Lipidomics reveals triacylglycerol accumulation due to impaired fatty

Article pubs.acs.org/jpr

Cite This: J. Proteome Res. 2019, 18, 2779−2790

Lipidomics Reveals Triacylglycerol Accumulation Due to Impaired Fatty Acid Flux in Opa1-Disrupted Fibroblasts Cinzia Bocca,*,† Mariame Selma Kane,† Charlotte Veyrat-Durebex,†,‡ Judith Kouassi Nzoughet,† Juan Manuel Chao de la Barca,†,‡ Stephanie Chupin,‡ Jennifer Alban,† Vincent Procaccio,†,‡ Dominique Bonneau,†,‡ Gilles Simard,‡,§ Guy Lenaers,† Pascal Reynier,†,‡ and Arnaud Chevrollier† †

Equipe Mitolab, Institut MITOVASC, CNRS 6015, INSERM U1083, Université d’Angers, 49933 Angers, France Département de Biochimie et Génétique, Centre Hospitalier Universitaire, 49933 Angers, France § INSERM U1063, Université d’Angers, 49933 Angers, France

Downloaded via MOUNT ALLISON UNIV on August 4, 2019 at 12:59:18 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

‡

S Supporting Information *

ABSTRACT: OPA1 is a dynamin GTPase implicated in mitochondrial membrane fusion. Despite its involvement in lipid remodeling, the function of OPA1 has never been analyzed by whole-cell lipidomics. We used a nontargeted, reversed-phase lipidomics approach, validated for cell cultures, to investigate OPA1-inactivated mouse embryonic fibroblasts (Opa1−/− MEFs). This led to the identification of a wide range of 14 different lipid subclasses comprising 212 accurately detected lipids. Multivariate and univariate statistical analyses were then carried out to assess the differences between the Opa1−/− and Opa1+/+ genotypes. Of the 212 lipids identified, 69 were found to discriminate between Opa1−/− MEFs and Opa1+/+ MEFs. Among these lipids, 34 were triglycerides, all of which were at higher levels in Opa1−/− MEFs with fold changes ranging from 3.60 to 17.93. Cell imaging with labeled fatty acids revealed a sharp alteration of the fatty acid flux with a reduced mitochondrial uptake. The other 35 discriminating lipids included phosphatidylcholines, lysophosphatidylcholines, phosphatidylethanolamine, and sphingomyelins, mainly involved in membrane remodeling, and ceramides, gangliosides, and phosphatidylinositols, mainly involved in apoptotic cell signaling. Our results show that the inactivation of OPA1 severely affects the mitochondrial uptake of fatty acids and lipids through membrane remodeling and apoptotic cell signaling. KEYWORDS: membrane/fusion, mitochondria, fatty acid oxidation, lipid droplets, dominant optic atrophy, optic neuropathy

■

INTRODUCTION The human OPA1 protein is a ubiquitous dynamin GTPase1 mainly involved in the fusion of the mitochondrial inner membrane and in structuring the cristae. OPA1 intervenes directly or indirectly in several functions essential for cell survival, i.e. the control of apoptosis,1,2 energetic metabolism,3 calcium flow,4 maintenance of mitochondrial DNA integrity,5−7 oxidative stress,8 mitophagy,9,10 aging,11−13 and inflammation.11 It also plays an important pathophysiological role because OPA1 mutations are responsible for a large spectrum of neurological disorders sharing optic nerve degeneration and impaired visual acuity.14−17 To obtain a better understanding of the consequences of OPA1 dysfunction at the clinical and experimental levels, we used metabolomics on several OPA1 models. A targeted study of a haploinsufficient mouse model (Opa1+/-) revealed that of the nine tissues analyzed, only the optic nerve and the plasma presented a discriminating metabolomic signature, further illustrating the specific vulnerability of the optic nerve to OPA1 dysfunction.18 The optic nerve signature, perceptible before © 2019 American Chemical Society

the clinical occurrence of optic neuropathy, was characterized by decreased concentrations of sphingomyelins and phosphatidylcholines (in the mono and diacyl forms), indicative of myelin sheath alterations. The signature also showed modifications in the concentration of metabolites involved in neuroprotection such as dimethyl arginine, carnitine, spermine, spermidine, carnosine, and glutamate, suggesting a concomitant axonal metabolic dysfunction. With the same experimental approach, fibroblasts from OPA1+/- patients showed no discriminant metabolomic signature,19 unlike our findings in Leber’s hereditary optic neuropathy, which is related to mitochondrial DNA mutations. In contrast, using a nontargeted metabolomics approach, the analysis of plasma from individuals carrying pathogenic variants of OPA1 showed a discriminant signature,20 revealing the unexpected involvement of purine metabolism in OPA1 pathophysiology as well as a deficiency of aspartate and glutamate related to possible Received: January 31, 2019 Published: June 4, 2019 2779

DOI: 10.1021/acs.jproteome.9b00081 J. Proteome Res. 2019, 18, 2779−2790

Article

Journal of Proteome Research

Figure 1. Lipidomics workflow. Each chronological step of the identification, processing, and statistical analysis is indicated with the corresponding figure or table number in bold type. Key elements of the principal features are shown in italics.

d62- (16:0/16:0), Cer d31- (16:0/16:0), SM d31- (16:0/ 16:0), PE d31- (16:0/18:1), PI d31- (16:0/18:1), PG d31(16:0/18:1), and PS d31- (16:0/18:1) with >98% purity were acquired from Coger SAS (Paris, France). Antibodies (ab42364, EP1332Y, ab186695, and ab186696) were obtained from Abcam (Paris, France), tris-Glycine Gel from Life Technologies (Illkirch, France), and DMEM-F12 from Jacques Boy Institute of Biotechnology (Reims, France). The DMEM medium supplemented with fetal bovine serum (FBS) was acquired from PAN-biotech (Wimborne, UK) and the Mitotracker green from Molecular Probes (Oregon, United States).

energetic defects. Finally, the nontargeted metabolomic analysis of the complete inactivation OPA1 in Opa1−/− MEFs confirmed the operation of bioenergetic remodelling with a crucial alteration in the levels of aspartate, the nucleotides, and related metabolites.21 The signatures of these different models share similarities such as the deficiency of aspartate and glutamate, which in itself may be responsible for the metabolic impairment of nucleotide metabolism. Modifications of phospholipids were also recurrently observed together with changes in sphingomyelins, which play a determinant role in axonal protection. To further these metabolomic studies, we used a nontargeted lipidomic approach to compare Opa1−/− MEFs with their wildtype counterparts.

■

Cell Cultures

Immortalized mouse embryonic fibroblasts (MEFs) from Opa1−/− knockout (KO) C57BL/6 mice and Opa1+/+ wildtype (WT) controls were cultivated in Dulbecco’s modified Eagle’s medium with a nutrient mixture F12 (DMEM-F12) supplemented with 10% FBS at 37 °C, 5% CO2. The analyses of Opa1−/− MEFs and Opa1+/+ MEFs were all performed within the same exponential growth stage, and a supplementary flask per condition was used for cell counts.

EXPERIMENTAL PROCEDURES

Chemicals and Reagents

Methanol (MeOH), water, isopropanol (IPA), acetonitrile (ACN), formic acid (Optima LC/MS grade), and BODIPY C12 558/568 were purchased from Fisher Scientific (Illkirch, France). Ammonium formate for LC/MS analysis and chloroform (SupraSolv quality) was bought from VWR International Llc (Fontenay-sous-Bois, France). Ammonium bicarbonate with >99% purity was acquired from SigmaAldrich (St. Quentin Fallavier, France). Isotopically labeled metabolite standards, including TG d5- (14:0/16:1/14:0), PC

Lipidomics Analyses

Samples (n = 10 for each MEF cell line) were randomly prepared as follows. After removal of the medium, the cell layer was rinsed twice with an aqueous solution containing 0.22% NaCl before being quenched with a cold mixture of 2780


Article

Journal of Proteome Research Table 1. List of Lipids Detected in Opa1-Disrupted Fibroblasts category

main class

glycerolipids

diradylglycerols triradylglycerols

glycerophospholipids

glycerophosphocholines

glycerophosphoethanolamines

glycerophosphoglycerols glycerophosphoinositols

sphingolipids

glycerophosphoserines phosphosphingolipids

sterol lipids

ceramides neutral glycosphingolipids acidic glycosphingolipids sterols

class key

subclass

Nb detected

DG TG TGp TGe LPC LPCe PC PCe PCp PE PEe PEp PG PI PIp PS SM SMpO Cer CerG1 GM3 ChE

diacylglycerols triacylglycerols 1-alkenyl,2,3-diacylglycerols 1-alkyl,2,3-diacylglycerols monoacylglycerophosphocholines monoalkylglycerophosphocholines diacylglycerophosphocholines 1-alkyl,2-acylglycerophosphocholines 1-alkenyl,2-acylglycerophosphocholines diacylglycerophosphoethanolamines 1-alkyl,2-acylglycerophosphoethanolamines 1-alkenyl,2-acylglycerophosphoethanolamines diacylglycerophosphoglycerols diacylglycerophosphoinositols 1-alkenyl,2-acylglycerophosphoinositols diacylglycerophosphoserines sphingomyelins sphingomyelins + hydroxyl group ceramides simple Glc series gangliosides steryl esters

7 27 7 8 2 2 37 20 10 15 1 9 1 9 1 5 18 1 13 8 3 8

35 000 fwhm resolution with an AGC target of 2 × 105, a maximum IT of 125 ms, and a general normalized collision energy (NCE) of 30 eV with a stepped NCE at 50%. Chromatography was carried out using a Dionex UltiMate 3000 UHPLC (Dionex, Sunnyvale, CA, United States) equipped with a Phenomenex Kinetex (1.7 μm EVO C18, 150 × 2.10 mm, 100 Å) HPLC column kept at a temperature of 45 °C. A multistep gradient (preceded by an equilibration time of 3 min), starting with 68% of mobile phase A consisted of ACN/H2O (60/40) and 32% of mobile phase B of IsoP/ ACN (90/10), both containing 0.1% formic acid and 10 mM ammonium formate, was used. The flow rate was maintained at 0.26 mL/min during a total runtime of 30 min. The UHPLC autosampler temperature was set at 10 °C, and only 10 μL of each sample was injected.

ammonium bicarbonate at 155 mM and MeOH (67.7/32.3). The cell suspension, estimated at four million cells, obtained after mechanical scraping was then collected and stored at −80 °C until analysis in aliquots of one million cells. Internal quality controls (QCs) were generated by mixing the samples together before the extraction protocol and treated similarly. We used a nontargeted reversed-phase (RP) lipidomic method previously validated for cell cultures (Supplementary Table S1). Briefly, 10 μL of the isotope metabolite standards mixture (10 μg/mL in MeOH) was added with chloroform to each cellular suspension. After 2 h of agitation at 4 °C and centrifugation (2000g at 4 °C for 5 min), the organic phase was evaporated to dryness. The aqueous phase was submitted to a second extraction with a mixture of chloroform and MeOH (66.7/33.3) and then shaken for 1 h at 4 °C before centrifugation. Each organic and aqueous phase was evaporated to dryness, reconstituted, and pooled with an ACN/ Isop/H2O (65/35/5) solution prior to the UHPLC−HRMS analysis. A Thermo Scientific Q Exactive mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a heated electrospray ionization source (HESI II) was used for this study, which was carried out in both the positive and negative modes. The HESI II was operated at a spray voltage of 3.5 kV in the positive (+) mode and at 3.0 kV in the negative (−) mode with capillary temperatures of 250 °C (+) or 320 °C (−), respectively, a heater temperature of 350 °C, sheath gas flow of 35 arbitrary units, auxiliary gas flow of 10 (+) or 5 (−) arbitrary units, and spare gas flow of 1 (+) or 0 (−) arbitrary units. During the full scan acquisition, spectra were acquired from 120 to 1800 m/z, and the instrument was operated at 70 000 full width at half maximum (fwhm) resolution with an automatic gain control (AGC) target of 3 × 106 and a maximum injection time (IT) of 250 ms. MS/MS fragmentations were performed on several QCs at the beginning and at the end of the sequence. The isolation window was set at 1 m/z, and the instrument was operated at

Lipidomics Data Processing

Data analyses were performed according to the workflow diagram shown in Figure 1. Thermo Scientific LipidSearch 4.1 software (Thermo Fisher Scientific, Bremen, Germany), which used a lipid database containing more than 1.5 million lipid ions and their expected fragments combined with an identification algorithm, was used for the identification process (lipid class detected: Table 1). This software enabled naming the lipids as recommended by the Metabolomics Standards Initiative for putatively annotated compounds at identification level 222 according to their class and the total number of carbons and double bonds detected on the fatty acid (FA) chains (e.g., TG (66:8): a triglyceride with 66 carbons and eight double bonds on the FA chains). When the quality of the MS/MS fragmentation spectra was sufficiently enhanced, i.e. when the number of matches with product ion peaks in the spectrum was higher than 5 and more than 50% of fragments could be identified, the exact composition of the FA chains was given for each molecule (e.g., TG(20:4/20:4/26:0): a triglyceride with 2 chains of 20 carbons and 4 double bonds and an FA chain of 26 carbons and 0 double bonds). 2781


Article


ECLIPSE Ti-E (Nikon, Amsterdam, Netherlands) equipped with a 100× oil-immersion objective (Nikon Plan Apo100x, N.A. 1.45) and an EMCCD-Evolved camera controlled by Metamorph 7.7 software (Molecular Devices, Sunnyvale, CA, United States). Thirty-five image planes were acquired along the Z-axis at 0.1 μm increments. Following acquisition, images were first iteratively deconvoluted using Huygens Essential software (Scientific Volume Imaging, Hilversum, The Netherlands) with the maximum iteration scored at 50 and a quality threshold at 0.01. Volumetric rendering and spot detection functions were then performed using Imaris 8.0 software (Bitplane, Zurich, Switzerland).

Following the identification step, a TraceFinder 4.1 (Thermo Fisher Scientific, Bremen, Germany) processing method based on the identification list created with LipidSearch was used for peak detection with the following criteria of acceptance: quality control CV area below 30%, an isotopic pattern matching the chemical formula, and a linearity of dilution with an r2 value close to 1. Identical conditions were applied for each isotope lipid standards used in the protocol. For unequivocal identification, the lipids retained were then examined by class, checking for each that the number of carbons and unsaturations was coherent with the retention time (RT). When available, the RT was verified with the corresponding isotope standard. Before performing any statistical analyses, data were normalized by the MS total useful signal (MSTUS) of each sample using Microsoft Excel software. Moreover, the data set was log10 transformed, mean-centered, and scaled by the square root of the standard deviation of each variable (Pareto scaling) to reduce the contribution of the most intense ions. Multivariate analyses were performed with Simca-P+ v. 14.0 (Umetrics, Umea, Sweden). Principal component analysis (PCA), an unsupervised method, was used to investigate the population structure and to emphasize spontaneous clustering or separation between samples on the basis of their global lipid profiles. To highlight molecules implicated in the lipidomic signatures, orthogonal partial least-squares discriminant analysis (OPLS-DA), a supervised method, was carried out, and only metabolites that showed a strong power of discrimination in the model and highly statistical reliability were retained. More precisely, variables were gradually excluded according to the results obtained from different plots: the S-plot (p(corr)[1] vs p[1]), the loading column plot with jack-knife confidence intervals, the coefficient plot, and the variable importance in the projection (VIP) plot. The purpose was to minimize the risk of overfitting and reduce the variability of prediction. OPLS-DA simulations were crossvalidated by leaving out one-third of the samples and replicating three times. The qualities and performances of OPLS-DA models were evaluated using the Q2Ycum (goodness of prediction), the R2Ycum (goodness of fit) values, the cross validation-analysis of variance (CV-ANOVA), and the permutation test (evaluation of the risk of overfitting). Only metabolites with a VIP value greater than 1 were considered relevant in the lipidomics footprint. Univariate analyses were performed on MetaboAnalyst 3.5 using the Volcano plot module.23 Thus, only metabolites with a fold-change greater than 1.5 and a Wilcoxon test at the threshold value of p < 0.05 were considered. To further minimize the error rate, the Bonferroni correction for the familywise error rate was applied, and only molecules that remained significant were retained.

■

RESULTS

Detected Lipids

Using a nontargeted lipidomics workflow in the positive and negative ionization modes (Figure 1), we accurately detected and identified 212 lipids (Supplementary Table S2), among which the most important classes belonged to the group of glycerophosphocholines (PC, 33.49%), followed by triacylglycerols (TG, 19.81%), glycerophosphoethanolamines (PE, 11.79%), phosphosphingolipids (8.96%), and ceramides (6.13%). Other classes such as glycerophosphoserines (PS) or sterols (ChE) were less represented (1, a threshold considered to be discriminant. Univariate analysis associated with the Bonferroni correction showed that the proportions of all these metabolites also differed significantly between the two cell lines with an FC > 1.5. The 69 discriminant lipids identified are presented in Table 2. Thirty-four (49%) were triacylglycerols (81% of total TG

The Greater Level of Triacylglycerols in Opa1−/− MEFs Is Caused by the Altered Flux of Fatty Acids

The greater level of 34 triacylglycerols in the Opa1−/− MEFs led us to investigate the general flux of fatty acids in cells stained with Mitotracker green to visualize mitochondrial network (Figure 5A) and with BODIPY 558/568 C12 (BODIPY C12), a saturated fatty acid with a 12-carbon chain, to visualize the recruitment of fatty acids (Figure 5B and 5C). Cells were first labeled overnight for 20 h with 1 mM of BODIPY C12 before being placed in fresh medium, and the kinetics of fatty acid distribution was followed at 48 and 72 h of wash-out using wild-field/deconvolution fluorescence microscopy. In Opa1+/+ MEFs (Figure 5B), the BODIPY C12 probe, incorporated in many spherical structures corresponding to lipid droplets, spread quickly across the 2783


Article


Figure 4. Lipidomics signature of Opa1−/− versus Opa1+/+ MEFs. (A) Unsupervised PCA score plot of the two Opa1−/− (blue squares) and Opa1+/+ (green circles) MEF cell lines. There is a spontaneous separation on the t[1] axis related to the genotype. (B) Supervised OPLS-DA score plot of the two Opa1−/− (blue squares) and Opa1+/+ (green circles) MEF cell lines. The model, constructed with 179 molecules shown in panel C, is able to discriminate the Opa1 genotypes with the first t[1] axis. (C) Volcano plot showing the contribution of each metabolite to the model described in panel B according to the VIP and the p(corr) values. Molecules with a VIP value greater than 1 (69 lipids) were considered the most important for defining the signature of Opa1−/− MEFs.

filamentous mitochondrial network. In contrast, the Opa1−/− cells showed slower assimilation of the BODIPY C12 probe into the lipid droplets together with a very weak signal in the fragmented mitochondria labeled by Mitotracker green (Figure 5A). Interestingly, after an incubation of 20 h with the BODIPY C12 probe, the phase-contrast images overlaid on the red fluorescent images highlighted an accumulation of the probe on the plasma membrane, especially in the Opa1−/− genotype (Supplementary Figure S2B), indicating the increased binding of fatty acids on membrane lipids. These results have been confirmed with a counterstaining of the plasma membrane using a plasma membrane targeting probe (Supplementary Figure S3). Quantification of the mitochondrial labeling by the BODIPY C12 probe (Figure 5C) confirmed that the fatty acid distribution across mitochondria was initially significantly reduced in Opa1−/− MEFs compared to Opa1+/+ MEFs even though it increased later (p < 0.001). The number of lipid droplets per cell in Opa1+/+ MEFs was not modified over time, whereas in Opa1−/− MEFs, a significant increase occurred after 48 h of wash-out (Figure 5B). The quantification of the size of

the lipid droplets at 72 h revealed smaller values in Opa1−/− MEFs (0.7 ± 0.4 μm) than in their Opa1+/+ (3.1 ± 0.9 μm) counterparts (p < 0.001). Thus, our results disclose delayed incorporation of fatty acids due to an important lipidmembrane binding and poor exchanges between lipid droplets and mitochondria in Opa1−/− MEFs compared to Opa1+/+ MEFs (Figure 5D), a difference that may be related to the alteration of the mitochondrial network and to the reduced oxidation of fatty acids by mitochondria.

■

DISCUSSION After having characterized the metabolomic profile of the polar metabolites of Opa1−/− MEFs,21 we used the same strategy to focus on a comprehensive lipidomic analysis of these cells. Two hundred and twelve lipids were correctly measured in whole-cell extracts by the approach we recently developed.25 Univariate and multivariate statistical analyses showed that 69 (31.8%) of the 212 lipids discriminated between the Opa1−/− and the Opa1+/+ genotypes, highlighting a dramatic increase in triglyceride levels at the foreground of this signature. Indeed, 2784


Article

Journal of Proteome Research Table 2. Seventy Lipids Found Relevant in the Opa1−/− MEF Cell Line Signaturea identification compound TG(48:1p) TG(54:5e) TG(51:1p) TG(50:1p) TG(53:2p) TG(50:4) TG(56:6e) TG(50:0p) TG(52:4) TG(50:0e) TG(53:1p) TG(52:1p) PI(38:4p) TG(57:6e) TG(56:6) TG(56:5e) TG(55:5) TG(52:1e) TG(51:3) TG(52:3) TG(51:2) TG(50:3) TG(58:6e) TG(58:6) TG(54:3) TG(53:2) TG(50:1) TG(52:1) PC(38:5p) TG(49:3) TG(48:3) TG(49:2) TG(54:2) TG(52:2) TG(58:5e) TG(53:1) PC(34:2e) PC(34:1p) LPC(16:0e) PC(32:0e) PC(30:0e) PC(32:0p) Cer(d41:2) PC(31:0e) PC(36:4e) PC(34:0p) GM3(d40:1) PI(34:1) PE(40:4) PS(40:4) DG(36:1) GM3(d34:1)

suggested identification TG(18:0p/14:0/16:1) mix of TG(16:0e/16:0/22:5) and TG (18:0e/16:0/20:5) TG(18:0p/15:0/18:1) mix of TG(16:0p/16:0/18:1) and TG (18:0p/14:0/18:1) TG(18:0p/17:1/18:1) mix of TG(14:0/18:1/18:1) and TG (16:0/16:1/18:3) TG(16:0e/18:1/22:5) TG(16:0p/18:0/16:0) TG(16:0/18:1/18:3) TG(16:0e/18:0/16:0) TG(18:0p/17:0/18:1) TG(18:0p/16:0/18:1) PI(38:4p) TG(18:0e/17:1/22:5) TG(16:0/18:1/22:5) mix of TG(16:0e/18:0/22:5) and TG (18:0e/16:0/22:5) TG(18:1/17:1/20:3) TG(18:0e/16:0/18:1) mix of TG(15:0/18:2/18:1) and TG (16:1/18:1/17:1) TG(16:1/18:1/18:1) mix of TG(16:0/17:1/18:1) and TG (15:0/18:1/18:1) mix of TG(16:1/16:1/18:1) and TG (18:2/14:0/18:1) TG(18:0e/18:1/22:5) TG(18:0/18:1/22:5) TG(18:1/18:1/18:1) TG(17:0/18:1/18:1) TG(16:0/16:0/18:1) TG(18:0/16:0/18:1) PC(18:0p/20:5) TG(15:0/16:1/18:2) Mix of TG(16:1/16:1/16:1) and TG (16:0/14:0/18:3) TG(15:0/16:1/18:1) TG(54:2) TG(16:0/18:1/18:1) TG(18:0e/18:1/22:4) TG(18:0/17:0/18:1) PC(34:2e) PC(18:0p/16:1) LPC(16:0e) PC(16:0e/16:0) PC(30:0e) PC(16:0p/16:0) or PC(16:0e/16:1) Cer(d18:2/23:0) PC(16:0e/15:0) PC(16:0e/20:4) PC(18:0p/16:0) GM3(d40:1) PI(34:1) PE(18:0/22:4) PS(40:4) DG(18:0/18:1) GM3(d18:1/16:0)

multivariate analysis

univariate analysis % MS/MS matched

FC

p-value (after the Bonferroni correction)

VIP value

73 70

17.93 17.20

Lipidomics reveals triacylglycerol accumulation due to impaired fatty

Recommend Documents