Enhanced oxidative phosphorylation, re-organized intracellular signaling, and epigenetic de-silencing as revealed by oligodendrocyte translatome analysis after contusive spinal cord injury

Reducing the loss of oligodendrocytes (OLs) is a major goal for neuroprotection after spinal cord injury (SCI). Therefore, the OL translatome was determined in Ribotag:Plp1-CreERT2 mice at 2, 10, and 42 days after moderate contusive T9 SCI. At 2 and 42 days, mitochondrial respiration- or actin cytoskeleton/cell junction/cell adhesion mRNAs were upregulated or downregulated, respectively. The latter effect suggests myelin sheath loss/morphological simplification which is consistent with downregulation of cholesterol biosynthesis transcripts on days 10 and 42. Various regulators of pro-survival-, cell death-, and/or oxidative stress response pathways showed peak expression acutely, on day 2. Many acutely upregulated OL genes are part of the repressive SUZ12/PRC2 operon suggesting that epigenetic de-silencing contributes to SCI effects on OL gene expression. Acute OL upregulation of the iron oxidoreductase Steap3 was confirmed at the protein level and replicated in cultured OLs treated with the mitochondrial uncoupler FCCP. Hence, STEAP3 upregulation may mark mitochondrial dysfunction. Taken together, in SCI-challenged OLs, acute and subchronic enhancement of mitochondrial respiration may be driven by axonal loss and subsequent myelin sheath degeneration. Acutely, the OL switch to oxidative phosphorylation may lead to oxidative stress that is further amplified by upregulation of such enzymes as STEAP3.


Introduction
Contusive spinal cord injury (SCI) SCI has a complex pathogenesis that involves time-dependent components including primary and secondary injuries as well as post-injury remodeling and plasticity 1 .
SCI associated white matter damage (WMD) is a major driver of functional de cits below the level of injury 2,3 . Death of OLs contributes to WMD post-SCI 4,5 . However, few OL-expressed genes such as p75/Ngfr 6 , Bax 7 or Klk8 8 have been implicated as mediators of SCI-induced OL death/WMD. This is at least partly due to limited insight into the OL gene expression programs after SCI.
Single cell (sc) RNASeq technology has been recently applied to study the SCI transcriptomic response at the cellular level 9,10 . Speci cally, scRNASeq-enabled transcriptomic phenotyping revealed region-speci c changes in OL subtype content at chronic timepoints after hemisection or contusion SCI 9 . However, few SCI-associated OL gene expression changes were detected 9 . Importantly, the currently available scRNAseq technology has signi cant limitations. Its reliance on successful sorting of suspensions of viable cells is a problem when dealing with cells that have complex morphologies, spatially-regulated transcriptomes and/or sustained signi cant damage due to injury 11 . Additional challenges include limited depths/low sensitivity, high level of stochastic variability of single cell transcriptomes, ambiguity in interpreting negative signals, and data set contamination with highly expressed mRNAs that were released from other cells that lysed during sample preparation and/or sorting 11,12 . Last but not least, scRNASeq is focused on overall cellular mRNA levels. Therefore, scRNASeq data lack information on gene expression regulation at the level of protein synthesis. Yet, such post-transcriptional regulation that is present in neurons and OLs plays a major role in response of those cells to pathologies [13][14][15] .
Isolation of translating ribosomes and quanti cation of their associated mRNAs offers insight into the cell translatome and accounts for translation initiation, which is the critical regulatory step of protein synthesis [13][14][15] . The Ribotag technology enables analysis of cell type-speci c translatomes in whole animal studies 16 . In Ribotag mice, cell type-speci c expression of the Cre recombinase results in removal of a stop codon to produce the RPL22 protein with a hemagglutinin (HA) tag at the C-terminus 16 . The large ribosomal subunit that contains RPL22-HA associates with the small ribosomal subunit during successful translation initiation, which enables immunoa nity isolation of translating ribosomes by targeting the HA tag. Then, cell type-speci c translatomes are determined by RNASeq 16 . Ribotag has been successfully applied to study astrocyte-or macrophage translatomes after contusive thoracic SCI 17,18 .
The current study has been initiated to determine the translatome of mature OLs at various stages of recovery following moderate contusive SCI at the T9 level. Ribotag identi ed hundreds of differentially expressed genes in SCI-challenged OLs. These newly described gene expression landscapes implicate axonal disconnection and loss of myelin sheaths as major drivers of the acute OL gene expression response to SCI. Unexpectedly, similar factors may contribute to OL gene expression regulation in the subchronic phase of the recovery.

Results
Isolation and sequencing of the OL translatome from the spinal cord. OL-Ribotag mice were treated with tamoxifen to activate Cre-mediated recombination of the Rpl22 (STOP) −HA allele (Fig. 1a). Four weeks later, HA immunostaining was observed in 75% or 5% of CC1 + or CC1 − cells throughout the thoracic spinal cord, respectively (Fig. 1b-c). Double-positive cells were present in the white and grey matter, as expected for mature, CC1 + OLs (Fig. 1b-c). Conversely, in vehicle-treated controls, HA + cells were rare (< 2% or < 0.5% CC1 + or CC1-cells, respectively, Fig. 1b-c). Therefore, tamoxifen treatment resulted in e cient and OLspeci c expression of the RPL22-HA in OL-Ribotag mice.
These time points were selected based on dynamics of spinal cord pathology after SCI including acute OL loss at the injury epicenter (dpi 2), peak of delayed OL apoptosis in the spared white matter (dpi 10) and limited remyelination (dpi 42) 4,5,19 . Uninjured, naïve OL-Ribotag mice were used as controls. After isolation of total spinal cord polysomes, OL polysomes were immunoprecipitated using anti-HA antibody. To verify their successful isolation, qPCR for neural cell marker transcripts was performed. As expected for translatomes from cells that are estimated to make about 20% of all spinal cord cells 20,21 , the average enrichment of an OL marker mRNA reached 5.89 fold change (FC) total RNA control (Fig. 1d). Conversely, several astrocytic, neuronal or microglial marker transcripts were depleted from OL translatome samples (median FC 0.52, 0.36 or 0.1, respectively, Fig. 1d). Such a differential enrichment pattern was observed across all samples indicating successful isolation of OL translatome from spinal cord tissue of OL-Ribotag mice.
RNASeq was then performed on all samples. The principal component analysis (PCA) of the resulting mRNA expression data revealed robust separation of samples that represented OL-enriched translatome vs. total RNA input with the PC1 accounting for 70% variance (Fig. 2a). Further separation was also evident including that between naïve and SCI samples (PC2, 19% variance, Fig. 2a) or dpi 2 vs. other samples (PC3, 4% variance, Fig. 2a).
Next, OL-enriched mRNAs were identi ed for each set of samples. As compared to total RNA inputs, OL enrichment of Log2FC(Total) > 0.5 was observed for 3,511, 3,302, 3,262, or 3,313 mRNAs in naïve, dpi 2, dpi 10 or dpi 42 samples, respectively (q < 0.05, Supplementary Table S1). While established mRNA markers of mature OLs were enriched, mRNA markers of neurons, astrocytes or microglia were depleted ( Fig. 2b). Gene ontology term enrichment analysis (GO) was performed for OL-enriched mRNAs ( Fig. 2c and Supplementary Table S2). The top enriched GO terms did not include OL-speci c biological processes (BPs) or cellular components (CCs, Fig. 2c). However, OL-speci c GO-BPs/CCs such as axon ensheathment, myelination, or myelin sheath were highly overrepresented when the analysis focused on 520 mRNAs with highly selective OL expression as de ned by Log 2 FC(Total) > 2, q < 0.05 (Fig. 2c). A similar pattern of GO term enrichment was observed when analyzing OL-speci c mRNAs from other samples sets including dpi 2, 10, or 42 (Supplementary Fig. S1 and Table S2). Therefore, OL translatomes were successfully isolated from both intact and injured spinal cords.
Identi cation of differentially expressed OL genes after SCI. To compare SCI-mediated changes in OLenriched translatomes, differentially expressed mRNAs were identi ed between OL SCI vs. OL naïve samples (|Log 2 FC/naïve/|>1, q < 0.05). However, several well-established markers of microglia/monocytederived macrophages were identi ed as OL-upregulated after SCI (Fig. 3a). Those included such mRNAs as Itgam/Cd11b, Cx3cr1, and Aif1/Iba1, whose post-SCI protein expression has been con rmed in many studies to be microglia/macrophage-speci c (as reviewed in 19,22 ). In support of being expressed mainly by in ammatory cells, all those marker transcripts remained OL-depleted after SCI and their OL deenrichment did not change signi cantly (Fig. 3a). Moreover, extensive overlaps were observed between SCI-upregulated mRNAs from OL translatomes and the top 500-microglia-enriched mRNAs (Fig. 3b). Such ndings suggest that some mRNAs that are present in OL tranlsatome represent a contamination from non-OL cells, as recently shown in heterologous culture systems 23 . Such a non-speci c co-puri cation with cell type-tagged ribosomes may be particularly relevant acutely after CNS injury when tissue cellularity changes and some mRNAs such as those expressed by the in ammatory cells become extremely abundant 10 . To reduce interference by those potential contaminating transcripts, a two-arm ltration procedure was applied to differentially expressed mRNAs from OL translatomes (Fig. 3c, see Material and Methods for further details). By taking into account not just the OL expression change but also differential-(arm 1) or constantly high (arm 2) OL enrichment, the ltration identi ed hundreds of mRNAs that represent high con dence components of OL gene expression response to SCI (Fig. 3c, Bioenergetic re-organization and reduced morphological complexity/connectivity as major components of OL-speci c gene expression response to SCI. Among 344 highly upregulated transcripts on dpi 2, the top-enriched GO-BP terms included several broad categories such as those related to development, nervous system development, regulation of biological quality or signaling ( Fig. 4a and Supplementary   Table S4). More speci city emerged among top enriched GO-CC, GO-MF, and KEGG pathway terms. Those included such mitochondrial function-associated GOs as inner mitochondrial membrane, mitochondrion, proton-transporting ATP synthase activity, oxidative phosphorylation, or thermogenesis. Interestingly, eight components of the mitochondrial respirasome were highly upregulated (Log 2 FC(naïve) > 1, Fig. 4b).
On dpi 10, 155 highly upregulated mRNAs showed greatest overrepresentation of several broad GO-BP terms that were related to development ( Fig. 4c and Supplementary Table S4). In addition, enrichment of synapse-related GO-CCs was observed (Fig. 4c). Upregulation of synapse-associated transcripts could represent an attempt to re-establish OL-axonal synapses that were likely lost during the acute phase of SCI-associated axonal injury 24 . Development remained a top-enriched GO-BP theme among 294 highly upregulated mRNAs on dpi 42 ( Fig. 4d and Supplementary Table S4). In addition, high enrichment of mitochondria-associated GOs was found including the GO-CC or KEGG pathway terms mitochondrial inner membrane, mitochondrion, oxidative phosphorylation and thermogenesis (Fig. 4d). On dpi 42, both the spectrum and scale of mitochondria-related mRNA upregulations appeared to be even greater than that on dpi 2. Speci cally, 48 mitochondrial respirasome genes were upregulated in OLs on dpi 42 (Log 2 FC(naïve) > 0.5, q < 0.05, Fig. 4b) with median Log 2 FC(naïve) = 0.95). These data suggest that on dpi 42, OL metabolism is again reorganized to favor oxidative phosphorylation.
On dpi 2, 278 highly downregulated mRNAs from the OL-enriched translatome showed high level enrichment for several broadly de ned GO-BP terms that are associated with development ( Fig. 5a and Supplementary Table S4). In addition, several top enriched GO-BP, GO-CC, and GO-MF terms were related to cytoskeleton organization, cell projection organization, cell periphery, cell projections, cell junctions and actin cytoskeleton. While GO-BP cell adhesion (GO:0007155) was not among top enriched terms, 40 out of 271 genes in that category were also down with Log 2 FC(naïve)<-1, -log(q) = 5.06). Such a functional pro le of downregulated genes suggests adaptive changes to reduced morphological complexity of OLs and/or OL disconnection from other cells/extracellular matrix. Myelin sheath loss and disconnection from axons may be major drivers of these changes.
Although only 35 genes were highly downregulated in OLs on dpi 10, they showed signi cant overrepresentation of GO terms related to sterol biosynthesis ( Fig. 5b and Supplementary Table S4).
When data for all 23 OL-expressed components of the cholesterol biosynthesis superpathway were analyzed, 15 genes were downregulated (median Log 2 FC(naïve)=-1.14, q < 0.05) and just one (Hmgcs2) was upregulated (Fig. 5c, Supplementary Fig. S2). Importantly, downregulated genes included two critical regulators of cholesterol biosynthesis Hmgcr and Sqle (Fig. 5c, Supplementary Fig. S2). Both genes remained downregulated on dpi 42 (Fig. 5c). Importantly, all downregulated cholesterol biosynthesis genes showed OL-enriched expression in naïve, dpi 2, or dpi 10 mice with median log 2 FC(Total) = 1.55, 1.61, or 1.39, respectively (q < 0.05, Fig. 5c). Such an expression pattern is consistent with a critical role of cholesterol synthesis in long term maintenance of myelin and survival of mature OLs 25 . Therefore, downregulation of cholesterol biosynthesis may represent an adaptive response to loss of myelinated axons/myelin sheaths acutely post-SCI. It may also contribute to OL apoptosis subacutely post-SCI 4,5 .
Top-enriched GOs for 224 transcripts that were highly downregulated on dpi 42 included several broadly de ned terms such as development, biological regulation or binding ( Fig. 5d and Supplementary Table   S4). In addition, cell adhesion, cell junction and glutamatergic synapse were also enriched ( Fig. 5d and Supplementary Table S4). This cell disconnection-like response resembles that observed on dpi 2.
Collectively, GO analysis of OL-speci c gene expression suggests a bioenergetic shift towards oxidative phosphorylation and reduction of morphological complexity/cell connectivity both acutely and subchronically post-SCI. At least subchronically, stress adaptation and OL survival is the likely outcome of such changes as most OL loss occurs during acute/subacute phases of the SCI recovery 4,5 . However, those surviving OLs appear to undergo protracted degeneration including putative disconnection from axons.
Identifying candidate regulators of acute OL loss. After SCI, most OL loss occurs at the injury epicenter 24-48 hours post SCI 4,5 . Subacutely (dpi 7-dpi 21), additional OL loss via apoptosis is found in areas rostral and caudal from the epicenter 4,5 . No major OL loss has been reported beyond dpi 28 4,5 . Thus, OL mRNAs with acute post-SCI upregulation on dpi 2, but decreasing expression on dpi 42, may include major regulators of SCI-associated OL loss. Of 344 highly upregulated OL translatome mRNAs on dpi 2, 148 were either not signi cantly upregulated on dpi 42, or their dpi 42 levels were at least two-fold lower than on dpi 2 ( Fig. 6a, Supplementary Table S5). GO analysis of those OL loss-associated mRNAs revealed enrichment for GO-BP terms related to cell signaling (Fig. 6b, Supplementary Table S5). Those included response to stimulus, regulation of cell communication, regulation of signaling, and signal transduction. Cell death-associated GO-BP terms were also moderately enriched (e.g. GO:1901214, regulation of neuron death, 12/381 genes, -log(q) = 2.80, Supplementary Table S5).
Several acutely upregulated genes were identi ed as potential regulators of the cell death machinery ( Fig. 6d-f). Activation of cell death may be promoted by: (i) reactive oxygen species (ROS) generating enzymes Pcyox1l and Pla2g3 43,44 , (ii) an inhibitor of the death receptor-driven, pro-survival gene transcription, Tifab 45 , (iii) a pro-excitotoxic inhibitor of JUN degradation, Prr7 46 and/or (iv) a positive regulator of pro-apoptotic MAP kinases JNK/p38, Steap3 47 . Conversely, cell death initiation may be negatively regulated by increased expression of (i) enzymes that antagonize accumulation of the proapoptotic second messenger ceramide (Acsl5, Sphk1, 48,49 ), (ii) Sphk1 which stimulates production of the cytoprotective lipid mediator sphingosine-1-phosphate 49 , (iii) negative regulators of ERK1/2 that may limit persistent, pro-necrotic activation of the ERK1/2 pathway in oxidative stress-exposed cells 42 , (iv) several enzymes that contribute to anti-oxidant defenses reducing ROS toxicity (Lpo 50 , Apod 51 , Bcat1/2 52 , Aldh18a1 53 , Pyccr1 54 , Pfkfb4 55 ) and (v) reduced ROS generation due to Trf (transferrin)-mediated chelation of iron 56 . Likewise, the lipid-peroxidation mediated effector phase of cell death cascades including ferroptosis may be antagonized by those positive regulators of anti-oxidant defenses. Opposite effects on oxidative mechanisms of cell death execution are expected from upregulation of the potentially pro-ferroptotic lysosomal/endosomal ferroreductase Steap3 which may promote lipid peroxidation [57][58][59] . Similar consequences may follow upregulation of the GSH-depleting enzyme Chac1 compromising GSH-dependent lipid repair 59 . In addition, upregulation of the pro-apoptotic gene Hrk may promote SCI-induced OL apoptosis 60 .
Several upregulated genes with a potential to modulate oxidative cell death (Pla2g3, Apod, Trf, Lpo) encode for secreted proteins 61 . In addition, on dpi 2, those transcripts are highly OL-enriched, suggesting OLs to be the main source of their respective protein products in the contused spinal cord tissue acutely post-SCI (Fig. 6e). Other acutely OL-upregulated genes showing a similar pattern include those for secreted serine proteases (Klk8, Klk9) and secreted serine protease inhibitors (Serpina3n, Serpina3c, Fig. 6e). Excessive extracellular serine protease activity may promote cell death and enhance white matter damage 8,62,63 . Therefore, OLs appear to activate autocrine/paracrine mechanisms that via regulation of oxidative stress and/or extracellular proteolysis may modify acute pathogenesis of SCI.
Such a concept is supported by a report that SCI-associated OL death and axonal damage was reduced in Klk8 −/− mice 8 .
Epigenetic de-silencing as a potential regulator of the OL response to SCI. To identify mechanisms that may contribute to SCI-associated OL gene expression changes, overlaps between OL-upregulated genes and public datasets of ChIPSeq-con rmed mouse genome operons were determined using the ChIPSeq database module of the X2K suite 64 . At each post-SCI timepoint, top enriched operons included SUZ12 and MTF2, two components of the PRC2 chromatin repressive complex (Supplementary Table S6) 65 . To further validate the speci city of these enrichments, z-scores were calculated for operon overlap gene counts of 344 dpi 2 upregulated genes vs. average overlap gene count for 10 random sets of 344 OL expressed genes. OL-upregulated genes showed overlaps with seven mouse genome operons that passed a stringent speci city criterion of z > 5 (Fig. 7A). Four of those operons (including three with top z-scores) were for the chromatin silencing factor SUZ12. Two other chromatin silencing factors including RCOR3 and MTF2 were also highly enriched. The speci city of those ndings was further con rmed by weak operon enrichment among 289 acutely downregulated genes with a z-score range for top ten overlaps of 0.94-2.57 (Supplementary Table S6).
SUZ12/PRC2-mediated repressive methylation of histones is required for OL differentiation by downregulating differentiation inhibitory genes 66 . RCOR3 is a negative regulator of the KDM1A-RCOR1/2/3-HDAC-driven gene silencing resulting in upregulation of genes that are repressed during cell differentiation 67 . Suz12, Mtf2, and Rcor3 transcripts were detected both in total RNA and OL translatomes (Fig. 7b,c). Their expression was unaffected by SCI. While Suz12 and Mtf2 showed similar expression in OLs and total RNA, Rcor3 was moderately OL-enriched in all groups except dpi 42 (Fig. 7b,c). Therefore, the identi ed operons are matched by continuous OL expression of their respective regulators.
Interestingly, the list of 129 unique SUZ12 target genes that were OL-upregulated acutely after SCI showed overrepresentation of GO terms related to signaling (Fig. 7d) and included several potential regulators of acute OL loss (Supplementary Table S6). Most those PRC2 targets overlapped with the RCOR3 operon (Supplementary Table S6). The current data suggest that de-silencing of the PRC2 and RCOR3 operons may contribute to acute upregulation of many SCI-response genes in OLs. Moreover, desilencing of PRC2-repressed genes may also play a role in OL gene upregulation on dpi 10 and dpi 42.
STEAP3 as a novel marker of acute OL response to SCI. STEAP3 is a membrane protein that acts as a major endosomal/lysosomal metalloreductase that reduces Fe 3+ /Cu 3+ to Fe 2+ /Cu 2+ 57,68 . As Fe 2+ is then exported from endosomes/lysosomes to cytoplasm, STEAP3 plays an important role in the cellular iron supply 57,58,69 , but may also promote lipid peroxidation and ferroptosis 70 . In addition, STEAP3 binds to cell death effectors and signaling mediators to promote apoptosis and/or reduce activity of the prosurvival pathways 33,47,71 . It is also required for stress-dependent exosome secretion 72,73 . Therefore, OLupregulated Steap3 may promote OL loss, but may also, via stimulation of exosome-mediated secretion, affect axons and/or other spinal cord cells.
In OLs, Steap3 was sharply up on dpi 2 and remained elevated, albeit at reduced levels throughout the recovery period with Log 2 FC(naive) ranging from 6.52 on dpi 2 to 3.67 on dpi 42 (Fig. 6e). In total RNA samples, Steap3 levels were also increased throughout the recovery period with an apparent peak on dpi 2 (Log 2 FC(naive) = 2.9, 2.7, or 1.89 on dpi 2, 10, or 42, respectively, q < 0.05). On dpi 2, Steap3 was highly OL-enriched (Fig. 6e). At that timepoint, increased expression of STEAP3 protein was also con rmed by western blotting (Fig. 8a, b, Supplementary Fig. S5a). Therefore, at least acute OL tranlsatome upregulation of Steap3 transcript correlates with increased bulk spinal cord expression of STEAP3 protein.
In naïve mice, faint STEAP3 immuno uorescence was observed in cell bodies of white matter OLs (CC1 + ,   Fig. 8c, Supplementary Fig. S3). On dpi 1 or 2, such staining was also present throughout the spared ventral white matter at the injury epicenter as well as rostrally or caudally (Fig. 8c, Supplementary Fig. S3 and S4). Seemingly nuclear localization of OL STEAP3 staining on transverse sections may represent perinuclear localization that is polarized mainly in the rostral-caudal axis (Fig. 8c). Unipolar, perinuclear STEAP3 immuno uorescence that re ects localizations to the endosomal/trans-Golgi network (TGN) has been reported in various cell types 57,68,72,73 .
Acutely after SCI (dpi 1 or 2), a strong STEAP3 signal was also found in ring-like structures throughout the spared white matter (Fig. 8c, d, Supplementary Fig. S3 and S4). Those structures were observed both at the injury epicenter as well as at the penumbra region rostrally and caudally from the injury site. They often showed partially overlapping and/or closely associated signals for the myelin marker MBP, as well as axonal marker phospho-NFH (p-NFH, Fig. 8d, Supplementary Fig. S4). Importantly, those STEAP3 + structures likely represent a speci c STEAP3 signal, as they were not observed with a control IgG (Supplementary Fig. S3 and S4). Therefore, acute upregulation of OL STEAP3 may be directly related to degeneration of myelinated axons.
Interestingly, STEAP3 was upregulated following treatment of cultured OL precursor cells (OPCs) or OPCderived OLs with a mitochondrial uncoupling drug FCCP, but not the ER stress inducer tunicamycin ( Fig. 8e-h, Supplementary Fig. S5b, c). As FCCP uncouples mitochondrial respiration from ATP synthesis and increases mitochondrial ROS generation 74 , upregulation of STEAP3 may represent a response of OL lineage cells to mitochondria dysfunction/mitochondrial oxidative stress. The latter form of cellular damage may upregulate expression of iron supply genes as mitochondrial biogenesis is a major driver of iron demand 75 .
Finally, OL expression of another acutely upregulated and OL-enriched candidate pro-oxidant transcript, Pcyox1l 43 was con rmed by immunostaining. Unlike STEAP3, the PCYOX1L signal was observed mainly in cell bodies and adjacent processes of CC1 + or CNP + OLs of the ventral white matter ( Supplementary  Fig. S6). Therefore, OLs may directly contribute to SCI-associated oxidative stress.

Discussion
The current data set represents the rst comprehensive transcriptomic/translatomic description of OL gene expression response to SCI. Acutely, SCI-challenged OLs show translatomic changes indicative of a metabolic shift towards mitochondrial respiration that coincides with morphological simpli cation/cellular disconnection. Unexpectedly, a similar pattern reemerges subchronically. Acutely, OLs appear to undergo extensive re-organization of survival signaling networks. In addition, the acute OL translatome changes suggest an active role in regulation of cytotoxic mechanisms that contribute to secondary injury including extracellular proteolysis and oxidative stress. Epigenetic de-silencing appears to be a major driver of the SCI-activated OL gene expression. Lastly, STEAP3 is a novel marker of OL injury response that may play a dual role as a positive, pleiotropic regulator of OL death/degeneration and an enhancer of exosome-mediated intercellular communication.
The current translatome analysis revealed OL-speci c, SCI-mediated upregulation of several genes that were also identi ed as OL-upregulated in a recent scRNASeq analysis of the mouse spinal cord tissue acutely/subacutely after moderate contusive SCI 10 . Those include Apod, Pla2g3, Trf, Klk8, Serpina3c, Serpina3n, Steap3, Pcyox1l, Itgad, Pfkfb4, Bact1/2 or Spry1. In addition, SCI-associated OL translatome changes show partial overlap with OL transcriptomic response to other types of CNS injury 76 . Recent analysis of multiple scRNASeq datasets from various mouse models of white matter damage including amyloidosis, tauopathy, EAE, and lysolecithin-induced demyelination has identi ed three major transcriptomic pro les of disease-associated OLs called DA1, DA2 and IFN. Of note, signature transcripts of all those pro les showed signi cant overlaps with SCI-upregulated OL mRNAs (Supplementary Fig.  S7). The greatest overlaps were observed for the persistent, pro-in ammatory DA1 cluster. The overlap peaked on dpi 10, but was clearly present also on dpi 2 and dpi 42. The overlap with the transient cell injury response pro le DA2 peaked on dpi 2. Interestingly, Steap3 has been identi ed as a marker of DA1 76 .
While the current OL translatome dataset captures a large set of OL-expressed genes including those that are regulated by injury, it has important limitations that should be considered when choosing analysis tools or interpreting data. First, we observed apparent contamination of OL polysomes with mRNAs from non-OL cells including microglia/macrophages. Such a contamination may originate from non-speci c interactions between highly abundant non-OL mRNAs and IgG-coupled magnetic beads that occurs during sample preparation. Similar contamination has been recently reported in co-cultures of human and mouse neural cells 23 . The two-fold ltration process that took into account differential or constantly high OL enrichment reduced that contamination and identi ed high con dence components of the OL translatome that are regulated by SCI. However, while speci city of detection has improved, sensitivity likely suffered with over 60% of differentially expressed OL translatome mRNAs not passing the ltration criteria (Fig. 3). Therefore, some components of OL response are likely missed in our analysis. Nevertheless, such a conservative analysis approach is justi ed in acute CNS injury models where dramatic changes in both tissue cellular composition and transcriptomes of in ammatory cells are major factors that determine bulk transcriptome readouts.
Another limitation is related to imperfect OL speci city of Plp-CreERT2 expression. The Plp-CreERT2 transgenic line that was used here to activate the Ribotag also drives Cre-ERT2 activity in Schwann cells 77 . As spinal cord samples my contain some short fragments of spinal nerve roots, it is possible that a small fraction of OL polysomes is of Schwann cell origin. Such a contamination of the current dataset is suggested by an apparent OL enrichment and SCI upregulation of marker transcripts of Schwann cell nerve injury response including Eg 8 or Gdnf (Supplementary Table S2) 78 . Hence, at least some SCIassociated effects on OL gene expression may represent Schwann cell responses. Potential classi cation mistakes can be best avoided by con rming OL translatome mRNA changes using single cell level analysis of protein expression such as immuno uorescent or in situ hybridization-base approaches.
In myelinating glia, mitochondrial respiration is critical for myelination during development as myelin synthesis requires a large amount of ATP 79,80 . Conversely, in mature OLs, mitochondrial respiration is dispensable for OL survival, myelin sheath maintenance and axonal function 79 . Instead, OLs support axonal energy needs by providing lactate that is generated via aerobic glycolysis 79 . As electrical activity of axons is a major consumer of OL-generated lactate 79 , one could expect that OL disconnection from axons and/or axonal damage would lower lactate demand. Consequently, aerobic glycolysis would be reduced and mitochondrial respiration would increase. Our observed concomitant upregulation of respirasome genes, together with downregulation of genes associated with morphological complexity/connectivity, supports such a scenario both acutely and subchronically after SCI. Indeed, while most SCI-associated axonal loss is acute, continuing subchronic degeneration of myelinated axons has been described 81,82 . In addition, subchronic reorganization of OL respiration may be a response to post-SCI reduction in activity of speci c axonal tracts and/or spinal circuitries.
Lastly, SCI-mediated changes in OL gene expression indicate their active role in modifying the tissue environment to modulate secondary injury. Such effects may be mediated both by canonical secretion of signal-peptide containing proteins and non-canonical secretion by exosomes, as STEAP3 is a major regulator of exosome-mediated secretion 72,73 . Of note, signi cant enrichment of secreted proteins among upregulated genes on dpi 10 or dpi 42 suggests that OLs regulate not only acute damage, but also subacute/subchronic repair of the contused spinal cord.
In summary, this study establishes the rst translatomic chart of OL response to thoracic contusive SCI. It uncovers previously unrecognized aspects of OL biology after SCI, including putative metabolic reprogramming, re-organization of intracellular signaling and epigenetic de-silencing as a major driver of OL gene expression response to injury. These SCI-associated OL translatome data will be useful for design and/or interpretation of mechanistic studies of SCI-associated white matter damage as well as other types of white matter pathology. Current data are available as a searchable database at SCI OL Gene Expression Database (scigenedatabase.com). SCI. Female WT or OL-Ribotag mice were used for SCI at 8-10 weeks of age. In tamoxifen-induced mice, SCI was performed 3 weeks after completion of the induction treatment. Anesthetized animals (400 mg/kg body weight 2,2,2-tribromoethanol i.p.) were shaved around the surgical site and disinfected using 4% chlorohexidine solution. Lacri-Lube ophthalmic ointment (Allergen, Madison, NJ) was applied to prevent eye drying. Following dorsal laminectomy at the T9 vertebrae, moderate contusive SCI was performed using the IH impactor (50 kdyn force/400-600 µm displacement, In nite Horizons, Lexington, KY) as previously described 83 . Sham controls only received T9 laminectomy. Starting immediately after surgery, postoperative care included 0.1 ml saline (s.c. daily for 7 days), 5 mg/kg gentamycin (s.c. daily for 7 days), 0.1 mg/kg buprenorphine (s.c. every 12 h for 2 days), and manual expression of bladders twice a day for seven to ten days or until spontaneous voiding returned. All surgical and post-surgery procedures were completed according to NIH and IACUC guidelines. All surgeries were performed without knowledge of group assignment or genotype.

Materials and Methods
Tissue collection. Anesthetized mice were transcardially perfused with phosphate buffered saline (PBS, 4°C). For immunostaining, this was followed by 4% paraformaldehyde (PFA in PBS, 4°C) perfusion. Then, a 5 mm portion of the spinal cord spanning the injury epicenter was dissected and (i) post-xed for 1 hr in 4% PFA at 4°C (immunostaining) or (ii) ash frozen in liquid nitrogen and stored at -80°C until further use (Ribotag polysome puri cation or RNA/protein isolation). Ribotag RNA puri cation and RNASeq. Frozen spinal cord samples from 2 mice were pooled to produce one biological replicate (3 biological replicates/group representing 6 animals). Pooling was done to increase RNA yield of the Ribotag immune-puri cation as determined in pilot experiments. SCI and naïve mouse spinal cord samples were processed for input polysome-associated mRNAs (total spinal cord mRNA) or immune-puri ed OL polysome mRNAs using anti-HA antibody and magnetic beads as previously described 16 85 (see Supplementary Methods for a detailed description of the protocol). RNA isolation, mRNA library preparation and RNA sequencing on the Illumina NextSeq 500 platform followed standard procedures. Before preparing RNASeq libraries, successful isolation of OL polysomes was veri ed by qPCR for OL-and non-OL cell marker transcripts. Pilot studies revealed that polysome immunoprecipitation using a control IgG produced low RNA yields as compared to the anti-HA antibody con rming the speci city of the latter reagent (average of 48.4 or 285.34 ng RNA/sample with IgG or anti-HA, respectively. Given such a disparity in RNA recovery between the control IgG and the anti-HA antibody, further analyses focused on anti-HA-puri ed OL translatomes. Quantitative real-time PCR. To prepare cDNA, the SuperScript IV system was used following manufacturer's recommendations (Thermo Fisher, Cat# 18091050). qPCR was run using a micro uidic Custom TaqMan Gene Expression Array Card (Thermo Fisher, Cat# 43442249) containing primers for CNS cell-type speci c marker mRNAs of oligodendrocytes (Mbp, Plp1, Mog, Cldn11, Mobp, Opalin, Mag, Fa2h, Gjb1, Ermn, Gjc2, Klk6, Sox10), astrocytes (Aldh1l1, Hgf), neurons (Reln, Snap25, Lhx5), and microglia/macrophages (Osm, Cd68, Tmem19). The card design is shown in Supplementary Methods. RNA levels were quanti ed using the ΔΔCT method with Hprt, Ppia, and 18S rRNA as reference transcripts. For each sample pair (OL translatome and total spinal cord RNA from which OL translatome was isolated), OL mRNA levels were determined as a fold change of their total spinal cord expression.
RNAseq data analysis. Analysis of sequenced RNA was performed by the Kentucky IDeA Networks of Biomedical Research Excellence (KY INBRE) Bioinformatics Core. A quality control analysis was performed using FastQC (v.0.10.1) and indicated good sequencing quality. The reads were directly aligned to the Mus musculus reference genome (mm10.fa) using the STAR aligner (version 2.6). The average number of sequenced reads per sample was ~ 37,500,000 with an average alignment rate of 98.06%. Raw read counts were generated using HTSeq (v.0.10.0) and input to DESeq2 for differential expression analysis. The raw counts were normalized using Relative Log Expression (RLE) and ltered to exclude genes with fewer than 10 counts across samples. Principal component analysis (PCA) was performed on all 24 samples to measure variance of the overall mRNA expression pattern across all groups/sample sets. As part of the differential expression analysis, a DESeq2 interaction term was used to analyze differential OL enrichment after SCI and can be de ned by delta Log 2 FC(OL/total) = Log 2 FC(OL/total)_SCI -Log 2 FC(OL/total)_naive. In this analysis, OL translatome-and total spinal cord RNA samples from the same tissue were paired for determination of differential OL enrichment. To identify genes with an RNA origin-speci c effect (OL vs. total) at one or more time points, the full regression model was compared to a reduced model using a likelihood ratio test. Group-speci c effects were then identi ed at individual timepoints using a Wald test for signi cance. As a signi cant contamination of non-OL transcripts has been detected in OL translatomes from SCI samples, a two arm ltration process was implemented to identify highly likely components of the OL translatome that are differentially expressed after SCI (Fig. 3a-c). In arm 1, the identi ed transcripts were ltered using the DESeq2 interaction function to identify those that also showed differential OL enrichment of the same direction/magnitude (change of OL enrichment in SCI vs. naïve samples, |Log 2 FC/Total/ SCI -Log 2 FC/Total/ naïve |>1, q < 0.05). This process eliminated mRNAs whose OL translatome changes are driven primarily by their parallel changes in the total RNA pool. However, differential enrichment analysis may miss those mRNAs whose OL translatome levels change in the same direction/magnitude as in total spinal cord samples. In case of constantly OL-enriched mRNAs that also show preferential OL expression (Log 2 FC(Total) > 1, q < 0.05), their exclusion is not justi ed as OLs are their major expressors and crosscontamination from other cell types is less likely. In total RNA samples, injury related changes of such OLenriched transcripts are still a re ection of OL-speci c response to SCI. Therefore, the second arm of the ltration procedure identi ed those highly regulated mRNAs that were also OL enriched regardless of passing the differential enrichment ltration. All RNASeq data are available in GEO (Accession number: GSE225308). A searchable, public database is also available at http://scigenedatabase.com/.
Gene Ontology (GO) functional annotation analysis was performed using g: Statistical analysis of immunostaining, qPCR and immunoblotting data. All ratiometric data including % Ribotag + cells, qPCR-determined total spinal cord-normalized OL mRNA levels and control treatmentnormalized protein expression were performed using the two-tailed non-parametric Mann-Whitney u-test. RNASeq con rms successful isolation of the OL translatome from intact or injured spinal cord tissue.

Declarations
Tamoxifen-induced OL-Ribotag mice received SCI and OL translatomes were immunopuri ed as described in Fig. 1. Following qPCR analysis of cell marker transcripts (Fig. 1d)  Determining OL gene expression response to SCI. Differential gene expression was analyzed in OL translatomes from SCI vs. naïve samples. (a,b) SCI OL translatome contamination with marker mRNAs of microglia/monocyte-derived macrophages. At each timepoint after SCI, the top 1000 highly upregulated transcripts (Log 2 FC/naïve/ >1, q<0.05) included several established markers of microglia and/or monocyte-derived macrophages (a). While sharply upregulated in total RNA samples, those microglial mRNAs were OL depleted and their OL de-enrichment was unaffected after SCI. (b) Signi cant overrepresentation of the top 500 microglia marker transcripts (Brain RnaSeq database) among OLupregulated mRNAs after SCI, representation factor = 1 if the number of overlapping genes is as expected by a random chance (OPC, oligodendrocyte precursor cells; EC, endothelial cells). (c) Flow chart of the ltration procedure for high con dence identi cation of the SCI-upregulated component of the OL translatome. The two-arm ltration process was based on (i) differential OL enrichment after SCI or (ii) constantly high OL enrichment before and after SCI. In each case, OL expression analysis took into account changes in total RNA (see Text for more details, all identi ed highly regulated OL SCI DEGs are listed in the Supplementary Table S3). (d) After ltration, no overrepresentation of microglial marker mRNAs is present among OL mRNAs that are upregulated after SCI. The OL response to SCI includes biphasic upregulation of mitochondrial respirasome genes. (a,c,d) Overrepresented GOs (top 10/category) among high con dence OL translatome mRNAs that are highly upregulated after SCI (Log 2 FC/naïve/>1, q<0.05, and two-arm ltration / Fig. 3c/). Several GOs related to mitochondria are overrepresented on dpi 2 and 42, but not dpi 10 (red boxes). MF-molecular function. (b) Fifty out of 68 mitochondrial respirasome genes that are OL-expressed are also OL upregulated on dpi 2 and/or 42 (Log 2 FC/naïve/>0.5, q<0.05, bold). In most cases, their total spinal cord expression is unaffected (q>0.05, grey cells). Therefore, the biphasic upregulation of mitochondrial respirasome genes is OL-speci c. In naïve mice, most of those upregulated genes are OL depleted (Log 2 FC/Total/<-0.5, q<0.05, bold), which is consistent with reduced activity of oxidative phosphorylation in mature OLs.  Identifying candidate regulators of OL loss after SCI. (a) To identify likely candidate gene regulators of OL loss, OL translatome mRNAs that were highly upregulated on dpi 2 were further analyzed to identify those whose expression normalized or declined at least 2-fold on dpi 42 when OL numbers stabilize and when little OL death has been reported (see text for details). (b) Overrepresented GOs (top 10/category) among candidate gene regulators of OL loss include several GO terms that are related to signaling (red boxes).
(c-d) Literature-based analysis of candidate genes identi ed upregulated regulatory components of the survival signaling pathways including ERK and PI3K-AKT (c) as well as cell death activators and effectors (d). OL-upregulated genes are marked by black boxes. (e) OL translatome expression and OL enrichment of potential OL loss regulators that are listed in (c-d). Their dpi 2 read counts are shown in (f). In (e), genes with SCI-associated expression change or OL enrichment/depletion |Log 2 FC|>0.5 are in bold (q<0.05); non-signi cant effects are marked by grey cells (q>0.05). In (f), log 10 averages ±SD are shown.