Male c57BL/6N wild-type mice at postnatal day 75 (P75) were used for all experiments except for the differential analysis of myelin purified from brains immediately frozen after dissection compared to a post-mortem delay of 6 h at room temperature (Figure 4), for which female c57BL/6N wild-type mice at P56 were used. Mice were bred and kept in the animal facility of the Max Planck Institute of Experimental Medicine and sacrificed by cervical dislocation. For the procedure of sacrificing mice for subsequent preparation of tissue, all regulations given in the German animal protection law (TierSchG §4) are followed. Since sacrificing of rodents is not an experiment on animals according to §7 Abs. 2 Satz 3 TierSchG, no specific authorization or notification is required for the present work.

A myelin-enriched light-weight membrane fraction was biochemically purified from mouse brains by sucrose density centrifugation and osmotic shocks as recently described in detail (Erwig et al., 2019a). Mice were sacrificed by cervical dislocation at the indicated ages as three biological replicates per condition (n = 3). Protein concentration was determined using the DC Protein Assay Kit (Bio-Rad). Initial quality control by gel electrophoresis and silver staining of gels was performed as described (de Monasterio-Schrader et al., 2013; Joseph et al., 2019). Briefly, samples were separated on a 12% SDS-PAGE gel (1 h at 200 V) using the Bio-Rad system, fixated overnight in 10% [v/v] acetic acid and 40% [v/v] ethanol and then washed in 30% ethanol (2 × 20 min) and ddH₂O (1 × 20 min). For sensitization, gels were incubated 1 min in 0.012% [v/v] Na₂S₂O₃ and subsequently washed with ddH₂O (3 × 20 s). For silver staining, gels were impregnated for 20 min in 0.2% [w/v] AgNO₃/0.04% formaldehyde, washed with ddH₂O (3 × 20 s) and developed in 3% [w/v] Na₂CO₃/0.04% [w/v] formaldehyde. The reaction was stopped by exchanging the solution with 5% [v/v] acetic acid.

Gel-electrophoretic separation of myelin proteins with different pre-cast gel systems (Serva) was performed essentially as recently described in detail (Erwig et al., 2019a). Briefly, 1D separations were performed with 5 μg protein load before (pre-wash) or after (post-wash) subjecting myelin to consecutive high-salt and high-pH washing/centrifugation cycles as previously described (Werner et al., 2007; Jahn et al., 2013). Automated tryptic in-gel digestion of proteins in gel bands (Schmidt et al., 2013) and protein identification by LC-MS was performed as described (Ott et al., 2015). For 2D separations, myelin was first delipidated by methanol/chloroform precipitation and 300 μg protein was loaded on a 24 cm immobilized non-linear pH-gradient 3-12 strip (Serva) by active rehydration (Erwig et al., 2019a). Automated tryptic in-gel digestion of proteins in gel spots and protein identification by MALDI-TOF mass spectrometry was performed as described (Jahn et al., 2006; Werner et al., 2007).

In-solution digestion of myelin proteins according to an automated filter-aided sample preparation (FASP) protocol (Erwig et al., 2019a) and LC-MS-analysis by different MS^E-type data-independent acquisition (DIA) mass spectrometry approaches was performed as recently established for PNS myelin (Siems et al., 2020). Briefly, protein fractions corresponding to 10 μg myelin protein were dissolved in lysis buffer (1% ASB-14, 7 M urea, 2 M thiourea, 10 mM DTT, 0.1 M Tris pH 8.5) and processed according to a CHAPS-based FASP protocol in centrifugal filter units (30 kDa MWCO, Merck Millipore). After removal of the detergents, protein alkylation with iodoacetamide, and buffer exchange to digestion buffer [50 mM ammonium bicarbonate (ABC), 10% acetonitrile], proteins were digested overnight at 37°C with 400 ng trypsin. Tryptic peptides were recovered by centrifugation and extracted with 40 μl of 50 mM ABC and 40 μl of 1% trifluoroacetic acid (TFA), respectively. Combined flow-through were directly subjected to LC-MS-analysis. For quantification according to the TOP3 approach (Silva et al., 2006), aliquots were spiked with 10 fmol/μl of yeast enolase-1 tryptic digest or Hi3 EColi standard (Waters Corporation), the latter containing a set of quantified synthetic peptides derived from E. coli. Chaperone protein ClpB.

Nanoscale reversed-phase UPLC separation of tryptic peptides was performed with a nanoAcquity UPLC system equipped with a Symmetry C18 5 μm, 180 μm × 20 mm trap column and a HSS T3 C18 1.8 μm, 75 μm × 250 mm analytical column (Waters Corporation) maintained at 45°C. Peptides were separated over 120 min at a flow rate of 300 nl/min with a gradient comprising two linear steps of 3–35% mobile phase B (acetonitrile containing 0.1% formic acid) in 105 min and 35–60% mobile phase B in 15 min, respectively. Mass spectrometric analysis of tryptic peptides was performed using a Synapt G2-S QTOF mass spectrometer equipped with ion mobility option (Waters Corporation). UDMS^E analysis was performed in the ion mobility-enhanced data-independent acquisition mode with drift time-specific collision energies as described in detail (Distler et al., 2014a, 2016). Continuum LC-MS data were processed using Waters ProteinLynx Global Server (PLGS) and searched against a custom database compiled by adding the sequence information for yeast enolase 1, E. coli Chaperone protein ClpB and porcine trypsin to the UniProtKB/Swiss-Prot mouse proteome and by appending the reversed sequence of each entry to enable the determination of false discovery rate (FDR). Precursor and fragment ion mass tolerances were automatically determined by PLGS and were typically below 5 ppm for precursor ions and below 10 ppm (root mean square) for fragment ions. Carbamidomethylation of cysteine was specified as fixed and oxidation of methionine as variable modification. One missed trypsin cleavage was allowed. Minimal ion matching requirements were two fragments per peptide, five fragments per protein, and one peptide per protein. The FDR for protein identification was set to 1% threshold.

For post-identification analysis including TOP3 quantification of proteins, the freely available software ISOQuant¹ was used (Kuharev et al., 2015). Only peptides with a minimum length of seven amino acids that were identified with scores above or equal to 5.5 in at least two runs were considered. FDR for both peptides and proteins was set to 1% threshold and only proteins reported by at least two peptides (one of which unique) were quantified as parts per million (ppm) abundance values (i.e., the relative amount (w/w) of each protein in respect to the sum over all detected proteins). The Bioconductor R packages “limma” and “q-value” were used to detect significant changes in protein abundance by moderated t-statistics as described (Ambrozkiewicz et al., 2018; Siems et al., 2020). For proteome profiling of wild-type myelin by MS^E and UDMS^E, three independent experiments were performed, each with three biological replicates and sample processing with duplicate digestion and injection, resulting in a total of 12 LC-MS runs per experiment. Abundance values in ppm are given as averages of the four technical replicates per biological replicate and only proteins quantified in at least two out of three experiments are reported in the proteome resource (Supplementary Table S1). Proteins identified as contaminants from blood (albumin, hemoglobin) or hair cells (keratins) were removed from the list. Proteome profiling comparing wild-type myelin without and with post-mortem delay (Figure 4 and Supplementary Table S2) was performed with three biological replicates and duplicate digestion, resulting in a total of 6 LC-MS runs per condition. Data acquisition was performed in the DRE-UDMS^E mode (Siems et al., 2020) i.e., a deflection device was cycled between full (100% for 0.4 s) and reduced (5% for 0.4 s) ion transmission during one 0.8 s full scan, thereby providing a compromise between identification rates and dynamic range.

Published single-cell RNA-sequencing (scRNA-seq) gene expression matrices from datasets GSE60361 (Zeisel et al., 2015), GSE75330 (Marques et al., 2016), and GSE113973 (Falcão et al., 2018), were obtained from Gene Expression Omnibus (GEO) and analyzed using R package Seurat v3.1.0 (Butler et al., 2018; Stuart et al., 2019). Mature oligodendrocyte cell populations were selected from each dataset as specified in the results section and normalized gene counts were used for calculating average expression profiles across single cells. Bulk proteome and transcriptome datasets specified in the results section were used as supplied in the Supplementary Tables to the respective publications.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with dataset identifier PXD020007. Pie chart, scatter plots, volcano plot and heatmap were prepared in Microsoft Excel 2013 and GraphPad Prism 8. Area-proportional Venn diagrams were prepared using BioVenn (Hulsen et al., 2008)². Trans-membrane domains were predicted using TMHMM Server v. 2.0 (Krogh et al., 2001)³ and Phobius (Käll et al., 2007)⁴.

To purify CNS myelin, we applied an established protocol (Erwig et al., 2019a) to prepare a light-weight membrane fraction from the brains of healthy c57Bl/6N-mice at P75. Aiming to systematically identify myelin-associated proteins we used five complementary approaches as summarized in Figure 1A. As the most straightforward way of preparing myelin for proteomic analysis, we separated proteins by 1D-SDS-PAGE and sectioned the lane into 24 equally sized slices (Figure 1B), which we subjected to automated tryptic in-gel digest followed by LC-MS-analysis, thereby identifying 788 proteins (Supplementary Table S1). When we subjected myelin to an additional washing step of high-pH and high-salt conditions (Figure 1B) to deplete soluble and peripheral membrane proteins before 1D-SDS-PAGE-separation and mass spectrometry, we identified 521 proteins (Supplementary Table S1). To establish a reference map of myelin proteins including proteoforms, we increased the resolving power of protein separation by subjecting myelin to 2D-gel electrophoresis with isoelectric focusing (IEF) in the first and horizontal SDS-PAGE in the second dimension (Figure 1C). We stained the comprised proteins with colloidal Coomassie (CBB250), picked CBB250-labeled gel-plugs (i.e., protein spots) for automated tryptic in-gel digestion and identified the proteins by peptide mass fingerprint (PMF) and MS/MS-fragment ion mass spectra, both acquired on a MALDI-TOF mass spectrometer. We identified 181 non-redundant proteins from 352 spots (Supplementary Table S1). Thereby we expanded our previous myelin protein map (131 non-redundant proteins from 217 spots, Werner et al., 2007), mainly owing to increased resolution in the first dimension by utilizing longer IEF-strips with a wider pH-range (Erwig et al., 2019a). When comparing the proteins identified using the three gel-based approaches we found a total of 930 proteins with a fair overlap (Figure 1D).

Considering that contemporary gel-free, label-free proteomic approaches allow the simultaneous identification and quantification of proteins (Neilson et al., 2011; Distler et al., 2014b) we subjected myelin to a workflow of solubilization using ASB-14 and high-urea conditions, automated tryptic in-solution digest by filter-aided sample preparation (FASP), fractionation of peptides by nanoUPLC, and ESI-QTOF mass spectrometry. This workflow was recently established for peripheral myelin (Siems et al., 2020). Importantly, the utilized data-independent acquisition (DIA)-strategy with data acquisition in the MS^E-mode allows the simultaneous quantification and identification of all peptides entering the mass spectrometer, and thereby, when signal intensities are correlated with a spike protein of known concentration (TOP3 method; Silva et al., 2006; Ahrné et al., 2013) the reliable quantification of proteins based on peptide intensities. When subjecting myelin to LC-MS-analysis using MS^E we quantified 393 proteins (Supplementary Table S1; labeled in orange in Figure 1E) with a false discovery rate (FDR) of <1% and an average sequence coverage of 38.6%. Notably, MS^E quantitatively covered myelin proteins with a dynamic range of over four orders of magnitude parts per million (ppm), thereby allowing quantification of the exceptionally abundant PLP and MBP. When using the ultra-definition (UD)-MS^E-mode, in which the ion mobility option provides an orthogonal dimension of peptide separation after liquid chromatography and before mass measurement, we identified and quantified 809 proteins (Supplementary Table S1; labeled in blue in Figure 1E) with an average sequence coverage of 35.0%. UDMS^E thus identified about twice as many proteins as MS^E. However, the larger number of proteins identified by UDMS^E went along with a compressed dynamic range of about three orders of magnitude ppm, which is insufficient to reliably quantify the most abundant myelin constituents including PLP, MBP, and CNP. The data acquisition mode-dependent differences in both numbers of quantified proteins and dynamic range are best explained by UDMS^E achieving more efficient precursor-fragment ion alignment and precursor fragmentation upon ion mobility separation of peptides (Distler et al., 2014a, 2016) which causes a ceiling effect for the detection of exceptionally intense peptide signals and thus a compressed dynamic range as previously observed for PNS myelin (Siems et al., 2020).

When comparing the proteins identified by MS^E, UDMS^E and gel-based approaches we found a reasonably high overlap (Figure 1F). Comparison of the 1155 proteins identified in CNS myelin in the present study with those 309 identified >10 years ago with the methodological standards of that time (Jahn et al., 2009) shows a remarkably high overlap as well as an about three-fold increase in the number of identified proteins (Figure 1G). Notwithstanding that a number of the identified proteins will originate from other cellular sources that contaminate purified myelin, we believe that many of them are indeed low-abundant constituents of the non-compact compartments of myelin.

A comparison of the proteins identified in CNS myelin with the recently established PNS myelin proteome (Siems et al., 2020) confirms that numerous proteins are present in both, but also that many proteins were identified exclusively in either CNS or PNS myelin (Figure 1H). Together, the evolving technical standards of in-solution sample preparation and MS^E-type DIA mass spectrometry allows to comprehensively identify and quantify proteins in myelin. However, only MS^E (but not UDMS^E) provides a dynamic range suited to address the relative abundance of the exceptionally abundant PLP, MBP, and CNP. As importantly, the evolution of gel-free methods shifts the major workload in myelin proteome analysis from manual sample handling to data analysis, with much less hands-on time required when compared to gel-based approaches.

As MS^E provides the best possible dynamic range (Figure 1E), we evaluated the relative abundance of all 393 proteins identified in myelin by MS^E (Figure 2 and Supplementary Table S1). As per this dataset PLP constitutes 38% of the total myelin protein [±1% relative standard deviation (RSD)]. MBP, CNP, and MOG constitute 30% (±1%), 5% (±0.2%), and 1% (±0.03%) of the total myelin protein, respectively (Figure 2). However, the present assessment of CNS myelin by MS^E extends well beyond the most abundant myelin constituents, thus quantifying known myelin proteins including the tetraspan-proteins CLDN11, CD81, TSPAN2, PLLP, CD9, CD82, GPM6B, and GJC3, the immunoglobulin-domain containing cell-surface proteins MAG, NFASC, CNTN1, RTN4, CNTN2, CADM4, HEPACAM, JAM3, CD47, and OMG, the enzymes SIRT2, CA2, and ASPA, the cytoskeletal and cytoskeleton-associated proteins TUBB4, SEPT2, SEPT4, SEPT7, SEPT8, TPPP, ANLN, GSN, CFL1, and PADI2 as well as MOBP, BCAS1, NDRG1, opalin, and CRYAB (Figure 2). By MS^E, 46 known myelin proteins account for approximately 80% of the total myelin protein (Figure 2). The remaining 27% is constituted by 347 proteins not yet validated as myelin constituents by independent methods.

An increasing number of studies provides mRNA or protein abundance profiles of myelin or oligodendrocytes. To systematically compare the CNS myelin proteome with these profiles, we correlated our MS^E-dataset (Figure 2 and Supplementary Table S1) via the gene name entries with related datasets for which quantitative information is publicly available (Figure 3).

We first plotted the present MS^E and UDMS^E-datasets (Supplementary Table S1) against each other (Figure 3A). Considering that the same starting material has been assessed it is not unexpected that the datasets correlate well, as reflected by a correlation coefficient of 0.90 (Figure 3A). Most visibly diverging from the linear regression line are the most abundant myelin proteins PLP, MBP, CNP, MOG, and CLDN11, reflecting that the dynamic range of UDMS^E is compressed in the high ppm-range compared to that of MS^E (also see Figure 1E). We then compared the present MS^E-dataset with an independent myelin proteome dataset previously established by MS^E (Jahn et al., 2009). We calculated a somewhat lower correlation coefficient of 0.74 (Figure 3B), probably owing to the previous use of a predecessor mass spectrometer generation that provided a considerably lower dynamic range. Yet, in conjunction with the high overlap between the proteins identified in the present and the previous study (Jahn et al., 2009) (Figure 1G), myelin proteome analysis emerges as fairly robust across independently purified starting material and different generations of mass spectrometers. We next compared the MS^E-dataset to the proteome of acutely isolated O4-immunopositive oligodendrocytes (Figure 3C) as determined by label-free quantification (LFQ) using data-dependent acquisition (DDA) on an orbitrap mass spectrometer and MaxQuant-software (Sharma et al., 2015). The O4-antibody preferentially immunolabels oligodendrocytes at the progenitor (OPC) and pre-myelinating stages (Sommer and Schachner, 1981; Bansal et al., 1989; Goldman and Kuypers, 2015); the correlation coefficient was calculated as 0.16 (Figure 3C). A correlation coefficient of 0.02 (Figure 3D) was found when comparing the MS^E-dataset with the LFQ-intensity profile of O1-immunopositive primary oligodendrocytes after 4 days in vitro (DIV) (Sharma et al., 2015). The myelin proteome as determined here is thus more closely related to the proteome of acutely isolated O4-immunopositive oligodendrocytes than to that of O1-immunopositive primary oligodendrocytes 4 DIV.

We then compared the MS^E-dataset with various available mRNA-abundance profiles. When comparing the MS^E-dataset to the transcriptome of purified CNS myelin as determined by RNA-seq (Thakurela et al., 2016) we calculated a correlation coefficient of 0.27 (Figure 3E). Interestingly, the comparison between the MS^E-dataset and the RNA-seq-based transcriptome of oligodendrocytes immunopanned from the cortex using antibodies against MOG (Zhang et al., 2014) revealed a roughly comparable correlation coefficient of 0.31 (Figure 3F). Notably, MOG-immunopositivity labels myelinating oligodendrocytes, implying that the stage of oligodendrocyte differentiation must be considered when judging dataset correlations. It is thus not surprising that a somewhat lower correlation coefficient of 0.10 (Figure 3G) was calculated when comparing the MS^E-dataset with the RNA-seq-based transcriptome of acutely isolated O4⁺- oligodendrocytes (Sharma et al., 2015). Finally, we compared the MS^E-dataset to several scRNA-seq-based transcriptome datasets (Zeisel et al., 2015; Marques et al., 2016; Falcão et al., 2018). To this aim we calculated the mean transcript abundance as average count reads per unique molecular identifier (UMI) of the cells in those clusters that reflect mature oligodendrocytes. When comparing the MS^E-dataset to mature oligodendrocytes sorted from the mouse cortex and hippocampus [all 484 cells in clusters Oligo5 and Oligo6 in Zeisel et al. (2015)], we find a correlation coefficient of 0.24 (Figure 3H). Importantly, we find a roughly similar correlation coefficient when comparing the MS^E-dataset to mature oligodendrocytes sorted from 10 regions of the mouse CNS [all 2748 cells in clusters MOL1–MOL6 in Marques et al. (2016)] (Figure 3I) or to mature oligodendrocytes sorted from the spinal cord of mice [all 617 cells in clusters MOL2-Ct and MOL5/6-Ct in Falcão et al. (2018)] (Figure 3J).

Together, when judging correlations between large datasets evaluating mRNA and protein abundance profiles of oligodendrocytes and myelin, aspects to be considered include the method of sample preparation, the stage of oligodendrocyte differentiation and the methodology of analysis. Yet, roughly similar correlation coefficients were calculated when comparing the myelin proteome with various proteomic and transcriptomic approaches to the molecular profiles of oligodendrocytes.

Autopsy material from human patients and healthy-appearing controls is increasingly evaluated by systematic molecular profiling, as exemplified by the recent snRNA-seq-based assessment of oligodendroglial transcriptional profiles in multiple sclerosis patients (Jäkel et al., 2019). Notably, the use of autopsy material involves a post-mortem delay between the death of a subject and the collection of a biopsy. However, post-mortem delay may affect sample integrity and thus data validity. Considering that proteomic analysis of myelin in mice is usually performed upon freezing of samples immediately after dissection, we asked whether the myelin proteome can also be assessed upon post-mortem delay. We thus purified myelin from the brains of c57Bl/6N-mice to compare the myelin proteome between mice after a post-mortem delay of 6 h at room temperature with that of mice upon sample freezing immediately after dissection. Upon SDS-PAGE-separation and silver staining, no signs of major degradation were evident and the band patterns appeared essentially similar (Figure 4A). We next subjected myelin to routine differential proteome profiling by UDMS^E with dynamic range enhancement (DRE-UDMS^E) (Supplementary Table S2). Using this data acquisition mode with intermediate features as to identification rates and dynamic range (for methodological details see Siems et al., 2020) we found that known myelin proteins displayed only minor differences as visualized in a volcano plot (red data points in Figure 4B) and a heatmap (Figure 4C). Indeed, no known myelin protein exceeded the threshold of a log₂-fold transformed fold-change (FC) of −1/+1, i.e., a 2-fold increased or 0.5-fold diminished relative abundance. Together, the myelin proteome displays only minor changes upon a post-mortem delay of 6 h, implying that proteomic assessment of myelin purified from autopsy samples appears feasible.