Most lipids in the central nervous system (CNS) reside in myelin membranes in the brain and spinal cord. Myelin sheaths surround and insulate axons to facilitate efficient, saltatory signal transduction. Myelin is formed by the extension of an oligodendrocyte lipid bilayer, which pinches off and wraps around the axon to form a multilamellar myelin sheath. Myelin lipids include cholesterol, phospholipids, glycolipids, and sphingomyelin. 1 Demyelination occurs when an oligodendrocyte or myelin is damaged and the myelin sheath degrades, which leads to a release of myelin debris including myelin lipids into the CNS parenchyma. In order for remyelination to occur, the myelin lipid debris is removed from the lesion site through microglia phagocytosis. 2 Little is known about how myelin lipids are metabolized after phagocytosis; it is likely that some lipids are directly recycled into new cellular or myelin membranes, while others may undergo hydrolysis or further metabolism. This study will begin to address the question of what happens to CNS lipids after demyelination by quantifying how the overall CNS lipidome changes with demyelination and remyelination.
Lipidomics is a powerful tool that enables the profiling of lipid dynamics in demyelination, which has the potential to provide insights into how lipids are regulated during myelin damage and repair. Previous studies performed in gliotoxin models of demyelination, such as the cuprizone model, have revealed changes in the brain lipidome during demyelination. [3][4][5] However, the interpretation of these previous studies is limited by fast remyelination, which can occur simultaneously with demyelination in the cuprizone model. 6 Rapid remyelination also makes it challenging to obtain longitudinal lipidomics data from distinct phases of demyelination and remyelination.
The goal of this study is to obtain longitudinal data on how the lipidomes of brain, spinal cord, and serum change during demyelination and remyelination. To achieve this goal, we have used a genetic mouse model of demyelination and remyelination (Plp1-iCKO-Myrf), 7 which enables us to directly compare changes in the multiple CNS tissues at the same time. In this model, tamoxifen induced ablation of Myrf (myelin regulatory factor) in mature oligodendrocytes (Plp1 promoter) induces widespread demyelination in the brain, spinal cord, and optic nerve. Myrf encodes a transcription factor required for the expression of myelin genes and the maintenance of a healthy myelin sheath, 8 and the induced loss of Myrf results in the degradation of the myelin. Myrf is not deleted in oligodendrocyte precursor cells (OPCs), and the OPCs are available to proliferate, differentiate into oligodendrocytes, and remyelinate the CNS. Newly formed oligodendrocytes are observed by 6 weeks in the optic nerve 7 and by 10 weeks in the brain, 9 and the formation of new cells is followed by a discrete phase of remyelination. 7,9 The Plp1-iCKO-Myrf model also features loss of motor function as measured by clinical scoring and rotarod analysis. Motor impairment begins around 5 weeks post-tamoxifen, peaks during 10-14 weeks, and shows gradual partial recovery over 14-24 weeks. 7,9 The clinical signs of motor loss and recovery track with the demyelination and remyelination that have been quantified in the brain by histology. 9 In this study, we used the Plp1-iCKO-Myrf mouse strain to perform unbiased lipidomics to profile brain, spinal cord, and serum during demyelination and remyelination. We also performed lipidomics on Plp1-iCKO-Myrf mice treated with Sob-AM2, which is a thyroid hormone agonist [10][11][12] that has improved remyelination and motor recovery in Plp1-iCKO-Myrf mice. 9 The lipidomics data have provided several new insights into CNS lipid dynamics during myelin damage and repair. The brain lipidome showed substantial changes that were the greatest near peak demyelination and showed partial normalization with remyelination. In contrast, the spinal cord lipidome was greatly altered at all time points including those associated with motor recovery and remyelination in the brain. The differences between the brain and spinal cord lipidomes correlated with histological analysis revealing that the spinal cord has poor remyelination in this model relative to the brain. In addition, we observed that demyelination phases were associated with both decreased and increased lipid species, but that later time points associated with remyelination did not have any elevated lipid levels. This study represents the first longitudinal profile detailing how CNS lipids change during the time course of myelin damage and repair in multiple tissues simultaneously.
Male and female C57BL6/J Plp1-iCKO-Myrf mice (Myrf f l/fl; Plp1-CreERT) 7 were bred at the University of Kansas by crossing Myrf(fl/f l) 8 mice with Plp1-CreERT 13 mice. All mice were housed in a climate-controlled room (24 ± 1 °C) with a constant 12 h light/dark cycle (12 h on, 12 h off) with food and water ad libitum. At 8 weeks of age, mice were injected intraperitoneally with 2 mg of tamoxifen (100 μL of 20 mg/mL tamoxifen) in corn oil daily for five consecutive days. Both Cre negative mice, which do not lose Myrf and undergo demyelination, and Cre positive mice, which do undergo demyelination, were used in all experiments. Mice were randomly assigned into the control or Sob-AM2 treatment groups. The control group received control chow (Envigo Teklad 2016 diet), and the Sob-AM2 treatment group received chow containing 420 μg/kg chow of Sob-AM2 (nominal daily dose of 84 μg/kg body weight) starting 2 weeks after tamoxifen injections. Mice on control chow were euthanized at four time points: 6, 12, 18, and 24 weeks post-tamoxifen injections. Mice treated with Sob-AM2 were euthanized at 12 and 18 weeks post-tamoxifen. Euthanasia was performed by carbon dioxide inhalation and cervical dislocation, and serum, brain, and spinal cord tissues were collected immediately, frozen on dry ice, and stored at -80 °C until processing. Mice for histological analysis were euthanized by intracardiac perfusion with HBSS (Hanks' Balanced Salt Solution) buffer followed by 4% paraformaldehyde in HBSS. All experiments were approved by the Institutional Animal Care and Use Committee at the University of Kansas.
All mice were trained on the rotarod 2-4 days before the first experiment. Training consisted of three 2 min trials at 8 rpm with 15 min rest intervals between trials. Motor function testing was performed at 6, 12, 18, and 24 weeks posttamoxifen using a rotarod (IITC Life Science). Testing was performed with three 5 min trials with the following program: 0:00-4:00, ramp from 8 to 40 rpm; 4:00-5:00, held at 40 rpm. The mice were allowed to return to their cage for at least 15 min between trials. If mice stayed on the rod for the full 5 min, then 300 s was recorded as the latency. If mice held onto the rod and rotated around once, then that was counted as a fall. Time to fall (latency) was recorded for the three trials and averaged for each mouse at each time point. Statistical analysis was performed to compare Cre negative (n = 20) and Cre positive (n = 24) mice at each time point using unpaired t tests corrected for multiple comparisons using the Holm-S ̌i ́daḱ method.
Free floating sections (40 μm) of brain and lumbar spinal cord were obtained using a Vibratome (Leica VT 1200S). Sections were mounted on positively charged slides and dried overnight at room temperature. Black-Gold staining was performed according to the manufacturer's protocol with the following modifications. Mounted sections were rehydrated in deionized water for 2 min. Then the slides were transferred into a 0.3% Black-Gold II (Histo-Chem) solution for 3-5 h until desired myelin impregnation was observed. Then the slides were rinsed with deionized water and transferred into a 1% sodium thiosulfate solution for 3 min. Slides were again rinsed with deionized water three times. Sections were dehydrated for 30 s in each ethanol solution (50, 75, 85, 95, 100%). Finally, the slides were immersed in xylene for 1-2 min and secured with coverslips using Cytoseal XYL mounting media (Thermo Fisher). Mounted slides were heated on a slide warmer for 2-3 h at 60 °C. Sections were imaged using a slide reader (BioTek Cytation 5) at 4× magnification. Multiple images were taken and stitched together to obtain the whole image.
Black-Gold images were analyzed by threshold analysis to determine a percent myelin staining within the tissue. Each time point was stained in a separate batch and represents n = 3 mice. Thresholds were set for each batch using Cre negative images to encompass the major white matter tracts (0-21 000 for week 12 and 0-17 500 for week 24 in brain and 0-26 000 for week 12 and 0-28 000 for week 24 in spinal cord). Direct comparisons were only performed on sections stained in the same batch. Statistical analysis was performed to compare Cre negative and Cre positive tissues at each time point using unpaired t tests corrected for multiple comparisons using the
Holm-S ̌i ́daḱ method. Brains were analyzed from Cre negative mice (n = 5 at week 12 and n = 3 at week 24) and Cre positive mice (n = 8 at week 12 and n = 6 at week 24). Spinal cords were analyzed from Cre negative mice (n = 6 at week 12 and n = 4 at week 24) and Cre positive mice (n = 13 at week 12 and n = 5 at week 24).
A modified Bligh-Dyer protocol was used to extract lipids from the brain and spinal cord tissue. 14 Brain homogenates (either 300 or 50 mg/mL) were prepared with ice cold water using a Bead Mill homogenizer (Bead Ruptor Elite, Omni International, USA). Spinal cord homogenates were prepared at 65 mg/mL in cold water. Immediately after homogenization, the brain homogenates were diluted with cold water (150 μL of 300 mg/mL brain homogenate was mixed with 800 μL of cold water. For the 50 mg/mL brain homogenates, 1 mL of homogenate was used directly. Spinal cord homogenates were diluted 10-fold (100 μL of the spinal cord homogenates with 900 μL of cold water). For all samples, 10 μL of the diluted homogenate was removed and stored at -80 °C for protein quantification using a BCA assay. The diluted tissue homogenates were combined with a mixture of chloroform (containing 0.01% butylated hydroxytoluene, BHT):methanol:water (3:2:1) in glass tubes. After shaking and vortexing thoroughly, the mixture was centrifuged (Sorvall ST 40R Centrifuge, Thermo Fisher Scientific) at 1300 rpm for 10 min. The lower layer was carefully removed and saved in a glass tube. The remaining top layer was further extracted twice with 1.25 mL of chloroform with 0.01% BHT; the lower layers were carefully removed and combined. The combined lower layer was then washed with 300 μL of 1 M KCl followed by 300 μL of water and vacuum-dried completely (Savant SpeedVac SPD130DLX vacuum concentrator, Thermo Fisher Scientific, USA) to obtain the dried lipid extract.
Serum (3 μL) was added directly to a vial containing 1.2 mL of chloroform:methanol:300 mM ammonium acetate in water (300:665:35). The contents were mixed thoroughly, and the vials were centrifuged for 5 min at low speed in a clinical centrifuge to pellet proteins before submission of samples to the mass spectrometer.
Quantification of phospholipids was performed by direct infusion triple quadrupole mass spectrometry on a Sciex 4000 QTrap at the Kansas Lipidomic Research Center at Kansas State University. Briefly, dried lipid samples (brain and spinal cord) were dissolved in 1 mL of chloroform and serum samples were used directly as prepared above. Aliquots were mixed with internal standards and solvents, and analysis was carried out as previously described. 15 Internal standards are indicated in Table S1. Lipid measurements were performed with the use of the acquisition and data processing parameters indicated in Table S1. LPE, PE, PA, PI, and PS were detected with neutral loss scans, and LPC and PC were detected with precursor ion scans (Table S1). For each lipid class, the fatty acid component was identified based on the number of total acyl carbons and total double bonds, but the individual fatty acids in diacylated lipids, their positions on the glycerol, and double bond positions were not determined. Sample mass spectra for each class are included in Figure S1.
Internal standards were from the same class as the analytes, and corrections for differences in response between the internal standards employed and SPLASH Lipidomix (Avanti Polar Lipids) were applied. For sphingomyelin (SM) and phosphatidylcholine (PC), PC internal standards were used and a correction for the response of SPLASH SM versus our PC internal standards was applied. No additional corrections for variation in response of the instrument to individual analytes versus their standards were applied. Thus, data were reported as normalized mass spectral intensity where a value of 1 indicates the intensity of 1 nmol of internal standard (Tables S2-S4).
The mass spectrometric analyses of the samples from the 6, 12, and 18 week groups (including control and Sob-AM2 treatment groups) were analyzed separately from the 24 week group. Brain, spinal cord, and serum were analyzed on separate days. To account for differences in the data due to the day of analysis, samples from the 12 and 18 week groups were analyzed again during the 24 week analysis and the levels were used to normalize the 24 week data. Serum samples from 24 weeks were not analyzed.
The number of samples analyzed were the following: for Cre negative, 6 weeks n = 3, 12 weeks n = 5, 18 weeks n = 6, and 24 weeks n = 4; for Cre positive, 6 weeks n = 8, 12 weeks n = 8, 18 weeks n = 8, and 24 weeks n = 7; for Cre negative with SobAM2, 12 weeks n = 6 and 18 weeks n = 5; for Cre positive with SobAM2, 12 weeks n = 8 and 18 weeks n = 9. Data were analyzed both as individual lipid species and as total lipids, in which the different fatty acyl derivatives were summed (Tables S2-S4). Quality controls were prepared by pooling a portion of all samples for a given analytical run. Then five quality control samples were analyzed at intervals during the sample run. The coefficient of variation (CV) of the quality control for each lipid species was determined (Tables S2-S4). Any lipid species with a CV higher than 0.3 was not included in the subsequent data analysis; this cutoff is based on FDA guidance for biomarker studies. 16 Similarly, all lipid totals with a CV higher than 0.3 were removed from further analysis except for total lysophosphatidylethanolamine (LPE), which had a sufficient CV in the brain, but not the spinal cord. Since the CV of total LPE was near 0.3 in the spinal cord (0.41 during weeks 6, 12, and 18, 0.38 for week 24), we included the data for comparison to the brain. The raw data will be available at the NIH Common Fund's National Metabolomics Data Repository (NMDR) website, the Metabolomics Workbench. 17 It has been assigned Study ID ST002958 and can be accessed directly via the Project DOI: 10.21228/M8W13F.
The data from five mice (for all tissues) were excluded from data analysis due to errors in genotyping. One spinal cord sample was excluded from data analysis due to a mistake during the lipid extraction procedure.
The total lipid levels for Cre negative and Cre positive mice were first compared at all four time points using the Holm-S ̌i ́daḱ correction method for multiple t tests. In addition, Cre negative mice on control chow were compared to Cre negative mice treated with SobAM2 chow and Cre positive mice on control chow were compared to Cre positive mice treated with SobAM2 chow using the Holm-S ̌i ́daḱ correction method for multiple t tests. The results of the statistical analyses including P values are contained in Tables S5-S7.
was also performed on all weeks 6-18 brain samples grouped by phenotype: healthy (all Cre negative), demyelination (weeks 6 and 12 Cre positive), and remyelination (week 18 Cre positive). To analyze the effects of Sob-AM2 treatment, PCA was performed with data sets that included (1) Cre negative mice administered control chow and Cre negative mice administered SobAM2 chow at week 12 or 18 and (2) Cre positive mice administered control chow and Cre positive mice administered SobAM2 chow at week 12 or 18. In addition, PCA was also performed on all week 12 samples (both genotypes and both treatment conditions) and all week 18 samples (both genotypes and both treatment conditions). The analyses were performed using R Statistical Software (ver. 4.2.3) and the factoextra package (ver. 1.0.7). 18,19 The code is available in the Supporting Information.
The individual lipid species were analyzed in the following groups: Cre negative versus Cre positive for all time points, Cre negative mice on control chow versus Cre negative mice on SobAM2 chow at week 12 or 18, and Cre positive mice on control chow and Cre positive mice on SobAM2 chow at week 12 or 18. The data was log 2 transformed prior to analysis, and lipid species that included values of "0" were dropped from the data at each time point. The fold change (FC) and P values from individual t tests were determined for all comparisons, and volcano plots were prepared plotting -Log 10 (P-value) on the y-axis versus Log 2 (FC) on the x-axis. The Log 2 (FC) was set at 0.5 and the P-value rejection threshold was determined for each time point using a permutation-based false discovery proportion estimate method. 20 The rejection thresholds were determined through the R package permFDP (ver. 0.1.0) 21 and are listed in Tables S8-S10. The volcano plots were generated using the R package enhanced Volcano (ver. 1.13.2). 22 The individual FC and P-values for all comparisons are included in Tables S8-S10, and the lipids that meet the FC and P-value thresholds are listed in Tables S11-S13.
Plp1-iCKO-Myrf mice were allowed to undergo normal developmental myelination prior to the induction of demyelination at 8 weeks of age. Demyelination was initiated by five daily doses of tamoxifen, which deleted the Myrf gene by conditionally activating Cre recombinase in mature oligodendrocytes (Plp1 promoter) 7 but not in OPCs. Previous studies in the model suggested that demyelination in the brain began around week 5 post-tamoxifen 7 with the greatest loss of brain myelin observed at week 10 post-tamoxifen. 9 Demyelination correlated with increasing loss of motor function, and from weeks 10-14 post-tamoxifen, the mice showed maximum motor loss with significant hindlimb weakness. Formation of new oligodendrocytes was observed in the brain by week 10 D post-tamoxifen with steady improvements in motor ability starting at week 14 post-tamoxifen. 9 The time course of demyelination and remyelination in the brains of Plp1-iCKO-Myrf mice was previously characterized by histology; 9 however, spinal cords were not examined. To determine the effects of demyelination in the spinal cord, myelin was stained in both brain and spinal cord tissue sections using Black-Gold (Figure 1A,B) from mice euthanized at peak motor disability (week 12) and motor recovery (week 24). Threshold analysis was used to quantify the percentage of Black-Gold staining for each tissue section (Figure 1C). Thresholds were set for each time point comparison based on inclusion of white matter tracts in the healthy (Cre negative) tissues. The brain showed demyelination at week 12 (39% of Cre negative levels) followed by remyelination at week 24 (66% of Cre negative levels). 9 In striking contrast to the brain, spinal cord showed similar levels of demyelination at week 12 and week 24 (30 and 27% of Cre negative levels, respectively), suggesting that remyelination is impaired in the spinal cord relative to the brain. The contrast between the brain and spinal cord provides a unique opportunity to compare the lipidomes of robust remyelination in the brain with impaired remyelination in the spinal cord.
Motor ability was assessed with the use of an accelerating rotarod motor test (Figure 1D) at weeks 6, 12, 18, and 24 posttamoxifen. Demyelination decreased motor function as mice had significantly reduced latency, or time spent on the rotarod, at all time points. The mice experienced the greatest loss of motor function at week 12 with partial recovery of motor function observed by week 24. These results were consistent with previous observations in this model. 9
On the basis of the histological results in Figure 1, we predicted that the brain and spinal cord would show different lipidomic profiles after induction of demyelination in the Plp1-iCKO-Myrf mice. Brains and spinal cords were isolated from both Plp1-iCKO-Myrf mice at 6, 12, 18, and 24 weeks posttamoxifen in two genotypes: Cre negative (healthy, no demyelination) and Cre positive (demyelination). These time points represented active demyelination (6 weeks), peak demyelination (12 weeks), and remyelination (18 and 24 weeks). The tissues were homogenized in water and extracted using a modified Bligh-Dyer protocol. 14,15 The tissues were analyzed by a lipidomics platform that included 12 major classes of lipids (Tables S2-S4). For each class, individual lipids were measured (ranging 14:0-22:0 for monoacylated lipids and 28:1-44:2 for diacylated lipids). Quality controls prepared from pooled samples were measured five times, and any individual lipids that had high coefficients of variation in the quality controls (>0.3) were excluded from subsequent analyses.
The individual lipids for each class were summed, and the totals were plotted for the four time points to visualize how the overall lipidomes of the brain (Figure 2) and spinal cord (Figure 3) change with demyelination and remyelination. The lipid levels for each sample are in Tables S2 andS3, and statistical analyses are available in Tables S5 andS6.
Analysis of Figures 2 and3 revealed major trends that are similar for both brain and spinal cord. First, ether-linked phosphatidylethanolamine (ePE), phosphatidylserine (PS), and phosphatidic acid (PA) were reduced at most or all time points. ePE, PS, and PA are signaling lipids in the central nervous system, 23,24 but little is known about the exact role of these lipids in myelin and how these lipids are affected with demyelination. In our data, demyelination caused chronic reductions in these lipid classes that were not restored even F with robust remyelination in the brain. Reduced PS and PA levels were observed in a recent study as markers of disease progression from primary progressive multiple sclerosis to secondary progressive multiple sclerosis. 25 However, further studies are required to understand the regulation of these signaling lipids during myelin repair.
Next, we observed that phosphatidylethanolamine (PE), phosphatidylinositol (PI), and sphingomyelin (SM) decreased with demyelination (weeks 12 and 18) and increased to normal levels at later time points (week 24). These are structural lipids that promote interactions within the myelin membranes to maintain compact and more stable myelin. 1,26 Recovery of structural lipids with remyelination provides further support for the important role of these lipids in myelin. In contrast, phosphatidylcholine (PC), lysophosphatidylcholine (LPC), and lyso-PE (LPE) were mostly unaffected by demyelination. This may imply that these lipids are less abundant in myelin relative to other lipid classes.
Finally, ether-linked PC (ePC) in the brain and ether-linked PS (ePS) in both the brain and spinal cord were elevated during active demyelination (week 6) had normalized or decreased levels at later time points. Ether-linked lipids, or plasmalogens, have relatively low abundance and may act as signaling molecules. 27 Plasmalogens have been implicated both as mediators of inflammation and as potential antioxidants, and the role of plasmalogens in neurological disease remains unclear. 28 Further studies will be needed to understand how ePC and ePS are modulating demyelination and remyelination in the Plp1-iCKO-Myrf model.
In addition to identifying trends in lipid classes, the total lipid analysis enabled us to directly define how the lipidomes of brain and spinal cord respond to demyelination (Figures 2 and3). Myelin lipids comprise a larger portion of the spinal cord, which is consistent with our observations that the spinal cord showed more dramatic fold changes in almost every lipid class as compared to the brain. Histological analysis revealed that the brain showed robust remyelination (Figure 1A) as compared to poor remyelination in the spinal cord (Figure 1B). The impaired remyelination in the spinal cord is also clear from the lipidomic results. In the brain, 8 of 11 lipid classes are at normal levels by week 24, whereas in the spinal cord only 4 of the 11 lipid classes are at normal levels at week 24. In general, it appears that structural lipids have returned to normal levels by week 24 in the brain (PC, PI, PE, and SM) and in the spinal cord (PC and SM). The structural lipids are major components of all cell types, and the results reflect the total CNS lipid content. In contrast, signaling lipids (PA, PE, and ePS) were still underrepresented at week 24 in the brain and spinal cord, which is intriguing and suggests that there may be some more lasting molecular signaling changes in the CNS following demyelination.
The total lipid data in Figures 2 and3 were further analyzed by PCA to determine if tissue that has undergone demyelination can be distinguished from healthy tissue based on changes to the lipidome. Unsupervised PCA was performed to classify the data at each time point in the brain (Figure 4, top) and spinal cord (Figure 4, bottom). The PCA score plots were annotated with 95% confidence clusters to group the samples. The healthy (Cre negative) and demyelination (Cre positive) groups clustered distinctly at all time points in both brain and spinal cord. The PCA score plot of brain at week 24 showed the least discrimination between the two groups, which is consistent with the robust remyelination and recovery of most lipid classes by this time point in the brain. PCA biplots shown in Figures S2 andS3 further illustrate lipid class trends, variability, and contribution to each principal component.
The data in Figures 1,3, and 4 provide support for the assertion that remyelination in the spinal cord was impaired relative to remyelination in the brain in the Plp1-iCKO-Myrf mouse model. It is important to note the Plp1-iCKO-Myrf mouse model features >90% demyelination in the brain and spinal cord simultaneously, and that this does not mimic the focal demyelination lesions observed in multiple sclerosis. The underlying mechanism for why the spinal cord shows poor remyelination relative to the brain is not yet clear. Despite the limitations of this mouse model, the observation of the impaired remyelination in the spinal cord may have clinical relevance. Brain and spinal cord remyelination differences have previously been observed in a study with primary and secondary multiple sclerosis patients. 25 Incomplete remyelination in the spinal cord was associated with disease severity, but S2 andS3). The genotypes cluster distinctly at all time points with the brain at week 24 showing the least discrimination. G this was not observed in the brain tissue. These observations align with the impaired spinal cord remyelination and lipidome recovery observed in our study.
Additional PCA was performed to determine if demyelination could be discriminated from remyelination in the brain. This analysis was performed only on weeks 6, 12, and 18 brain samples as they were all analyzed by mass spectrometry on the same day, and the brain had more robust remyelination than the spinal cord. The samples were divided into the following three groups based on phenotype: 1, healthy (all Cre negative, weeks 6, 12, and 18); 2, demyelination (Cre positive, weeks 6 and 12); and 3, remyelination (Cre positive, week 18). PCA analysis revealed that the three groups clustered separately with clear contrast between healthy, demyelinating, and remyelinating lipidomes (Figure 5). The demyelination group showed slightly more spread, which is consistent with the fact that this group contains mice at two different demyelination time points (weeks 6 and 12). Interestingly, the remyelinating lipidome grouped halfway between the normal and demyelinated lipidome groups.
To verify that the PCA clustering was not an artifact of mass spectrometry sample order, the same set of samples was grouped into three sets based on sample order during the analysis. As shown in Figure S4, PCA did not discriminate between the three groups and there was greater overlap. These results support the assertion that differences observed between phenotypes in Figures 4 and5 are based on biological differences and not based on mass spectrometry sample order.
To identify changes to individual lipid species at each time point, we prepared volcano plots showing all lipids that met the coefficient of variation cutoff. The Log 2 (fold change) was set at 0.5, and the -Log 10 (P-value) cutoff thresholds (Tables S8 andS9) were determined using a permutation-based false discovery proportion estimation method. 20 The individual lipids identified as significantly changing are listed in Tables S11 andS12.
Examining the volcano plots revealed that increased levels of individual lipids were primarily observed at earlier time points during demyelination whereas decreased levels were observed S2). S8 andS9. at all time points (Figure 6). These trends also supported the observation that the lipidome of the brain showed changes consistent with remyelination, whereas the lipidome of the spinal cord had more chronic changes associated with demyelination. In the brain, lipids with increased levels were only observed in weeks 6 and 12, whereas in the spinal cord elevated lipids were observed at all time points (Table 1). The higher number of increased lipids is consistent with the more persistent demyelination observed in the spinal cord and suggests that the elevated lipid species may be biomarkers associated with active demyelination states. Further support for remyelination in the brain comes from the number of reduced lipids. In the brain, 41 lipids were reduced at peak demyelination (week 12), whereas 79 lipids were reduced in the spinal cord at the same time point. By week 24, the brain had only 21 differentially identified lipids, whereas the spinal cord still had more than 60.
To better catalog which lipids were changing and how they were changing, we prepared Table 2, which contains the fold change for all individual lipids that are identified as significant in the volcano plot analyses and change in both the brain and the spinal cord. The brain has 62 individual lipids that change in at least one of the four time points, and the spinal cord has 121 individual lipids that change in at least one of the four time points. Of the 62 lipids that change in the brain, 58 also change in the spinal cord data set, and these are included in Table 2. The high overlap between the identities of the altered lipids between brain and spinal cord suggests that the observed changes are biologically relevant and related to the demyelination phenotype. Figure S5 contains the fold change data for the lipids that are changed only in the brain or the spinal cord.
The volcano plots (Figure 6) and Table 2 provide a different perspective that complements the analysis performed on the total lipid levels (Figures 234). For example, the total level of ePS was elevated in both the brain and spinal cord at 6 weeks (Figures 2 and3); however, no individual ePS lipids were identified in the brain at week 6 and only two were identified in the spinal cord (ePS 36:1 and 38:2). Another example is that several PS species were elevated at week 6 in the brain (1.4-2.3 fold) and in the spinal cord (1.7-10.1 fold), but total PS was not significantly elevated. Table 2 also provides further insights into the observation that certain lipid species are increased during active demyelination. At week 6, almost all lipids that were increasing are unsaturated, and many are highly unsaturated with three, four, or five double bonds (observed in PC, ePC, PS, and ePS). It has been recently recognized that polyunsaturated fatty acids may serve as precursors for a class of lipids known as specialized proresolving mediators that reduce oxidative stress and inflammation, and lead to improved tissue regeneration. 29,30 Further studies will be needed to determine if unsaturated lipids are mediating remyelination in the Plp1-iCKO-Myrf mice.
Several observations from our data set are consistent with results from other studies in different demyelination models. A limited lipidomics panel focusing on phosphatidylcholine (PC) in cuprizone treated mice revealed a decrease in PC 36:1 in the corpus callosum and cortex, and this decrease was also observed in post-mortem multiple sclerosis brain tissue. 5 We also observed that PC 36:1 was decreased at demyelinating time points in the brain (weeks 6 and 12) and spinal cord (weeks 6, 12, and 18). Another study used stereotaxic injection of lysolecithin into the corpus callosum to induce demyelination and analyzed the lipids in the lesion site with desorption electrospray ionization-MS imaging. They primarily observed changes to PCs and PEs. For example, PC 36:1 is also decreased in this study, but they observe changes in several other PC and PE species that are distinct from our results. 4 One confounding factor is that the lysolecithin model relies on an injection of lysolecithin (or LPC), which may lead to lipid alterations that are unique to the lysolecithin detergent used in this model.
Other studies in the cuprizone model have examined more lipid classes during remyelination revealing persistent changes in the prefrontal cortex after demyelination and remyelination. The largest changes included increased PS and decreased LPE. 3,31 We also observed increases in PS during demyelination, but these increases diminished at later time points. We did not observe major changes in LPE, and overall, we observed few changes in lysolipids, which contain a single acyl chain. Most changes were in the diacylated phospholipids (PS, PE, etc.) and ether-linked phospholipids (ePS, ePE, etc.).
Sob-AM2 is a CNS-penetrating thyroid hormone agonist that has been shown to promote remyelination in the brains of Plp1-iCKO-Myrf mice. 9 To define how the lipidome is altered by treatment with this remyelinating agent, the Plp1-iCKO-Myrf mice were administered chow containing Sob-AM2 (84 μg/kg nominal daily dose) starting at week 2 post-tamoxifen, a Sob-AM2 treatment method that was used and validated previously. Mice were euthanized at week 12 and week 18, and brain and spinal cord tissues were analyzed by the lipidomics panel at the same time as the untreated Cre negative and Cre positive samples.
To identify drug-related changes to the lipidome, the total lipid levels were analyzed with pairwise comparisons (Cre negative on control chow versus Cre negative on SobAM2 chow and Cre positive on control chow versus Cre positive on SobAM2 chow) at weeks 12 and 18. Sob-AM2 treatment elicited significant changes in several brain lipid classes (Figure 7) but had a minimal impact on lipids in the spinal cord (Figure S6).
Our goal was to identify Sob-AM2 induced lipid changes that may be relevant to the remyelinating effects of the drug. Four classes (PE, ePE, PS, and PI) all showed similar and interesting trends. Above, we noted that all four classes were decreased in the brains experiencing demyelination during week 12 and week 18 (Figure 2). With the addition of Sob-AM2 treatment, we observed increases in the total levels of PE, ePE, PS, and PI at week 18 (Figure 7). Critically, this only occurred in Cre positive mice that had undergone demyelination; drug treatment had no effect in the absence of In the brain, weeks 6 and 12 represent demyelination time points and weeks 18 and 24 represent remyelination time points. In the spinal cord, all weeks have significant demyelination. Lipids in rows shaded blue were increased during at least one time point, whereas lipids in unshaded rows were decreased at all time points.
demyelination. Sob-AM2 treatment effectively normalizes the amounts of PE, PS, and PI, which correlates with Sob-AM2 inducing remyelination in the brain by 18 weeks posttamoxifen. This also suggests that total PE, PS, and PI levels may correlate with both demyelination and remyelination. This is further corroborated by the results in the spinal cord (Figure 3), which showed substantial drops in PE and PS during all measured time points.
To further confirm the drug-induced changes on the lipidome and their relevance to drug-induced remyelination, we performed PCA on the four groups (Cre positive and Cre negative with and without Sob-AM2) in brain and spinal cord at week 12 and week 18 (Figure 8 and Figure S7). In the PCA score plots, the lipidomes of brains from Cre positive mice treated with Sob-AM2 cluster separately from those of Cre positive mice that received control (Figure 8A). This supports the assertion that remyelination induced by Sob-AM2 correlates with an altered lipidomic profile.
In contrast to the brain, Sob-AM2 had a limited effect on the lipidome of the spinal cord. This was supported both by the total lipid analysis (Figure S6) and by the grouped PCA score plot (Figure 8B and Figure S7). The PCA showed that control and Sob-AM2 treated groups clustered identically for Cre negative and Cre positive mice. Previously, it was shown that Sob-AM2 showed good efficacy in brain remyelination in the Plp1-iCKO-Myrf mice, 9 but the effect of Sob-AM2 on the spinal cord was not thoroughly characterized. The lack of a Sob-AM2 treatment effect in the spinal cord is explained in part by the extensive demyelination observed in the spinal cords of Plp1-iCKO-Myrf mice (Figure 1), which could limit the efficacy of any remyelination therapy. Further research is necessary to understand why the remyelination in the spinal cord is impaired relative to the brain in the Plp1-iCKO-Myrf model.
Although changes were observed in the brain with Sob-AM2 treatment at the total lipid level (Figure 7) and by PCA groupings of the total lipid levels, volcano plot analysis on individual lipid data revealed very few (<5) lipids that were altered with Sob-AM2 treatment in the brain or spinal cord (Figure S8).
Changes in serum lipid levels have potential therapeutic applications as peripheral biomarkers of CNS disease. To determine if any serum lipids were altered during demyelination, serum samples were collected and analyzed from the groups at weeks 6, 12, and 18. In the total lipid analysis, only one significant change was observed: LPC was reduced by 17% at week 12 (Figure 9A and Figure S9). In addition, PCA score plots poorly discriminated between demyelination and S2 andS3). remyelination in the serum at week 12 but showed no discrimination at week 6 or 18 (Figure 9B and Figure S10). The weak discrimination at week 12 is due primarily to the differences observed in the total LPC levels. However, no changes to individual lipids were significant in the volcano plot analysis (Figure S11). The reduction in total serum LPC should be further validated in future studies; however, there is precedent in the literature for alterations to lysolipids in inflammatory demyelinating disease. LPC can be converted to lysophosphatidic acid (LPA) by the action of the enzyme autotaxin, 32 which is known to be upregulated in multiple sclerosis. 33 Elevations in LPA have also been observed in serum and cerebrospinal fluid samples from people with multiple sclerosis. 34 LPA was not measured in this study, but the observed reduction in LPC correlates with previous findings and supports the further investigation of LPC/LPA/ autotaxin as peripheral biomarkers for demyelinating neurological diseases.
Finally, we analyzed how treatment with Sob-AM2 affected the serum lipidome. Sob-AM2 treatment led to many more changes in the serum lipidome (Figures S12-S14) as compared to the brain and spinal cord. Sob-AM2 is a thyroid hormone agonist, and thyroid hormone is a global regulator of lipid homeostasis in the periphery. 35 Therefore, it is not surprising that Sob-AM2 would have greater effects in serum. A recent study examined the serum lipidomes of patients with hyper-and hypothyroidisms, and they identified lipids related to glycerophospholipid metabolism as potential biomarkers associated with thyroid dysfunction. 36 We observed that Sob-AM2 treatment primarily altered the levels of the PC, ePC, LPC, and PI classes in both the total lipid analysis (Figures S12 andS13) and in the individual lipid volcano plots (Figure S14 and Table S13). Most changes observed with Sob-AM2 treatment are observed in both Cre negative (healthy) and Cre positive (demyelination) samples, suggesting that these changes are likely not directly related to the therapeutic action of Sob-AM2 in the CNS as a remyelinating drug.
This study represents the first longitudinal study that measures the lipidomes of demyelination and remyelination in multiple CNS tissues (brain and spinal cord) and in the periphery (serum). The brain and spinal cord tissues have distinct lipidomic profiles consistent with robust remyelination in the brain and impaired remyelination in the spinal cord. In contrast, the serum showed very few alterations with demyelination. The differences between brain and spinal cord were observed through PCA score plots using total lipid levels and through volcano plot analysis of individual lipids.
Early stages of demyelination were marked by increased total levels of ePS and ePC and by increases in lipid species containing unsaturated fatty acids. Chronic demyelination, in contrast, was associated with persistent reductions in several lipid classes including PE and PS. Treatment with Sob-AM2, a drug known to promote remyelination, normalized the levels of PE and PS in the brain, suggesting that these lipids may be potential biomarkers for tracking demyelination and remyelination in the CNS. Future studies will focus on elucidating the biological mechanisms underlying lipid alterations that were observed in this study with the goal of revealing how lipids are regulated during demyelination and remyelination.
Downloaded via UNIV OF KANSAS on April 23, 2024 at 20:21:49 (UTC).See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
https://doi.org/10.1021/acs.jproteome.3c00443J. Proteome Res. XXXX, XXX, XXX-XXX
J. Proteome Res. XXXX, XXX, XXX-XXX
https://doi.org/10.1021/acs.jproteome.3c00443J. Proteome Res. XXXX, XXX, XXX-XXX I