CNS myelin protein 36K regulates oligodendrocyte differentiation through Notch

In contrast to humans and other mammals, zebrafish can successfully regenerate and remyelinate central nervous system (CNS) axons following injury. In addition to common myelin proteins found in mammalian myelin, 36K protein is a major component of teleost fish CNS myelin. Although 36K is one of the most abundant proteins in zebrafish brain, its function remains unknown. Here we investigate the function of 36K using translation‐blocking Morpholinos. Morphant larvae showed fewer dorsally migrated oligodendrocyte precursor cells as well as upregulation of Notch ligand. A gamma secretase inhibitor, which prevents activation of Notch, could rescue oligodendrocyte precursor cell numbers in 36K morphants, suggesting that 36K regulates initial myelination through inhibition of Notch signaling. Since 36K like other short chain dehydrogenases might act on lipids, we performed thin layer chromatography and mass spectrometry of lipids and found changes in lipid composition in 36K morphant larvae. Altogether, we suggest that during early development 36K regulates membrane lipid composition, thereby altering the amount of transmembrane Notch ligands and the efficiency of intramembrane gamma secretase processing of Notch and thereby influencing oligodendrocyte precursor cell differentiation and further myelination. Further studies on the role of 36K short chain dehydrogenase in oligodendrocyte precursor cell differentiation during remyelination might open up new strategies for remyelination therapies in human patients.


| INTRODUCTION
Myelin sheath is the extension of plasma membrane of glial cells wrapping around neuronal axons (Sherman & Brophy, 2005). These glial cells are called oligodendrocytes in the central nervous system (CNS). Disruption of myelin sheaths in CNS can occur in various demyelinating diseases like multiple sclerosis, myelopathies, leukodystrophies (Franklin & Ffrench-Constant, 2008) and has implications for other neurodegenerative diseases, including secondary axon injury (Edgar & Nave, 2009). As myelination with nodes of Ranvier is a vertebrate specific adaptation (Zalc, 2016), zebrafish is an evolutionary early and more simple model organism for studying myelination in vivo. Still zebrafish have emerged to be a powerful and useful model over the past decade to study myelination and remyelination (Brösamle & Halpern, 2002;Kirby et al., 2006;Preston & Macklin, 2015) due to the translucent nature of their embryos and their fast ex-utero embryonic development.
In addition to the common conserved myelin proteins like myelin basic protein (Mbp) (Nawaz, Schweitzer, Jahn, & Werner, 2013) or proteolipid protein (Plp1 or DM20) (Schweitzer, Becker, Schachner, Nave, & Werner, 2006), 36K (flj13639, Zebrafish International Network [ZFIN] ZDB-GENE-030131-8,104) is a major component of teleost CNS myelin protein (Jeserich, 1983). The closest human homologue to zebrafish 36K is SDR12 (National Center for Biotechnology Information [NCBI] Accession Number: NP_001026889.1), with 61% amino acid identity (Uniprot alignment). The mRNA of SDR12, also referred to as FLJ13639, has been shown to be present in human white matter in the cerebral cortex, while its existence at protein level still remains hypothetical (Morris et al., 2004). In zebrafish, 36K is a membrane associated protein present in both myelin cytoplasm as well as in membrane (Morris et al., 2004). Other proteomic studies have found 36K to be one of the most abundant proteins in zebrafish brain (Gebriel et al., 2014). Despite its abundance, the function of 36K remains unknown.
Further, it remains unknown if other enzymatic or structural myelin proteins might replace its function in rodents or humans.
36K has been suggested to belong to the short chain dehydrogenase family, based on homology studies and the conserved cofactor binding domain (Morris et al., 2004). 36K is a highly basic protein similar to Mbp and is also present in compact myelin (Morris et al., 2004), which might suggest a role in myelin compaction similar to that of Mbp (Jeserich & Rauen, 1990). In this study, we used MO ® knockdown to further investigate the function of 36K in zebrafish CNS myelin.

| Antibody characterization
As part of the characterization of our custom-made anti-36K antibodies, CHO-K1 cells were over-expressed with a plasmid containing 36K, with EGFP as transfection control. Immunostaining was performed and 36K was detected only in 36K transfected cells. Western blot was also performed which further confirmed specific 36K expression in transfected cells. Further, in larval zebrafish, 36K was found to colocalize with some olig2+ OPCs. In Western blot with 36K Morpholino knockdown, less expression of 36K was observed which also confirms the specificity of the antibodies.

| Reverse transcription (rt) polymerase chain reaction and quantitative real time (q-rt) PCR
For rt-PCR and q-rt-PCR, total RNA was extracted from 30 larvae using TRIzol ® RNA Isolation (Invitrogen 15596018). 1 μg total RNA was reverse transcribed using iScript Reverse Transcription Supermix for rt-PCR (Biorad 1708841). q-rt-PCR was performed using DyNAmo Flash Probe qPCR kit (Thermofisher F455L). After 7 min (min) denaturation at 95 C 40 amplification cycles were carried out: 5 s at 95 C and 30 s at 60 C.
cDNA was diluted 1:20. iTaq Universal SYBR Green Supermix (Biorad) was used for q-rt-PCR in the following composition: 1 μl primer mix, 1 μl water, 5 μl master mix, 3 μl 1:20 cDNA. After 10 min of initial denaturation at 95 C, 40 amplification cycles were carried out: 10 s at 95 C and 1 min at 60 C with a ramp rate of 1.6 C/s. q-rt-PCR analysis was performed using CFX Manager™ and qBase software. All data were normalized to the reference genes and scaled to average hence generating calibrated normalized relative quantities (Hellemans, Mortier, Paepe, Speleman, & Vandesompele, 2007

2.11
By default, a Welch's t-test was used and p-values were adjusted according to the adaptive Benjamini-Hochberg procedure to decrease the number of "false-negatives" (Benjamini & Hochberg, 1995.
Only probes which had a |log 2 fold| > 0.5 and a p-value of <.05 were considered. Heatmaps from microarray analysis were generated using Heatmapper (Babicki et al., 2016).

| Behavior experiments
For behavior experiments, larvae from different experimental groups were arranged in 12-well plates with one larva per well. Each 12-well plate had three larvae from each of the four experimental groups. Following acclimatization for 2 hr in the well plates, recordings were done with the same lighting and temperature conditions in the afternoons between 14:00 and 16:00.
For unprovoked behavior test, swim trajectories of the larvae were recorded for 10 min. Recordings were analyzed from the second to ninth min. For stimulated behavior test, a marble weighing~20 g was dropped from a height of 9.5 cm along a slanted board 15 angle always fixed at the same position. The vibration caused by the marble drop stimulated the larvae to swim (escape) and the first swim episode following the marble drop was analyzed. The test was repeated twice for each larva and the average values were taken so as to exclude the problem of one trial being over representative in the value.
For unprovoked tests, image sequences were reduced so that 96 frames with 5 s separation were analyzed. For stimulated response tests, 30 frames at 30 fps (=1 s) were analyzed from the time point of the marble drop. By manually tracking the position of the larva in each well in all frames using MTrackJ plugin (Meijering, Dzyubachyk, & Smal, 2012) in Fiji, distance of swimming was measured.
2.14 | Thin layer chromatography/analysis of sphingolipids 2.14.1 | Lipid extraction Whole fish larvae (~100) were homogenized in 1 ml of ice-cold 20 mM Tris-HCl (pH 8.5) using an Ultraturrax T10 basic homogenizer (IKA, Staufen, Germany), followed by centrifugation at 100,000g for 1 hr. The pellet was transferred into Pyrex glass tubes and extracted with 6 ml of chloroform/methanol (2:1; v:v) at 60 C for 4 hr (under constant stirring). After centrifugation (1,000g for 10 min) supernatant was transferred into a Pyrex glass tube and the pellet was reextracted with 6 ml of chloroform/methanol (1:1; v:v) overnight at 60 C. Samples were again centrifuged and the supernatant from second extraction was combined with the first supernatant to obtain the total lipid extract. The lipid extract was filtered through glass wool and dried under a stream of nitrogen (45 C) and dissolved in 0.5 ml of chloroform/methanol (1:1). Aliquots of these total lipid extracts were examined by TLC.

| Alkaline hydrolysis and desalting
The total lipid extract was dried under a stream of nitrogen, dissolved in 2.5 ml of methanol, and sonicated in an ultrasonic bath for 10 min. After

| Thin layer chromatography
Lipids were separated on HPTLC silica gel 60 plates (Merck, Darmstadt, Germany). Lipids isolated from identical numbers of fish larvae and lipid standards were applied by hand or using an automatic TLC sampler (ATS4; CAMAG, Berlin, Germany v:v) to obtain a better separation of free fatty acids and cerebrosides.
TLC plates were scanned using a commercial scanner (Epson Perfection V700 Photo) within 30 min after staining.
2.15 | Mass spectrometry 2.15.1 | Lipid extraction As described in (Rabionet et al., 2008), fish larvae (~100) were homogenized in 1 ml of ice-cold 20 mM Tris-HCl (pH 8.5) using an Ultraturrax T10 basic homogenizer (IKA, Staufen, Germany), followed by centrifugation at 100,000g for 1 hr. The pellet was resuspended in 100 μl of water and transferred into Pyrex glass tubes. After adding 1 ml of chloroform and 1 ml of methanol (to obtain a ratio of chloroform:methanol:water 10:10:1; v:v:v), samples were sonicated in an ultrasonic bath for 10 min. Samples were then centrifuged for 5 min at 4,000g and the supernatants were transferred into fresh Pyrex glass tubes. Two milliliter of Chloroform/methanol/water (10:10:1; v:v:v) was added to the pellet and the extraction was repeated. Pellet was reextracted with 2 ml of chloroform/methanol/water (

| LC-MS/MS analysis of sphingolipids
Extracted lipids were separated on Waters I class UPLC equipped with Waters CSH C18 column (length 100 mm, diameter 2.1 mm, particle size 1.7 μm) using a gradient starting with 57% solvent A (50% methanol, 50% water) and 43% solvent B (99% 2-propanol, 1% methanol), both containing 10 mM ammonium formiate, 0.1% formic acid and 5 μM sodium citrate (Table S2). UPLC was coupled to an ESI-(QqQ)tandem mass spectrometer (Waters Xevo TQ-S) for compound detection in +ESI MRM mode except for sulfatides in −ESI MRM mode (Table S3). Samples were injected and processed using MassLynx software, mass spectrometric peaks were quantified according to their peak area ratio with respect to the internal standard using TargetLynx software (both v 4.1 SCN 843) both from Waters Corporation (Manchester, UK). Subsequently, quantification of ceramides, hexosylceramides, lactosylceramides, sulfatides (SM4s), and sphingomyelins was adjusted to the length of the acyl-chain and dihydo(G)SL as well as phyto-(G)SL quantification were further adjusted by a factor calculated between the intensities of external ceramides and dihydroceramide or phytoceramide standards of same concentration.

| Statistical analysis
All graphs generated and the statistical analyses were performed using GraphPad Prism version 6.0. All experiments were repeated at least three times unless otherwise mentioned. In general, N represents the total number of experiments and n represents the number fish larvae used in total.   In adult zebrafish, we performed immunohistochemistry in paraffin sections where the antibody specifically recognized cells in brain and spinal cord (Figure 1e,f). 36K expression could be observed to co-localize with some olig2+ oligodendrocytes and oligodendrocyte precursor cells (OPCs) in both brain and spinal cord CNS regions at 3 dpf (Figure 1g). At 5 dpf co-localization of 36K with Mbp + myelinating oligodendrocytes could be clearly observed in these regions ( Figure S1e).

| 36K MO specifically and efficiently knocks down 36K protein expression
In order to investigate the function of 36K, we designed two independent antisense translation blocking MOs (MO1 and MO2) targeting different sites of the 5 0 initial regions of the 36K mRNA. Knockdown of 36K protein with 36K MO1 was confirmed by western blot (Figure 2a). MO1 injected fish showed~80% less 36K protein at 4 dpf in comparison to non-injected (referred to as uninjected) and control Morpholino (control MO) injected fish ( Figure S2a). 36K expression level could be rescued when MO1 was co-injected with a 36K mRNA (MO1 + RNA), in which the MO1 binding site was mutated so that MO1 cannot bind to the injected mRNA. In zfl injected with MO1, a reduction in body length was observed from 3 dpf till 5 dpf, which was rescuable by MO1 + RNA injections (Figure 2b,c). This reduction in body length was also observed with the second translation blocking MO to 36K (MO2) (Figure 2c). The human orthologue to zebrafish 36K is SDR12 (Morris et al., 2004). The length phenotype was not rescuable when MO1 was co-injected with the mRNA of SDR12, which MO1 should not be able to bind ( Figure S2b). An enlarged pericardial sac and slower yolk consumption were observed in some MO injected zfl (Figure 2b). Somite numbers were unaltered although the body length was reduced. In MO1 injected zfl, the expression of 36K in the  (Figure 2d,e) and spinal cord (Figure 2d,f) was virtually absent in comparison to the control MO group.

| 36K knockdown fish are less responsive than controls
We then examined if behavior is functionally altered in the MO1 injected zfl. For this purpose, we performed unprovoked swimming tests at 5 dpf where the zfl swim for 8 min and the total distance of swimming was measured. We found that the fish in MO1 group swam 60% shorter total distance in comparison to the controls (Figure 3a show this in more detail in Figure S3e-h. Finally, to investigate for effects on myelin sheath compaction, we compared electron micrographs between MO1 and control MO zfl at 5 dpf. We observed unusual cytoplasmic vacuolated structures in between the myelin lamellae of MO1 (Figure 4h).

| 36K regulates OPC numbers
To assess the number of OPCs following 36K down-regulation, we used Tg(olig2:GFP) zfl (Shin, Park, Topczewska, Mawdsley, & Appel, 2003). OPCs that differentiate into oligodendrocytes migrate from the pMN domain of the spinal cord toward the dorsal region (Ravanelli & Appel, 2015). Myelination starts from late 3 dpf and is highly dynamic at 5 dpf (Brösamle & Halpern, 2002;Kirby et al., 2006). We observed significantly fewer dorsally migrated olig2+ cells in the spinal cord at 3 dpf in the MO1 and MO2 groups in comparison to the control MO and non-injected groups but no differences in cell numbers within the ventral spinal cord (Figure 5a-c, Figure S4a). This effect could be rescued in MO1 + RNA group zfl suggesting that the effect on OPC numbers is specific to 36K down-regulation We then asked if there is 36K expression when there is no differentiation toward oligodendrocytes. We treated zfl from 36 hpf until 4 dpf, subsequent to neurogenesis (Ravanelli & Appel, 2015), with TrichostatinA (TSA) a histone deacetylase (HDAC) inhibitor, which has been formerly shown to inhibit migration ( Figure 5g) and differentiation of OPCs to oligodendrocytes completely (Takada & Appel, 2010).
We saw no 36K expression any more at 4 dpf in the TSA treated larvae at all (Figure 5f), suggesting 36K expression to be specific for differentiating and later also mature oligodendrocytes. We hypothesize that at an early time-point, a meagre amount of 36K is sufficient for enzymatic activity through which oligodendrocyte differentiation is influenced, and only at a later time-point of the larvae development, a large amount of 36K is present in differentiated oligodendrocytes and compact myelin.

| 36K acts on OPCs through Notch signaling
Several pathways have been described to be involved in neurogenesis and gliogenesis during development of the spinal cord (Briscoe & Novitch, 2008). We commissioned a microarray chip analysis to investigate mRNA expression levels of around 60,000 transcripts in the MO1 versus control MO zfl groups at 3 dpf. From this, we assessed for differences in expression levels of genes, involved in signaling pathways known to influence neurogenesis and gliogenesis. Sonic hedgehog (Shh) pathway has been shown to have a featured role in dorsoventral patterning (Fuccillo, Joyner, & Fishell, 2006). Since Notch signaling is known to influence the general fate of glial cells (Artavanis-Tsakonas, 1999;Fortini et al., 1993;Kim et al., 2008;Louvi & Artavanis-Tsakonas, 2006), TgBAC(gfap:gfap-GFP) zfl (Chen et al., 2009) were used to visualize astrocytes and radial glia.
We found increased Glial fibrillary acidic protein (Gfap) expression linked to EGFP in the MO1 injected zfl in comparison to control MO injected zfl (Figure 6a,e,f). This suggests increased differentiation toward astrocytes instead of oligodendrocytes due to up-regulated Notch. The fate of these cells, which are generated by the enhanced proliferation mentioned earlier (Figure 5h,j) can therefore be explained by increased proliferation and differentiation toward astrocytes. This is also supported by an increase (2.0-fold; p = .001; ZFIN ZDB-GENE-001103-2) in astrocyte specific sox9 (Pompolo & Harley, 2001) expression as shown by microarray analysis (Table S1). To confirm that the increased Gfap expression in MO1 zfl is not because of reactive gliosis, we analyzed macrophages and microglial genes from the microarray analysis and found no change here (Table S1). Dorsal root ganglion (DRG) cells are negatively regulated by Delta-Notch signaling (Cornell & Eisen, 2002). Using Tg(-8.4ngn1:GFP) zfl (Blader et al., 2003), we observed significantly fewer (~50%) DRG cells in the MO1 group at 3 dpf as well as at 5 dpf in four analyzed spinal segments (Figure 6g-j). However, motor neurons and myelination in PNS remained unaffected in MO1 zfl ( Figure S5a and Figure S6a, Table S1). In order to examine if axonal presynaptic innervation is affected, zfl were stained with Sv2 antibody, a marker for presynaptic terminals (Jonz & Nurse, 2003;Wan et al., 2010) and no obvious difference was found in the MO1 group ( Figure S6b). As F3/Contactin is a binding partner for oligodendrocyte Notch (Hu et al., 2003), we also analyzed the expression of contactin from the microarray but found it not to be altered (Table S1). To investigate if Notch up-regulation could be due to miR-132 regulation through the previously described miR-132/Ctbp2 circuit (Salta et al., 2014), we scanned for miR-132, Ctbp2, Sirt1 from our microarray data and also analyzed miR-132 by q-rt-PCR ( Figure S6c,d), but found no significant changes here.
MO injected zfl were bath immersed with DAPT in embryo medium from 24 hpf onward (Figure 7a) so that the period of gliogenesis is covered although initial motor neuron genesis is unaffected as explained in (Salta et al., 2014). The reduced body length phenotype could be rescued partially with 5 μg/ml DAPT (Figure 7b). The number of dorsally migrated olig2 positive cell was fully rescued in the DAPT treated MO1 group in comparison to DMSO treated MO1 group zfl ( Figure 7c,d).
Since the 36K knockdown phenotypes can be rescued with a gamma secretase inhibitor, we aimed to further validate increased access/activity of intra membrane gamma secretase cleavage in our MO1 zfl. Due to the lack of suitable antibodies to directly study increased NEXT cleavage by gamma secretase to release NICD in these zfl, we assessed the cleavage of Amyloid precursor protein (App) instead (Krishnaswamy, Verdile, Groth, Kanyenda, & Martins, 2009). Following alpha or beta secretase cleavage of App, gamma secretase cleaves the remaining C terminal fragments (CTFs) further to release the App intracellular domain (AICD). We found decreased CTF and increased AICD in MO1 zfl (Figure 7e,f). This further supports enhanced gamma secretase cleavage in 36K knockdown fish, which is in addition to the upregulated Notch ligands (deltaa) upstream of gamma secretase cleavage as described above (Figure 6c, Figure S5, Table S1). (j) Suggested schematic: When 36K is present in endogenous amounts, lipids in the membrane are unaltered, Notch ligands are unaffected, access and activity of gamma secretase is not increased. Hence, there is regular gamma secretase cleavage of Notch, NICD is at low levels and Notch targets are off. Oligodendrocytes differentiate normally. When 36K is knocked down, lipids in the membrane are altered. This might lead to an up-regulation of Notch ligands as well as increased access and activity of gamma secretase leading to an increased cleavage of NICD, and therefore up-regulating Notch signaling. Notch up-regulation inhibits oligodendrocyte differentiation while favoring further proliferation and development of glial astrocytes Oppermann, 2008;Persson et al., 2009). Since 36K belongs to the SDR family (Morris et al., 2004), we further hypothesized that the distribution of transmembrane Notch ligands as well as the activity of intra membrane gamma secretase cleavage of diverse transmembrane proteins like Notch or APP is altered due to a change in membrane lipids. We therefore performed thin layer chromatography (TLC) to assess changes in lipid composition between control MO and MO1 zfl groups. In total lipids, we did not observe obvious differences (data not shown), however, when sphingolipids were analyzed at 5 dpf, in 36K knockdown zfl we reproducibly observed a weaker band for a lipid comigrating with cerebrosides standard (GalC; Figure 7g, Figure S7a).

| DISCUSSION
In this study, we investigate the function of 36K during developmental myelination in the CNS of zebrafish. 36K is one of the most abundant proteins in zebrafish brain white matter (Gebriel et al., 2014). In contrast to humans and other mammals, zebrafish can successfully and functionally regenerate and remyelinate CNS axons following injury (Becker & Becker, 2008;Münzel et al., 2012;Münzel, Becker, Becker, & Williams, 2014). Although the presence of 36K protein in humans is only hypothetical, its mRNA has been found in CNS white matter (Morris et al., 2004). Thus, it is important to understand the specific role of 36K in zebrafish myelin. We suggest two functions for 36K, one at an early time-point in oligodendrocyte differentiation as described in this study, and possibly another subsequent yet uncharacterized structural function in compaction of CNS myelin. In addition to the expression of 36K in mature oligodendrocytes and compact myelin (Morris et al., 2004), we could detect early expression of 36K in some olig2+ OPCs, in the developing brain-stem region already at 3 dpf and in the spinal cord at 5 dpf. We speculate that for the early developmental function of 36K, only a little amount of protein/enzyme is necessary, whereas for its structural function a large amount of protein might be needed in compact myelin. Since 36K belongs to the short chain dehydrogenase family (Morris et al., 2004), and some SDR have been shown to regulate lipid metabolism (Kavanagh et al., 2008;Persson et al., 2009), we propose that 36K is involved in lipid metabolism as 36K knockdown larvae show alterations in membrane lipid composition. We assume that this lipid alteration, together with the increase in Notch ligands, is responsible for the enhanced activity of prerequisite intramembrane cleavage of gamma secretase substrates and the final cleavage by gamma secretase itself. This causes an up-regulation of Notch signaling and hence influencing oligodendrocyte differentiation as shown in this study ( Figure 7j).
In 36K knockdown larvae, the ratio of hexosylceramides containing sphinganine to sphingosine is decreased. This could be either due to an increase in hexosylceramides containing sphingosines or a reduction of hexosylceramides containing sphinganines. The actual molecular mechanism underlying this ratio change is yet to be charac- Disruptions in genes involved in ganglioside synthesis have been shown to cause myelination defects (Sheikh et al., 1999). The myelination defects in 36K knockdown zfl suggest a possible deficiency in axonal action potential conduction due to which, the zfl are less responsive than controls in stimulated behavior tests. Reduced startle responses have also been observed in hypo-myelinated mice models (Poggi et al., 2016;Tanaka et al., 2009), which would further support our hypothesis. It might be due to various reasons that only in about 30% of the larvae this startle response discrepancy is observed, for example, myelination is strongly impaired but there is still always some remaining myelination present in our knockdown model larvae.
Furthermore 36K knockdown zfl swim shorter distance in unprovoked swim tests. However, if 36K knockdown larvae do respond in our stimulated swim tests, they swim equal initial distances as their controls. Our observations therefore neither point toward changes in motor neurons as also shown from microarray analysis nor presynaptic axonal innervation as depicted by anti-Sv2 staining. If motor neurons would have been affected, the larvae would not swim equal distances like controls responding to a stimulus (Sonnack et al., 2015).
Altogether this argues in favor of neural conduction and myelination defects rather than muscular defects being causative in 36K knockdown.
Our observations suggest that 36K plays a major role in oligoden-  Tsakonas, 1999;Fortini et al., 1993;Grandbarbe, 2003;Kim et al., 2008). In accordance with this, we observe increased glial Gfap and sox9 expression in 36K knockdown larvae. As we do not find an upregulation in genes related to immune response, this increase in Gfap is most likely due to a cell fate change leading to increased astrocytes or radial glia rather than reactive astrogliosis.
Likewise the development of oligodendrocytes from olig2+ OPCs is known to be favored by down-regulation of Delta Notch signaling (Louvi & Artavanis-Tsakonas, 2006;. DAPT has been shown to indirectly down-regulate Notch by inhibiting gamma secretase (Geling, 2002;Salta et al., 2014). When larvae were treated with DAPT at a concentration known to inhibit Notch up-regulation (Salta et al., 2014), dorsally migrated olig2+ OPC numbers and shorter body length phenotype in 36K knockdown zfl could be rescued while control MO injected larvae remained unaffected. This further suggests an involvement of endogenous 36K early in oligodendrocyte development by inhibiting Notch signaling at the time of initial oligodendrocyte differentiation. A direct or indirect involvement of miRNA-132 in the regulation of oligodendrocyte or glial fate as described in (Salta et al., 2014) linked to the function of 36K could not be found in our study.
Mutations in pescadillo, a gene involved in ribosome biogenesis and proliferation and further described to be required for oligodendrocyte differentiation, have been associated with a shorter body length phenotype (Simmons & Appel, 2012). The length phenotype could be partially rescued when 36K knockdown larvae were treated with DAPT. Hence it might also be attributed to Notch upregulation, as Notch is known to be involved in many other developmental processes.
Custom-made antibodies raised specifically against zebrafish 36K could confirm the localization of 36K mainly in CNS tissue, which is in accordance with previous findings (Morris et al., 2004).  (Robu et al., 2007;Rossi et al., 2015). We did not observe an induction of p53 expression in the MO1 injected larvae, minimizing the possibility of off-target effects. To our knowledge, there is currently no knock out fish line for 36K available. We tried generating CRISPR-Cas9 mutants for 36K, but already failed to prove initial genome editing in the injected larvae. Generating a knockout mutant for 36K would have possibly allowed us to follow the function of 36K at later time points in compaction of myelin as well as in remyelination. We speculate that a complete knockout of 36K might not be viable and would suggest an inducible cell type specific knockout strategy for the future. This however would involve the challenge of identifying a CNS specific OPC and oligodendrocyte promoter, or rather the 36K promoter itself. Although injecting the Morpholino in a null mutant background to completely rule out any off-target effects would have further confirmed the specificity of the MOs, the usage of multiple MOs against the same target and rescue experiments strongly confirm the specificity of our observations.
Alternatively, to get hands on a zfl model without differentiated oligodendrocytes, we treated the larvae with TSA, a histone deacetylase inhibitor which inhibits migration and differentiation of oligodendrocytes (Takada & Appel, 2010). In this model we could confirm absence of migrating and differentiating OPCs, as well as absence of 36K in the treated larvae as shown by Western blot. Altogether, presence of 36K in compact myelin (Morris et al., 2004), its significantly high abundance in adult zebrafish brain (Gebriel et al., 2014), and complete absence of 36K when there are no differentiated oligodendrocytes after TSA treatment, suggest a later structural function for 36K possibly in compaction of myelin even after the stage of initial development. This structural function in compaction of myelin could be similar to the role of Mbp in the major dense line as previously suggested by (Jeserich & Rauen, 1990) since both of them are highly basic proteins (Morris et al., 2004).
In conclusion, our observations demonstrate a role for 36K in oligodendrocyte differentiation through Notch signaling pathway, in addition to a yet unknown possible structural function later in compact myelin. Notch signaling is affected by 36K knockdown in two possibly interconnected ways. First, this involvement is most likely due to an enzymatic role of 36K in lipid metabolism. Since membrane lipid composition is altered in 36K knockdown zfl, gamma secretase activity is enhanced and Notch signaling is up-regulated. Second, this is due to an up-regulation of Notch ligands in 36K knockdown zfl, although the direct cause for the transcriptional up-regulation of these ligands remains unknown. Both of these will lead to an upregulation of Notch signaling. In turn, fewer oligodendrocytes differentiate, leading to less myelin in 36K knockdown larvae. Understanding the function of 36K in zebrafish myelin improves the background knowledge for usage of zebrafish as a model for CNS myelination.
This could eventually provide further opportunities to better understand demyelinating diseases and remyelination in human CNS.

DECLARATION OF INTERESTS
The authors declare no competing interests.

DATA AVAILABILITY STATEMENT
Microarray data and all other relevant data are available within the article or from the authors upon reasonable request.