Fbxw7 is a critical regulator of Schwann cell myelinating potential

Myelin insulates and protects axons in vertebrate nervous systems. In the central nervous system (CNS), oligodendrocytes (OLs) make numerous myelin sheaths on multiple axons, whereas in the peripheral nervous system (PNS) myelinating Schwann cells (SCs) make just one myelin sheath on a single axon. Why the myelinating potentials of OLs and SCs are so fundamentally different is unclear. Here, we find that loss of Fbxw7, an E3 ubiquitin ligase component, enhances the myelinating potential of SCs. Fbxw7 mutant SCs are seen myelinating multiple axons in a fashion reminiscent of OLs as well as aberrantly myelinating large axons while simultaneously ensheathing small unmyelinated axons - typically distinct roles of myelinating SCs and non-myelinating Remak SCs, respectively. We found that several of the Fbxw7 mutant phenotypes, including the ability to generate thicker myelin sheaths, were due to dysregulation of mTOR. However, the remarkable ability of mutant SCs to either myelinate multiple axons or myelinate some axons while simultaneously encompassing other unmyelinated axons is independent of mTOR signaling. This indicates distinct roles for Fbxw7 in regulating multiple aspects of SC behavior and that novel Fbxw7-regulated mechanisms control modes of myelination previously thought to fundamentally distinguish myelinating SCs from non-myelinating SCs and OLs. Our data reveal unexpected plasticity in the myelinating potential of SCs, which may have important implications for our understanding of both PNS and CNS myelination and myelin repair.


INTRODUCTION
In vertebrates, oligodendrocytes (OLs) and Schwann cells (SCs) are specialized glial cells that generate the myelin sheaths of the central nervous system (CNS) and peripheral nervous system (PNS), respectively. In both the CNS and PNS myelin enables fast action potential propagation and protects the axon that it surrounds [1]. Myelin sheaths of the CNS and PNS are broadly similar in composition and structure, and there is a large degree of overlap in the molecular control of myelination by OLs and myelinating SCs [2,3]. However, there are important differences in how OLs and SCs interact with axons. One major fundamental difference between OLs and myelinating SCs is the ratio by which they myelinate axons. In peripheral nerves SCs select specific axons in a process called radial sorting, whereby they are thought to extend exploratory processes into a bundle of unmyelinated axons and select just one axon, greater than 1 µm in diameter, for myelination. Upon completion of radial sorting, promyelinating SCs are associated 1:1 with large-caliber axons (<1 µm in diameter), whereupon they initiate formation of a myelin sheath. In contrast, multiple small-caliber axons in peripheral nerves are ensheathed by non-myelinating Remak SCs into Remak bundles. The mechanisms controlling the Remak SC vs. myelinating SC fate are unclear. In the CNS, OLs also extend exploratory processes that dynamically interact with potential target axons [4], but unlike SCs, OLs are able to myelinate many axon segments [5]. Given the broad molecular similarities between myelinating SCs and OLs, it is unclear why SCs and OLs exhibit such differences in myelinating capacity.
The E3 ubiquitin ligase component F-box and WD-repeat domain containing 7 (Fbxw7) is an important regulator of OL development and CNS myelination [6][7][8]. Here, using SC-specific knockout approaches in mice, we demonstrate that Fbxw7 plays distinct and novel roles in SCs. Of principal interest is that, in the absence of Fbxw7, SCs gain the ability to myelinate multiple axons in a fashion reminiscent of OLs. Also surprising was the ability of Fbxw7 mutant SCs to generate myelin around large caliber axons while simultaneously ensheathing many additional small caliber axons. Electron microscopy and immunofluorescence (IF) analyses confirm that these cells are indeed SCs and not OLs that may have infiltrated the PNS. Additionally, Fbxw7 mutants display early increases in SC number, smaller Remak bundles, and hypermyelination which are ameliorated upon loss of mTOR.
However, even in the absence of mTOR, Fbxw7 mutant SCs retain the remarkable ability to myelinate multiple axons, as well as simultaneously myelinate large axons while ensheathing small unmyelinated axons. This suggests that the molecular mechanisms that regulate the fundamental differences in myelinating potential between SCs and OLs are independent of mTOR signaling.
Taken together, our findings show that the restriction of myelinating SCs to myelinate a single axon is not immutable and that Fbxw7 is a critical player in regulating the myelinating potential of SCs.

Fbxw7 cell-autonomously regulates SC development
Fbxw7 is a substrate recognition component of SKP1-Cullin-F-box (SCF) ubiquitin ligase complexes, which catalyze addition of ubiquitin moieties on certain proteins to target them for proteasomal degradation [9]. We and others have previously reported that fbxw7 zebrafish mutants display striking overexpression of myelin basic protein (mbp) in the CNS, enhanced OL numbers and thicker CNS myelin [6][7][8].
A role for Fbxw7 in the PNS has never been described. However, given the role of Fbxw7 in OLs, we hypothesized that Fbxw7 is also required in SCs. To test this, we employed a conditional knockout strategy in mice. The Dhh cre transgene results in Cre recombinase expression under the Desert hedgehog promoter at approximately embryonic day (E) 12.5 in SC precursors [10]. To delete Fbxw7 specifically in SCs, we crossed Dhh cre with an Fbxw7 fl/fl transgenic line in which loxP sites flank exons 5 and 6 of Fbxw7 [11]. This creates a frameshift upon Cre activity resulting in a null allele, which we confirmed by the absence of Fbxw7 mRNA by RT-PCR ( Figure S1A). In all cases, Dhh cre (-) ;Fbxw7 fl/+ and Dhh cre(-) ;Fbxw7 fl/fl siblings were used as controls.
We analyzed sciatic nerves by transmission electron microscopy (TEM) and found that at postnatal day 3 (P3), heterozygous Dhh cre(+) ;Fbxw7 fl/+ (Het) and homozygous Dhh cre(+) ;Fbxw7 fl/fl (cKO) mutant mice displayed an increase in SC nuclei as well as in the proportion of myelinated axons relative to controls (Figure 1 A-E). However, by P42, SC numbers were equivalent in mutant and wild type nerves ( Dhh cre(+) ;Fbxw7 fl/+ and Dhh cre(+) ;Fbxw7 fl/fl nerves relative to controls (Figure S1 B-P). These data demonstrate that Fbxw7 functions cell-autonomously to regulate multiple aspects of SC development.

Fbxw7 mutant Schwann cells can myelinate multiple axons
We also found that loss of Fbxw7 dramatically increased the myelinating potential of SCs ( highlight that these phenotypes are distinct from "polyaxonal myelination" in which a bundle of axons is myelinated together as though it was one larger axon (Supplementary Figure 2K), which has been previously reported in mutants with radial sorting defects and occurs at low frequency even in wildtype nerves [12].
In total, by cross-sectional TEM analyses, approximately 12% of Fbxw7 mutant SCs display enhanced myelinating potential, and this proportion remained consistent throughout the life of the animal (Figure 2 F, I). However, we found that in many cases multiple myelinated axons ( Figure 2I; Importantly, in both types of aberrant SC-axon interactions in Fbxw7 mutants, we were able to trace continuous SC cytoplasm and basal lamina on the abaxonal surface of the mutant SC ( Figure   2H, inset, white arrowhead). Since SCs secrete a basal lamina and OLs do not [13], the presence of a basal lamina strongly supports the notion that these cells are indeed SCs and not OLs that might have infiltrated the PNS. To further confirm that the mutant cells are not OLs, we analyzed Fbxw7 mutant sciatic nerves by immunofluorescence for markers of the OL lineage. We did not find any evidence of Olig2-positive cells (immature OL lineage cells) in either control or Fbxw7 mutant nerves (data not shown). These data suggest that the enhanced myelinating potential observed in Fbxw7 mutants is not due to OL infiltration. Instead, it suggests that mutant SCs are extending multiple cytoplasmic projections and that more than one of these processes can go on to make myelin.

Fbxw7 regulates mTOR to control SC number, myelination, and Remak bundle organization
To our knowledge, no genetic or pharmacological manipulation in vivo has been reported to increase the myelinating potential of SCs in the manner observed in Fbxw7 mutants. However, the hypermyelination and oversorted Remak bundle defects observed in Fbxw7 mutants have also been described following SC-specific deletion of Pten [14] or overactivation of Akt [15], which both result in elevated mTOR signaling. Previous reports show loss of Fbxw7 function enhances levels of mTOR and its targets [16], and Fbxw7 was recently shown to regulate mTOR in CNS myelination [7]. Therefore, we examined levels of mTOR and found that total mTOR protein levels are approximately 2-fold higher in Fbxw7 mutants relative to controls ( Figure 3A-B). Similarly, consistent with previous findings that multiple feedback loops regulate the PI3K/mTOR pathway [17], we found that the mRNA levels of mTOR as well as several targets of mTOR were also significantly elevated in Fbxw7 mutant nerves ( Figure 3C).

SC myelinating potential is independent of mTOR signaling
Strikingly, however, loss of mTOR in Fbxw7 mutant SCs was unable to restore normal SC myelinating potential. Despite the fact that Dhh cre(+) ;Fbxw7 fl/+ ;mTOR fl/fl nerves had normal numbers of SCs, delayed radial sorting, larger Remak bundles, and thinner myelin, loss of mTOR was insufficient to suppress the enhanced myelinating potential observed in Fbxw7 mutant SCs. Double mutant SCs were still seen myelinating multiple axons, as well as displaying myelinating/Remak "hybrid" phenotypes ( Figure  . These data suggest that Fbxw7 plays a novel role in controlling SC myelinating potential that is independent of mTOR signaling ( Figure 5). It is remarkable that SC myelinating capacity remains enhanced in Dhh cre(+) ;Fbxw7 fl/+ ;mTOR fl/fl mutants despite the fact that deletion of mTOR has caused reduced myelin thickness, delayed radial sorting, and defects in Remak SC ensheathment. This suggests that the morphological processes controlling the myelinating potential of SCs are distinct from the cellular behaviors involved in radial sorting, myelination, or Remak SC ensheathment.

DISCUSSION
In stark contrast to OLs in the CNS, which can myelinate dozens of axons simultaneously, myelinating SCs are restricted to myelinating a single axonal segment in the PNS. The molecular mechanisms controlling the differences in myelinating potential between SCs and OLs remain mysterious. Here we show that SCs are capable of myelinating multiple axons in vivo and challenge the notion that the ability of a SC to extend multiple processes is mutually exclusive with the capacity to make myelin. Remak SCs are also capable of extending multiple processes and interacting with many axons, but unlike OLs they do not myelinate axons. However, upon loss of Fbxw7, single SCs gain the ability to both myelinate some axons, as well as encompass many unmyelinated axons as if Fbxw7 mutant SCs are myelinating SC and Remak SC "hybrids." The multi-axonal myelination observed in Fbxw7 mutant SCs is reminiscent of an OL's ability to simultaneously extend multiple cytoplasmic processes and myelinate multiple axon segments. However, the presence of basal lamina and lack of OL-lineage markers suggests that OLs have not infiltrated Fbxw7 mutant nerves.
Perhaps Fbxw7 plays a role in inhibiting the myelinating potential of SCs by limiting the extent to which they can generate multiple processes, thus restricting myelinating SCs to a single axonal segment.
Although neither of the aberrant Fbxw7 mutant SC-axon interaction phenotypes have been previously described in vivo, several of the other phenotypes observed in Fbxw7 mutant nerves resemble phenotypes described in mutants where mTOR signaling is enhanced such as in Pten mutants [14] and constitutively active Akt mutants [15]. It is well documented, from these studies and others, that mTOR levels must be tightly regulated in SCs such that any type of manipulation results in defective PNS myelination [14,15,[19][20][21][22][23][24][25]. mTOR is a bona fide target of Fbxw7 in other contexts [16] and Fbxw7 was recently shown to control OL myelination through mTOR [7]. Thus, we tested the hypothesis that Fbxw7 regulates mTOR to control SC development. We showed that mTOR and some of its targets are upregulated in Fbxw7 mutants, suggesting that mTOR activity is elevated.
Double transgenic analysis demonstrated that mTOR is epistatic to Fbxw7 and is responsible for regulating early SC numbers, appropriate myelin thickness, and Remak bundle organization.
Notably, however, loss of mTOR in Fbxw7 mutant SCs was unable to restore typical SC:axon ratios. Despite the fact that Dhh cre(+) ;Fbxw7 fl/+ ;mTOR fl/fl nerves had normal numbers of SCs, delayed radial sorting, larger Remak bundles, and thinner myelin, loss of Fbxw7 was nevertheless sufficient to drive increased myelinating potential of SCs. This suggests that the myelinating potential of SCs is independent of mTOR signaling, and importantly, that the mechanisms controlling the typical SC:axon ratio is also controlled independently from other morphological behaviors of SCs, including radial sorting, Remak SC ensheathment, and membrane wrapping.
This highlights the complicated nature of removing a gene like Fbxw7, which is a master regulator of master regulators. Given the role of Fbxw7 in E3 ubiquitin ligase complexes, it is likely that there is an as yet unidentified target protein that is dysregulated in Fbxw7 mutants that is responsible for the enhanced myelinating potential. However, it is equally possible that the complex phenotypes observed in Fbxw7 mutant nerves result from combinatorial interactions amongst multiple misregulated targets, or that this represents a novel function of Fbxw7.
A potentially interesting future direction would be to explore the hypothesis that Fbxw7 mutant SCs are behaving similarly to repair SCs. After PNS injury, SCs readily become repair SCs, which clear axon and myelin debris, recruit macrophages, aid in axon regrowth, and finally remyelinate axons [26]. A recent study on the length and morphology of SCs throughout development, differentiation, and remyelination found that repair SCs become extremely long and branched after injury [27]. Given the observation that myelinated axons were joined by long thin processes of SC cytoplasm in Fbxw7 mutants it is possible that Fbxw7 mutant SCs have also assumed a branched morphology. Although dispensable during SC development, the key switch controlling the transition to the repair SC phenotype is upregulation of the transcription factor c-Jun [28,29]. Importantly, c-Jun is a well-known target of Fbxw7 in other contexts [30], and loss of Fbxw7 has previously been shown to elevate c-Jun levels [31]. Recent evidence also suggests that mTOR is transiently reactivated after nerve injury to promote the elevation of c-Jun [21]. Thus, future efforts should assess c-Jun levels in In contrast to SC remyelination, OL remyelination and CNS recovery after injury is limited in mammals, and OLs do not de/redifferentiate to aid in recovery [32]. One interesting hypothesis is that the distinction between the ability of SCs vs. OLs to facilitate repair after injury is rooted in fundamental qualities that distinguish these cells, such as the number of axons they associate with and myelinate. It is now clear that OLs interpret and respond to neuronal activity by selective myelination and that OLs play a critical role in circuit development and/or maintenance [33][34][35][36]. It is intriguing to speculate that OLs may link circuits through their interactions with and myelination of multiple axons. However, in the PNS, where circuits are less complex, there is not strong evidence for SC participation in circuit function. Rather, more emphasis seems to be on the ability of SCs to rapidly and faithfully respond to injuries, which occur more readily in the PNS. Presumably this rapid response would be more difficult if SCs were required to demyelinate more than one axon, especially if only some of those axons were injured, while others were intact. Given that Fbxw7 mutant SCs provide the first in vivo description of multi-axonal myelination by SCs, Fbxw7 mutants represent a unique and useful tool with which to investigate the impact of differences between myelinating SCs, Remak SCs, and OLs on nervous system repair.      Error bars depict S.D. One-way ANOVA for all measures except g-ratio, where we used two-way ANOVA in order to compare by both genotype and axon diameter; * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001. All mouse lines were genotyped as previously described [10,11,37].

Transmission electron microscopy (TEM)
TEM was performed on mouse sciatic nerves at P3, P21, P42 and 6 months as previously described [38]. Briefly, nerves were drop-fixed in modified Karnovsky's fixative (4% PFA, 2% glutaraldehyde, 0.1M sodium cacodylate, pH 7.4) at least overnight at 4°C. Samples were then washed with 0.1M sodium cacodylate to remove fixative, and then post-fixed for 1 hour in 2% osmium tetroxide in 0.1M sodium cacodylate. Nerves were then dehydrated with increasing concentrations of ethanol followed by propylene oxide (PO). Samples were then infiltrated for 1-2 hours in 2:1 PO:EPON, and then overnight in 1:1 PO:EPON with gentle agitation at room temperature. Samples were then transferred to 100% EPON while residual PO was allowed to fully evaporate (>4 hours). Sectioning, grid staining, imaging, and image analyses were performed as described for zebrafish. Oregon Health & Science University. Images were acquired every 50 nm using an FEI Teneo VolumeScope Microscope. Sections were annotated using FIJI and then the movie was composed using Microscopy Image Browser software.

RNA isolation and reverse transcription
Total RNA was extracted from flash-frozen P21 mouse sciatic nerves (N=3 Dhh Cre(-) ;Fbxw7 fl/fl littermate controls and N=3 Dhh Cre(+) ;Fbxw7 fl/fl animals), using a standard TRIzol extraction protocol (Life Technologies, ThermoFisher Scientific, Waltham, MA). Briefly, TRIzol was added to the frozen tissue samples, which were then allowed to thaw at room temperature for 10 min. During this incubation time, and while still in TRIzol, nerves were cut into much smaller pieces using microdissection scissors. Samples were homogenized via disruption with a plastic-tipped electric homogenizer, followed by passage through syringe and 22.5g needle at least ten times, and then a 27g needle at least ten more times until no lumps of tissue were observed. Once the nerves had been homogenized, we proceeded as usual with the standard TRIzol RNA extraction procedure as per manufacturer instructions.
Total RNA (500 ng) was then reverse transcribed in 20 µl using the Superscript III First Strand cDNA Synthesis Kit (Invitrogen, ThermoFisher Scientific, Waltham, MA) using random hexamers, as per manufacturer instructions. All cDNA products were diluted 1:5 prior to use in qPCR reactions.

Quantitative reverse transcription PCR
To assay mRNA expression levels of mTOR and members of the mTOR signaling pathway, we used the RT 2 Profiler PCR Array for Mouse mTOR Signaling (Qiagen, PAMM-098ZA, Valencia, CA). A complete gene list can be found on the manufacturer's website. All assays were performed on a ViiA7 Real-Time PCR system (Applied Biosystems, ThermoFisher Scientific, Waltham, MA), in a total volume of 10 µl using 2X SsoFast Evagreen Supermix (BioRad, Hercules, CA) and 50 ng of cDNA per reaction. Standard qPCR settings were used: 95°C for 10 min followed by 40 cycles of 95°C for 15 seconds (sec) then 60°C for 30 sec, followed by melt curve analysis. As suggested by the RT 2 profiler manual, we adjusted the ramp rate to 1°C/sec. All controls including housekeeping genes, positive controls for amplification, and controls for genomic DNA contamination were included as standards in the array. qPCR data was analyzed using Microsoft Excel. Relative expression was calculated using the ΔΔCt method [39]. Genomic contamination was negligible in all samples. To control for input variations, ΔCt was calculated by comparing the Ct of each gene of interest (GOI) to the average Ct of the 5 housekeeping genes (Actb, B2m, Gapdh, Gusb, Hsp90ab1) for that sample. ΔΔCt was then calculated relative to expression compared to that seen in the littermate control. Average relative expression (RQ), or fold change (2-ΔΔCt), over controls is shown in Figure S3. All error bars depict RQmax and RQmin, which represent the maximum and minimum limits of possible RQ values based on the standard error of the ΔCt values. The grey line at y = 1 represents the controls.
QUANTIFICATION AND STATISTICAL ANALYSIS All data are reported as mean + standard deviation (S.D.). Statistically significant differences were determined using one-way ANOVA for all experiments with more than two groups but only one dependent variable. Similarly, two-way ANOVA was used for experiments with multiple groups and two dependent variables. All experiments with only two groups and one dependent variable were compared using an unpaired t-test with Welch's correction, which assumes unequal variance. Figure   legends specify which test was used for specific experiments. In all cases * = p<0.05; ** = p < 0.01; *** = p<0.001; and **** = p<0.0001; NS = not significant. A "#" symbol was used to highlight data that was trending towards significance (p<0.1). In all cases, asterisks immediately above a bar indicate the significance of that sample relative to the control sample. If any other comparisons, such as Dhh Cre(+) ;Fbxw7 fl/+ (Het) to Dhh Cre(+) ;Fbxw7 fl/fl (cKO), were significant, this is indicated with a bar spanning above the two samples being compared with the appropriate asterisks. If not indicated otherwise, the comparison was not significant. In most cases, Dhh Cre(+) ;Fbxw7 fl/+ samples were not statistically distinguishable from Dhh Cre(+) ;Fbxw7 fl/fl .