- Email: [email protected]
Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy
Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Contents lists available at ScienceDirect
Biochemical and Biophysical Research Communications journal homepage: www.elsevier.com/locate/ybbrc
Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy Dong-Kyu Lee a, Nguyen Phuoc Long b, Juwon Jung c, Tae Joon Kim b, Euiyeon Na b, Yun Pyo Kang b, Sung Won Kwon a, b, *, Jiho Jang d, ** a
Research Institute of Pharmaceutical Sciences, Seoul National University, Seoul, 08826, Republic of Korea College of Pharmacy, Seoul National University, Seoul, 08826, Republic of Korea c Institute of Reproductive Medicine and Population, Medical Research Center, Seoul National University, Seoul, 03080, Republic of Korea d Department of Physiology and Brain Korea 21 PLUS Project for Medical Science, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 6 November 2018 Accepted 20 November 2018 Available online xxx
Precise pathophysiology with respect to the phenotypic variations and severity of X-ALD, specifically between adrenomyeloneuropathy (AMN) and childhood cerebral adrenoleukodystrophy (CCALD), has not been fully discovered. Herein, a systematic analysis using multi-layered lipidomics and transcriptomics was conducted to elucidate distinctive metabolic biosignatures among healthy control, AMN, and CCALD. Significant alterations regarding the accumulation of very long chain fatty acids were found in various lipid species such as phospholipids, glycerolipids, and sphingolipids. Remarkably, TG and CER that are physiologically essential were markedly down-regulated in CCALD than AMN. Transcriptomic analysis further supported the robustness of our findings by providing valuable information on the gene expressions of the regulatory factors. For instance, regulators of sphingolipid catabolism (SMPD1, CERK, and SPHK1) and TG anabolism (GPAM, GPAT2, and MBOAT2) were more up-regulated in AMN than in CCALD. These observations, among others, were in line with the recognized alterations of the associated lipidomes. In conclusion, the homeostatic imbalance of the complex lipid networks may be pathogenically important in X-ALD and the particular dysregulations of TG and CER may further influence the severity of CCALD among X-ALD patients. © 2018 Published by Elsevier Inc.
Keywords: X-linked adrenoleukodystrophy Adrenomyeloneuropathy Childhood cerebral adrenoleukodystrophy Lipidomics Transcriptomics
1. Introduction X-linked adrenoleukodystrophy (X-ALD) is a devastating metabolic disorder, of which saturated very long-chain fatty acids (VLCFAs) accumulate in all tissues, particularly brain, adrenal cortex, testis, and dermal fibroblast [1,2]. VLCFA accumulation is due to inherited defects in the peroxisomal transporter, which is encoded by ABCD1 and plays a key role in entering VLCFA into peroxisome for its degradation [3]. Two prevalent phenotypes among various
* Corresponding author. Research Institute of Pharmaceutical Sciences, Seoul National University, Seoul, 08826, Republic of Korea. ** Corresponding author. Department of Physiology, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea. Tel.: 82 2 2228 1729; fax: 82 2 393 0203. E-mail addresses: [email protected] (S.W. Kwon), [email protected] (J. Jang).
phenotypes are the acute inflammatory cerebral demyelinating form of X-ALD (CCALD), which either in young boys or more rarely in adults [4], and adrenomyeloneuropathy (AMN) that slowly manifests a noninflammatory distal axonopathy in the spinal cord between 20 and 30 years of ages [5]. The cerebral inflammatory response, which is the major difference between AMN and CCALD, has long been thought to begin after the massive accumulation of VLCFA in cholesterol esters, gangliosides, phospholipids, or proteolipid [6,7]. However, only ABCD1 defects cannot adequately explain the molecular basis behind the phenotypic variability of X-ALD to date [8]. A myriad of pathophysiological studies of X-ALD were limited to VLCFA, especially saturated and monounsaturated species, and its related biochemical pathways [9]. Only few papers investigated cholesteryl esters and phospholipids, and yet their findings were
https://doi.org/10.1016/j.bbrc.2018.11.123 0006-291X/© 2018 Published by Elsevier Inc.
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
2
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Abbreviations AMN CCALD PUFA X-ALD VLCFA ABCD1 HDF DG TG PC
Adrenomyeloneuropathy Childhood cerebral adrenoleukodystrophy Polyunsaturated fatty acid X-linked adrenoleukodystrophy Very long-chain fatty acid ABC transporter subfamily D member 1 Human dermal fibroblast Diacylglycerol Triacylglycerol Phosphatidylcholine
only concerning the accumulation of VLCFA in each species [10,11]. Ultimately, the underlying mechanisms that lead to fetal progressive inflammatory demyelination remains unresolved. Of note, complex lipids, their corresponding enzymes, and other regulatory factors of the lipid metabolic pathways afford the flexibility against the pathological factors by being interactive to other alterations [12]. Thus, a more intensive and comprehensive method that is capable to characterize the whole lipidome alterations and associated regulatory factors is needed to explain the pathological mechanisms of X-ALD. With the development of novel high-throughput technologies, the systematic analysis of multi-dimensional omics data could accelerates the discovery of hidden interconnections between molecular signatures at different levels [13]. Herein, we conducted integrative lipidomic and transcriptomic platforms to assess the association of lipid metabolic pathways as concerns distinct phenotypes of X-ALD. Major lipid classes including phospholipids, glycerolipids, and sphingolipids were overall quantified based on high-throughput lipid profiling using liquid chromatography-mass spectrometry (LC-MS). Its application with primary dermal fibroblasts that were derived from both phenotypes compared the whole lipid abundances among healthy control (human dermal fibroblast, HDF), AMN, and CCALD. Furthermore, the gene expression involved in metabolic pathways of those lipid candidates were obtained using high-throughput transcriptome analysis to compare AMN and CCALD. Collectively, our functional omics approach provides informative evidence and reveals the hidden metabolic signatures that enormously influence the pathophysiology of the mild and severe forms of X-ALD. 2. Materials and methods 2.1. Chemicals and reagents High performance liquid chromatography (HPLC) grade chloroform, methanol, 2-propanol, acetonitrile and water were from J.T.Baker (Phillipsburg, NJ, USA). Eluent buffers, ammonium acetate and formic acid, were from Sigma-Aldrich (St. Louis, MO, USA). Diacylglycerol (DG, 12:0/12:0), triacylglycerol (TG, 17:0/17:0/17:0), phosphatidylcholine (PC, 10:0/10:0), phosphatidylethanolamine (PE, 10:0/10:0), sphingomyelin (SM, 18:1d/17:0), and ceramide (CER, C17) for internal standards were from Avanti Polar Lipids (Alabaster, AL, USA).
PE SM CER LPE LPC PlsPC PlsPE PLS-DA VIP EPA DHA
Phosphatidylethanolamine Sphingomyelin Ceramide Lysophosphatidylethanolamine Lysophosphatidylcholine Plasmenylcholine Plasmenylethanolamine Partial least square-discriminant analysis Variable importance for projection Eicosapentaenoic acid Docosahexaenoic acid
agreed and signed an informed consent prior to the sample collection. 2.3. Establishment of X-ALD patients-derived cells Human X-ALD-fibroblasts (GM04496 (CCALD), GM04934 (CCALD), GM07530 (AMN) and GM17819 (AMN)) and control human dermal fibroblast (neonatal HDF and adult HDF) cells were purchased from Coriell Institute (Camden, NJ, USA) and Invitrogen (Carlsbad, CA, USA), respectively. Human X-ALD-fibroblasts were also derived and established from skin punch biopsies taken from two Korean CCALD patients and Korean AMN patient, as described in the previous report with the patients' clinical information [14]. Different passages of the patients-derived cells and commercial cell lines were used to represent biological and technical variants of the samples in the experiments. Fibroblasts were cultured in fibroblast medium made up of Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS; GIBCO, Grand Island, NY, USA), 1 mM glutamine (GIBCO), 1% nonessential amino acids (Invitrogen), and penicillin/streptomycin (Invitrogen) as previously described [14]. 2.4. Gene expression profiling and microarray analysis In the present study, global gene expression analysis using Affymetrix GeneChip® HT HG-U133 þ PM Array was conducted (DNA Link, Seoul, Korea), and isolated RNA samples (300 ng) were used as input into the Affymetrix procedure (Affymetrix, Santa Clara, CA, USA), as shown in Supplementary Information. Scanned image data was extracted through Affymetrix Command Console software 1.1. The raw data generated through above procedure meant expression intensity data and was used for the next step. For the normalization, Affymetrix microarray suite 5 (MAS5) algorithm implemented in Affymetrix Expression Console software (version 1.1) was used. In order to remove noise in the search of significant genes, samples were excluded from the analysis when MAS5 detection calls were not determined as present calls. Highly expressed genes as comparing the signal value of control and test sample, were selected for the further study. Lipid metabolism related genes were named following HUGO and classified based on the information of gene function in Gene ontology and KEGG pathway database [15]. 2.5. Lipidomics data acquisition and analysis
2.2. Ethical statement The protocol for the generation and analysis of fibroblasts from ALD patients was approved by Yonsei University Institutional Review Board (IRB 4-2016-0194). Every patient or patient's parent
Lipids in harvested primary cells were extracted using freezethaw method with some modification as described before [16]. Detailed reagents and procedures for sample preparation are described in Supplementary Information. Lipid extracts were
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
injected and separated using a 1260 HPLC system (Agilent Technologies, Santa Clara, CA, USA) equipped with Agilent 6530 QTOFMS. Electrospray ionization with positive mode was operated with chromatographic and spectrometric parameters for lipidomics (Supplementary Information). After obtaining chromatograms by untargeted LC-MS analysis with scan mode, targeted MS/MS with the same parameters aforementioned confirmed fragment pattern of each lipid. MS/MS spectrum of each species were shown in Supplementary Table S1- S10. The quantitative result of every lipid were extracted using data alignment in MZmine (Supplementary Information). The aligned data was used for multivariate statistical analysis using SIMCA-Pþ (version 11.0, Umetrics, Umeå Sweden) and MetaboAnalyst (version 3.0, http://www.metaboanalyst.ca) [17]. For PLS-DA and heatmap analysis, data was normalized by the area of internal standard for each lipid species, log-transformed, and Pareto-scaled. Holm-Sidak multiple comparisons of HDF, AMN, and CCALD were conducted using GraphPad Prism (version 7.00, GraphPad software, USA). Data visualization was conducted using Adobe Illustrator CS6 (Adobe Systems Inc., San Jose, CA, USA). 2.6. Data availability The data generated from microarray analysis is archived on Gene Expression Omnibus. The accession number is GSE117647. 3. Results 3.1. Lipid profiling uncovers the significant alteration between HDF and two X-ALD groups Total 61 samples (14 HDF, 21 AMN, and 26 CCALD) were finally analyzed by high-throughput lipidomics approach to discover altered lipids among HDF (healthy control), AMN, and CCALD. Our high-throughput profiling was capable to profile 227 lipids including 68 triacylglycerol (TG), 18 diacylglycerol (DG), 17 sphingomyelin (SM), 26 phosphatidylethanolamine (PE), 54 phosphatidylcholine (PC), 5 lysophosphatidylethanolamine (LPE), 13 lysophosphatidylcholine (LPC), 13 Plasmenylethanolamine (PlsPE), and 13 Plasmenylcholine (PlsPC). For data exploration, a heatmap of all the lipid species was used to visualize the lipid profiles of HDF, AMN, and CCALD (Fig. 1a). Noticeably, the longer the acyl chains of TG, SM, PE, PC, LPC, and CER, the bigger the differences between control and X-ALD groups were. Partial least square-discriminant analysis (PLS-DA) was then applied to distinguish three groups based on whole lipid profiles (Fig. 1b). In the score scatter plot, component 1 (26.8% variables) separated HDF from the other groups while component 2 (17.3% variables) separated AMN and CCALD. QC samples were strictly clustered in the center of the plot. The goodness-of-fit (R2) of 0.77 and the goodness-of-prediction (Q2) of 0.58 of the PLS-DA model revealed the robustness of itself, together with a 1000-time permutation test that had p-value under 0.001 (Supplementary Fig. S1). TG, SM, PE, PC, LPC, and CER species, which were found distinctly different among three groups in heatmap analysis, had particular altered patterns of each lipid with different acyl chain length and number of unsaturated bonds (Fig. 1c and Supplementary Figs. S2eS11). Lipid species that made up of VLCFAs appeared with extreme fold changes between HDF and both X-ALD groups. TG that afforded three fatty acids showed predominant differences when increasing the carbon number over C58. In case of monounsaturated TGs, TG58:1 had a 2.4 fold change, TG60:1 had a 4.5 fold change, and TG62:1 had a 5.6 fold change. PCs, PEs with two fatty acid available showed a similar pattern of over C42. SMs, CERs, and LPCs with one fatty acid showed the same pattern in
3
C26:0-included lipids. Unlike the other lipids, PlsPC and PlsPE, which are the output of peroxisomal catabolism of lipids containing VLCFA, did not showed any big changes. These lipid differences between control and X-ALD groups suggests that ABCD1 defect induces the synthesis of VLCFA into various kinds of lipid species. In addition, unsaturated fatty acids were found to be highly abundant in HDF compared to both AMN and CCALD (Supplementary Fig. S12). A comparison regarding the unsaturated fatty acids between control and X-ALD groups can be found in Supplementary Information. 3.2. AMN highly activates the sphingolipid metabolism from sphingomyelin to sphingosine-1-phosphate Multivariate analysis of the lipid levels between AMN and CCALD was conducted to reveal the compositions of lipid molecules that played an important role in underlying lipidomic mechanisms. As shown in Fig. 2a, we obtain a PLS-DA score plot, in which the component 1 (26.6% variables) and 2 (11.6% variables) exhibit a relatively well separated two groups. The goodness-of-fit (R2) of 0.96 and the goodness-of-prediction (Q2) of 0.89 of the model and a 1000-time permutation test with p-value under 0.001 revealed the robustness of the current analysis (Supplementary Fig. S1). Significantly, 13 out of 15 CERs turned out to differentiate two phenotypes, having variable importance for projection (VIP) value over 1 (Supplementary Table S2). In loading plot of PLS-DA, CERs were highly headed toward AMN group, which showed a higher expression of CERs in AMN (Fig. 2b). This observation once again was confirmed by the univariate analysis (Fig. 2c). It is also worth mentioning that, although C24DH CER (dihydroceramide C24) and C26 CER (Ceramide C26) did not have VIP value over 1, they were also differentially expressed (adjusted p-value < 0.05). Altogether, these results suggest that CER and the related sphingolipid metabolism may be the highly relevant signatures and are responsible for the phenotypic variations in XALD. The metabolic regulators that are responsible for sphingolipid metabolism are described in Fig. 2d. In our transcriptome analysis, those regulators were classified into two categories: regulators related to the synthesis of complex structures from original backbone, such as sphingosine-1-phosphate (sphingolipid anabolism in Fig. 2e) and regulators related to the metabolism of glycosphigolipid and SM to its backbone (sphingolipid catabolism to sphingolipid-1-phosphate). In the anabolism, SGMS2, UGCG, CERT, LASS6, LASS1, and UGT8 were more down-regulated in AMN than CCALD. Especially, SGMS2, UGCG, and UGT8, which involved in the utilization of CER for the synthesis of SM and glycosphingolipid, showed a profound difference. This suggested that the higher levels of CERs in AMN might come from the alleviated expression levels of these two anabolic regulators. In contrast, SMPD1 and SMPDL3A, which are responsible for the hydrolyzation of SM to generate CER, were relatively lower in AMN compared to CCALD. Other metabolic regulators that belong to the catabolism showed higher expressions compared to CCALD whether those regulators are up- or downregulated in comparison to HDF. Collectively, transcriptome expression from the simple to complex structure of sphingolipid was highly decreased in AMN, and especially SGMS2, UGCG, and UGT8 might cause the accumulation of CER. 3.3. The anabolic pathway of triacylglycerol is up-regulated in AMN In contrary to the up-regulation in the anabolism of sphingolipid pathway (from sphingosine-1-phosphate to sphingomyelin), CCALD possessed a considerable deficiency in the anabolic pathway of TG species. Indeed, as being shown in the loading plot of previous
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
4
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Fig. 1. The alterations of lipidomes in 10 lipid species in control (HDF) and two X-ALD forms (AMN and CCALD). a) Heatmap of the expression of all lipid species (total 227 lipids). The normalized expression of each lipid in HDF (n ¼ 14), AMN (n ¼ 21), and CCALD (n ¼ 26) are shown in scale range from green to red. b) PLS-DA score plot of the whole lipidome data. The observations of HDF, AMN and CCALD are colored as gray, green and red, respectively. c) The relative alterations of distinctive lipid species (PE, SM, TG, CER, LPC, and PC). The fold change was shown as mean with s.e.m. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
PLS-DA of Fig. 2a, TGs with the shorter the FA chains were headed to AMN (Fig. 3a). It implied that TGs with shorter-chain FAs, which typically exist and be utilized in healthy human cells, were highly abundant in the mild phenotype of X-ALD. Those statistically significant alterations were selectively found in C48 to C58 TG species and the expression level of TG was generally higher in AMN than that of CCALD (Fig. 3b). Furthermore, given that the significant alterations in X-ALD groups were started from C60, properly owing to the VLCFA accumulation, the above observed patterns suggested that the expression of typical TGs was specifically impaired in CCALD, the severe phenotype of X-ALD. To further elucidate the modulating mechanisms in TG anabolism and catabolism, we assessed the metabolic enzymes and other
regulators of TG pathway (Fig. 3c). The expression levels of those transcriptomes were categorized into the anabolism and the catabolism, respectively (Fig. 3d). Most of genes involved in TG synthesis showed relatively lower expressions in CCALD. GK, which plays an essential role in coupling phosphate groups to glycerol, showed a relatively small diminution than CCALD; and GPAM, GPAT2, and GPAT4, which binds one acyl chain on glycerol-3phosphate, were highly expressed in AMN. LCLAT1, MBOAT1, MBOAT2, AGPAT2, AGPAT3, and AGPAT4, which help attach more acyl chain on 1-acyl-sn-glycerol-3-phosphate, were up-regulated in AMN. The gene expression levels of MOGAT2, MOGAT3, and LPIN1, three out of six genes related to DG synthesis, were increased. DGAT2 and MOGAT3, which generate TG from DG, were
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
5
Fig. 2. Ceramides as distinctive markers between AMN and CCALD, and altered expression of transcriptomes in the sphingolipid pathway. a) PLS-DA score plot shows a relative clear separation between two phenotypes of X-ALD. The observations of AMN (n ¼ 21) and CCALD (n ¼ 26) are colored green and red, respectively. b) Loading plots of PLS-DA of two groups. Ceramides are colored red. c) Relative abundances (HDF as 100%) of every ceramide in AMN and CCALD. Data are shown as mean with s.e.m.; *p < 0.05, **p < 0.01, and ***p < 0.001. d) Sphingolipid pathway and corresponding regulators. e) The expression of enzymes and metabolic regulators of the sphingolipid pathway. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
fairly up-regulated in AMN. Collectively, AMN had higher expressions of the key metabolic regulators involved in the TG anabolic pathway. Reversely, CCALD had a clear deficiency in the TG synthesis. Nonetheless, in TG catabolism, the pathway from TG to its backbone, revealed very slight alterations of the regulators. Some of them had relatively lower expressions (CEL, PNPLA2, PNLIP, PLRP1, PLRP2, DGKA, DGKB, DGKG, DGKQ, and AGK), in line with the accumulation of TG. Therefore, our systematic analysis revealed the impairment of TG anabolism in CCALD, triggered by the downregulated expression of transcriptomes in TG synthesis, and the impairment of the typical TGs might biochemically be associated with the severity in CCALD patient. 4. Discussion We conducted large-scale and systematic omics analysis to investigate the abnormal cellular metabolism at the lipidome and transcriptome levels. As a proof-of-concept, we successfully found distinctive candidates, PUFAs, CERs, and TGs, among the three phenotypes and elucidated the relationship between the abnormal metabolism of X-ALD and the lipid regulatory pathways from multi-layered datasets. Our findings provide a profound information-rich background regarding the patterns of lipidome abnormalities and altered expressions of corresponding metabolic regulators in different forms of X-ALD. At first, considerable aberrants in VLCFAcontaining phospholipids (PE, PC, and LPC), sphingolipids (CER and SM) and glycerolipids (DG and TG) were revealed. VLCFA turned out to be coupled to various lipids species, except PlsPC and PlsPE, since these two species are synthesized by peroxisomal lipid metabolism using fatty alcohol as substrate unlike the
other species [18]. This observation shows a good agreement with previous studies and considerably expands the combinable lipid species. Novel signatures that distinguish AMN and CCALD and their associations to sphingolipid and triacylglycerol metabolism were, to the best of our knowledge, firstly elucidated in our study. Interestingly, most of the markers in those metabolic pathways are decreased in CCALD, and are not related to the lipids containing VLCFA, rather to the lipids abundantly existing and physiologically essential in healthy control. These results suggest that the dysregulations of abundant and typical lipids utilized for sustaining lipid metabolic homeostasis may be involved in the severity of X-ALD. Transcriptome data further supported the investigation for those pathways. In sphingolipid pathway, the anabolism toward the more complex sphingolipids including sphingomyelin and glycosphingolipid in AMN, were down-regulated, and reversely, the catabolism were up-regulated in AMN. These lipidomic and transcriptomic alterations together revealed that sphingolipid pathways do not maintain the homeostasis in CCALD due to the reduction of the lipid catabolism. This imbalance in sphingolipid homeostasis were known as sphingolipidoses, a class of lipid metabolic disorder. Interestingly, our results about this pathway were quantitatively similar to the one of the sphingolipidoses, namely Farber disease, a type of disease showing ceramide accumulation by ASAH1 mutation [19]. In our transcriptome data, the ASAH1 expression was also decreased accordingly to the severity of the X-ALD type (Fig. 2d). The lipidomic similarity with Farber disease may help explain the different level of neurological toxicity between AMN and CCALD. Moreover, ceramide accumulation in AMN can be explained by the reduced anabolism from ceramide to glycosphingolipid (UGCG) or sphingomyelin (SGMS2), and reduced
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
6
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Fig. 3. The alterations of triacylglycerol differentiating AMN and CCALD. a) Loading plot of PLS-DA model. Triacylglycerols are the boxes colored in red (C48) to yellow (C64). b) Relative abundances (HDF as 100%) of every triacylglycerol in AMN and CCALD. Data are shown as mean with s.e.m.; *p < 0.05, **p < 0.01, and ***p < 0.001. c) Triacylglycerol metabolism and corresponding regulators. e) The expression of enzymes and metabolic regulators of the triacylglycerol metabolism. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
sphingolipid trafficking process by slight decrease in CERT expression [20,21]. Upregulated triacylglycerol metabolism in AMN is also a distinctive phenomenon in our study. TG species are well known as a protective effect against fatty acid-induced lipotoxicity [22]. Physiologically, TG keeps some excess saturated or monounsaturated fatty acids in lipid droplets. This process protects the cells against the lipotoxicity of VLCFA, including saturated and mono-unsaturated types, and prevents the induction of necroptosis in neural cells like oligodendrocytes and astrocytes [23,24]. Furthermore, the accumulation of TG containing PUFA (PUFA-TG) and their localization in lipid droplet limits the toxicity by preventing the PUFA-phospholipid from undergoing oxidation and increasing oxidative stress [25,26]. Our results revealed that PUFATGs with typical acyl chains are predominantly significant and having bigger fold change than saturated lipid with same carbon number. The PUFA-TG accumulation in AMN might help reduce cellular toxicity from ABCD1 defects, but this process might not be active in CCALD. However, specific roles of PUFA-TGs in X-ALD should be explored in future studies. In conclusion, systematic analysis combining lipidomics and transcriptomics revealed the difference between major X-ALD phenotypes by focusing on the metabolism of typical lipids, rather than VLCFA alone. Surprisingly, we found novel dysregulated metabolism of polyunsaturated fatty acids in X-ALD groups. Furthermore, the quantitative alterations of lipids and transcriptomes between AMN and CCALD largely derived from distinctive pathways in sphingolipid and triacylglycerol metabolism. Our study comprehensively describes the collapse of the lipid homeostasis, especially in typical lipids that are
physiologically essential for healthy control. The findings may partly explain why X-ALD patients suffer from a wide spectrum of the symptoms. Our systematic approach could greatly facilitates the understanding of the pathophysiology of X-ALD. Future studies are warranted to overcome the current limitations and extend our observation for linking essential lipids and their related metabolic regulators. Conflicts of interest The authors declare no conflict of interest. Acknowledgement This work was supported by Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (NRF2018R1D1A1B07046207), and the National Research Foundation of Korea grant funded by the Korean government (MSIP) (NRF2018R1A5A2024425). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbrc.2018.11.123. Transparency document Transparency document related to this article can be found online at https://doi.org/10.1016/j.bbrc.2018.11.123.
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
References [1] J.M. Powers, Y. Liu, A.B. Moser, H.W. Moser, The inflammatory myelinopathy of adreno-leukodystrophy: cells, effector molecules, and pathogenetic implications, J. Neuropathol. Exp. Neurol. 51 (1992) 630e643. [2] J. Singh, S. Giri, Loss of AMP-activated protein kinase in X-linked adrenoleukodystrophy patient-derived fibroblasts and lymphocytes, Biochem. Biophys. Res. Commun. 445 (2014) 126e131. [3] I. Singh, A.E. Moser, S. Goldfischer, H.W. Moser, Lignoceric acid is oxidized in the peroxisome: implications for the Zellweger cerebro-hepato-renal syndrome and adrenoleukodystrophy, Proc. Natl. Acad. Sci. U.S.A. 81 (1984) 4203e4207. [4] D. Trompier, C. Gondcaille, G. Lizard, S. Savary, Regulation of the adrenoleukodystrophy-related gene (ABCD2): focus on oxysterols and LXR antagonists, Biochem. Biophys. Res. Commun. 446 (2014) 651e655. [5] J. Berger, S. Forss-Petter, F.S. Eichler, Pathophysiology of X-linked adrenoleukodystrophy, Biochimie 98 (2014) 135e142. ~ iga, M. Lees B, Fatty acid composition of human myelin [6] O. Bizzozero, A.,G. Zun proteolipid protein in peroxisomal disorders, J. Neurochem. 56 (1991) 872e878. [7] M. Igarashi, H.H. Schaumburg, J. Powers, Y. Kishimoto, E. Koilodny, K. Suzuki, Fatty acid abnormality in adrenoleukodystrophy, J. Neurochem. 26 (1976) 851e860. [8] F. Kallabi, E. Ellouz, M. Tabebi, G. Ben Salah, N. Kaabechi, L. Keskes, C. Triki, H. Kamoun, Phenotypic variability in a Tunisian family with X-linked adrenoleukodystrophy caused by the p.Gln316Pro novel mutation, Clin. Chim. Acta 453 (2016) 141e146. [9] S. Kemp, F. Valianpour, S. Denis, R. Ofman, R.-J. Sanders, P. Mooyer, P.G. Barth, R.J.A. Wanders, Elongation of very long-chain fatty acids is enhanced in Xlinked adrenoleukodystrophy, Mol. Genet. Metabol. 84 (2005) 144e151. [10] C. Theda, A.B. Moser, J.M. Powers, H.W. Moser, Phospholipids in X-linked adrenoleukodystrophy white matter: fatty acid abnormalities before the onset of demyelination, J. Neurol. Sci. 110 (1992) 195e204. [11] P. Sharp, D. Johnson, A. Poulos, Molecular species of phosphatidylcholine containing very long chain fatty acids in human brain: enrichment in X-linked adrenoleukodystrophy brain and diseases of peroxisome biogenesis brain, J. Neurochem. 56 (1991) 30e37. [12] S. Beloribi-Djefaflia, S. Vasseur, F. Guillaumond, Lipid metabolic reprogramming in cancer cells, Oncogenesis 5 (2016) e189. [13] I. Stagljar, The power of OMICs, Biochem. Biophys. Res. Commun. 479 (2016) 607e609. [14] J. Jang, S. Park, H. Jin Hur, H.-J. Cho, I. Hwang, Y. Pyo Kang, I. Im, H. Lee, E. Lee,
[15]
[16]
[17] [18]
[19] [20]
[21]
[22]
[23] [24]
[25] [26]
7
W. Yang, H.-C. Kang, S. Won Kwon, J.-W. Yu, D.-W. Kim, 25-hydroxycholesterol contributes to cerebral inflammation of X-linked adrenoleukodystrophy through activation of the NLRP3 inflammasome, Nat. Commun. 7 (2016) 13129. N.P. Long, K.H. Jung, S.J. Yoon, N.H. Anh, T.D. Nghi, Y.P. Kang, H.H. Yan, J.E. Min, S.S. Hong, S.W. Kwon, Systematic assessment of cervical cancer initiation and progression uncovers genetic panels for deep learning-based early diagnosis and proposes novel diagnostic and prognostic biomarkers, Oncotarget 8 (2017) 109436e109456. ~ averas, S. Lo pez, J.V. Castell, M.T. Donato, A. Lahoz, Extending J.C. García-Can metabolome coverage for untargeted metabolite profiling of adherent cultured hepatic cells, Anal. Bioanal. Chem. 408 (2016) 1217e1230. J. Xia, I.V. Sinelnikov, B. Han, D.S. Wishart, MetaboAnalyst 3.0dmaking metabolomics more meaningful, Nucleic Acids Res. 43 (2015) W251eW257. H.R. Waterham, S. Ferdinandusse, R.J.A. Wanders, Human disorders of peroxisome metabolism and biogenesis, Biochim. Biophys. Acta Mol. Cell Res. 1863 (2016) 922e933. Y.-H. Xu, S. Barnes, Y. Sun, G.A. Grabowski, Multi-system disorders of glycosphingolipid and ganglioside metabolism, J. Lipid Res. 51 (2010) 1643e1675. K. Kumagai, C.A. Elwell, S. Ando, J.N. Engel, K. Hanada, Both the N- and Cterminal regions of the Chlamydial inclusion protein D (IncD) are required for interaction with the pleckstrin homology domain of the ceramide transport protein CERT, Biochem. Biophys. Res. Commun. 505 (2018) 1070e1076. Y. Ishibashi, M. Ito, Y. Hirabayashi, Regulation of glucosylceramide synthesis by Golgi-localized phosphoinositide, Biochem. Biophys. Res. Commun. 499 (2018) 1011e1018. L.L. Listenberger, X. Han, S.E. Lewis, S Cases, R.V. Farese, D.S. Ory, J.E. Schaffer, Triglyceride accumulation protects against fatty acid-induced lipotoxicity, Proc. Natl. Acad. Sci. U.S.A. 100 (2003) 3077e3082. L.R. Parisi, N. Li, G.E. Atilla-Gokcumen, Very long chain fatty acids are functionally involved in necroptosis, Cell Chem. Biol. 24 (2017) 1445e1454, e1448. €nfeld, S. Kahlert, G. Reiser, Toxic effects of X-linked S. Hein, P. Scho adrenoleukodystrophy-associated, very long chain fatty acids on glial cells and neurons from rat hippocampus in culture, Hum. Mol. Genet. 17 (2008) 1750e1761. N. Li, Y. Sancak, J. Frasor, G.E. Atilla-Gokcumen, A protective role for triacylglycerols during apoptosis, Biochemistry 57 (2018) 72e80. E. Jarc, A. Kump, P. Malavasi c, T.O. Eichmann, R. Zimmermann, T. Petan, Lipid droplets induced by secreted phospholipase A2 and unsaturated fatty acids protect breast cancer cells from nutrient and lipotoxic stress, Biochim. Biophys. Acta Mol. Cell Biol. Lipids 1863 (2018) 247e265.
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
Contents lists available at ScienceDirect
Biochemical and Biophysical Research Communications journal homepage: www.elsevier.com/locate/ybbrc
Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy Dong-Kyu Lee a, Nguyen Phuoc Long b, Juwon Jung c, Tae Joon Kim b, Euiyeon Na b, Yun Pyo Kang b, Sung Won Kwon a, b, *, Jiho Jang d, ** a
Research Institute of Pharmaceutical Sciences, Seoul National University, Seoul, 08826, Republic of Korea College of Pharmacy, Seoul National University, Seoul, 08826, Republic of Korea c Institute of Reproductive Medicine and Population, Medical Research Center, Seoul National University, Seoul, 03080, Republic of Korea d Department of Physiology and Brain Korea 21 PLUS Project for Medical Science, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 6 November 2018 Accepted 20 November 2018 Available online xxx
Precise pathophysiology with respect to the phenotypic variations and severity of X-ALD, specifically between adrenomyeloneuropathy (AMN) and childhood cerebral adrenoleukodystrophy (CCALD), has not been fully discovered. Herein, a systematic analysis using multi-layered lipidomics and transcriptomics was conducted to elucidate distinctive metabolic biosignatures among healthy control, AMN, and CCALD. Significant alterations regarding the accumulation of very long chain fatty acids were found in various lipid species such as phospholipids, glycerolipids, and sphingolipids. Remarkably, TG and CER that are physiologically essential were markedly down-regulated in CCALD than AMN. Transcriptomic analysis further supported the robustness of our findings by providing valuable information on the gene expressions of the regulatory factors. For instance, regulators of sphingolipid catabolism (SMPD1, CERK, and SPHK1) and TG anabolism (GPAM, GPAT2, and MBOAT2) were more up-regulated in AMN than in CCALD. These observations, among others, were in line with the recognized alterations of the associated lipidomes. In conclusion, the homeostatic imbalance of the complex lipid networks may be pathogenically important in X-ALD and the particular dysregulations of TG and CER may further influence the severity of CCALD among X-ALD patients. © 2018 Published by Elsevier Inc.
Keywords: X-linked adrenoleukodystrophy Adrenomyeloneuropathy Childhood cerebral adrenoleukodystrophy Lipidomics Transcriptomics
1. Introduction X-linked adrenoleukodystrophy (X-ALD) is a devastating metabolic disorder, of which saturated very long-chain fatty acids (VLCFAs) accumulate in all tissues, particularly brain, adrenal cortex, testis, and dermal fibroblast [1,2]. VLCFA accumulation is due to inherited defects in the peroxisomal transporter, which is encoded by ABCD1 and plays a key role in entering VLCFA into peroxisome for its degradation [3]. Two prevalent phenotypes among various
* Corresponding author. Research Institute of Pharmaceutical Sciences, Seoul National University, Seoul, 08826, Republic of Korea. ** Corresponding author. Department of Physiology, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea. Tel.: 82 2 2228 1729; fax: 82 2 393 0203. E-mail addresses: [email protected] (S.W. Kwon), [email protected] (J. Jang).
phenotypes are the acute inflammatory cerebral demyelinating form of X-ALD (CCALD), which either in young boys or more rarely in adults [4], and adrenomyeloneuropathy (AMN) that slowly manifests a noninflammatory distal axonopathy in the spinal cord between 20 and 30 years of ages [5]. The cerebral inflammatory response, which is the major difference between AMN and CCALD, has long been thought to begin after the massive accumulation of VLCFA in cholesterol esters, gangliosides, phospholipids, or proteolipid [6,7]. However, only ABCD1 defects cannot adequately explain the molecular basis behind the phenotypic variability of X-ALD to date [8]. A myriad of pathophysiological studies of X-ALD were limited to VLCFA, especially saturated and monounsaturated species, and its related biochemical pathways [9]. Only few papers investigated cholesteryl esters and phospholipids, and yet their findings were
https://doi.org/10.1016/j.bbrc.2018.11.123 0006-291X/© 2018 Published by Elsevier Inc.
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
2
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Abbreviations AMN CCALD PUFA X-ALD VLCFA ABCD1 HDF DG TG PC
Adrenomyeloneuropathy Childhood cerebral adrenoleukodystrophy Polyunsaturated fatty acid X-linked adrenoleukodystrophy Very long-chain fatty acid ABC transporter subfamily D member 1 Human dermal fibroblast Diacylglycerol Triacylglycerol Phosphatidylcholine
only concerning the accumulation of VLCFA in each species [10,11]. Ultimately, the underlying mechanisms that lead to fetal progressive inflammatory demyelination remains unresolved. Of note, complex lipids, their corresponding enzymes, and other regulatory factors of the lipid metabolic pathways afford the flexibility against the pathological factors by being interactive to other alterations [12]. Thus, a more intensive and comprehensive method that is capable to characterize the whole lipidome alterations and associated regulatory factors is needed to explain the pathological mechanisms of X-ALD. With the development of novel high-throughput technologies, the systematic analysis of multi-dimensional omics data could accelerates the discovery of hidden interconnections between molecular signatures at different levels [13]. Herein, we conducted integrative lipidomic and transcriptomic platforms to assess the association of lipid metabolic pathways as concerns distinct phenotypes of X-ALD. Major lipid classes including phospholipids, glycerolipids, and sphingolipids were overall quantified based on high-throughput lipid profiling using liquid chromatography-mass spectrometry (LC-MS). Its application with primary dermal fibroblasts that were derived from both phenotypes compared the whole lipid abundances among healthy control (human dermal fibroblast, HDF), AMN, and CCALD. Furthermore, the gene expression involved in metabolic pathways of those lipid candidates were obtained using high-throughput transcriptome analysis to compare AMN and CCALD. Collectively, our functional omics approach provides informative evidence and reveals the hidden metabolic signatures that enormously influence the pathophysiology of the mild and severe forms of X-ALD. 2. Materials and methods 2.1. Chemicals and reagents High performance liquid chromatography (HPLC) grade chloroform, methanol, 2-propanol, acetonitrile and water were from J.T.Baker (Phillipsburg, NJ, USA). Eluent buffers, ammonium acetate and formic acid, were from Sigma-Aldrich (St. Louis, MO, USA). Diacylglycerol (DG, 12:0/12:0), triacylglycerol (TG, 17:0/17:0/17:0), phosphatidylcholine (PC, 10:0/10:0), phosphatidylethanolamine (PE, 10:0/10:0), sphingomyelin (SM, 18:1d/17:0), and ceramide (CER, C17) for internal standards were from Avanti Polar Lipids (Alabaster, AL, USA).
PE SM CER LPE LPC PlsPC PlsPE PLS-DA VIP EPA DHA
Phosphatidylethanolamine Sphingomyelin Ceramide Lysophosphatidylethanolamine Lysophosphatidylcholine Plasmenylcholine Plasmenylethanolamine Partial least square-discriminant analysis Variable importance for projection Eicosapentaenoic acid Docosahexaenoic acid
agreed and signed an informed consent prior to the sample collection. 2.3. Establishment of X-ALD patients-derived cells Human X-ALD-fibroblasts (GM04496 (CCALD), GM04934 (CCALD), GM07530 (AMN) and GM17819 (AMN)) and control human dermal fibroblast (neonatal HDF and adult HDF) cells were purchased from Coriell Institute (Camden, NJ, USA) and Invitrogen (Carlsbad, CA, USA), respectively. Human X-ALD-fibroblasts were also derived and established from skin punch biopsies taken from two Korean CCALD patients and Korean AMN patient, as described in the previous report with the patients' clinical information [14]. Different passages of the patients-derived cells and commercial cell lines were used to represent biological and technical variants of the samples in the experiments. Fibroblasts were cultured in fibroblast medium made up of Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS; GIBCO, Grand Island, NY, USA), 1 mM glutamine (GIBCO), 1% nonessential amino acids (Invitrogen), and penicillin/streptomycin (Invitrogen) as previously described [14]. 2.4. Gene expression profiling and microarray analysis In the present study, global gene expression analysis using Affymetrix GeneChip® HT HG-U133 þ PM Array was conducted (DNA Link, Seoul, Korea), and isolated RNA samples (300 ng) were used as input into the Affymetrix procedure (Affymetrix, Santa Clara, CA, USA), as shown in Supplementary Information. Scanned image data was extracted through Affymetrix Command Console software 1.1. The raw data generated through above procedure meant expression intensity data and was used for the next step. For the normalization, Affymetrix microarray suite 5 (MAS5) algorithm implemented in Affymetrix Expression Console software (version 1.1) was used. In order to remove noise in the search of significant genes, samples were excluded from the analysis when MAS5 detection calls were not determined as present calls. Highly expressed genes as comparing the signal value of control and test sample, were selected for the further study. Lipid metabolism related genes were named following HUGO and classified based on the information of gene function in Gene ontology and KEGG pathway database [15]. 2.5. Lipidomics data acquisition and analysis
2.2. Ethical statement The protocol for the generation and analysis of fibroblasts from ALD patients was approved by Yonsei University Institutional Review Board (IRB 4-2016-0194). Every patient or patient's parent
Lipids in harvested primary cells were extracted using freezethaw method with some modification as described before [16]. Detailed reagents and procedures for sample preparation are described in Supplementary Information. Lipid extracts were
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
injected and separated using a 1260 HPLC system (Agilent Technologies, Santa Clara, CA, USA) equipped with Agilent 6530 QTOFMS. Electrospray ionization with positive mode was operated with chromatographic and spectrometric parameters for lipidomics (Supplementary Information). After obtaining chromatograms by untargeted LC-MS analysis with scan mode, targeted MS/MS with the same parameters aforementioned confirmed fragment pattern of each lipid. MS/MS spectrum of each species were shown in Supplementary Table S1- S10. The quantitative result of every lipid were extracted using data alignment in MZmine (Supplementary Information). The aligned data was used for multivariate statistical analysis using SIMCA-Pþ (version 11.0, Umetrics, Umeå Sweden) and MetaboAnalyst (version 3.0, http://www.metaboanalyst.ca) [17]. For PLS-DA and heatmap analysis, data was normalized by the area of internal standard for each lipid species, log-transformed, and Pareto-scaled. Holm-Sidak multiple comparisons of HDF, AMN, and CCALD were conducted using GraphPad Prism (version 7.00, GraphPad software, USA). Data visualization was conducted using Adobe Illustrator CS6 (Adobe Systems Inc., San Jose, CA, USA). 2.6. Data availability The data generated from microarray analysis is archived on Gene Expression Omnibus. The accession number is GSE117647. 3. Results 3.1. Lipid profiling uncovers the significant alteration between HDF and two X-ALD groups Total 61 samples (14 HDF, 21 AMN, and 26 CCALD) were finally analyzed by high-throughput lipidomics approach to discover altered lipids among HDF (healthy control), AMN, and CCALD. Our high-throughput profiling was capable to profile 227 lipids including 68 triacylglycerol (TG), 18 diacylglycerol (DG), 17 sphingomyelin (SM), 26 phosphatidylethanolamine (PE), 54 phosphatidylcholine (PC), 5 lysophosphatidylethanolamine (LPE), 13 lysophosphatidylcholine (LPC), 13 Plasmenylethanolamine (PlsPE), and 13 Plasmenylcholine (PlsPC). For data exploration, a heatmap of all the lipid species was used to visualize the lipid profiles of HDF, AMN, and CCALD (Fig. 1a). Noticeably, the longer the acyl chains of TG, SM, PE, PC, LPC, and CER, the bigger the differences between control and X-ALD groups were. Partial least square-discriminant analysis (PLS-DA) was then applied to distinguish three groups based on whole lipid profiles (Fig. 1b). In the score scatter plot, component 1 (26.8% variables) separated HDF from the other groups while component 2 (17.3% variables) separated AMN and CCALD. QC samples were strictly clustered in the center of the plot. The goodness-of-fit (R2) of 0.77 and the goodness-of-prediction (Q2) of 0.58 of the PLS-DA model revealed the robustness of itself, together with a 1000-time permutation test that had p-value under 0.001 (Supplementary Fig. S1). TG, SM, PE, PC, LPC, and CER species, which were found distinctly different among three groups in heatmap analysis, had particular altered patterns of each lipid with different acyl chain length and number of unsaturated bonds (Fig. 1c and Supplementary Figs. S2eS11). Lipid species that made up of VLCFAs appeared with extreme fold changes between HDF and both X-ALD groups. TG that afforded three fatty acids showed predominant differences when increasing the carbon number over C58. In case of monounsaturated TGs, TG58:1 had a 2.4 fold change, TG60:1 had a 4.5 fold change, and TG62:1 had a 5.6 fold change. PCs, PEs with two fatty acid available showed a similar pattern of over C42. SMs, CERs, and LPCs with one fatty acid showed the same pattern in
3
C26:0-included lipids. Unlike the other lipids, PlsPC and PlsPE, which are the output of peroxisomal catabolism of lipids containing VLCFA, did not showed any big changes. These lipid differences between control and X-ALD groups suggests that ABCD1 defect induces the synthesis of VLCFA into various kinds of lipid species. In addition, unsaturated fatty acids were found to be highly abundant in HDF compared to both AMN and CCALD (Supplementary Fig. S12). A comparison regarding the unsaturated fatty acids between control and X-ALD groups can be found in Supplementary Information. 3.2. AMN highly activates the sphingolipid metabolism from sphingomyelin to sphingosine-1-phosphate Multivariate analysis of the lipid levels between AMN and CCALD was conducted to reveal the compositions of lipid molecules that played an important role in underlying lipidomic mechanisms. As shown in Fig. 2a, we obtain a PLS-DA score plot, in which the component 1 (26.6% variables) and 2 (11.6% variables) exhibit a relatively well separated two groups. The goodness-of-fit (R2) of 0.96 and the goodness-of-prediction (Q2) of 0.89 of the model and a 1000-time permutation test with p-value under 0.001 revealed the robustness of the current analysis (Supplementary Fig. S1). Significantly, 13 out of 15 CERs turned out to differentiate two phenotypes, having variable importance for projection (VIP) value over 1 (Supplementary Table S2). In loading plot of PLS-DA, CERs were highly headed toward AMN group, which showed a higher expression of CERs in AMN (Fig. 2b). This observation once again was confirmed by the univariate analysis (Fig. 2c). It is also worth mentioning that, although C24DH CER (dihydroceramide C24) and C26 CER (Ceramide C26) did not have VIP value over 1, they were also differentially expressed (adjusted p-value < 0.05). Altogether, these results suggest that CER and the related sphingolipid metabolism may be the highly relevant signatures and are responsible for the phenotypic variations in XALD. The metabolic regulators that are responsible for sphingolipid metabolism are described in Fig. 2d. In our transcriptome analysis, those regulators were classified into two categories: regulators related to the synthesis of complex structures from original backbone, such as sphingosine-1-phosphate (sphingolipid anabolism in Fig. 2e) and regulators related to the metabolism of glycosphigolipid and SM to its backbone (sphingolipid catabolism to sphingolipid-1-phosphate). In the anabolism, SGMS2, UGCG, CERT, LASS6, LASS1, and UGT8 were more down-regulated in AMN than CCALD. Especially, SGMS2, UGCG, and UGT8, which involved in the utilization of CER for the synthesis of SM and glycosphingolipid, showed a profound difference. This suggested that the higher levels of CERs in AMN might come from the alleviated expression levels of these two anabolic regulators. In contrast, SMPD1 and SMPDL3A, which are responsible for the hydrolyzation of SM to generate CER, were relatively lower in AMN compared to CCALD. Other metabolic regulators that belong to the catabolism showed higher expressions compared to CCALD whether those regulators are up- or downregulated in comparison to HDF. Collectively, transcriptome expression from the simple to complex structure of sphingolipid was highly decreased in AMN, and especially SGMS2, UGCG, and UGT8 might cause the accumulation of CER. 3.3. The anabolic pathway of triacylglycerol is up-regulated in AMN In contrary to the up-regulation in the anabolism of sphingolipid pathway (from sphingosine-1-phosphate to sphingomyelin), CCALD possessed a considerable deficiency in the anabolic pathway of TG species. Indeed, as being shown in the loading plot of previous
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
4
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Fig. 1. The alterations of lipidomes in 10 lipid species in control (HDF) and two X-ALD forms (AMN and CCALD). a) Heatmap of the expression of all lipid species (total 227 lipids). The normalized expression of each lipid in HDF (n ¼ 14), AMN (n ¼ 21), and CCALD (n ¼ 26) are shown in scale range from green to red. b) PLS-DA score plot of the whole lipidome data. The observations of HDF, AMN and CCALD are colored as gray, green and red, respectively. c) The relative alterations of distinctive lipid species (PE, SM, TG, CER, LPC, and PC). The fold change was shown as mean with s.e.m. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
PLS-DA of Fig. 2a, TGs with the shorter the FA chains were headed to AMN (Fig. 3a). It implied that TGs with shorter-chain FAs, which typically exist and be utilized in healthy human cells, were highly abundant in the mild phenotype of X-ALD. Those statistically significant alterations were selectively found in C48 to C58 TG species and the expression level of TG was generally higher in AMN than that of CCALD (Fig. 3b). Furthermore, given that the significant alterations in X-ALD groups were started from C60, properly owing to the VLCFA accumulation, the above observed patterns suggested that the expression of typical TGs was specifically impaired in CCALD, the severe phenotype of X-ALD. To further elucidate the modulating mechanisms in TG anabolism and catabolism, we assessed the metabolic enzymes and other
regulators of TG pathway (Fig. 3c). The expression levels of those transcriptomes were categorized into the anabolism and the catabolism, respectively (Fig. 3d). Most of genes involved in TG synthesis showed relatively lower expressions in CCALD. GK, which plays an essential role in coupling phosphate groups to glycerol, showed a relatively small diminution than CCALD; and GPAM, GPAT2, and GPAT4, which binds one acyl chain on glycerol-3phosphate, were highly expressed in AMN. LCLAT1, MBOAT1, MBOAT2, AGPAT2, AGPAT3, and AGPAT4, which help attach more acyl chain on 1-acyl-sn-glycerol-3-phosphate, were up-regulated in AMN. The gene expression levels of MOGAT2, MOGAT3, and LPIN1, three out of six genes related to DG synthesis, were increased. DGAT2 and MOGAT3, which generate TG from DG, were
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
5
Fig. 2. Ceramides as distinctive markers between AMN and CCALD, and altered expression of transcriptomes in the sphingolipid pathway. a) PLS-DA score plot shows a relative clear separation between two phenotypes of X-ALD. The observations of AMN (n ¼ 21) and CCALD (n ¼ 26) are colored green and red, respectively. b) Loading plots of PLS-DA of two groups. Ceramides are colored red. c) Relative abundances (HDF as 100%) of every ceramide in AMN and CCALD. Data are shown as mean with s.e.m.; *p < 0.05, **p < 0.01, and ***p < 0.001. d) Sphingolipid pathway and corresponding regulators. e) The expression of enzymes and metabolic regulators of the sphingolipid pathway. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
fairly up-regulated in AMN. Collectively, AMN had higher expressions of the key metabolic regulators involved in the TG anabolic pathway. Reversely, CCALD had a clear deficiency in the TG synthesis. Nonetheless, in TG catabolism, the pathway from TG to its backbone, revealed very slight alterations of the regulators. Some of them had relatively lower expressions (CEL, PNPLA2, PNLIP, PLRP1, PLRP2, DGKA, DGKB, DGKG, DGKQ, and AGK), in line with the accumulation of TG. Therefore, our systematic analysis revealed the impairment of TG anabolism in CCALD, triggered by the downregulated expression of transcriptomes in TG synthesis, and the impairment of the typical TGs might biochemically be associated with the severity in CCALD patient. 4. Discussion We conducted large-scale and systematic omics analysis to investigate the abnormal cellular metabolism at the lipidome and transcriptome levels. As a proof-of-concept, we successfully found distinctive candidates, PUFAs, CERs, and TGs, among the three phenotypes and elucidated the relationship between the abnormal metabolism of X-ALD and the lipid regulatory pathways from multi-layered datasets. Our findings provide a profound information-rich background regarding the patterns of lipidome abnormalities and altered expressions of corresponding metabolic regulators in different forms of X-ALD. At first, considerable aberrants in VLCFAcontaining phospholipids (PE, PC, and LPC), sphingolipids (CER and SM) and glycerolipids (DG and TG) were revealed. VLCFA turned out to be coupled to various lipids species, except PlsPC and PlsPE, since these two species are synthesized by peroxisomal lipid metabolism using fatty alcohol as substrate unlike the
other species [18]. This observation shows a good agreement with previous studies and considerably expands the combinable lipid species. Novel signatures that distinguish AMN and CCALD and their associations to sphingolipid and triacylglycerol metabolism were, to the best of our knowledge, firstly elucidated in our study. Interestingly, most of the markers in those metabolic pathways are decreased in CCALD, and are not related to the lipids containing VLCFA, rather to the lipids abundantly existing and physiologically essential in healthy control. These results suggest that the dysregulations of abundant and typical lipids utilized for sustaining lipid metabolic homeostasis may be involved in the severity of X-ALD. Transcriptome data further supported the investigation for those pathways. In sphingolipid pathway, the anabolism toward the more complex sphingolipids including sphingomyelin and glycosphingolipid in AMN, were down-regulated, and reversely, the catabolism were up-regulated in AMN. These lipidomic and transcriptomic alterations together revealed that sphingolipid pathways do not maintain the homeostasis in CCALD due to the reduction of the lipid catabolism. This imbalance in sphingolipid homeostasis were known as sphingolipidoses, a class of lipid metabolic disorder. Interestingly, our results about this pathway were quantitatively similar to the one of the sphingolipidoses, namely Farber disease, a type of disease showing ceramide accumulation by ASAH1 mutation [19]. In our transcriptome data, the ASAH1 expression was also decreased accordingly to the severity of the X-ALD type (Fig. 2d). The lipidomic similarity with Farber disease may help explain the different level of neurological toxicity between AMN and CCALD. Moreover, ceramide accumulation in AMN can be explained by the reduced anabolism from ceramide to glycosphingolipid (UGCG) or sphingomyelin (SGMS2), and reduced
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
6
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Fig. 3. The alterations of triacylglycerol differentiating AMN and CCALD. a) Loading plot of PLS-DA model. Triacylglycerols are the boxes colored in red (C48) to yellow (C64). b) Relative abundances (HDF as 100%) of every triacylglycerol in AMN and CCALD. Data are shown as mean with s.e.m.; *p < 0.05, **p < 0.01, and ***p < 0.001. c) Triacylglycerol metabolism and corresponding regulators. e) The expression of enzymes and metabolic regulators of the triacylglycerol metabolism. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
sphingolipid trafficking process by slight decrease in CERT expression [20,21]. Upregulated triacylglycerol metabolism in AMN is also a distinctive phenomenon in our study. TG species are well known as a protective effect against fatty acid-induced lipotoxicity [22]. Physiologically, TG keeps some excess saturated or monounsaturated fatty acids in lipid droplets. This process protects the cells against the lipotoxicity of VLCFA, including saturated and mono-unsaturated types, and prevents the induction of necroptosis in neural cells like oligodendrocytes and astrocytes [23,24]. Furthermore, the accumulation of TG containing PUFA (PUFA-TG) and their localization in lipid droplet limits the toxicity by preventing the PUFA-phospholipid from undergoing oxidation and increasing oxidative stress [25,26]. Our results revealed that PUFATGs with typical acyl chains are predominantly significant and having bigger fold change than saturated lipid with same carbon number. The PUFA-TG accumulation in AMN might help reduce cellular toxicity from ABCD1 defects, but this process might not be active in CCALD. However, specific roles of PUFA-TGs in X-ALD should be explored in future studies. In conclusion, systematic analysis combining lipidomics and transcriptomics revealed the difference between major X-ALD phenotypes by focusing on the metabolism of typical lipids, rather than VLCFA alone. Surprisingly, we found novel dysregulated metabolism of polyunsaturated fatty acids in X-ALD groups. Furthermore, the quantitative alterations of lipids and transcriptomes between AMN and CCALD largely derived from distinctive pathways in sphingolipid and triacylglycerol metabolism. Our study comprehensively describes the collapse of the lipid homeostasis, especially in typical lipids that are
physiologically essential for healthy control. The findings may partly explain why X-ALD patients suffer from a wide spectrum of the symptoms. Our systematic approach could greatly facilitates the understanding of the pathophysiology of X-ALD. Future studies are warranted to overcome the current limitations and extend our observation for linking essential lipids and their related metabolic regulators. Conflicts of interest The authors declare no conflict of interest. Acknowledgement This work was supported by Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (NRF2018R1D1A1B07046207), and the National Research Foundation of Korea grant funded by the Korean government (MSIP) (NRF2018R1A5A2024425). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbrc.2018.11.123. Transparency document Transparency document related to this article can be found online at https://doi.org/10.1016/j.bbrc.2018.11.123.
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123
D.-K. Lee et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
References [1] J.M. Powers, Y. Liu, A.B. Moser, H.W. Moser, The inflammatory myelinopathy of adreno-leukodystrophy: cells, effector molecules, and pathogenetic implications, J. Neuropathol. Exp. Neurol. 51 (1992) 630e643. [2] J. Singh, S. Giri, Loss of AMP-activated protein kinase in X-linked adrenoleukodystrophy patient-derived fibroblasts and lymphocytes, Biochem. Biophys. Res. Commun. 445 (2014) 126e131. [3] I. Singh, A.E. Moser, S. Goldfischer, H.W. Moser, Lignoceric acid is oxidized in the peroxisome: implications for the Zellweger cerebro-hepato-renal syndrome and adrenoleukodystrophy, Proc. Natl. Acad. Sci. U.S.A. 81 (1984) 4203e4207. [4] D. Trompier, C. Gondcaille, G. Lizard, S. Savary, Regulation of the adrenoleukodystrophy-related gene (ABCD2): focus on oxysterols and LXR antagonists, Biochem. Biophys. Res. Commun. 446 (2014) 651e655. [5] J. Berger, S. Forss-Petter, F.S. Eichler, Pathophysiology of X-linked adrenoleukodystrophy, Biochimie 98 (2014) 135e142. ~ iga, M. Lees B, Fatty acid composition of human myelin [6] O. Bizzozero, A.,G. Zun proteolipid protein in peroxisomal disorders, J. Neurochem. 56 (1991) 872e878. [7] M. Igarashi, H.H. Schaumburg, J. Powers, Y. Kishimoto, E. Koilodny, K. Suzuki, Fatty acid abnormality in adrenoleukodystrophy, J. Neurochem. 26 (1976) 851e860. [8] F. Kallabi, E. Ellouz, M. Tabebi, G. Ben Salah, N. Kaabechi, L. Keskes, C. Triki, H. Kamoun, Phenotypic variability in a Tunisian family with X-linked adrenoleukodystrophy caused by the p.Gln316Pro novel mutation, Clin. Chim. Acta 453 (2016) 141e146. [9] S. Kemp, F. Valianpour, S. Denis, R. Ofman, R.-J. Sanders, P. Mooyer, P.G. Barth, R.J.A. Wanders, Elongation of very long-chain fatty acids is enhanced in Xlinked adrenoleukodystrophy, Mol. Genet. Metabol. 84 (2005) 144e151. [10] C. Theda, A.B. Moser, J.M. Powers, H.W. Moser, Phospholipids in X-linked adrenoleukodystrophy white matter: fatty acid abnormalities before the onset of demyelination, J. Neurol. Sci. 110 (1992) 195e204. [11] P. Sharp, D. Johnson, A. Poulos, Molecular species of phosphatidylcholine containing very long chain fatty acids in human brain: enrichment in X-linked adrenoleukodystrophy brain and diseases of peroxisome biogenesis brain, J. Neurochem. 56 (1991) 30e37. [12] S. Beloribi-Djefaflia, S. Vasseur, F. Guillaumond, Lipid metabolic reprogramming in cancer cells, Oncogenesis 5 (2016) e189. [13] I. Stagljar, The power of OMICs, Biochem. Biophys. Res. Commun. 479 (2016) 607e609. [14] J. Jang, S. Park, H. Jin Hur, H.-J. Cho, I. Hwang, Y. Pyo Kang, I. Im, H. Lee, E. Lee,
[15]
[16]
[17] [18]
[19] [20]
[21]
[22]
[23] [24]
[25] [26]
7
W. Yang, H.-C. Kang, S. Won Kwon, J.-W. Yu, D.-W. Kim, 25-hydroxycholesterol contributes to cerebral inflammation of X-linked adrenoleukodystrophy through activation of the NLRP3 inflammasome, Nat. Commun. 7 (2016) 13129. N.P. Long, K.H. Jung, S.J. Yoon, N.H. Anh, T.D. Nghi, Y.P. Kang, H.H. Yan, J.E. Min, S.S. Hong, S.W. Kwon, Systematic assessment of cervical cancer initiation and progression uncovers genetic panels for deep learning-based early diagnosis and proposes novel diagnostic and prognostic biomarkers, Oncotarget 8 (2017) 109436e109456. ~ averas, S. Lo pez, J.V. Castell, M.T. Donato, A. Lahoz, Extending J.C. García-Can metabolome coverage for untargeted metabolite profiling of adherent cultured hepatic cells, Anal. Bioanal. Chem. 408 (2016) 1217e1230. J. Xia, I.V. Sinelnikov, B. Han, D.S. Wishart, MetaboAnalyst 3.0dmaking metabolomics more meaningful, Nucleic Acids Res. 43 (2015) W251eW257. H.R. Waterham, S. Ferdinandusse, R.J.A. Wanders, Human disorders of peroxisome metabolism and biogenesis, Biochim. Biophys. Acta Mol. Cell Res. 1863 (2016) 922e933. Y.-H. Xu, S. Barnes, Y. Sun, G.A. Grabowski, Multi-system disorders of glycosphingolipid and ganglioside metabolism, J. Lipid Res. 51 (2010) 1643e1675. K. Kumagai, C.A. Elwell, S. Ando, J.N. Engel, K. Hanada, Both the N- and Cterminal regions of the Chlamydial inclusion protein D (IncD) are required for interaction with the pleckstrin homology domain of the ceramide transport protein CERT, Biochem. Biophys. Res. Commun. 505 (2018) 1070e1076. Y. Ishibashi, M. Ito, Y. Hirabayashi, Regulation of glucosylceramide synthesis by Golgi-localized phosphoinositide, Biochem. Biophys. Res. Commun. 499 (2018) 1011e1018. L.L. Listenberger, X. Han, S.E. Lewis, S Cases, R.V. Farese, D.S. Ory, J.E. Schaffer, Triglyceride accumulation protects against fatty acid-induced lipotoxicity, Proc. Natl. Acad. Sci. U.S.A. 100 (2003) 3077e3082. L.R. Parisi, N. Li, G.E. Atilla-Gokcumen, Very long chain fatty acids are functionally involved in necroptosis, Cell Chem. Biol. 24 (2017) 1445e1454, e1448. €nfeld, S. Kahlert, G. Reiser, Toxic effects of X-linked S. Hein, P. Scho adrenoleukodystrophy-associated, very long chain fatty acids on glial cells and neurons from rat hippocampus in culture, Hum. Mol. Genet. 17 (2008) 1750e1761. N. Li, Y. Sancak, J. Frasor, G.E. Atilla-Gokcumen, A protective role for triacylglycerols during apoptosis, Biochemistry 57 (2018) 72e80. E. Jarc, A. Kump, P. Malavasi c, T.O. Eichmann, R. Zimmermann, T. Petan, Lipid droplets induced by secreted phospholipase A2 and unsaturated fatty acids protect breast cancer cells from nutrient and lipotoxic stress, Biochim. Biophys. Acta Mol. Cell Biol. Lipids 1863 (2018) 247e265.
Please cite this article as: D.-K. Lee et al., Integrative lipidomic and transcriptomic analysis of X-linked adrenoleukodystrophy reveals distinct lipidome signatures between adrenomyeloneuropathy and childhood cerebral adrenoleukodystrophy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2018.11.123