Abstract
White matter abnormalities are an emerging pathological feature of schizophrenia. However, their attributions to the disease remain largely elusive. ErbB receptors and their ligands, some of which are essential for peripheral myelination, confer susceptibility to schizophrenia. By synergistically manipulating ErbB receptor activities in a oligodendrocyte-stage-specific manner in mice after early development, we demonstrate the distinct effects of ErbB signaling on oligodendrocytes at various differentiation states. ErbB overactivation, in mature oligodendrocytes, induces necroptosis causing demyelination, whereas in oligodendrocyte precursor cells, induces apoptosis causing hypomyelination. In contrast, ErbB inhibition increases oligodendrocyte precursor cell proliferation but induces hypomyelination by suppressing the myelinating capabilities of newly-formed oligodendrocytes. Remarkably, ErbB inhibition in mature oligodendrocytes diminishes axonal conduction under energy stress and impairs working memory capacity independently of myelin pathology. This study reveals the etiological implications of oligodendrocyte vulnerability induced by ErbB dysregulation, and elucidates the pathogenetic mechanisms for variable structural and functional white matter abnormalities.
Introduction
Adolescence is the critical period for the central nervous system (CNS) to completely develop and mature. In particular, CNS myelin generated by oligodendrocytes (OLs) is one of the most developmentally active component in the adolescent brain. This may lead to CNS myelin being a highly susceptible target in psychiatric disorders such as schizophrenia which typically develops during adolescence (Fields, 2008; Hoistad et al., 2009; Kessler et al., 2007; Peters and Karlsgodt, 2015). A growing body of literature points to abnormalities in the structure, component proteins, or regulating molecules of CNS myelin in schizophrenic patients (Douaud et al., 2007; Fields, 2008; Hof and Schmitz, 2009; Hoistad et al., 2009; Kelly et al., 2018; Uranova et al., 2011). Schizophrenia is increasingly viewed as a spectrum disorder based on varied symptom severity and genetic risk. Especially, white-matter microstructural changes as examined by structural brain imaging techniques are sensitive to the symptom severity or genetic loading of schizophrenic patients (Karlsgodt, 2020). Therefore, understanding schizophrenia related myelin pathogenesis is crucial for the development of diagnostic standards or therapeutic targets given that it is one of the most promising features whose progression can be examined periodically in patients.
Tyrosine kinase receptors ErbB(1-4) mediate the signaling of numerous growth factors which are categorized into the neuregulin (NRG) family and the epidermal growth factor (EGF) family (Iwakura and Nawa, 2013; Mei and Nave, 2014). The NRG and EGF family members bind differentially to the four ErbB receptors. Due to the indispensable function of NRG1-ErbB signaling in peripheral myelination (Nave and Salzer, 2006), it was expected that NRG-ErbB signaling played a role in CNS myelin development. However, the contradictory results from different research groups have silenced any significance of this previous postulate (Brinkmann et al., 2008; Makinodan et al., 2012; Schmucker et al., 2003; Taveggia et al., 2008). Genetic ablation of NRG1 or ErbB4, the ligand and receptor that have received extensive attention from researchers in schizophrenia field (Harrison and Law, 2006; Mei and Nave, 2014), induces neither developmental alteration nor pathogenesis in white matter of mutant mice (Brinkmann et al., 2008). However, studies combining genetic linkage analysis and brain imaging techniques have associated NRG1 or ERBB4 variability with reduced white matter density or integrity in human subjects (McIntosh et al., 2008; Winterer et al., 2008; Zuliani et al., 2011).
Notably, in addition to NRG1 and ErbB4, many molecules in the ErbB signaling pathways exhibit single nucleotide polymorphisms (SNPs) or aberrant expression that are implicated in schizophrenia or other psychiatric disorders. Both gain and loss of ErbB signaling have been indicated by genetic and biochemical studies (Harrison and Law, 2006; Mei and Nave, 2014). Particularly, NRG1 and ErbB4 have been revealed to increase the mRNA levels, protein levels, or receptor activity in the schizophrenic brain (Chong et al., 2008; Hahn et al., 2006; Joshi et al., 2014; Law et al., 2006; Law et al., 2012). It is noteworthy that EGFR (ErbB1), which only binds the EGF family ligands, is elevated in the brain of schizophrenic patients (Futamura et al., 2002) and shows potential in regulating oligodendrogenesis in developmental and pathological conditions (Aguirre et al., 2007). Thus, NRG-ErbB and EGF-ErbB signaling may be synergistic in the regulation of CNS myelin integrity.
In the CNS, OL precursor cells (OPCs) after terminal mitosis differentiate into newly-formed OLs (NFOs). NFOs then span differentiation states from pre-myelinating OLs to newly myelinating OLs. Myelinating OLs effectively generate myelin sheaths in a short time window before further differentiating into mature OLs (MOs) that maintain the myelin sheath (Bergles and Richardson, 2015; Czopka et al., 2013; Hughes et al., 2018; Tripathi et al., 2017; Watkins et al., 2008; Xiao et al., 2016). To study whether ErbB receptors, through mediating NRG-ErbB and EGF-ErbB signaling, cooperate to regulate OLs and CNS myelin, we adopted an inducible pan-ErbB strategy and manipulated ErbB receptor activities specifically in OL lineage cells in vivo. This strategy allowed us to avoid characterizing the complex composition of ErbB ligands or receptors in OL lineage cells and helped us focus on their cellular function. The results reveal that ErbB dysregulation differentially affects OPCs, NFOs, and MOs, leading to CNS demyelination, hypomyelination, and even OL dysfunction that causes cognitive deficits independently of myelin pathology.
Results
ErbB overactivation in OLs induces demyelination and hypomyelination
Studies on OL-specific knock-out mice have validated the expression of ErbB3 and ErbB4 in OLs (Brinkmann et al., 2008), while phosphorylated EGFR is detected in OLs by immunostaining (Palazuelos et al., 2014). We characterized the expression of ErbB receptor members in subcortical white matter regions at different postnatal days. Note that the subcortical white matter regions isolated from mice before P5 had few myelin components. Our results indicate that ErbB2 is barely expressed in mouse CNS myelin, whereas EGFR, ErbB3 and ErbB4 are expressed with relatively stable levels during juvenile to adolescent development (Figure 1-figure supplement 1A,B).
To manipulate ErbB receptor activities in the OL lineage of mice, we employed tetracycline-controlled systems whose induction or blockade depends on the presence of doxycycline (Dox). We first characterized the impact of elevated ErbB receptor activities on CNS myelin by generation of Plp-tTA;TRE-ErbB2V664E (Plp-ErbB2V664E) bi-transgenic ‘Tet-Off’ mice (Figure 1A). Notably, Plp-ErbB2V664E mice around P35 exhibited severe ataxia while walking on a grid panel (Figure 1B). Moreover, Plp-ErbB2V664E mice showed difficulty in rolling over, indicating severely impaired motor coordination.
With the expression and phosphorylation of ectopic ErbB2V664E, endogenous ErbB receptors (EGFR, ErbB3 and ErbB4) were strikingly phosphorylated and activated in the white matter of Plp-ErbB2V664E mice (Figure 1C,D). Overactivation of ErbB receptors caused lower myelin staining intensity as exhibited in the corpus callosum of Plp-ErbB2V664E mice at 9 days post Dox-withdrawal (dpd) after LFB staining. Notably, at 14 dpd, myelin loss became more evident throughout the corpus callosum, suggesting that Plp-ErbB2V664E mice were undergoing CNS demyelination after Dox withdrawal (Figure 1E,F). Consistently, western blotting revealed loss of myelin basic protein (MBP), an indicator for MOs and myelin, in the brain of Plp-ErbB2V664E mice (Figure 1G). Moreover, the electron microscopic examination (EM) of the myelin ultrastructure revealed that myelin sheaths of some axons in Plp-ErbB2V664E mice ruptured or underwent breakdown (Figure 1H,I and Figure 1-figure supplement 2A), consistent with the idea of demyelination. Due to demyelination, only a few intact axons were detected in the midline of the corpus callosum of Plp-ErbB2V664E mice at 14 dpd (Figure 1H,I). When axonal tracts in the corpus callosum were immunostained by TuJ1, the antibody recognizing neuronal specific β-tubulin III, the immunoreactivity dramatically reduced in Plp-ErbB2V664E mice at 14 dpd (Figure 1-figure supplement 2B). In addition, as a pathological condition, demyelination is usually complicated and aggravated by the pathological responses from nearby astrocytes and microglia. Indeed, in the white matter of Plp-ErbB2V664E mice, astrogliosis and microgliosis were revealed (Figure 1J,K).
Interestingly, despite the demyelination, the detectable axons, throughout the corpus callosum, optic nerve, and prefrontal cortex in Plp-ErbB2V664E mice, were hypermyelinated. Myelinated axons detected in the brain of Plp-ErbB2V664E mice had significantly smaller g-ratio (axon diameter/fiber diameter), a quantitative indication of myelin thickness for individual axons with different diameters (Figure 1I). Myelin thickness showed no difference between TRE-ErbB2V664E and littermate Plp-tTA mice after Dox withdrawal (Figure 1-figure supplement 2C). Therefore, hypermyelination of detectable axons in Plp-ErbB2V664E mice was a result of the overexpression of ErbB2V664E, which was detected by an antibody against ErbB2 (Figure 1C,G).
Hypermyelination of individual axons in Plp-ErbB2V664E mice phenocopied that observed in NRG1-overexpressing mice (Brinkmann et al., 2008). Notably, EM examination of the ultrastructure in Plp-ErbB2V664E mice at 9 dpd revealed that most axons were intact in the midline of the corpus callosum, although they have been significantly hypermyelinated (Figure 1-figure supplement 2D,E). These results confirmed that hypermyelination occurs early in Plp-ErbB2V664E mice, and demyelination and axonal degeneration in Plp-ErbB2V664E mice are pathological events induced secondarily by continuous ErbB activation.
Next, we examined the effects of ErbB activation on CNS myelin by a ‘Tet-on’ system generated in Sox10+/rtTA;TRE-ErbB2V664E (Sox10-ErbB2V664E) mice (Figure 1L). Sox10-ErbB2V664E mice with Dox feeding from P21 developed severe motor dysfunction, including ataxia and tremors, and died around P35. As a result, Sox10-ErbB2V664E and littermate control mice were investigated at P30 after 9 days with Dox-feeding (dwd). These mice had smaller body sizes at P30 and walked with difficulty on a grid panel (Figure 1M). Western blotting revealed the expression and phosphorylation of ectopic ErbB2V664E accompanied with increases in phosphorylation of ErbB3 and ErbB4, but not that of EGFR, in the white matter of Sox10-ErbB2V664E mice (Figure 1N,O). Brain slices stained by LFB exhibited lower staining intensity in the white matter of Sox10-ErbB2V664E mice (Figure 1P,Q), consistent with the lower MBP levels detected by western blotting (Figure 1N,O). The examination of the ultrastructure by EM revealed that the axons in the corpus callosum and optic nerve of Sox10-ErbB2V664E mice exhibited thinner myelin with significantly increased g-ratio (Figure 1S). Sox10+/rtTA is a knock-in mouse line, so that the allele with Sox10-rtTA would not transcribe Sox10 mRNA (Ludwig et al., 2004). We analyzed the ultrastructure of myelinated axons in Sox10+/rtTA and littermate TRE-ErbB2V664E mice at P30 and did not observe any differences (Figure 1-figure supplement 2F), therefore we can exclude the possible effect of haploinsufficiency of Sox10 on late postnatal myelin development.
It is notable that in Sox10-ErbB2V664E mice, the numbers of myelinated axons were not altered (Figure 1R), and myelin sheaths exhibited normal morphology (Figure 1S). Moreover, neither microgliosis nor astrogliosis was detected in the white matter of Sox10-ErbB2V664E mice (Figure 1T,U). Because there was no indication of inflammatory pathogenesis, we can conclude that thinner myelin in Sox10-ErbB2V664E white matter are caused by developmental deficits not pathological conditions. Therefore, ErbB activation in Sox10-ErbB2V664E mice induces CNS hypomyelination rather than demyelination.
Plp-tTA targets mainly MOs whereas Sox10+/rtTA targets OPC-NFOs
The finding that Plp-ErbB2V664E and Sox10-ErbB2V664E mice had no overlaps in histological and biochemical phenotypes was unexpected considering that Sox10 is reported to express throughout the OL lineage, and Sox10+/rtTA knock-in mice have been used to investigate all OL lineage cells (Wegener et al., 2015). Because the induction of the ‘Tet-on’ or ‘Tet-off’ system by Dox feeding or Dox withdrawal has a delayed effect on gene expression, and reporter proteins could accumulate to label the consecutive cellular stages, the results obtained by using TRE-controlled reporter mice fail to accurately reveal the original cells targeted by tTA or rtTA. To circumvent this problem, we delivered a TRE-controlled fluorescence reporter carried by an adeno-associated virus (AAV) into the mouse brains at P14 or P35, and examined tTA- or rtTA-targeted cells as well as their derivatives within 2 days.
Plp-tTA mice were raised with no Dox feeding, whereas Sox10+/rtTA were fed with Dox for 3 days before the stereotactic injection (Figure 2A). One or 2 days after virus injection, the reporter-expressing (YFP+) cells were all immunopositive for Olig2 in both mouse lines at either age (Figure 2-figure supplement 1A-D), confirming their OL lineage specificity. To analyze the differentiation stage specificity, we immunostained AAV-TRE-YFP infected brain slices with an antibody for NG2 or PDGFRα that labels OPCs, or the antibody CC1 that labels post-mitotic OLs. The results showed that very few (4-7%) YFP+ cells were OPCs, while 92-97% of them were post-mitotic OLs 1 or 2 days after virus injection in Plp-tTA mice at either age, as well as in Sox10+/rtTA mice at P14 (Figure 2B,C and Figure 2-figure supplement 1A-D). However, for Sox10+/rtTA mice at P35, approximately 26% of YFP+ cells were OPCs 1 day after virus injection, and it decreased to 8% after 1 more day (Figure 2C). It is known that OPCs can differentiate into NFOs as quick as 2.5 hours (Xiao et al., 2016). These results may suggest that most OPCs targeted by Sox10-rtTA at P35 are undergoing terminal differentiation (tOPCs), and Sox10-rtTA increasingly targets tOPCs from P14 to P35.
There are also OL lineage cells belonging to the NFO stage that includes a transition from CC1- to CC1+. β-catenin effector TCF4 is specifically expressed in the NFO stage (Fancy et al., 2009; Fu et al., 2009; Ye et al., 2009), which is present in a subset of Olig2+ cells but is absent in OPCs (PDGFRα+) in mice at P30 (Figure 2D). In Sox10+/rtTA mice at P14, immunostaining revealed that approximately 49% of YFP+ cells were NFOs (TCF4+) 1 day after virus injection, but it reduced to 28% after 1 more day (Figure 2E,G). TCF4+ cells found in the corpus callosum at P35 were far fewer than P14, and these cells appeared as clusters (Figure 2-figure supplement 1E). Interestingly, in Sox10+/rtTA mice at P35, YFP+ cells were mostly found in regions with TCF4+ cell clusters (Figure 2-figure supplement 1E), where approximately 56% of YFP+ cells were TCF4+ 1 day after virus injection and it reduced to 29% after 1 more day (Figure 2F,G). The half reduction of NFO percentage in YFP+ cells from day 1 to day 2 was consistent with the previous report that NFOs differentiate into MOs in 1 or 2 days (Xiao et al., 2016). There was another possibility that the transcriptional activity of Sox10-rtTA was low in MOs, and thus took more days to generate detectable YFP levels. We analyzed the densities of TCF4+YFP+ cells and found that they reduced to half from day 1 to day 2 after viral infection in Sox10+/rtTA mice at either P14 or P35 (Figure 2H). The results excluded the possibility that the reduction of NFO ratio in YFP+ cells was due to the increase of targeted MOs, and confirmed the maturation of labeled NFOs from day 1 to day 2. A similar transition rate is applicable for targeted NFOs from day 0 to day 1. Therefore, these results corroborate that the majority of cells targeted by Sox10-rtTA at the time of AAV-TRE-YFP infection were tOPCs and NFOs (Figure 2I).
AAV-TRE-YFP in Plp-tTA mice also labeled some TCF4+ cells, which comprised only 7-12% of YFP+ cells 1 or 2 days after virus injection at either age (Figure 2E-H). It was noticeable that Plp-tTA did not specifically target TCF4+ cell clusters in the corpus callosum at P35 (Figure 2-figure supplement 1F). These results implied that Plp-tTA did not specifically target the tOPC or NFO stage but randomly expressed in the OPC-NFO period at a low ratio. Conversely, 90% of the YFP+ cells were TCF4- and 92-97% were CC1+ in Plp-tTA mice at either P14 or P35, suggesting that Plp-tTA steadily targets MOs after early development (Figure 2I).
ErbB overactivation causes MO necroptosis and OPC apoptosis
With an understanding of the differentiation stage-specific targeting preferences of Sox10+/rtTA and Plp-tTA mice, we investigated the cellular mechanisms that determine the different myelin responses in Plp-ErbB2V664E and Sox10-ErbB2V664E mice. We found intact post-mitotic OLs (CC1+) decreased in the corpus callosum of Plp-ErbB2V664E mice starting from 6 dpd (Figure 3A-C). The CC1+ cells with striking number reduction were MOs because the densities of NFOs (TCF4+) did not show significant change (Figure 3-figure supplement 1B,C). Meanwhile, OPCs (NG2+Olig2+) pathologically regenerated (Ki67+Olig2+) in the trunk of corpus callosum, indicating that the pathogenetic factor was released in the myelin-enriched region (Figure 3-figure supplement 1A-E).
MO number reduction occurred earlier than the time when demyelination was obviously observed in the corpus callosum of Plp-ErbB2V664E mice, suggesting that oligodendropathy may be the cause of demyelination. We examined the corpus callosum of Plp-ErbB2V664E mice by TdT-mediated dUTP nick end labeling (TUNEL) assay, and observed as little apoptotic signaling as that in controls (Figure 3D,E). This result reveals that the degenerating CC1+ cells were necrotic rather than apoptotic. Consistently, the OL nuclei associated with the destroyed myelin sheaths in Plp-ErbB2V664E mice were regular nuclei without apoptotic chromatin condensation (Figure 1-figure supplement 2A). In support of this theory, MLKL, the protein mediating cell necroptosis (Cai et al., 2014) as well as the peripheral myelin breakdown after nerve injury (Ying et al., 2018), demonstrated an increased expression in the callosal CC1+ cells in Plp-ErbB2V664E mice from 6 dpd (Figure 3F-H). Necroptosis is a programmed form of necrosis. RIP3 is the kinase at the upstream of MLKL in this programmed death signaling pathway (Ofengeim et al., 2015; Sun et al., 2012). Notably, the expression of RIP3 was also elevated in the callosal CC1+ cells in Plp-ErbB2V664E mice from 6 dpd as revealed by both immunostaining and western blotting (Figure 3F-H). Based on the timeline, MO necroptosis was the primary defect induced in Plp-ErbB2V664E mice, followed by myelin breakdown, OPC regeneration, axonal degeneration, and other pathological events as reported in multiple sclerosis (Bradl and Lassmann, 2010; Ofengeim et al., 2015).
In contrast, for Sox10-ErbB2V664E mice, a dramatic increase in cell apoptosis (TUNEL+) in the corpus callosum was observed (Figure 3E,I,J). These apoptotic nuclei were localized in NG2+ cells (Figure 3K), indicating apoptosis of OPCs. On the other hand, no increase of MLKL or RIP3 was detected (Figure 3G,H,L), indicating there was no necroptosis. Notably, both the NG2+ cells with and without TUNEL+ nuclei were hypertrophic in Sox10-ErbB2V664E mice (Figure 3K). This phenomenon was not revealed for NG2+ cells in Plp-ErbB2V664E mice (Figure 3-figure supplement 1B).
ErbB inhibition in OPC-NFOs, but not in MOs, induces hypomyelination
Next, to investigate whether ErbB receptors are functionally required for OLs, we examined the effects of inhibiting ErbB activities in different OL stages on CNS myelin. To this end, we first generated Sox10+/rtTA;TRE-dnEGFR (Sox10-dnEGFR) mice (Figure 4A and Figure 4-figure supplement 1A,B). In line with the inhibitory effect of dnEGFR on endogenous ErbB activities (Chen et al., 2017), phosphorylation of ErbB3 and ErbB4 was reduced in white matter from Sox10-dnEGFR mice at P35 (Figure 4B,C). The phosphorylation of EGFR was not altered in Sox10-dnEGFR mice at P35 (Figure 4B,C), consistent with the finding in Sox10-ErbB2V664E mice (Figure 1N,O). Myelin thickness and ultrastructures in the white matter of Sox10-dnEGFR and littermate control mice at P35 with 14 dwd did not show significant differences (Figure 4-figure supplement 1C). Therefore, we raised these mice to adulthood with continuous Dox feeding. Phosphorylation of EGFR, instead of ErbB3 or ErbB4, was apparently reduced in the white matter of Sox10-dnEGFR mice at P65 (Figure 4B,C). This change of ErbB receptors targeted by dnEGFR in Sox10-dnEGFR mice implied a switch of functional NRG-ErbB signaling to EGF-ErbB signaling in Sox10-rtTA-targeted cells from adolescence to adulthood.
For Sox10-dnEGFR mice at P65 with 44 dwd, myelin stained by LFB exhibited reduced intensity in the trunk of the corpus callosum (Figure 4D). Moreover, axons in the corpus callosum and optic nerve of Sox10-dnEGFR mice at P65 were hypomyelinated in comparison with that of littermate controls (Figure 4E). Consistently, MBP was reduced in the white matter of Sox10-dnEGFR mice at P65 (Figure 4B,C). Therefore, ErbB inhibition in OPC-NFOs starting from P21 results in hypomyelination in adulthood.
On the other hand, we crossed Plp-tTA and TRE-dnEGFR to generate Plp-tTA;TRE-dnEGFR (Plp-dnEGFR) mice (Figure 4F and Figure 4-figure supplement 1D,E). Western blotting revealed significant suppression on the phosphorylation of EGFR, as well as a mild suppression on that of ErbB3 and ErbB4 (Figure 4G,H), consistent with their overactivation in Plp-ErbB2V664E mice (Figure 1C,D). No CNS myelin differences were observed in Plp-dnEGFR and littermate control mice at P35 with 14 dpd (Figure 4-figure supplement 1F).
We extended our investigation to P65 when dnEGFR still functionally suppressed ErbB receptor activities in the white matter of Plp-dnEGFR mice (Figure 4G,H). Even in the adult mice, when dnEGFR had been expressed in MOs for 44 days, the brains of Plp-dnEGFR and littermate mice exhibited no difference in LFB-stained myelin (Figure 4I), MBP protein levels (Figure 4G,H), or myelin ultrastructures (Figure 4J). These results suggest that the dual blockade of endogenous NRG-ErbB and EGF-ErbB signaling in MOs does not affect CNS myelin integrity, and that ErbB activities are not required for the maintenance of CNS myelin after maturation.
ErbB activation blocks OPC proliferation and survival, whereas promotes NFO differentiation and myelination
It is intriguing to note that both ErbB inhibition and overactivation in OPC-NFOs result in hypomyelination in the CNS. This implies that ErbB signaling can regulate CNS myelin development by different mechanisms. To investigate the mechanisms, we first compared the states of OL lineage cells in Sox10-ErbB2V664E and Sox10-dnEGFR mice. In line with the finding that OPCs underwent apoptosis, the numbers of Olig2+, NG2+, and CC1+ cells significantly decreased in the corpus callosum of Sox10-ErbB2V664E mice (Figure 5A and Figure 5-figure supplement 1A), while there was no OPC or OL differences between Sox10+/rtTA and TRE-ErbB2V664E littermates at P30 with 9 dwd (Figure 5-figure supplement 1A). Further, we found proliferating OPCs (Ki67+Olig2+) significantly decreased in the corpus callosum of Sox10-ErbB2V664E mice (Figure 5D,G). In contrast, the densities of proliferating OPCs (Ki67+Olig2+) and Olig2+ cells increased in Sox10-dnEGFR mice at P65 (Figure 5B,E,G and Figure 5-figure supplement 1B), despite the fact that these increases were not observed in Sox10-dnEGFR mice at P35 (Figure 5-figure supplement 2A-C). No pathological signs were observed and the increased Olig2+ cells comprised of both NG2+ and CC1+ cells (Figure 5B and Figure 5-figure supplement 1B). It could not be determined whether apoptosis decreased in Sox10-dnEGFR white matter, as the apoptotic cells (TUNEL+) were minimal in white matter of both Sox10-dnEGFR and control mice (Figure 5H). The consistent results from gain-of-function (Sox10-ErbB2V664E) and loss-of-function (Sox10-dnEGFR) studies support a negative regulation of OPC proliferation and survival by ErbB signaling.
Neither differences in Olig2+, CC1+, NG2+, or Ki67+Olig2+ cell densities (Figure 5C,F,G; Figure 5-figure supplement 1C; Figure 5-figure supplement 2D-F), nor in TUNEL+ cells (Figure 5I), were observed in white matter of Plp-dnEGFR mice and littermate controls at P35 or P65. Different cellular and histological phenotypes in Sox10-dnEGFR and Plp-dnEGFR mice consolidated again the different targeting specificities of Sox10+/rtTA and Plp-tTA. Moreover, the conflicting observations that the numbers of post-mitotic OLs (CC1+) increased but myelin thickness reduced in the brain of Sox10-dnEGFR mice (Figure 4E and Figure 5B), suggested that ErbB inhibition in OPC-NFOs had significantly impaired the myelinating capacity of OLs.
Given that Sox10-dnEGFR and Sox10-ErbB2V664E mice both exhibited hypomyelination, they should share a molecular or cellular deficit in myelination. We performed RNA-seq analyses of subcortical white matter tissues and identified 68 genes which exhibited similar expression tendencies in Sox10-ErbB2V664E and Sox10-dnEGFR mice (Figure 6A). This group of genes have potential to regulate CNS myelination. Notably, in addition to Gsn and Itgb4 that have been identified as characteristic genes for myelinating OLs (Zhang et al., 2014), Enpp6, Itpr2, and Slc12a2 as characteristic genes for NFOs also exhibited significantly reduced expression in both mouse lines, supporting the notion that NFO deficiency contributes to hypomyelination. We examined the distribution of Enpp6-expressing cells by in situ hybridization (Xiao et al., 2016), and found that Enpp6+ cell numbers were indeed reduced in the corpus callosum of Sox10-dnEGFR mice at P35, although the reduction became indistinguishable for mice at P65 (Figure 6C,E). We further examined NFOs by immunostaining for TCF4, and found TCF4+ cell numbers were also reduced in the corpus callosum of Sox10-dnEGFR mice at P35, although the reduction became indistinguishable for mice at P65 (Figure 6G,I). Therefore, NFO differentiation was impaired shortly after ErbB inhibition, although it took 44 days to result in obvious hypomyelination in Sox10-dnEGFR mice. TCF4+ and Enpp6+ cell numbers were also reduced in Sox10-ErbB2V664E mice (Figure 6B,F,E,I), which were due to the shortage of OPCs for differentiation.
Interestingly, we also observed lowered TCF4+ and Enpp6+ cell numbers in Plp-dnEGFR mice at P35 (Figure 6D,E,H,I). Plp-tTA targeted a fraction of NFOs (Figure 2E-I). The different myelination states and similar NFO number reduction in Sox10-dnEGFR and Plp-dnEGFR mice suggest that, besides regulating myelinating capability of NFOs, ErbB signaling separately regulates another aspect of NFO differentiation, i.e., the transition time from NFO to MO stage. Indeed, in contrast to that in Sox10-dnEGFR and Plp-dnEGFR mice, the ratio of TCF4+ to CC1+ cell densities increased in Sox10-ErbB2V664E mice (Figure 6J), which may suggest a prolonged transition to the MO stage for NFOs with ErbB activation.
ErbB inhibition in MOs disrupts cognitive function in the absence of myelin alteration
OLs also offer essential trophic support to neurons in addition to forming myelin (Nave and Werner, 2014). We compared the behavioral performance of Sox10-dnEGFR mice, with hypomyelination in most brain regions, and Plp-dnEGFR mice, with normal myelin, to investigate whether disrupting ErbB signaling in MOs induces deficits other than dysmyelination. Sox10-dnEGFR mice performed worse than control mice in the rotarod test (Figure 7A), and were slightly hypoactive in the open field test (Figure 7B). Nevertheless, they performed normally in the central/peripheral zone analysis for assessment of anxiety, stereotyped behavior analysis and social interaction analysis for potential autistic-like phenotype, prepulse inhibition analysis for sensory gating, as well as forced swim and tail suspension tests for depression (Figure 7-figure supplement 1A-E). Interestingly, Plp-dnEGFR mice performed normally, similar to the controls in most tests, except exhibiting a subtle hyperactivity in the open field test (Figure 7C,D and Figure 7-figure supplement 1F-J). The different results from the two mouse lines implied that the impaired motor coordination could be attributed to the hypomyelination in the CNS.
We further tested these mice in the eight-arm radial water maze, a paradigm analyzing working memory capacity. It is known that myelin integrity is fundamental to cognitive performance of patients (Kujala et al., 1997). Moreover, although ErbB3/ErbB4 double knock-out does not induce myelin alteration in the CNS during early postnatal development (Brinkmann et al., 2008), a study of specifically depleting ErbB3 in mice from P19 has associated CNS hypomyelination with working memory deficits in adult mice (Makinodan et al., 2012). However, not only Sox10-dnEGFR mice, which had CNS hypomyelination, but also Plp-dnEGFR mice, which did not have myelin alteration, had significantly more working memory errors than control mice (Figure 7E,F and Figure 7-Video 1). Note that they had normal eyesight as performed in the visible platform test, as well as similar reference memory errors that indicated unaltered spatial recognition and memory (Figure 7E,F). This phenotype in Plp-dnEGFR mice reveals that working memory deficiency can be caused directly by ErbB inhibition in MOs through a myelination-independent mechanism.
ErbB inhibition in MOs suppresses axonal conduction under energy stress
To determine what kind of function was impaired in white matter tracts of Plp-dnEGFR mice, we acutely isolated the optic nerves from adult mice (P90-P110) and recorded electrical stimulus-evoked compound action potentials (CAPs). The areas under CAPs, which are proportional to the total number of excited axons, indicate the nerve conduction. We found comparable areas under CAPs in Plp-dnEGFR optic nerves and control nerves responding to stimuli of the same intensity (Figure 7G). The maximal CAPs, which represent that all axons in the nerves are excited, were similar in Plp-dnEGFR optic nerves and control nerves (Figure 7G). In contrast, they were reduced in Sox10-dnEGFR optic nerves as compared with littermate controls (Figure 7H). These results reflected that the basic axonal conduction was not affected in Plp-dnEGFR white matter tracts, whereas it was impaired in Sox10-dnEGFR white matter tracts that exhibited hypomyelination.
In addition to myelin, macroglial metabolites are important for axonal conduction maintenance under conditions of energy deprivation (Funfschilling et al., 2012; Saab et al., 2016; Trevisiol et al., 2017). We challenged the optic nerves by incubating them in the oxygen-glucose deprivation (OGD) condition for 60 min. CAPs fell gradually in control optic nerves, and finally fell to 30% of the initial levels (Figure 7I,J). However, for Plp-dnEGFR optic nerves in the OGD condition, CAP failure was slightly accelerated and aggravated (Figure 7I). Contrarily, for Sox10-dnEGFR optic nerves under the same condition, CAP failure was decelerated and attenuated (Figure 7J). When the glucose and oxygen levels in the bathing medium were restored, CAPs in control optic nerves and Plp-dnEGFR optic nerves recovered to 60% of the initial levels (Figure 7I,J). However, in Sox10-dnEGFR optic nerves, CAPs recovered to 80% of the initial levels (Figure 7J).
It is notable that continuous electrical stimulation caused a baseline CAP decline in Plp-dnEGFR optic nerves, whereas a baseline CAP enhancement in Sox10-dnEGFR optic nerves, before the OGD (Figure 7I,J). Therefore, we further examined the axonal conduction under a physiological condition with increasing energy demands generated by neuronal activities (Saab et al., 2016; Trevisiol et al., 2017). By stimulating the control optic nerves with several trains of short bursts with frequency increased from 1 to 100 Hz, we confirmed that the low frequency stimulation (5-25Hz) has only minor influence on the CAPs, whereas the high frequency stimulation (50-100Hz) exhausts axonal energy and results in CAP decline (Figure 7K,L). For Sox10-dnEGFR optic nerves, intriguingly, the 5-25Hz electrical stimuli amplified CAPs and the 50-100Hz stimuli induced smaller CAP decay than that in control nerves (Figure 7L). In contrast, in Plp-dnEGFR optic nerves, either group of stimuli significantly aggravated the CAP decay (Figure 7K).
These results showed that Sox10-dnEGFR white matter tracts exhibited resistance to energy stress induced by both pathological (OGD) and physiological (neuronal activities) conditions. This may be ascribed to increased OL numbers, as that OLs are an essential venue for glycolysis and energy substrate production in support of axonal conduction (Funfschilling et al., 2012). In contrast, Plp-dnEGFR white matter was deficient in the maintenance of axonal conduction, especially under physiological energy stress. It is notable that MO numbers were not altered in Plp-dnEGFR mice that have ErbB inhibition in MOs, whereas ErbB receptors were not inhibited in MOs of Sox10-dnEGFR mice that have increased MOs (Figure 5B,C). Therefore, the opposite results of Sox10-dnEGFR and Plp-dnEGFR optic nerves in the energy challenging studies reveal that ErbB inhibition in MOs impairs the glia-axon energy coupling efficiency within electrically active neural circuits, which can compromise the cognitive function in Plp-dnEGFR mice in the absence of myelin alteration (Figure 7F).
Discussion
Our results demonstrate that both ErbB3/ErbB4 receptors binding to the NRG family ligands and EGFR binding to the EGF family ligands are functional in adolescent and adult OLs. With the discovery of two valuable in vivo research mouse tools that differentially target OLs at MO and OPC-NFO stages, we reveal that NRG-ErbB and EGF-ErbB signaling cooperate in OPCs, NFOs, and MOs to simultaneously regulate myelination and axonal energy supporting functions. Aberrant ErbB activation or inhibition causes white matter abnormalities with distinct pathological characteristics and biological markers (Figure 7M).
ErbB overactivation is pathogenetic in MOs through inducing myelin overproduction and MO necroptosis, which results in demyelination followed by pathological changes including axon degeneration, OPC regeneration, astrogliosis and microgliosis. Notably, ErbB overactivation in OPCs induces apoptosis, without stimulating inflammatory pathological responses in the brain (Figures 1 and 3). Caspase-8 activation has been reported to be the key event to determine apoptotic fate of cells (Oberst et al., 2011), and defective activation of caspase-8 is critical for RIP1/RIP3/MLKL signaling to induce OL necroptosis in multiple sclerosis (Ofengeim et al., 2015). A cell-type specific RNA-sequencing transcriptome analysis suggests that caspase-8 is minimally expressed in post-mitotic OLs but is detectable in OPCs (Zhang et al., 2014), which may determine MO necroptosis but OPC apoptosis under continuous ErbB activation. Interestingly, studies on genetically modified mice that overexpressing hEGFR in OL lineage cells, or overexpressing NRG1 Type I or Type III in neurons, did not report myelin pathogenesis (Aguirre et al., 2007; Brinkmann et al., 2008). Nevertheless, mice with overactivation of the ErbB downstream signaling in OL lineage cells exhibit myelin and axonal pathology (Harrington et al., 2010; Ishii et al., 2016). Olig2-cre;Ptenflox/flox mice that overactivate PI3K/Akt signaling in OL lineage cells have loosened myelin lamellae in the spinal cord at 14 weeks and axonal degeneration in the cervical spinal cord fasciculus gracilis at 62 weeks (Harrington et al., 2010). Plp-CreER;Mek/Mek mice, which overexpress a constitutively activated MEK, a MAPK kinase, in OL lineage cells with tamoxifen induction, show demyelination in the spinal cord 3 months after MAPK (Erk) overactivation is induced (Ishii et al., 2016). Devastating effects of ErbB2V664E in OLs may be due to its potent promotion of endogenous ErbB activation and multiple downstream signaling. Nevertheless, observations in the present study and in Olig2-cre;Ptenflox/flox and Plp-CreER;Mek/Mek mice corroborate the concept that constitutively activating ErbB signaling in OL lineage cells is pathogenetic, even though it may take a long time for moderate activation to result in pathological symptoms.
The profound demyelination or hypomyelination in Plp-ErbB2V664E and Sox10-ErbB2V664E mice should have disrupted many brain functions, although we could not evaluate these functions in the two strains by behavioral tests due to their severe motor dysfunction (Figure 1B,M). Intriguingly, a battery of behavioral tests for Sox10-dnEGFR and Plp-dnEGFR mice only revealed working memory deficits for both of them, except for the impaired motor coordination in Sox10-dnEGFR mice that have moderate hypomyelination (Figure 7A-F and Figure 7-figure supplement 1A-J). The further analyses emphasize that, although endogenous ErbB activation is required for both NFOs and MOs, it is used for the control of myelination and glia-axon energy coupling, respectively. Thus, ErbB inhibition in OLs impairs cognitive functions via myelination-dependent and -independent mechanisms. Plp-dnEGFR mice are a good model to affirm the myelination-independent contributions of OLs to higher brain function. As exhibited in Plp-dnEGFR mice, dual inhibition of NRG-ErbB and EGF-ErbB signaling in MOs does not affect myelin or OL numbers in the adolescent and adult brains. However, endogenous ErbB activities in MOs are indispensable for the maintenance of axonal conduction under physiological energy stress (Figure 7I,K), which are important for neuronal circuit efficiency as well as cognitive performance (Figure 7F). It is interesting that axonal conduction under energy stress was enhanced in Sox10-dnEGFR white matter tracts (Figure 7J,L), although it remains unclear whether the improved energy supplementation alleviates or aggravates the cognitive deficits in Sox10-dnEGFR mice (Figure 7E). Multiple questions remain unanswered, such as ErbB signaling regulates glucose metabolism in MOs or the transportation of energy metabolites from MOs to axons. ErbB dysregulation disrupts glutamatergic synaptic transmission in neurons (Luo et al., 2014; Ting et al., 2011; Woo et al., 2007). Glutamatergic synaptic transmission onto OLs was recently discovered to be essential for their energy substrate supply to axons (Saab et al., 2016), as well as for OL development (Kougioumtzidou et al., 2017). Therefore, it would be worth pursuing whether ErbB signaling regulates OL development or the trophic support from MOs to neurons by modulating glutamatergic synaptic transmission on OLs.
The roles of ErbB signaling in CNS myelination have long been debated because of the contradictory findings reported by different research groups. For example, ErbB3 has been reported to be dispensable for OL development (Schmucker et al., 2003), and ErbB3/ErbB4 double knockout does not result in CNS myelin alteration (Brinkmann et al., 2008). However, there are other reports that indicate inducing ErbB3 depletion by Plp-CreER in OL lineage cells from P19, not P36, results in adult hypomyelination (Makinodan et al., 2012). It is notable that ErbB3 has peaked expression during P15-P30 (Figure 1-figure supplement 1), and ErbB receptor manipulation in OPC-NFOs alters ErbB3/4 activities in white matter at P30-35 but not at P65 (Figure 1N and Figure 4B). Note that Plp-CreER is a mouse tool that can target OPCs and their progeny (Guo et al., 2009). Therefore, the hypomyelination in Plp-CreER;ErbB3flox/flox mice is in line with our findings in Sox10-dnEGFR mice and reflects the positive role of ErbB signaling in NFO myelination during late postnatal development. EGFR is expressed stably in white matter during P20-P40 (Figure 1-figure supplement 1). The phosphorylation of EGFR is altered in white matter in all four mouse strains, which have ErbB receptor manipulation either in MOs or in OPC-NFOs (Figure 1C,N and Figure 4B,G). This suggests the general involvement of EGFR in OL function and development. The role of EGFR in CNS myelin development is supported by the report that transgenic mice with overexpression of hEGFR in all OL lineage cells (CNP-hEGFR) have enhanced myelin maturation, and hypomorphic EGFR mice (wa2) have delayed myelin maturation (Aguirre et al., 2007). CNP-hEGFR mice exhibit enhanced oligodendrogenesis in the subventricular zone, reflecting the function of EGFR in promoting neural progenitors to differentiate into OPCs during the early development (Aguirre et al., 2007). This is different from the pathological OPC regeneration revealed in the corpus callosum of Plp-ErbB2V664E mice. It is notable that CNP-hEGFR increases the numbers of myelinated axons but not myelin thickness (Aguirre et al., 2007), which is different from the hypermyelination phenotype revealed in the CNS of NRG1-overexpressing mice (Brinkmann et al., 2008), suggesting that EGFR unlikely participates in myelin overproduction in MOs. It will be interesting to know whether the active EGFR, as revealed in Plp-dnEGFR mice (Figure 4G), is required for the trophic supportive function of MOs.
The negative regulation of ErbB activation on OPC proliferation and survival is unexpected because many in vitro studies have suggested otherwise. However, there is an interesting observation in transgenic mice CNP-dnErbB4, which are designed to overexpress a dominant negative ErbB4 mutant that specifically blocks the activities of ErbB3 and ErbB4 in all OL lineage cells. In this strain, post-mitotic OL numbers increase 40% in the corpus callosum although axons are hypomyelinated (Roy et al., 2007). Moreover, Olig2-cre;Ptenflox/flox mice, which have activation of the PI3K/Akt pathway in all OL lineage cells, exhibit hypermyelination but decreased OL densities in the developing corpus callosum (Harrington et al., 2010). These previously enigmagic observations are now well-explained by our findings that ErbB signaling plays different roles in OPCs and NFOs.
The white matter abnormalities observed in our mouse models are reminiscent of diverse myelin-related clinical and pathological characteristics in schizophrenic brains, including reduced white matter volume, decreased OL densities, reduced myelin gene products, apoptotic OLs, and damaged myelin (Douaud et al., 2007; Fields, 2008; Hoistad et al., 2009; Uranova et al., 2011; Uranova et al., 2007). Elevated ErbB activation has been repeatedly implicated in schizophrenia, and the increase could be caused by genetic factors (Harrison and Law, 2006; Law et al., 2012). To our knowledge, we are the first to reveal that ErbB overactivation can primarily induce oligodendropathy and myelin pathogenesis in white matter, providing a possible predisposition of a genetic variability in ErbB receptors or ligands to the white matter lesion. Notably, SNP8NRG243177 with T-allele, which increases NRG1 Type IV production (Law et al., 2006), is associated with the reduced white matter integrity (McIntosh et al., 2008) as well as increased psychotic symptoms (Hall et al., 2006) in schizophrenic patients.
Further, ErbB receptors and their ligands have been reported to reduce expression or lose function in some schizophrenic patients (Harrison and Law, 2006; Mei and Nave, 2014). Specific working memory deficits in Sox10-dnEGFR and Plp-dnEGFR mice firmly support that oligodendropathy can be a primary cause for the cognitive symptoms of schizophrenia. Moreover, myelin is very sensitive to environmental insults. Modest disruption of ErbB signaling by genetic mutation or SNPs is able to render myelin vulnerable to such insults, aggravating focal loss of connections under conditions of stress, ischemia, sleeplessness, trauma, etc. Therefore, OL dysfunction in patients, which is difficult to measure with current techniques, may eventually evolve into a detectable structural alteration in white matter that could contribute to another type of brain dysfunction. Collectively, our study provides novel insights into the pathophysiology of diseases initiated or aggravated within white matter.
Materials and methods
Animals
Plp-tTA transgenic mice (Inamura et al., 2012) were from the RIKEN BioResource Center (Stock No. RBRC05446). Sox10+/rtTA mice were from Dr. Michael Wegner (University Erlangen-Nurnberg, Germany). Transgenic mice TRE-ErbB2V664E (Stock No. 010577) and TRE-dnEGFR (Stock No. 010575) were from the Jackson Laboratory (Chen et al., 2017). Among ErbB1-4 receptors, ErbB2 that does not bind to any known ligand is the preferred partner to other ligand-bound ErbB members. ErbB2V664E contains an amino acid mutation (Vla664/Glu664) within the transmembrane domain facilitating its dimerization with other ErbB receptors and potentiating their downstream signaling (Chen et al., 2017). DnEGFR, a dominant negative mutant of EGFR, is a truncated form of EGFR, losing the intracellular kinase domain but retaining the ability to form dimers with other ligand-bound ErbB members. When overexpressed, dnEGFR efficiently blocks the activation of any endogenous ErbB receptors under either NRG or EGF stimulation (Chen et al., 2017). Unless indicated, mice were housed under SPF conditions before experiments, in a room with a 12-hr light/dark cycle with access to food and water ad libitum. For biochemical and histological experiments, Plp-tTA;TRE-dnEGFR (Plp-dnEGFR), Plp-tTA;TRE-ErbB2V664E (Plp-ErbB2V664E), Sox10+/rtTA;TRE-dnEGFR (Sox10-dnEGFR), or Sox10+/rtTA;TRE-ErbB2V664E (Sox10-ErbB2V664E) mice with either sex and their littermate control mice with matched sex were used. For indicated behavioral tests, only male mice were used, while both male and female mice were used for the other behavioral tests because the results were not affected by sex difference. Animal experiments were approved by the Institutional Animal Care and Use Committee of the Hangzhou Normal University. For genotyping, the following primers were used: for Plp-tTA (630bp), PLPU-604 5’-TTT CCC ATG GTC TCC CTT GAG CTT, mtTA24L 5’-CGG AGT TGA TCA CCT TGG ACT TGT; for Sox10+/rtTA (618bp), sox10-rtTA1 5’-CTA GGC TGT CAG AGC AGA CGA, sox10-rtTA2 5’-CTC CAC CTC TGA TAG GT CTT G; for TRE-dnEGFR (318bp), 9013 5’-TGC CTT GGC AGA CTT TCT TT, 7554 5’-ATC CAC GCT GTT TTG ACC TC; for TRE-ErbB2V664E (625bp), 9707 5’-AGC AGA GCT CGT TTA GTG, 9708 5’-GGA GGC GGC GAC ATT GTC.
Tet-Off or Tet-On treatment of mice
Mice with Tet-system contain genes of tetracycline-controlled transcriptional activator (tTA) or reverse tetracycline-controlled transcriptional activator (rtTA) driven by cell-specific promoters. When fed with Dox, these mice are able to switch on or off expression of a gene under the control of tetracycline-responsive element (TRE), specifically in rtTA- or tTA-expressing cells, which are called ‘Tet-on’ or ‘Tet-off’, respectively. The offspring of Sox10+/rtTA during the indicated periods were fed with Dox (2 mg/mL × 10 mL/day from P21 to P35, and 1 mg/mL × 10 mL/day from P35 to indicated test day) in drinking water to induce the expression of ErbB2V664E or dnEGFR in Sox10-ErbB2V664E and Sox10-dnEGFR mice, respectively (Tet-On). For the offspring of Plp-tTA, Dox was given (Tet-off) from the embryonic day (through pregnant mothers) to their weaning day at P21 to inhibit the expression of ErbB2V664E or dnEGFR during this period in Plp-ErbB2V664E or Plp-dnEGFR mice (0.5 mg/mL × 10 mL/day of Dox before P21). Water bottles were wrapped with foil to protect Dox from light. All used littermate control mice were treated the same.
Stereotactic injection of AAV viruses
pAAV-TRE-EYFP plasmids (Addgene) were packaged as AAV2/9 viruses, and produced with titers of 2.0E+13 particles per mL by OBio (Shanghai, China). Mice were anesthetized by 1% pentobarbital (50 mg/kg, i.p.) and mounted at stereotaxic apparatus (RWD68025). AAV-TRE-EYFP (2 μL) was injected into the corpus callosum (from bregma in mm, M-L: ±1.2, A-P: +0.5, D-V: -2.2) under the control of micropump (KDS310) at a speed of 0.05 μL/min. Injecting needles (Hamilton NDL ga33/30 mm/pst4) were withdrawn 10 min after injection. Infected brains were isolated 1 or 2 days later and brain slices were immunostained with anti-GFP antibody to enhance the visualization of the reporter protein.
Electron Microscopy
Mice were anesthetized and transcardially perfused with 4% sucrose, 4% paraformaldehyde (PFA) and 2% glutaraldehyde in 0.1 M phosphate buffer (PB, pH 7.4). The brains, optic nerves, or sciatic nerves were isolated carefully. The corpora callosa and prefrontal cortices were further dissected carefully under stereoscope. Tissues were post-fixed overnight at 4°C in 1% glutaraldehyde in 0.1 M PB. Samples were washed by 0.1 M PB 24 hr later, and osmicated with 2% osmium tetroxide 30-60 min at 4°C, washed by 0.1 M PB and by deionized H2O at 4°C, and dehydrated in graded (50-100%) ethanol. Samples were incubated with propylene oxide and embedded with embedding resins. Ultrathin sections were stained with 2% uranyl acetate at 4°C for 30 min, and then photographed with Tecnai 10 (FEI). EM images were analyzed by Image J (NIH). To eliminate the bias on circularity, g-ratio of each axon was calculated by the perimeter of axons (inner) divided by the perimeter of corresponding fibers (outer). Axonal diameters were normalized by perimeters through equation: diameter = perimeter/π. This procedure allows for inclusion of irregularly shaped axons and fibers and helps to eliminate biased measurement of diameters based on circularity. For quantitative analysis, cross sections of each neural tissue were divided into 5 areas, and more than two images, randomly selected from each area, were examined.
Immunofluorescence staining
Deeply anesthetized mice were transcardially perfused with 0.01 M PBS and then 4% PFA in 0.01 M PBS. Mouse brains were isolated and post-fixed in 4% PFA in 0.01 M PBS overnight at 4 °C, and then transferred into 20% and subsequently 30% sucrose in PBS overnight at 4 °C. Brains were then embedded in OCT (Thermo Fisher scientific) and sectioned into 20 μm on a cryostat sectioning machine (Thermo Fisher scientific, Microm HM525). Brain slices were incubated with blocking buffer (10% fetal bovine serum and 0.2% Triton-X-100 in 0.01 M PBS) for 1 hr at room temperature, and then incubated at 4 °C overnight with primary antibodies diluted in blocking buffer. The primary antibodies used were: GFP (1:500, Abcam, ab13970), CC1 (1:500, Abcam, ab16794), NG2 (1:200, Abcam, ab50009), Ki67 (1:400, Cell Signaling Technology, 9129), GFAP (1:2000, Millipore, MAB360), Iba1 (1:1000, Millipore, MABN92), TCF4 (1:500, Millipore, 05-511), Olig2 (1:500, Millipore, AB9610), TUJ1 (1:500, Sigma, T5076), RIP3 (1:500, QED, 2283), MLKL (1:500, Abgent, AP14272B). After washing three times with 0.1% Triton-X-100 in 0.01 M PBS, samples were incubated at room temperature for 1 hr with Alexa-488 or -594 secondary antibody, and then washed and mounted on adhesion microscope slides (CITOTEST) with fluorescent mounting medium. Nuclear labeling was completed by incubating slices with DAPI (0.1 μg/mL, Roche) at room temperature for 5 min after incubation with secondary antibodies. Except for the antibody against NG2, antigen retrieval in 0.01 M sodium citrate buffer (pH 6.0) at 80-90 °C for 10 min was necessary before primary antibody incubation for brain slices to achieve definitive signals. Images were taken by a Zeiss LSM710 confocal microscope or a Nikon Eclipse 90i microscope. For cell counting based on immunostaining results, soma-shaped immunoreactive signals associated with a nucleus was counted as a cell. The immunostaining intensity was measured by Image J with background subtraction.
Luxol fast blue (LFB) staining
After sufficient washing with 0.01 M PBS, PFA-fixed brain slices were transferred into a mixture of trichloromethane and ethanol (volume ratio 1:1) for 10 min and then 95% ethanol for 10 min. They were next incubated in 0.2% Luxol fast blue staining solution (0.2 g Solvent blue 38, 0.5 mL acetic acid, 95% ethanol to 100 mL) at 60 °C overnight. In the next day, tissues were incubated for 5 min each in turn in 95% ethanol, 70% ethanol and ddH2O for rehydration, followed by incubation alternatively in 0.05% Li2CO3, 70% ethanol and ddH2O for differentiation until the contrast between the gray matter and white matter became obvious. After that, tissues were incubated for 10 min each in 95% and 100% ethanol to dehydrate, and then 5 min in dimethylbenzene to clear, before quickly mounting with neutral balsam mounting medium (CWBIO). All steps were operated in a ventilation cabinet. The LFB intensity in the corpus callosum was measured by Image J with background subtraction, and normalized to that of controls.
TUNEL assay
Apoptotic cells were examined with terminal deoxynucleotidyl transferase (TdT)-mediated deoxyuridine triphosphate (dUTP) nick-end labeling (TUNEL) assay according to the manufacturer’s instructions (Vazyme; Yeasen). In brief, PFA-fixed brain slices were digested for 10 min by proteinase K (20 μg/mL) at room temperature. After washing twice with PBS, brain slices were incubated with Equilibration Buffer for 30 min at room temperature, and subsequently with Alexa Fluor 488-12-dUTP Labeling Mix for 60 min at 37°C. After washing with PBS three times, brain slices were stained with DAPI before being mounted under coverslips. For co-labeling of apoptotic nuclei in slices with immunofluorescence staining, TUNEL assay was performed after washing of the secondary antibody.
Western blotting
Subcortical white matter tissues were isolated and homogenized. Homogenates in lysis buffer (10 mM Tris-Cl, pH 7.4, 1% NP-40, 0.5% Triton-X 100, 0.2% sodium deoxycholate, 150 mM NaCl, 20% glycerol, protease inhibitor cocktail) at a ratio of 2 mL per 100 mg tissue were lysed overnight in 4°C. Lysates were centrifuged at 12,000 g and 4°C for 15 min to get rid of the unsolved debris. Concentration of the supernatant was measured by BCA assay. Proteins in samples were separated by 6-12% SDS-PAGE, transferred to a Immobilon-P Transfer Membrane (Millipore), and then incubated with indicated primary antibodies diluted in blocking buffer at 4°C overnight after blocking by 5% non-fat milk solution in TBST (50 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Tween 20) for 1 hr at room temperature. The primary antibodies used were: pErbB3 (1:2500, Abcam, ab133459), pErbB4 (1:2500, Abcam, ab109273), pErbB2 (1:2500, Abgent, AP3781q), EGFR (1:5000, Epitomics, 1902-1), pEGFR (1:2500, Epitomics, 1727-1), GAPDH (1:5000, Huabio, EM1101), MBP (1:1000, Millipore, MAB382), Olig2 (1:1000, Millipore, MABN50), ErbB3 (1:200, Santa Cruz Biotechnology, sc-285), ErbB4 (1:200, Santa Cruz Biotechnology, sc-283), ErbB2 (1:200, Santa Cruz Biotechnology, sc-284), RIP3 (1:2000, QED, 2283), MLKL (1:2000, Abgent, AP14272B). For antibodies against phosphorylated proteins, 10% fetal bovine serum was used as blocking buffer. Next day, the membranes were washed by TBST for three times and incubated with the secondary antibodies for 1 hr at room temperature. Membranes were washed again and incubated with Immobilon
Western Chemiluminescent HRPSubstrate (Millipore) for visualization of chemiluminescence by exposure to X-ray films or Bio-Rad GelDOCXR+ Imaging System. Intensities of protein bands were measured by Image J, and statistical analysis was performed after subtraction of the background intensity and normalization with controls in each batch of experiments.
In situ hybridization
RNA in situ hybridization was performed as previously described (Schaeren-Wiemers and Gerfin-Moser, 1993) with minor modifications. Briefly, the 14 μm PFA-fixed brain sections were post-fixed in 4% PFA in PBS for 20 min, incubated in 2 μg/mL Proteinase K in 50 mM Tris-Cl (pH 7.4) with 5 mM EDTA at room temperature for 10 min, re-fixed in 4% PFA in PBS for another 10 min, and then acetylated in 1.33% triethanolamine and 0.25% acetic anhydride solution at room temperature for 10 min. The acetylated sections were washed and incubated in hybridization buffer (50% formamide, 0.25 mg/mL yeast RNA, 0.5 mg/mL herring sperm DNA, 5x Denhard’s, 5x SSC, Invitrogen) at room temperature for 1 hr, and then hybridized with 0.5 ng/μL digoxigenin-labeled Enpp6 riboprobe in hybridization buffer at 65°C for 16 hr. The hybridized sections were washed three times in 0.2x SSC at 65°C for total 1 hr, and then were blocked with 10% sheep serum (Sigma-Aldrich) in solution I containing 100 mM pH 7.5 Tris-Cl with 0.15 M NaCl at room temperature for 1 hr, followed by incubation with alkaline phosphatase-conjugated anti-digoxigenin antibody (Roche) in the same solution at 4°C overnight. After washing three times with solution I for total 1 hr, and twice with developing buffer containing 100 mM pH 9.5 Tris-Cl, 0.1 M NaCl, 50 mM MgCl2 and 0.1% Tween-20, the sections were incubated with 2% NBT/BCIP solution (Roche) in the developing buffer at room temperature in the dark. The reaction was stopped by immersing the sections in PBS with 5 mM EDTA when appropriate signals were detected. To obtain mouse Enpp6 riboprobes, a 1.3 kb fragment corresponding to Enpp6 mRNA (1400–2700 nt of NM_177304.4) was cloned into pBluescript II KS(-). The linearized plasmids were used as templates for in vitro transcription with T3 RNA polymerase (Promega) according to the manufacturer’s instructions.
Real-time reverse transcription-PCR (RT-PCR)
Total RNA was extracted from isolated mouse white matter using TRIzol following manufacturer’s protocol. cDNA was synthesized by using the 5x All-In-One RT MasterMix (abmGood). Real-time PCR was performed in four repeats for each sample by using BrightGreen 2x qPCR MasterMix (abmGood) with the Bio-Rad CFX96 real-time PCR system as previously described (Chen et al., 2017). Relative mRNA levels were analyzed by software Bio-Rad CFX Manager 1.5. Transcripts of targeted genes were normalized to those of mouse 18S rRNA gene in the same samples. Primers for 18S rRNA were 5’-CGG ACA CGG ACA GGA TTG ACA and 5’-CCA GAC AAA TCG CTC CAC CAA CTA with a 94 bp PCR product. Primers for mouse EGFR gene and transgene dnEGFR were 5’-TCC TGC CAG AAT GTG AGC AG and 5’-ACG AGC TCT CTC TCT TGA AG with a 500 bp PCR product.
RNA-Seq Analyses
Subcortical white matter tissues isolated from Sox10-dnEGFR and littermate controls, or Sox10-ErbB2V664E mice and littermate controls, were used (three pairs for each group) for global transcriptome analysis by LC-Bio Co (Hangzhou, China). The final transcriptome was generated by Histat and StringTie. StringTie was used to estimate the expression level for mRNAs by calculating FPKM (Fragments Per Kilobase of exon model per Million mapped reads). Differentially expressed genes were identified by comparing FPKM of the mRNA reads from three sample pairs between Sox10-dnEGFR, or Sox10-ErbB2V664E mice, and their littermate controls, by paired Student t test via MeV (MultiExperiment Viewer). Gene lists with significant difference (P < 0.05) in expression between Sox10-dnEGFR and littermate controls, or between Sox10-ErbB2V664E and littermate controls, were compared, and genes with similar expression tendencies in Sox10-dnEGFR and Sox10-ErbB2V664E mice were identified. Z value of these genes was calculated according to their FPKM by an equation “Z sample-i = [(log2(Signal sample-i)-Mean (Log2(Signal) of all samples)][Standard deviation (Log2(Signal) of all samples)]” and plotted as heat map by MeV. Gene Ontology (GO) term enrichment was analyzed by PANTHER Overrepresentation Test (Released 20171205) through http://geneontology.org with the significance estimated by Fisher’s Exact Test (FDR, false discovery rate).
Behavioral Tests
Plp-ErbB2V664E mice at P35 and Sox10-ErbB2V664E mice at P30 after indicated Dox treatment were used in grid walking tests for motor function analysis. Behavioral analyses for Sox10-dnEGFR and littermate controls with Dox-feeding from P21, or Plp-dnEGFR mice and littermate controls with Dox-withdrawal from P21, were carried out with 12- to 16-week-old animals by investigators unaware of their genotypes. Tested mutant mice had littermate control mice with same sex. For PPI, social interaction, eight-arm radial water maze, forced swim and tail suspension, all tested mice were male. Animals were tested at a sequence of open field, social interaction, rotarod, PPI, eight-arm radial water maze, and then forced swim and tail suspension, to minimize the influence of stress on their behavioral performance. There were 2-day gaps between tests.
Open field and stereotyped behavior analysis
Animals were placed in a chamber (30 cm × 30 cm × 34.5 cm) and their movements were monitored and traced by a tracking software EthoVision XT 12 (Noldus, The Netherland). Locomotive activity was measured and summated at 5-min intervals over a 30-min period. Frequency and cumulative duration of stereotyped behaviors observed during 30-min traveling in the open field, including grooming, hopping, rearing supported, and sniffing, were determined by EthoVision XT 12 and statistically analyzed. Anxiety of the animals was assessed by the differences of time that they spent in the central zone and peripheral zone during the 30 min.
Rotarod
To evaluate the sensorimotor coordination, mice were placed on an accelerating rotarod (Mouse Rota-Rod NG, Harikul Science, UB47650, Italy) and assessed for ability to maintain balance on the rotating bar that accelerated from 4 to 40 rpm over a 5-min period. Mice were tested for 4 trials in the first day with 30-min gap between trials, and were tested for another 4 trials 24 hr later. Latency before fall from the rod was recorded.
Prepulse inhibition (PPI) test
These tests were conducted in a sound-attenuated chamber (Panlab, LE116, Spain). Mice were placed in a Plexiglas restrainer mounted on a grid floor, and their startle responses were captured by a movement sensor and analyzed by a software Startle v1.2. Before the test, mice were allowed to habituate to the chamber with a 60 dB background white noise for 5 min, and to 4 times of auditory-evoked startle stimulating pulse (10000 Hz at 120 dB, 20 ms) with random 5-30 sec intervals. In the PPI test, mice were subjected to startle pulse trials (120 dB, 20 ms) or prepulse/pulse trials (20 ms 10000 Hz at 75, 80, or 85 dB with 100 ms interval before a 20 ms 120dB startle stimulus) with random 5-30 sec intervals between trials. Different trial types were presented randomly with each trial type presented 9-12 times, and no two consecutive trials were identical. Max startle response within 300 ms after onset of startle stimulus was recorded. PPI (%) was calculated according to the equation: [100 - (startle amplitude on prepulse-pulse trials/startle amplitude on pulse alone trials) ×100].
Social interaction
Mice were placed in a square chamber (50 cm × 50 cm × 50 cm) with two small transparent boxes at two opposite corners. The chamber was dark and mouse movement was monitored by infrared camera and traced by Anymaze software (Stoelting, Wood Dale, Illinois). After 5 min habituation in the chamber, mice were returned to their home cages. Mice were placed into the same chamber 2 hr later, with one of the boxes holding a stranger mouse that could be seen and smelled through multiple holes in the box. Social interaction ability of the tested mice was determined by their traveling distances in the quadrant with the stranger mouse as compared with the traveling distances in the quadrant with an empty box.
Eight-arm radial water maze
According to the previous report with modification (Penley et al., 2013), animals were trained in eight-arm radial water maze for two weeks, with four trials each day to search a hidden platform in each trial for escaping from the water at 20-22 °C. Four hidden platforms were placed at the end of a same set of arms for all the training and tests, as illustrated in Figure 7-figure supplement 1K. After a trial that mice reached a hidden platform, mice were returned to their home cages with towel and warming pads. There was a 30-sec gap between trials, and the visited platform was removed before the next trial. Mice aborted swimming during training were discarded. Two weeks later, the trained mice were tested for their working memory capacities that were represented by avoiding arms with visited platforms in previous trials. In the last trial of the test day, the animals had highest working memory load for they had to avoid swimming into three arms with platforms removed. First and repeat entries into any arm that previously had a platform were counted as working memory errors, and first entries into any arm that never had a platform were counted as reference memory errors which represent deficits in spatial recognition or long term memory. The day after test day, all the animals were tested in a simple visible platform task with 5 trials in a round pool, and each trial contained a visible platform placed at a different position 0.5-1.0 meter away from tested mice. The latency of mice reached the visible platform in each trial was averaged to assess their eyesight.
Forced swim and Tail suspension
In the forced swim test, mice were forced to swim for 15 min in a cylinder with diameter at 11 cm, water depth at 30 cm and temperature at 22-24 °C. One day later, mice were forced to swim again for 5 min in the same cylinder and mouse movement was recorded and analyzed by Anymaze software. The tail suspension test was carried out 2 days later after the forced swim test, in which mice were suspended by using adhesive tape applied to the tail and videotaped for 5 min. Mouse movement during the 5 min was traced and analyzed by Anymaze software. For both tests, immobile period was defined by 70% of mouse bodies were motionless and lasted for at least 1 sec. Summation of immobile periods for each mouse was taken into statistical analysis.
Grid walking test
Mice were placed on an elevated metal grid panel with each grid cell 5 × 0.8 cm2, and their movements were videotaped. The velocity of mouse movement and percentage of foot-slip steps in total steps were calculated to assess locomotor function of mice. Scores of foot slips reflect precise stepping, coordination of the four limbs, and accurate paw placement, indicating ability of animals in sensorimotor coordination.
Electrophysiology
Following anesthesia and decapitation, optic nerves were isolated from mice and superfused with oxygenated artificial cerebrospinal fluid (ACSF) containing (in mmol/L): 119 NaCl, 2.5 KCl, 2.5 CaCl2,1.3 MgCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 glucose; pH 7.4. Optic nerve CAP recording methods were adopted and modified from the previous reports (Saab et al., 2016). Briefly, two ends of the optic nerves were attached by suction electrodes, which backfilled with ACSF and connected to an IsoFlex (AMPI) for stimulation or a MultiClamp 700B (Molecular Device) for recording. The recorded signal was amplified 50 times, filtered at 30 kHz, and acquired at 20-30 kHz. Data were collected and analyzed by pClamp 10.3 software (Molecular Devices). The optic nerves were equilibrated for at least 30 min in the perfusion chamber in normal ACSF at room temperature before experiments. All experiments were performed at room temperature.
Maximal CAP recording
For each recorded nerve, stimulus pulse (100 µs duration) strength was adjusted with a stepped increase and finally to evoke the maximal CAP. CAPs were elicited 5 times at every step of the stimulus strength. After reaching the maximal CAP, the stimulus was increased an additional 25% for supramaximal stimulation to ensure the activation of all axons in the nerve. Note the supramaximal stimulation did not further change the CAPs. The areas under CAPs were calculated to determine the nerve conduction.
Oxygen-glucose deprivation (OGD) assay
The assay was performed as previously described with modification (Saab et al., 2016; Trevisiol et al., 2017). During experiments, CAPs were evoked by the supramaximal stimulus every 20 sec. After 60-min stimulation of the baseline CAP in normal condition, OGD was induced for the nerves by switching bathing solution from oxygenated ACSF (saturated with 95% O2/5% CO2) to glucose-free ACSF (replaced with equimolar sucrose to maintain osmolarity) that was saturated with 95%N2/5%CO2. After 60-min OGD, oxygenated ACSF was restored and CAPs were recorded for another hour. CAPs recorded after 30-min baseline stimulation was taken as the initial CAPs. The effects of OGD on the nerve conduction and recovery were determined by normalizing the areas of CAPs recorded during OGD or recovery sessions to that of initial CAPs.
Neural activity dependence assay
The protocol was modified from published reports (Saab et al., 2016; Trevisiol et al., 2017). Before the experiments, CAPs were recorded every 30 sec to obtain baseline with the stimulus pulse strength set at the supramaximal levels. To evaluate the conduction of optic nerves under increasing energy demands, we gradually increased the stimulating frequency from 1 to 100 Hz. Each stimulating frequency was applied for 30-60 sec. For 1 and 5 Hz stimuli, CAPs were continuously recorded and CAP areas were measured for each CAP. For 10 to 100 Hz, nerves were stimulated by a train of 100 stimuli, and rest for 1 sec before the next train of stimuli. CAP areas were sampled for the last four stimuli of each train and averaged as one data point. For the statistical analysis, CAP areas were normalized to the initial levels.
Statistical analysis
Statistical analyses other than for RNA-seq data (described separately above) were performed using Prism (Graphpad) and presented as mean ± s.e.m.. For western blotting and LFB staining results, statistical analyses were performed after subtraction of the background intensity and normalization with controls in each batch of experiments to minimize the influences of batch-to-batch variations. Two-tailed unpaired Student’s t test was used for analysis between two groups with one variable, one-way ANOVA test was used for analysis among three or more groups with one variable, and two-way ANOVA test was used to determine difference among groups with two variables. Statistical significance was set at *P < 0.05, **P < 0.01, ***P < 0.001.
Data Availability
All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files have been provided for all manuscript figures. Source data have been provided online at datadryad.org (https://doi.org/10.5061/dryad.jq2bvq87c). The accession number for the RNA-Seq data presented in this article is GEO: GSE123491.
Competing interests
The authors declare no competing interests.
Other Supplemental Files
Figure 7-Video 1. The performances recorded for Sox10-dnEGFR mice and Plp-dnEGFR mice, as well as their controls, in the 4th trial of eight arm radial water maze at the test day.
Figure 6-Source data 1. The Excel file contains the processed RNA-seq results of genes with differential expression in white matter tissues between Sox10-ErbB2V664E mice and littermate TRE-ErbB2V664E mice at P30 with 9 dwd.
Figure 6-Source data 2. The Excel file contains the processed RNA-seq results of genes with differential expression in white matter tissues between Sox10-dnEGFR mice and littermate TRE-dnEGFR mice at P35 with 14 dwd.
Source data for graphs. The zip file includes all raw numerical data in Prism files for graphs in each figure.
Acknowledgements
We thank Wanwan He, Kaiwei Zhang, Youguang Yang, and Shasha Zhang in Hangzhou Normal University for the assistance in EM image analyses, and Dr. Wanhua Shen for the assistance in electrophysiological experiments. We also thank Dr. Woo-Ping Ge in UT Southwestern for comments on the manuscript.
This work was supported by grants from National Natural Science Foundation of China (31371075, 31671070, and 31871030 to YT).
Footnotes
↵4 These authors contribute equally.