Conservation and divergence of myelin proteome and oligodendrocyte transcriptome profiles between humans and mice

Human myelin disorders are commonly studied in mouse models. Since both clades evolutionarily diverged approximately 85 million years ago, it is critical to know to what extent the myelin protein composition has remained similar. Here, we use quantitative proteomics to analyze myelin purified from human white matter and find that the relative abundance of the structural myelin proteins PLP, MBP, CNP, and SEPTIN8 correlates well with that in C57Bl/6N mice. Conversely, multiple other proteins were identified exclusively or predominantly in human or mouse myelin. This is exemplified by peripheral myelin protein 2 (PMP2), which was specific to human central nervous system myelin, while tetraspanin-2 (TSPAN2) and connexin-29 (CX29/GJC3) were confined to mouse myelin. Assessing published scRNA-seq-datasets, human and mouse oligodendrocytes display well-correlating transcriptome profiles but divergent expression of distinct genes, including Pmp2, Tspan2, and Gjc3. A searchable web interface is accessible via www.mpinat.mpg.de/myelin. Species-dependent diversity of oligodendroglial mRNA expression and myelin protein composition can be informative when translating from mouse models to humans.

To systematically identify and quantify the protein constituents of human CNS myelin, we 124 biochemically purified a myelin-enriched light-weight membrane fraction from the normal-125 appearing white matter of five human subjects post-mortem. By electron microscopic 126 assessment of the myelin fraction, constituents other than multilamellar myelin sheaths were 127 largely absent (Figure 1-supplement 1), confirming that other membrane fractions had been 128 efficiently removed.

130
We then subjected both the myelin fraction and the corresponding brain homogenate to  (Figure 1c). Markers for other cell types 153 or compartments were either reduced in abundance in the myelin fraction compared to brain 154 lysate or not identified at all (Figure 1-supplement 3). This indicates that the fraction is suited 155 for proteomic analysis of human myelin.

156
We used the MS E dataset to calculate the relative abundance of myelin proteins in the human white matter (Figure 1d), considering that quantification of exceptionally abundant proteins 160 requires a high dynamic range. The most abundant myelin proteins were the structural 161 constituents proteolipid protein (PLP), myelin basic protein (MBP) and cyclic nucleotide 162 phosphodiesterase (CNP), which accounted for 44.8%, 28.4% and 4.5% of the total myelin 163 protein, respectively. In addition, numerous known myelin proteins were identified and 164 quantified at lower abundance (Figure 1d). Previously known myelin proteins constituted 165 approximately 82% of the total human myelin protein (Figure 1d), while the remaining 18% 166 were accounted for by other proteins, including occasional contaminants from other cellular 167 sources (Figure 1-supplement 3).

169
Comparison to the mouse myelin proteome

170
We hypothesized that the protein composition of human and mouse myelin displays some 171 degree of divergence. To compare human and mouse myelin, we first separated myelin of both 172 species by SDS-PAGE. By silver staining, the band patterns were roughly comparable but not 173 identical (Figure 2a), supporting the hypothesis that some differences exist. To elucidate 174 differences at the molecular level, we compared the present human mass spectrometric data 175 with those of our recent proteomic analysis of myelin purified from the brains of C57Bl/6N mice 176 using the same workflow and methodology 9 (ProteomeXchange Consortium PRIDE partner 177 repository, dataset identifier PXD020007). As expected, the majority of known myelin proteins 178 was identified in myelin of both species (Figure 2b). However, a subset of known myelin 179 proteins was identified only in either human or mouse myelin (Figure 2b), in agreement with 180 the hypothesis that the protein composition of myelin is not identical across these species.

182
For example, we noted that peripheral myelin protein 2 (PMP2, also termed P2 or fatty acid 183 binding protein (FABP8)) was identified in human CNS myelin (Figure 2b). PMP2 has long 184 been known as a constituent of myelin in the peripheral nervous system (PNS) synthesized by datasets, against those identified in mouse myelin as recently established using the same 196 methodology 9 (ProteomeXchange Consortium PRIDE partner repository, dataset identifier 197 PXD020007). Indeed, the datasets correlated well with correlation coefficients of 0.88 (MS E ) 198 and 0.59 (UDMS E ) (Figure 2c,d) but clearly diverged to some extent. We therefore cross-199 compared the abundance of individual myelin proteins in human and mouse myelin by MS E 200 using heatmap visualization (Figure 2e). We found that major structural myelin proteins 201 including PLP, MBP, CNP, SEPTIN2, SEPTIN7 and SEPTIN8 displayed a similar relative 202 abundance in myelin of both species. However, several other myelin proteins were 203 comparatively more abundant in human myelin, as exemplified by crystallin-aB (CRYAB), CD9

204
(also termed tetraspanin-29/TSPAN29), and peptidyl arginine deiminase (PADI2), or in mouse 205 myelin, including myelin-associated oligodendrocyte basic protein (MOBP), sirtuin-2 (SIRT2), 206 and carbonic anhydrase 2 (CA2). Importantly, when detecting these proteins by 207 immunoblotting in myelin of both species (Figure 2f) these results were generally consistent 208 with the mass spectrometric comparison (Figure 2e). Yet, quantitative mass spectrometry 209 emerged as more straightforward than immunoblotting when comparing the relative 210 abundance of proteins across species if species-dependent differences in splice isoforms 211 exist. This is exemplified by MBP, which -owing to species-dependent alternative splicing 22 -212 displays three main isoforms (14.0, 17.0 and 18.5 kDa) in mouse CNS myelin but only one 213 dominant isoform (18.5 kDa) in human CNS myelin, in agreement with previous observations 214 23,24 . Taken together, the protein composition of human and mouse CNS myelin is similar with 215 respect to the relative abundance of major structural proteins but displays remarkable 216 qualitative and quantitative differences regarding many other myelin proteins.

218
Integrated scRNA-seq profile of human and mouse mature oligodendrocytes (MOL)

219
To identify species-dependent transcriptional differences that may underlie the diversity of the 220 myelin proteome, we utilized high-resolution mRNA-abundance profiles to assess the 221 oligodendrocyte lineage in both humans and mice. To this aim, we retrieved previously 222 published scRNA-seq datasets from the CNS of humans 17,18,25-28 and mice 16,28-33 and 223 evaluated all cells designated as oligodendrocyte progenitor cells (termed OPC in the 224 following), newly formed oligodendrocytes (termed NFO) and mature oligodendrocytes 225 (termed MOL) from non-diseased subjects (Figure 3-supplement 1a

236
We noted that cells expressing NFO markers (BMP4, ENPP6, FYN, GPR17) clustered well in 237 the mouse but not the human integrated dataset (Figure 3-supplement 1d), probably owing 238 to the low number of NFO in the latter. Indeed, only 132 NFO were comprised in the human 239 scRNA-seq datasets, considerably fewer compared to 10,391 NFO recovered from the mouse

280
Taken together, the integrated scRNA-seq profiles of MOL generally correlated well between 281 humans and mice, in particular with respect to most known myelin-related mRNAs. However, 282 multiple distinct transcripts were found with qualitatively or quantitatively divergent abundance 283 when compared between the species. It is also noteworthy that the degree of correlation is

304
MOL in clusters 2, 3 and 4 are associated with GO terms grouped as protein synthesis, electron 305 transport and immune activation, respectively. However, their functional specialization and 306 relevance remain to be shown. Less speculatively, both human and mouse MOL comprise all 307 five subpopulations to an approximately similar extent (Figure 4c), implying that none of these 308 MOL subpopulations is restricted to either one of these species.

313
We performed quantitative proteome analysis to determine the protein composition of human 314 CNS myelin. Subjecting myelin biochemically purified from human subcortical white matter to 315 label-free mass spectrometry allowed identifying hundreds of proteins with very high 316 confidence. More importantly, the method involves quantifying peptide intensities without 317 prefractionation, thereby providing direct information about the relative abundance of myelin 318 proteins. The latter provides a considerable advancement compared to previous approaches 319 involving prefractionation at the protein level via 1D-gels 24,34 or at the peptide level via 2D-320 liquid chromatography 35,36 , which yielded lists of proteins identified in human CNS myelin but 321 without information about their relative abundance.

323
Knowing the relative abundance of myelin proteins enables both considering their 324 stoichiometric relationships and cross-species comparisons. For example, the filament-325 forming septins SEPTIN2, SEPTIN4, SEPTIN7 and SEPTIN8 displayed a molar stoichiometry 326 of about 1:1:2:2 in human CNS myelin. Notably, the same septin subunits are also comprised 327 in myelin of mice with a similar molar stoichiometry 9 , a likely prerequisite for their assembly 328 into similar core multimers and higher order structures. Indeed, experiments in mice have 329 previously shown that these septin subunits assemble into membrane-associated filaments 330 that stabilize the adaxonal compartment of CNS myelin 37 . Integrating the current view on 331 septin assembly 38 and the relative abundance of septin subunits in myelin it is possible to 332 deduce that the predominant core multimer in myelin is a hexamer of septins 2/4-8-7-7-8-2/4.

333
The comparatively low abundance of SEPTIN9 in myelin implies that core octamers occur less 334 frequently. Assessing the relative abundance of myelin proteins also allows deducing that for 335 each core hexamer up to one molecule of the adaptor protein anillin that facilitates septin 336 assembly 39 is present in CNS myelin. Together, the relative abundance and multimer 337 composition of myelin septins emerges as conserved between human and mouse CNS myelin, 338 similar to that of other structural myelin proteins, including PLP, MBP and CNP.

340
On the other hand, we also found considerable qualitative and quantitative differences 341 between the protein composition of human and mouse CNS myelin. For example, the 342 tetraspan-transmembrane proteins TSPAN2 40-42 and GPM6B/M6B/Rhombex29 43,44 were 343 previously established as myelin proteins in mice and rats, and they were readily identified in 344 CNS myelin of mice by both mass spectrometry 9 and immunoblot. However, these proteins 345 were of very low abundance or virtually undetectable in human myelin by both techniques. It 346 has been established in experimental mice that TSPAN2 and GPM6B contribute to 347 immunomodulation 42 and myelin biogenesis 44 , respectively. The present data thus imply that 348 protein GJC3/CX29 has also been established as a myelin protein in mice and rats and is mouse but not human myelin. Interestingly, though, deletion of the Gjc3-gene in mice did not 353 have evident morphological or functional consequences for CNS myelin or oligodendrocytes 354 47,48 . The benefit for mouse CNS myelin to comprise GJC3 thus remains unknown at this time.

356
On the other hand, peripheral myelin protein 2 (PMP2, previously termed P2 or FABP8), a 357 membrane-phosphoinositide-binding protein 49 , has long been known as a constituent of

365
or what the evolutionary constraints may be that led to its dropout from rodent CNS myelin. A 366 lead may come from the investigation of the PNS of Pmp2-deficient mice, which displayed an 367 altered myelin lipid profile associated with reduced motor nerve conduction velocity 54 . It is 368 tempting to speculate that presence of PMP2 in human but not mouse CNS myelin may affect 369 the composition or organization of its lipids, and, possibly, conduction velocity. The PMP2 gene 370 causes -when mutated -the peripheral neuropathy Charcot-Marie-Tooth (CMT) disease type 371 1G 55,56 . A subset of these patients has been tested by brain MRI; however, no major pathology 372 of the white matter was found that would be typical of a leukodystrophy 55 . Yet, our finding that 373 PMP2 is a myelin protein in the human CNS indicates that further testing these and other 374 CMT1G patients for central involvement may find yet-overlooked impairments, possibly more can not formally rule out that differences in the sex or age of specimen, brain region, sample 380 preparation or data analysis may affect the degree of correlation. However, we note that both 381 male and female donors are represented in the human samples, and that machinery, 382 methodology and data analysis were the same in establishing the mouse myelin proteome 9 383 and the human myelin proteome assessed here. The post mortem-delay unavoidable for 384 sampling human specimen is unlikely to affect the present comparison when considering that human samples used here and that an experimental post mortem-delay of six hours did not 387 considerably affect the myelin proteome in C57Bl/6N mice 9 . Finally, we believe that the high 388 degree of cross-species similarity regarding the abundance of structural myelin proteins 389 between humans and mice allows trust in the overall comparison of myelin protein 390 composition, including for proteins displaying cross-species dissimilarity. Thus, individual 391 myelin proteins displaying species-dependent differences may be owing to species-dependent 392 differences in intracellular trafficking and incorporation into the myelin sheath, stability and 393 turnover rate, mRNA-to-protein translation efficiency, or actual mRNA expression.    The Netherlands). The diagnoses were confirmed by a neuropathologist.

442
The following subjects were used for myelin purification and proteome analysis of lysate (L)

515
software freely available at www.isoquant.net) was used as described previously 9,61 . Only 516 proteins represented by at least two peptides (minimum length six amino acids, score ≥ 5.5, 517 identified in at least two runs) were quantified as parts per million (ppm), i.e. the relative amount 518 (w/w) of each protein in respect to the sum over all detected proteins. FDR for both peptides 519 and proteins was set to 1% threshold and at least one unique peptide was required. Human 520 myelin fractions and the corresponding white matter homogenates were assessed as five 521 biological replicates (n = 5) each. The proteome analysis was repeated as an independent 522 replicate experiment from the same protein fractions, resulting in ten LC-MS runs per condition.

523
The mass spectrometry proteomics data have been deposited to the ProteomeXchange

564
Original immunoblots are provided in Figure 2-

986
a Immunoblot analysis of myelin purified from human normal-appearing white matter (CNS),

987
C57Bl/6N mouse brains (CNS) and C57Bl/6N mouse sciatic nerves (PNS) using antibodies 988 specific for PMP2. Blot shows two biological replicates per condition. Note that PMP2 was 989 readily detected in human CNS myelin and mouse PNS myelin but not in mouse CNS myelin.

990
SIRT2 was detected as a control. LGI3