clinical spectrum of X-ALD is initiated by mutations in a single gene, the ABCD1, the pathomechanisms for demyelination, the inflammatory process, the axonal degeneration and the adrenal insufficiency clearly differ. Therefore, additional pathogenic factors critically shaping the clinical manifestation of X-ALD /X-AMN ought to exist (24). Some of the molecular pathology present in nervous tissue can be reproduced in fibroblasts or other cells by incubation with excess of C26:0, including oxidative stress or dysfunction of proteasomal and mitochondrial compartments. Despite this finding, it is not clear how a single defect in peroxisomal fatty acid metabolism could elicit such a varied neurologic phenotype. Therefore, it is proposed that other metabolic functions depending directly or indirectly on the ABCD1 gene function may operate to modulate disease onset and severity.
To further investigate the consequences of ABDC1 loss, we applied an integrated biology approach that comprised the use of functional genomics in mouse spinal cords and untargeted metabolomic and lipidomic analyses of patients' blood samples—to characterize AMN-associated molecular profiles (Fig. 1). These techniques have been previously used to discover new biomarkers for disease diagnosis and prognosis and monitor therapeutic interventions (25–27). The inferred AMN signature was experimentally validated using several techniques: (i) a quantitative mass spectrometry approach for measuring lipid mediators of inflammation and lipid peroxidation markers; (ii) a Q-PCR array for assessing the expression of inflammatory cytokines and their cognate receptors in peripheral blood mononuclear cells (PBMC); (iii) a targeted Q-PCR analysis for gene expression of key dysregulated metabolic nodes in PBMC and (iv) an immunoassay using MILLIPLEX™ technology for quantifying cytokines and adipokines in plasma. The results provide a novel concept of disturbed glycolipid and glycerolipid signaling linked to complex inflammatory networks as potential disease modifying nodes in AMN, offering novel opportunities for biomarker identification and development of tailored therapeutics.
A procedural diagram that indicates the process followed to obtain the signature in AMN patients.
Results
We set out to identify a molecular signature for AMN by conducting a metabolomic/lipidomic analysis on PBMC and plasma obtained from patients and controls. The results were combined with transcriptomic data from spinal cords from the Abcd1− mouse model at different stages of disease progression, using an integrated bioinformatic analysis as depicted in Figure 1. We pinpointed several dysregulated key pathways that were subsequently experimentally validated by independent, complementary techniques.
Metabolomic and lipidomic analysis in plasma and PBMC from AMN patients
Plasma and PBMC from AMN patients and healthy, gender and age-matched controls were collected. Lipidomic and metabolomic analysis using mass spectrometry was used to characterize their molecular profile. As there is no single universal method for metabolite extraction, two independent protocols were used to evaluate a wide range of molecules, from polar (metabolome) to apolar (lipidome) molecules. Thus, to characterize AMN-associated changes in plasma and PBMC, we performed non-targeted metabolomic and lipidomic analyses using a LC-Q-TOF system (Figs 2 and 3, respectively).
Metabolomic and lipidomic profiles in plasma from AMN patients (n = 13 AMN and 13 healthy individuals). Heat map representation of hierarchical clustering of molecular features found in each plasma sample from control and AMN subjects in metabolomic (A) and lipidomic (B) studies. Each line in this graphic represents an accurate mass ordered by retention time and colored according to its abundance intensity normalized to the internal standard with baselining to the median/mean across the samples. The scale from blue to red represents this normalized abundance in arbitrary units. Tridimensional PCA and PLS-DA graphs demonstrating a differential plasma profile for control and AMN subjects in metabolomic (C) and lipidomic (D) studies. Blue spots represent controls, and red spots represent AMN subjects. X, Principal component 1; Y, Principal component 2; Z, Principal component 3. Positive and negative ionization refers to the settings of the mass spectrometer used for the analysis of the compounds.
Metabolomic and lipidomic profiles in PBMC from AMN patients (n = 13 AMN and 13 healthy individuals). Heat map representation of hierarchical clustering of molecular features found in each PBMC sample from control and AMN subjects in metabolomic (A) and lipidomic (B) studies. Each line in this graphic represents an accurate mass ordered by retention time and colored according to its abundance intensity normalized to the internal standard with baselining to the median/mean across the samples. The scale from blue to red represents this normalized abundance in arbitrary units. Tridimensional PCA and PLS-DA graphs demonstrating a differential PBMC profile in control and AMN subjects in metabolomic (C) and lipidomic (D) studies. Blue spots represent controls, and red spots represent AMN subjects. X, Principal component 1; Y, Principal component 2; Z, Principal component 3.
For plasma, using those molecular features present in at least 50% of the samples within the same group (3008 and 5579 from metabolomic and lipidomic analyses, respectively), we performed a clustering analysis (Fig. 2A and B). This approach demonstrated that AMN was the main factor determining both the metabolomic and lipidomic profiles of plasma. To further explore the metabolic differences between control and AMN individuals, a multivariate statistical analysis was employed, including PCA (an unsupervised technique) and PLS-DA (a supervised technique) (Fig. 2C and D) (28). Both techniques uncovered AMN-specific plasma metabolomic and lipidomic signatures. The statistical analysis of plasma samples revealed that 348 molecules including metabolites and lipid species were significantly different between the genotypes, and some of these molecules were identified (Supplementary Material, Table S1).
Pathway analyses combining the molecules with a putative identity revealed several targets and included lipid-driven inflammation-associated pathways such as ceramide degradation and sphingomyelin metabolism (Supplementary Material, Table S2).
Metabolites involved in the bile acid biosynthesis pathway (cholic acid, glycocholic acid and 5-β-cholestane-3α,7α,12α-triol) were significantly different between AMN and control subjects, as found as well in the PBMC analyses. We also found diminished levels of three free fatty acids (n-hexanoic, eicosapentaenoic and hexadecanedioic acids) in the AMN group (Supplementary Material, Table S1). Finally, succinic semialdehyde, an intermediate in the catabolism of γ-aminobutyrate that is implicated in neurotransmission (29), was upregulated in samples from AMN patients (Supplementary Material, Table S2).
Applying a similar approach to that used for analyzing plasma, we observed that the PBMC metabolomic profile (Fig. 3A and C) discriminated better between groups than did the lipidomic profile (Fig. 3B and D). The statistical analyses revealed that 793 molecules were significantly different between the groups. Supplementary Material, Table S3 lists those molecules with a putative identification. Notably, higher histamine concentrations were present in the AMN PBMC samples than in control samples, and hypoxanthine, the product of xanthine oxidase, was also present at an increased level. Supporting the suggestion that there is disturbed bile acid metabolism in AMN, 5β-cholestane-3α,7α,12α,26-tetrol levels were significantly higher in the PBMC samples of AMN patients than in control samples. Intriguingly, glycolipids such as lactosylceramide (LacCer), and most of the glycerophosphatidylethanolamine pathway metabolites (phosphatidylserine, phosphatidylethanolamine and CDP-ethanolamine), and triglyceride species were increased in the AMN patients (Supplementary Material, Table S3). Pathway analyses of the molecules with a putative identity using the Consensus-Path platform revealed several nodes which included proinflammatory cascades driven by bioactive lipids pathways such as ceramide degradation, sphingomyelin metabolism and eicosanoid biosynthesis including metabolites from the leukotriene, prostaglandin and thromboxane subfamilies (Supplementary Material, Tables S2 and S4). Collectively, the vast majority of all identified metabolites can be associated with inflammation and/or redox homeostasis (30,31).
Integrated analysis of ‘-omics’ data in X-ALD
We next integrated the metabolomic and lipidomic results with transcriptome data from the spinal cords of Abcd1− mice (32), with the aim of uncovering core molecular footprints of the disease across different species and cell types, which could be of relevance for disease pathogenesis. In Figure 4, we show the common dysregulated pathways between the transcriptomes of X-ALD mouse spinal cords at 3.5, 12 and 22 months and the metabolomes of AMN patients, in PBMC (Fig. 4A) and plasma (Fig. 4B). Three shared dysregulated pathways stand out: (i) the metabolism of lipids and lipoproteins, (ii) signaling by GPCR (G-protein-coupled receptors) and (iii) sphingolipid metabolism. Pathways (i) and (ii) are also dysregulated in plasma. The number of dysregulated pathways is shown in Figure 4C and D.
Integrated functional enrichment analysis of Abcd1− mice and AMN patients. Common dysregulated pathways in metabolomic data from (A) PBMC and (B) plasma from AMN patients with at least one transcriptomic data condition (at 3.5, 12 or 22 months) from Abcd1− spinal cords (SC) at 3.5, 12 and 22 months of age. This corresponds to 18 of 47 dysregulated pathways in PBMC (A) and 6 of 14 in plasma (B). Significant pathways were identified by using the hypergeometric distribution function. Black spots represent dysregulated pathways with P < 0.05 according to the hypergeometric test. A venn diagram showing the number of significant pathways and the overlap between spinal cords (SC) from Abcd1− mice at 3.5, 12 and 22 months and the metabolomic data from (C) PBMC and (D) plasma.
For precise visualization of the metabolic reactions, we used a metabolic map from Kegg (http://www.genome.jp/kegg-bin/show_pathway?map01100v), to build our own AMN metabolome map (Fig. 5). This integrated analysis approach yielded several nodes of disturbance in which both the enzyme's expression (from Abcd1− mouse spinal cords) and their reaction products or substrates (from AMN patient's plasma or PBMC) were trending in the same direction. Four areas of concerted dysregulation of enzyme-metabolite pairs are highlighted in Figure 5. The first is the synthesis of lactosylceramide (LacCer) from glucosylceramide via β-1,4-galactosyltransferase (B4GALT6). Expression of this enzyme was raised in the transcriptomic analysis and also in Q-PCR validatory assays in spinal cords (Fig. 6A). The second node highlights a dysregulation of glycerophospholipid metabolism, with CDP-ethanolamine as a substrate, phosphatidylethanolamine as a product and the ethanolamine phosphotransferase CEPT1 as the catalytic enzyme. Furthermore, phosphatidylethanolamine is converted to phosphatidylserine by phosphatidylserine synthase. CEPT1 also catalyzes the final step in the synthesis of phosphatidylcholine by transferring phosphocholine from CDP-choline to diacylglycerol. The resultant phosphatidylcholine is metabolized to arachidonic acid by the calcium-independent, cytosolic phospholipase 2γ, cPLA2γ (PLA2G4C). This is an enzyme family that hydrolyzes glycerophospholipids to produce free fatty acids and lysophospholipids, both of which serve as precursors in the production of signaling molecules and second messengers of inflammatory processes such as eicosanoids and diacylglycerol (33). Both the Pla2g4c and Cept1 transcripts were raised in the transcriptomic analysis of spinal cords and were confirmed in the validatory Q-PCR analysis of Abcd1− spinal cords that ensued (Fig. 6A), already very early in life, at 3 months of age (Fig. 6A). The third area of concerted dysregulation is the formation of retinal (retinaldehyde) by retinol dehydrogenase 11 (RDH11), on the pathway of retinoic acid biosynthesis from retinol (vitamin A). This increase in retinal is correlated to increased expression of the Rdh11 transcript levels in the transcriptome of Abcd1− spinal cords; and in the validatory Q-PCR analysis of AMN patients's mononuclear cells (Fig. 6B) and in the spinal cords of Abcd1− mice (Fig. 6A). RDH11 is a membrane-bound enzyme induced by sterol-regulatory element binding proteins (SREBPs), and proposed to reduce in addition to retinal, other toxic fatty aldehydes such as the oxidation product 4-hydroxynonenal (4-HNE) (34), which has been found in human X-ALD samples (35). RDH11 malfunction causes a human syndrome involving brain and retina development (36), which highlights the importance of the enzyme. The fourth node puts the accent on the biosynthesis of bile acids, with lower levels of cholic and glycocholic acids, which are final products of peroxisomal β-oxidation. This is concordant with dysregulated expression of the peroxisomal bifunctional protein and the racemase AMACR enzymes in the Abcd1− mouse spinal cords. Upstream in the same pathway, we find an accumulation of the 5-β-cholestane-3α,7α,12α-triol and 5β-cholestane-3α,7α,12α,26-tetrol, cholesterol intermediates in the bile synthesis pathway. The enzyme that catalyzes the oxidation step from that of 5-β-cholestane-3α,7α,12α-triol to 5β-cholestane-3α,7α,12α,26-tetrol is the sterol 27 hydroxylase (CYP27A1), a mitochondrial member of the cytochrome P450 superfamily. CYP27A1 transcript levels were increased in PBMC (Fig. 6B), as predicted in the transcriptome analysis of spinal cords. Notably, the inactivation of this enzyme is responsible for cerebrotendinous xanthomatosis (OMIM 213700), a rare autosomal recessive lipid storage disease associated with central demyelination, ataxia and spastic paraplegia (37). In addition, we observed that several enzymes of fatty acid metabolism (synthesis and oxidation) were dysregulated in the transcriptome analysis, including acetyl-CoA acyltransferase 2 (mitochondrial 3-oxoacyl-Coenzyme A thiolase) (Acaa2); acetyl-CoA carboxylase alpha (Acaca); acyl-CoA dehydrogenase; short/branched chain (Acadsb); hydroxyacyl-CoA dehydrogenase (Hadh) and the hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase (trifunctional protein), alpha subunit (Hadha). These enzymes are shown on the map as black lines and dots (Fig. 5). We believe that the concomitant alteration in the levels of several fatty acids, such as hexanoic and myristic acids, as identified by metabolomics, is also of relevance although they are not displayed in the original Kegg map and therefore could not be depicted in Figure 5.
An AMN metabolomic signature derived from the integrated ‘omics’ analysis of Abcd1− mice and AMN patients. (A) A metabolic map (KEGG:01100) displaying genes and metabolites as lines and spots, respectively. Dysregulated genes in Abcd1− mice at least two of the three ages (3.5, 12 and 22 months) are indicated as black lines, and the metabolites dysregulated in AMN patients in plasma or PBMC are indicated as red spots. The four areas with more marked dysregulation of concerted metabolite-enzyme routes are displayed. (B) Reactions highlighted in (A) in more detail. B4GALT6: UDP-Gal:betaGlcNAc beta 1,4- galactosyltransferase, polypeptide 6; PLA2G4C: phospholipase A2, group IVC (cytosolic, calcium-independent); Cept1: choline/ethanolamine phosphotransferase 1; RDH11: Retinol dehydrogenase 11, AMACR: alpha-methylacyl-CoA racemase; and HSD17B4: hydroxysteroid (17-beta) dehydrogenase 4 (peroxisomal bifunctional enzyme). As the identity of 5-beta-cholestane-3-alpha-7-alpha-12-alpha-26-tetrol was established on the basis of exact mass, it cannot be differentiated from other isobaric bile acid synthesis intermediates.
Eicosanoids, oxidized polyunsaturated fatty acids and adipokine levels in body fluids from AMN patients (n = 13 AMN and 13 healthy age- and sex-matched individuals). (A) B4galt6, Pla2g4c, Cept1, Rdh11, Cyp27a1, Pparα, Pparβ/δ, Pparγ and Gpx4 gene expression in Abcd1− spinal cords at 3.5 and 12 months of age. Gene expression was normalized to the reference control gene mouse Rpl0. (B) B4GALT6, PLA2G4C, CEPT1, RDH11, CYP27A1, PPARα, PPARβ/δ, PPARγ and GPX4 gene expression in PBMC from AMN patients and healthy controls. Gene expression was normalized to the reference control gene human RPL0. (C) Relative levels of the inflammation-associated lipids arachidonic acid (AA), docosahexaenoic acid (DHA), prostaglandins D2, E2 and F2α (PGD2, PGE2, PGF2α), 6-keto-PGF1α, (±) 9-hydroxy-10E, 12Z octadecadienoic acid (9S-HODE), (±) 13(S)-hydroxy-9Z, 11E octadecadienoic acid (13S-HODE), (±)12-, and 15-hydroxy-5Z, 8Z, 11Z, 13E-eicosatetraenoic acid (12S-HETE and 15S-HETE) and thromboxane B2 (TXB2). (C) The levels of HGF (hepatocyte growth factor), IL6, IL8, MCP-1 (monocyte chemoattractant protein-1 or CCL2), NGF (nerve growth factor), TNFα, leptin, adiponectin, total PAI-1 (plasminogen activator inhibitor-1) and resistin were quantified in serum from AMN patients and controls by using Milliplex technology. The values are means ± SEM. Significant differences have been determined by one-tail Student's t-test (*P < 0.05 and **P < 0.01) or Wilcoxon rank sum test (#P < 0.05, ##P < 0.01 and ###P < 0.001) according to Shapiro–Wilk normality test.
Inflammation and oxidative stress in blood from AMN patients
To validate our hypothesis derived from the integrative analysis of –omics data, pointing to dysregulated glycosphingolipid and glycerophospholipid metabolism, which govern inflammatory cascades, we used several complementary approaches. These were (i) a quantitative mass spectrometry panel to assess lipid mediators of inflammation and lipid peroxidation markers, mostly derivatives of arachidonic acid, in plasma (Biocrates); (ii) a Q-PCR array to measure the expression of inflammatory cytokines and their cognate receptors in PBMC and (iii) an immunoassay using MILLIPLEX™ technology to identify adipokines and cytokines in plasma. The Biocrates MS/MS analyses pinpointed that three inflammatory products of arachidonic acid metabolites increased in AMN samples: the eicosanoids thromboxane B2, and 12- and 15- hydroxyeicosatetraenoic acids (TXB2, 12S-HETE and 15S-HETE) (Fig. 6C). As the products of arachidonic acid metabolism such as leukotriene B4, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, 4-hydroxy-2-nonenal (4-HNE), 15S-HETE and 9/13-HODEs are potent ligands of peroxisome proliferator-activated receptor PPARα, PPARβ/δ and PPARγ, we analyzed the expression of these master regulators of lipid metabolism and inflammation (38) by Q-PCR. Our results show that PPARβ/δ but not PPARα or PPARγ were upregulated in PBMC from AMN patients (Fig. 6B) and PPARγ was lowered in Abcd1− spinal cords at 12 months of age (Fig. 6A).
We next set out to investigate correlative levels of cytokines that would underscore the inflammatory process suggested by the ‘omics’ integrative analyses. Using Milliplex technology, we observed raised levels of some inflammatory cytokines [hepatocyte growth factor (HGF), IL6, IL8, MCP-1 and tumor necrosis factor-α (TNFα)] and lower levels of adiponectin in plasma from AMN patients (Fig. 6D). This is suggestive of a manifest proinflammatory profile in AMN plasma. The latter adipokine is anti-inflammatory hormone secreted by the adipose tissue, of newly uncovered relevance in neurodegeneration (39), which we also found decreased in the plasma of the Abcd1− mouse model (32). Next, we examined the expression of several cytokines and their receptors in PBMC from AMN patients. We used a cytokine array and showed that several molecules involved in T helper 1 (Th1) (proinflammatory) and/or Th2 (more protective) polarization were increased in PBMC from AMN patients (Fig. 7A). We complemented the inflammatory profile data by using targeted Q-PCR to assess the expression of members of the suppressor of cytokine signaling (SOCS) and signal transducer and activator of transcription (STAT) families (STAT1, STAT6 and SOCS3) (Fig. 7B) (40,41). In addition to controlling the expression of inflammatory cytokines, the SOCS and STAT family genes also control the polarization of Th cells toward a Th1, Th2 or Th17 phenotype, thus playing a central role in adaptive immune responses with autocrine and paracrine immunomodulatory capacities (42–45). Intriguingly, in PBMC from AMN patients, we found that STAT1, which drives a proinflammatory Th1 differentiation response by immune cells, was upregulated, whereas SOCS3 and STAT6, which are involved in Th2 maturation, together with IL4, also had increased levels (Fig. 7B). Cytokine profile expression of PBMC showed activation of inflammation via the IL36 pathway (IL36A, IL36B and IL36G), instead of the classical IL1/TNFα/IL6 pathway. Furthermore, we also detected upregulation of the IL9/IL9R as well as IL4, IL5/IL5R, IL10 and IL13, which are Th2 markers (Fig. 7A). In conclusion, the cytokine gene profile in AMN patients was not strongly directed toward a Th1, Th17 or Th2 response but was suggestive of a more generalized inflammatory imbalance.
Inflammatory cytokines, chemokines and receptors related signaling in PBMC from AMN patients (n = 13 AMN and 13 healthy age- and sex-matched individuals). (A) Gene expression of 84 genes of the Inflammatory Cytokines and Receptors Signaling Pathway RT2 Profiler Q-PCR Array (Qiagen). Gene expression was normalized to internal controls. (B) IL4, IL6, STAT6, SOCS3, and STAT1 gene expression in PBMC from AMN patients and healthy controls. Gene expression was normalized to the reference control human RPL0. Genes have been classified according to their roles in Th2 and Th1/Th17 polarization. The values represent mean ± SEM. Significant differences have been determined by one-tail Student's t-test (*P < 0.05 and **P < 0.01) or Wilcoxon rank sum test (#P < 0.05 and ##P < 0.01) according to Shapiro–Wilk normality test.
Discussion
In this study, we used integrated systems biology analysis as a tool for uncovering novel dysregulated pathways underlying molecular pathology and of candidate biomarker metabolites and molecules for the monogenic disease AMN. This study showed that ABCD1 dysfunction impacts on the levels of metabolites from several peroxisomal pathways (as precursors or final products) involved in very long-chain fatty acids, bile acid or purine metabolism (46). Thus, decreased availability of substrates for peroxisomal β-oxidation, owing to deficient transport of the very long-chain fatty acids substrate by ABCD1 (6), can contribute to decreased levels of n-hexanoic acid, one of the final products of very long-chain fatty acids β-oxidation, which is then degraded in mitochondria (47). Furthermore, it has been reported that the structurally related hexadecanedioic acid, a dicarboxylic fatty acid, accumulates in the blood in peroxisomal disorder patients but not in X-ALD patients, indicating that degradation of dicarboxylic acids is taking place inside peroxisomes independently of ABCD1 (48). Intriguingly, we observed in this study that the levels of hexadecanedioic acid were lower in X-ALD patients than in control individuals, which could be due to decreased synthesis through ω-oxidation or to increased degradation in peroxisomes via ABCD3.
Among the lipidic markers disclosed in PBMC, we could pinpoint molecules with putative identities as methyl substituted oleate and linoleate. Recent data indicate the negative association of other methyl substituted fatty acids in plasma, such as methyl palmitate and methyl stearate, with healthy diet indexes (49). Similarly, methyl palmitate has been found in serum as a biomarker of metabolism derangement in obesity (50). Interestingly, this molecule is considered as a signaling intermediate with potent vascular tone regulation properties (51). Several other methyl-derived fatty acids are part of the volatiles present in organic compartments (52). It is tempting to speculate that the role of these molecules as immune modulators could contribute to the differences observed in the inflammatory profiles of PBMC from AMN patients. Noteworthy, the function of other monomethyl branched fatty acids, playing key roles in post-embryonic development in Caenorhabditis elegans, is intimately related to peroxisomal function, where peroxisomes could be involved in breaking down these fatty acids (53). Again, it may be also proposed that accumulation of these methyl derivatives in PBMC, besides a potential pathogenic role, could indicate that the ABCD1 transporter is involved in the degradation of these compounds, or that ABCD1-dependent peroxisomal functions are compromised in AMN. Thus, these compounds may have potential as biomarkers for AMN.
Moreover, we identified a new pathway alteration in AMN, associated with bile acid biosynthesis. Indeed, cytosolic bile acid precursors (5β-cholestane-3α,7α,12α-triol in plasma and 5β-cholestane-3α,7α,12α,26-tetrol in PBMC) accumulated, whereas the final products of this route generated by peroxisomal β-oxidation, cholic and glycocholic acid were lower in X-ALD patients than in control individuals. This latter finding is in agreement with previous data showing that very long-chain acyl-CoA synthetase activity–crucial in the first steps of bile acid peroxisomal β-oxidation (54)—is impaired in X-ALD patients (55). It is tempting to propose that the ensuing accumulation of the cytotoxic C27-bile acid synthesis intermediates (56) could contribute to disturbed mitochondrial function and redox homeostasis. Both noxious factors have been described in the Abcd1− mice nervous tissue, in patient's brain or in the fibroblasts, blood cells and/or plasma of X-ALD patients (14–19,24,57–60). The markedly raised levels of the purine hypoxanthine that we found in PBMC (Supplementary Material, Table S5) may also be a net contributor to the increased oxidative damage, because hypoxanthine has well-known pro-oxidant effects (61,62). Oxidative damage plays a main contributing factor to the pathogenesis of this disease, as it is directly caused by C26:0 excess, appearing early in the disease course (58), and its neutralization leads to disease arrest (21).
The integration of transcriptome data from the spinal cords of Abcd1− mice with the metabolomic data obtained from peripheral tissues of AMN patients has been instrumental in revealing several dysregulated pathways shared between mice and humans, with concerted dysregulation of metabolites and gene expression as represented in Figure 5. This underlines the notion that a very specific disturbance in the peroxisomal degradation of a particular type of fatty acids, those with very long-chains, exerts a major impact on the metabolism of apparently unrelated bioactive lipids such as the glycosphingolipids and the glycerophospholipids, of relevance for neurological disease (63). Indeed, the increased levels of several phosphatidylserine and phosphatidylethanolamine species, the main components of cellular membranes (64), may suggest a disturbed lipid composition and structural defects in the myelin sheaths in this disease and could subserve as peripheral proxies for central nervous system lipid dearrangement. Of note, we only detected the very long or ultra-long species of phosphatidylethanolamine and phosphatidylserine being increased. This may be a reflection of the excess of C26:0, which could outcompete the more abundant long-chain PUFAs and MUFAs which occupy the lateral chains of these glycerophospholipid species in physiological conditions. It is well known that the length of the glycerophospholipid acyl chain and the degree of saturation are important determinants of physical properties such as membrane microviscosity and have been found altered in X-ALD erythrocytes and adrenocortical cells (65,66). In particular, C26:0 has been predicted to exert disruptive effects on membrane structure and functions by interfering with the lateral chains of phospholipids (67). This may be a main reason why excess of C26:0 is noxious to mitochondria OXPHOS (16), to the point of interfering with the assembly of OXPHOS supercomplexes in a murine model of Zellweger syndrome (which lacks peroxisomes thus accumulate C26:0) (68). Far beyond the structural consequences of C26:0 accumulation, we should keep in mind that neural membrane phospholipids serve as reservoir for second messengers such as the eicosanoids prostaglandins and thromboxanes, which are derived from arachidonic acid after the breakage of phosphatidylserine and phosphatidylethanolamine by the phospholipase cPLA2γ. The levels of this enzyme are elevated in patients PBMC and mouse spinal cords (Fig. 6A–C), very early in life at 3 months of age, when first signs of lipo-oxidative damage (MDAL) appear (58). We therefore posit that a derangement in the bioactive glycerophospholipid composition may be the initial driver of an inflammatory signaling cascade, to which mitochondria ROS production may decisively contribute and possibly exacerbate (16,69). In a similar manner, and in addition to their structural roles in the regulation of structural properties of membrane bilayers, glycosphingolipids have emerged as crucial players governing signal transduction pathways implicated in several diseases (70,71). Indeed, the increased levels of the lactosylceramide (LacCer) we observed in PBMC may be of direct relevance to neuroinflammation. A recent seminal report revealed that both LacCer levels and B4GALT6 expression are raised in multiple sclerosis lesions (72). The authors found that LacCer was excreted by astrocytes and induced the activation of microglia and recruitment of CNS-infiltrating monocytes by regulating the production of the chemokines CCL2 and GM-CSF (72). Also, LacCer is itself a producer of free radicals (73) and can directly activate phospholipase cPLA2γ (74), thus fueling the generation of the proinflammatory arachidonic acid and its derivatives thromboxane A2 (TBX2), 12S-HETE, 15S-HETE as shown in Figure 6C. Notably, PLA2G4C can be directly activated by TNFα and other proinflammatory cytokines through NFκB signaling (75), which underlines the intertwining of glycerolipid and glycosphingolipid signaling with general inflammatory cascades.
Accordingly, we observed higher levels of cytokines and chemokines, such as IL6, IL8, TNFα, MCP-1 and CCL11, in plasma from AMN patients compared with healthy controls (Fig. 6D). The results from this study agree with earlier reports investigating affected and non-affected brain areas in patients with cerebral inflammatory disease (cALD), in which IL6, TNFα, MCP-1 and CCL11 levels were higher than in controls (24,35,76–78). Moreover, the IL8 and MCP-1 levels were found to be elevated in the CSF from cerebral childhood adrenoleukodystrophy patients (79), stressing the similarities between discordant disease phenotypes. Intriguingly, IL6, IL8 and MCP-1 are cytokines belonging to the senescence-associated secretory phenotype (SASP) (80) and accumulate in the body fluids of patients with other neurodegenerative disorders in which chronic inflammation is of relevance, such as multiple sclerosis, Alzheimer's (AD) and Parkinson's diseases (81–83).
In contrast, the cytokine expression profile obtained from PBMC in AMN does not reflect this proinflammatory scenario (Fig. 7), perhaps explaining why pure AMN patients do not exhibit cerebral inflammatory demyelination. We detected increased levels of IL10, IL36RN and IL37 mRNA expression as well as increased IL4, IL5 and IL13 mRNA expression (Fig. 7A), all of which are Th2 cytokines mostly known to play a role in limiting inflammatory responses (84–86). Therefore, this scenario suggests that a negative feedback mechanism for counterbalancing the proinflammatory phenomena occurring either in the central nervous system or in the periphery (such as in myeloid cells, endothelial cells, liver, adipose or other tissues as ABCD1 is rather ubiquitously expressed) is in place. An example of counter regulation is the expression of mRNA encoding members of the IL1 and interferon (IFN) families. The absence of changes in IL1, the best characterized member of the family, may indicate that the mechanism underlying the upregulation of these cytokines is different from the classical inflammasome-mediated increase in IL1 expression. This is highlighted by the absence of changes of the proinflammatory IL6, IL8, MCP-1 and TNFα in PBMC, in contrast to plasma (Figs 6D and 7B). Inflammatory cytokines, toll-like receptor activation and other stresses have been described to induce the expression of these cytokines. The upregulation of IFNA2 and STAT1 mRNA, genes known to be promoted by toll-like receptor and inflammatory stimuli, could facilitate the response to IFN. However, the enhanced SOCS3 levels may limit the responses to these cytokines (Fig. 7B). A further example involves the elevated production of the inflammatory cytokine IL36 by innate immune cells and lymphocytes, which may be alleviated by the concomitant expression of IL-36RN and/or IL37 (87,88). To illustrate the complexity of the scenario, some cytokines may exert dual roles promoting or counteracting inflammation. For instance, it has been reported that IL10, initially described as a Th2 cytokine able to inhibit IFNγ production by Th1 cells (89), is also produced by monocytes, dendritic cells and other lymphocyte subsets in response to inflammatory insults. Intriguingly, IL13, recently proposed to play a neuroprotective role (90), is implicated in allergy and anti-helminths immunity. IL9 is a mediator of Th17-driven inflammatory disease, and its inactivation attenuates the development and progression of experimental autoimmune encephalomyelitis, a mouse model of multiple sclerosis (91). Our findings suggest that a fine balance between pro- and anti-inflammatory molecules may be the key to determining disease prognosis and progression toward an overt cerebral inflammatory phenotype. Dysfunctional balancing is likely to be very sensitive to genetic and environmental/ stochastic factors, which could explain the large observed variation in disease onset, progression, and severity spanning the broad phenotypic expression in X-ALD. Altogether, we provide compelling evidence to shift the paradigm of AMN being considered as a non-inflammatory disorder to that of a disease state in which specific, lipid-mediated proinflammatory cascades are relevant and under tight control. Moreover, our data unravel several key regulators of glycolipid- and glycerolipid-mediated inflammation as therapeutic targets, thus suggesting that therapies aimed at blocking cPLA2γ, B4GALTN6, IL36A, IFNA2 or IL9R could arrest disease progression to more severe phenotypes. The study demonstrates the value of using integrated systems analysis in the study of human disease.
Materials and Methods
Participants and ethics
Samples from pure AMN male patients (i.e. those with an absence of gadolinium enhancement in T1/T2-weighted images in 1.5T MRIs and no clinical progression in the last 12 months) and healthy male volunteers were collected after an overnight fast (n = 13, age range 24–51 years for AMN and n = 13, age range 23–49 years for controls). Patients were considered as pure AMN when no opening of blood–brain barrier as determined with gadolinium enhancement was observed, at the time of taking the samples. Blood was processed by centrifugation within 2 h of collection using a gradient of Histopaque to separate plasma, erythrocytes and PBMC. Plasma and PBMC were stored at −80°C. The use of all samples was approved by the Clinical Research Ethics Committee of the Bellvitge University Hospital (AC130/10). Informed written consent was obtained from all patients and control individuals.
QTOF-based metabolomic and lipidomic analysis
Polar metabolites from plasma and PBMC were extracted as previously described (92). Briefly, three volumes of cold methanol were added to 30 µl of plasma or 20 µl of PBMC lysate (containing 90 µg of protein), which was incubated for 1 h at −20°C and then centrifuged for 3 min at 12 000 g. The supernatant was recovered, evaporated using a Speed Vac (Thermo Fisher Scientific, Barcelona, Spain) and resuspended in aqueous 0.4% acetic acid and 2 ng/ml of 9-anthracene carboxylic acid in methanol as an internal standard (1:1, v/v) (93). Total lipids from plasma or cell lysates were extracted with chloroform:methanol (2:1, v/v) in the presence of 0.01% butylated hydroxytoluene to avoid artefactual oxidation, as previously described (94) and, after drying, were resuspended in chloroform:methanol (1:3).
Polar metabolite and lipid extracts were analyzed using a HPLC 1290 series coupled to an ESI-Q-TOF MS/MS 6520 (Agilent Technologies, Santa Clara, CA, USA). For the LC/MS metabolomics method, 2 µl of sample extract was applied onto a reversed phase column (Zorbax SB-Aq 1.8 µm 2.1 × 50 mm; Agilent Technologies) equipped with a precolumn (Zorba-SB-C8 Rapid Resolution Cartridge 2.1 × 30 mm 3.5 µm) with a column temperature of 60°C. The flow rate was 0 6 ml/min. Solvent A was composed of water containing 0.2% acetic acid, and solvent B was composed of methanol containing 0.2% acetic acid. The gradient started at 2% B and increased to 98% B over 13 min and was held at 98% B for 6 min. Posttime was established as 5 min, as previously described (93). The LC/MS lipidomic analysis was based on a previously described method (95). Two microliters of lipid extract was injected onto an XBridge BEH C18 shield column (100 mm L × 2.1 mm ID × 1.7 µm; Waters, Milford, MA, USA) kept at 80°C. The mobile phases, delivered at 0 5 ml/min, consisted of ammonium formate (20 mm at pH5) (A) and methanol (B). The gradient started at 50% B and reached 70% B in 14 min and was followed by a slow gradient of 70–90% B over 50 min and an isocratic separation at 90% B for 15 min. The mobile phase B subsequently reached 100% over 5 min and was maintained for an additional 5 min. This method allows the orthogonal characterization [based on exact mass (<10 ppm) and on retention time] of lipids. When combined with deuterium or C13-labeled internal standards, this strategy is useful for attributing potential identities with low uncertainty (95).
Custom cDNA microarray for Abcd1− mice
The cDNA microarray used in this study was manufactured at the IGBMC (Institut de Génétique et de Biologie Moléculaire et Cellulaire) as reported earlier (32). The generation and genotyping of Abcd1− mice have been previously described (23,96). The microarray experiment was deposited in the Array Express Database under accession number E-MTAB-79. Mice used for microarray experiments were on a pure C57BL/6J background. Total RNA for microarray analysis was extracted using the RNeasy Kit (Qiagen), from n = 4–5 mice per genotype and condition. All samples were hybridized in duplicate and analyzed using a dye swapping design. All methods employed in this study are in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publications No. 85–23, revised 1996).
Pathway analysis
Metabolomic and lipidomic data were scaled using an auto-scaling algorithm and hierarchical clustering and principal component analyses (PCA). Partial least discriminant data (PLS-DA) for metabolomics and lipidomics were obtained by using Mass Profiler Professional software (Agilent Technologies). The number of components chosen for PLS-DA was 4. Validation of the model was achieved in silico with an N-fold validation type with 3-folds and 10 repeats as validation parameters.
To evaluate which pathways or functional categories were enriched in differentially expressed genes and metabolites, we computed a hypergeometric distribution function using Bioconductor packages in the R programming environment. For metabolomic-derived pathways, the ConsensusPathDB-human platform (97) integrating interaction networks in Homo sapiens metabolome was used. Briefly, this platform collates pathways from several public databases of protein interactions and signaling and metabolic pathways, as well as gene regulation, in humans. In this study, we used several databases to reduce bias by enhancing coverage: KEGG, Reactome, Netpath, Biocarta, HumanCyc and the pathway interaction database (PID), Signalink, Inoh, Wikipathways, Pharmgkb, Humancyc and Ehmn. Only pathways showing two metabolites or more in the over-representation analyses and a P-value cutoff of <0.01 were taken into account.
Eicosanoids and oxidized polyunsaturated fatty acids levels
For an independent confirmation of the above described, non-targeted approach, eicosanoids and other oxidized polyunsaturated fatty acids were extracted from plasma samples with aqueous acetonitrile that contained deuterated internal standards. The metabolites were determined by HPLC-tandem mass spectrometry (LC-MS/MS) with multiple reaction monitoring (MRM) in negative mode using a SCIEX API 4000 QTrap mass spectrometer with electrospray ionization. The LC-MS/MS method used for the analytical determination of eicosanoids has been published (98). The plasma concentrations of 9S-HODE, 13S-HODE, 14(15)-EpETE, 12S-HETE, 15S-HETE, 15S-HpETE, LTB4, 5S-HpETE, TXB2, LTD4, DHA, PGE2, 8-iso PGF2α, PGF2α, 6-keto-PGF1α, PGD2 and AA were measured by a triple quadrupole mass spectrometry-based metabolite quantification assay (Biocrates Life Science AG) (Table 1).
Abbreviations of eicosanoids and oxidized polyunsaturated fatty acids
| Abbreviation | Name |
|---|---|
| 9S-HODE | (±)9-hydroxy-10E,12Z-octadecadienoic acid |
| 13S-HODE | 13(S)-hydroxy-9Z,11E-octadecadienoic acid |
| 14(15)-EpETE | (±)14(15)-epoxy-5Z,8Z,11Z,17Z-eicosatetraenoic acid |
| 12S-HETE | 12(S)-hydroxy-5Z,8Z,10E,14Z-eicosatetraenoic acid |
| 15S-HETE | 15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| 15S-HpETE | 15(S)-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| LTB4 | Leukotriene B4 |
| 5S-HpETE | 5(S)-hydroperoxy-6E,8Z,11Z,14Z-eicosatetraenoic acid |
| TXB2 | Tromboxane B2 |
| LTD4 | Leukotriene D4 |
| DHA | Docosahexaenoic acid |
| PGE2 | Prostaglandin E2 |
| 8-iso PGF2a | 8-iso-Prostaglandin F2alpha |
| PGF2a | Prostaglandin F2alpha |
| 6-keto-PGF1a | 6-keto-Prostaglandin F1alpha |
| PGD2 | Prostaglandin D2 |
| AA | Arachidonic acid |
| Abbreviation | Name |
|---|---|
| 9S-HODE | (±)9-hydroxy-10E,12Z-octadecadienoic acid |
| 13S-HODE | 13(S)-hydroxy-9Z,11E-octadecadienoic acid |
| 14(15)-EpETE | (±)14(15)-epoxy-5Z,8Z,11Z,17Z-eicosatetraenoic acid |
| 12S-HETE | 12(S)-hydroxy-5Z,8Z,10E,14Z-eicosatetraenoic acid |
| 15S-HETE | 15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| 15S-HpETE | 15(S)-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| LTB4 | Leukotriene B4 |
| 5S-HpETE | 5(S)-hydroperoxy-6E,8Z,11Z,14Z-eicosatetraenoic acid |
| TXB2 | Tromboxane B2 |
| LTD4 | Leukotriene D4 |
| DHA | Docosahexaenoic acid |
| PGE2 | Prostaglandin E2 |
| 8-iso PGF2a | 8-iso-Prostaglandin F2alpha |
| PGF2a | Prostaglandin F2alpha |
| 6-keto-PGF1a | 6-keto-Prostaglandin F1alpha |
| PGD2 | Prostaglandin D2 |
| AA | Arachidonic acid |
Abbreviations of eicosanoids and oxidized polyunsaturated fatty acids
| Abbreviation | Name |
|---|---|
| 9S-HODE | (±)9-hydroxy-10E,12Z-octadecadienoic acid |
| 13S-HODE | 13(S)-hydroxy-9Z,11E-octadecadienoic acid |
| 14(15)-EpETE | (±)14(15)-epoxy-5Z,8Z,11Z,17Z-eicosatetraenoic acid |
| 12S-HETE | 12(S)-hydroxy-5Z,8Z,10E,14Z-eicosatetraenoic acid |
| 15S-HETE | 15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| 15S-HpETE | 15(S)-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| LTB4 | Leukotriene B4 |
| 5S-HpETE | 5(S)-hydroperoxy-6E,8Z,11Z,14Z-eicosatetraenoic acid |
| TXB2 | Tromboxane B2 |
| LTD4 | Leukotriene D4 |
| DHA | Docosahexaenoic acid |
| PGE2 | Prostaglandin E2 |
| 8-iso PGF2a | 8-iso-Prostaglandin F2alpha |
| PGF2a | Prostaglandin F2alpha |
| 6-keto-PGF1a | 6-keto-Prostaglandin F1alpha |
| PGD2 | Prostaglandin D2 |
| AA | Arachidonic acid |
| Abbreviation | Name |
|---|---|
| 9S-HODE | (±)9-hydroxy-10E,12Z-octadecadienoic acid |
| 13S-HODE | 13(S)-hydroxy-9Z,11E-octadecadienoic acid |
| 14(15)-EpETE | (±)14(15)-epoxy-5Z,8Z,11Z,17Z-eicosatetraenoic acid |
| 12S-HETE | 12(S)-hydroxy-5Z,8Z,10E,14Z-eicosatetraenoic acid |
| 15S-HETE | 15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| 15S-HpETE | 15(S)-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid |
| LTB4 | Leukotriene B4 |
| 5S-HpETE | 5(S)-hydroperoxy-6E,8Z,11Z,14Z-eicosatetraenoic acid |
| TXB2 | Tromboxane B2 |
| LTD4 | Leukotriene D4 |
| DHA | Docosahexaenoic acid |
| PGE2 | Prostaglandin E2 |
| 8-iso PGF2a | 8-iso-Prostaglandin F2alpha |
| PGF2a | Prostaglandin F2alpha |
| 6-keto-PGF1a | 6-keto-Prostaglandin F1alpha |
| PGD2 | Prostaglandin D2 |
| AA | Arachidonic acid |
Quantitative RT-PCR
To offer an orthogonal approach to the integrated biology (-omic) methods, total RNA from PBMC was extracted using an RNeasy Kit (Qiagen) according to the manufacturer's instructions. One microgram of RNA was transcribed into cDNA using Superscript II reverse transcription reagents in a final volume of 25 μl (Invitrogen). Gene expression was measured using 0.1–0.2 μl of cDNA. TaqMan real-time PCR was performed with an ABI PRISM 7300HT sequence detection system using the TaqMan® Universal PCR master mix and standardized primers for mouse B4GALT6 (Hs00191135_m1), CEPT1 (Hs01035232_m1), CYP27A1 (Hs01017992_g1), GPX4 (Hs00989768_g1), IL4 (Hs00174122_m1), IL6 (Hs00174131_m1), PLA2G4C (Hs01003743_m1), PPARα, (Hs00947539_m1), PPARβ/δ (Hs00606407_m1), PPARγ, (Hs01115513_m1), RDH11 (Hs00211283_m1), SOCS3 (Hs01000485_g1), STAT1 (Hs01013996_m1) and STAT6 (Hs00598625_m1). Expression of the genes of interest was normalized to that of the reference control, human RPL0 (Hs99999902_m1). Each sample was run in duplicate, and the mean value of the duplicates was used to calculate the mRNA expression using the comparative (2−ΔCt) method, according to the manufacturer's instructions (17).
Q-PCR array for inflammatory cytokines and receptors
Similarly, to study the pathways suggested by the omic measurements, the Human Inflammatory Cytokines and Receptors RT² Profiler Q-PCR Array (PAHS-011, Qiagen) contains gene-specific primer sets for 84 relevant genes encoding inflammatory cytokines and receptors, 5 housekeeping genes and 2 negative controls. After performing thermal cycling, real-time amplification data were analyzed by using ABI Prism 7300HT software. Gene expression was normalized to internal controls (housekeeping genes: β2-microglobulin, hypoxanthine phosphoribosyltransferase 1, ribosomal protein L13a, glyceraldehyde-3-phosphate dehydrogenase, β-actin) to determine the fold change in gene expression between healthy controls and AMN patients.
Analyses of inflammation mediators
All adipokine levels in plasma were determined by immunoassays using the LINCOplex kit (Luminex xMAP Technology). Human serum adipokine panel A (HADK1–61K-A) was used for adiponectin, resistin and total plasminogen activator inhibitor-1 (PAI-1), and human serum adipokine panel B (HADK2–61K-B) was used for interleukin 6 (IL6), IL8, TNFα, monocyte chemoattractant protein-1 (MCP-1), hepatocyte growth factor (HGF), leptin and nerve growth factor (NGF). All values were interpolated from standard curves generated with the relevant recombinant human proteins provided with the commercial kits (Millipore).
Statistical analysis
Quantitative mass spectrometry (Biocrates), Q-PCR array expression and immunoassay using MILLIPLEX data were examined for normality using the Shapiro–Wilk test. Significant differences were determined by using a one-tailed Student’'s t-test if the data were normally distributed (*P < 0.05, **P < 0.01, ***P < 0.001) or the Wilcoxon rank sum test otherwise (#P < 0.05, ##P < 0.01, ###P < 0.001), according to the Shapiro–Wilk normality test. Data are presented as the mean ± standard error of the mean (SEM). Statistical analyses were performed using the software program SPSS 12.0. For metabolomics and lipidomics, Student's t-test, P < 0·05, with Benjamini–Hochberg Multiple Testing Correction was used.
Funding
We are indebted to Dr Arndt G. Benecke for critically reading the manuscript. This study was supported by grants from the Hesperia Foundation, the Asociación Española contra las Leucodistrofias (ALE-ELA España), the Spanish Ministry for Health and Social Policy (MSPSI EC10-137), the European Commission (FP7-241622), the Spanish Institute for Health Carlos III and ‘Fondo Europeo de Desarrollo Regional (FEDER), Unión Europea, una manera de hacer Europa’ (FIS PI11/01043, FIS PI14/00410, FIS ICI14/00076), and the Autonomous Government of Catalonia AGAUR (2009SGR85, 2014SGR1430 to A.P.); the Spanish Institute for Health Carlos III and ‘Fondo Europeo de Desarrollo Regional (FEDER), Unión Europea, una manera de hacer Europa’ (FIS PI14/00581) to C.C. and the Spanish Institute for Health Carlos III Miguel Servet program CP11/00080 to S.F. The studies conducted at the Department of Experimental Medicine were supported in part by R+D grants from the Spanish Ministry of Science and Innovation (BFU2009-11879/BFI), the Spanish Ministry of Health (PI111543; PI1300584, PI140115), the Autonomous Government of Catalonia (2009SGR735), the ‘La Caixa’ Foundation, and COST B35 The CIBER on Rare Diseases (CIBERER) and on Physiopathology of Obesity and Nutrition (CIBEROBN) are initiatives of the ISCIII.
Conflict of Interest statement. None declared.
References
Author notes
These authors contributed equally to this work.
