HIFα Regulates Developmental Myelination Independent of Autocrine Wnt Signaling

Transient HIFα stabilization during early CNS development

HIF1α and HIF2α share high sequence homology and function in a similar manner through partnering with HIF1β and subsequently activating HIFα target genes (Semenza, 2012). Therefore, we used antibodies against the representative HIF1α to determine whether HIFα protein is stabilized in the perinatal and early postnatal CNS, where the vasculature is still in immature state (Harb et al., 2013; Yuen et al., 2014). We found that HIF1α protein was detected by immunohistochemistry (IHC) in Sox10⁺ oligodendroglial lineage cells in the white matter areas of the spinal cord (Fig. 1A) and the forebrain (Fig. 1C) at P5. As positive controls, HIF1α-immunoreactive signals were markedly increased in age-matched transgenic Sox10-Cre:Vhl^fl/fl mice (Fig. 1B,D), in which Sox10-Cre-mediated VHL disruption prevents HIFα from degradation and causes HIFα accumulation specifically in oligodendroglial lineage cells. HIF1α signals became barely detectable in the forebrain white matter by P10 (Fig. 1E), which temporally approximates to the peak of CNS angiogenesis (Harb et al., 2013; Yuen et al., 2014), and remarkably upregulated in mice that had been challenged by hypoxia/ischemia injury at P6 (Fig. 1F; Shen et al., 2010; Liu et al., 2011; Chung et al., 2013). Our quantification data showed that the density of HIF1α⁺Sox10⁺ cells were significantly increased in P10 H/I-injured forebrain white matter compared with sham controls (mean ± SEM; sham, 4.2 ± 4.2; H/I, 312 ± 9.7; n = 3; t₍₄₎ = 29.3, p < 0.0001). Consistent with HIF1α histologic upregulation, Western blot assays of microdissected subcortical white matter demonstrated a substantial elevation of HIF1α and its canonical target gene PKM2 in H/I-injured brain (Fig. 1G), which was also confirmed by PKM2 IHC (Fig. 1H). Concomitant with HIFα activation, myelin staining by MBP showed severe disturbance of developmental myelination in H/I-injured white matter compared with sham controls (Fig. 1I1–I3). These results demonstrate that HIFα is transiently stabilized in the early postnatal CNS but rapidly downregulated in the second postnatal week and that hypoxia/ischemia injury results in sustained HIFα stabilization in the brain white matter.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

HIFα stabilization in the early postnatal murine CNS. A, B, IHC of HIF1α and oligodendroglial lineage marker Sox10 in the spinal cord ventral white matter (Spc-WM, marked by dotted lines) of HIFα intact mice (A) and HIFα stabilized (Sox10-Cre:Vhl^fl/fl) mice (B) at P5. C, D, IHC of HIF1α and Sox10 in the forebrain periventricular white matter (Fbr-PWM; marked by dotted lines) of HIFα intact mice (C) and HIFα stabilized (Sox10-Cre:Vhl^fl/fl) mice (D) at P5. E, F, IHC of HIF1α and Sox10 in the Fbr-PWM of P10 mice that had been subjected to sham operation (E) and H/I injury on P6 (i.e., 4 d post-H/I or sham; F; for details, see Materials and Methods). Arrowheads in A–F point to HIF1α⁺Sox10⁺ cells. G, Western blot of microdissected subcortical white matter for HIF1α and canonical HIFα target PKM2 at 6 h post-H/I or sham. GAPDH, Internal loading control. H, PKM2 IHC in the Fbr-PWM at P10, 4 d post-H/I or sham. I1–I3, IHC and quantification of MBP staining in the Fbr-PWM at P10, 4 d post-H/I or sham. MBP intensity was normalized to that of the contralateral brain hemispheres to the occluded artery of the same mouse. Scale bars, 20 µm.

HIFα hyperactivity impairs OPC differentiation but not oligodendrocyte maturation

Pathologic hypoxia–ischemia insults caused aberrant HIFα activation in the white matter (Fig. 1F–H) and hypomyelination (Fig. 1I1–I3). To determine whether aberrant HIFα stabilization disturbs normal developmental myelination, we induced HIFα hyperactivity by conditionally disrupting the negative regulator VHL (Fig. 1B,D). Constitutive Sox10-Cre:Vhl^fl/fl (Sox10:VHL cKO) pups were much smaller than littermate controls (Vhl^fl/+, or Vhl^fl/fl) at birth and rarely survived beyond P7. Our prior data show that Sox10-Cre-mediated gene disruption initiates in early OPCs during embryonic CNS development (Zhang et al., 2018a). Sox10-Cre-elicited HIFα hyperactivity (see Fig. 6A) caused severe hypomyelination in the developing postnatal spinal cord (Fig. 2A). The number of CC1⁺ mature OLs (Fig. 2B,C) and the mRNA level of major myelin protein (Fig. 2D) were reduced by ∼40–60% in Sox10:VHL cKO mutants compared with littermate controls. However, the number of PDGFRα⁺ OPCs was unaffected in Sox10:VHL cKO mutants (Fig. 2C). These data suggest that constitutive HIFα activation in OPCs inhibits OPC differentiation and disturbs developmental myelination.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

HIFα activation impairs developmental myelination by inhibiting OPC differentiation but not subsequent OL maturation. A, Myelin staining of MBP in the spinal cord of Sox10-Cre:Vhl^fl/fl (Sox10:VHL cKO) and littermate Ctrl (Vhl^fl/+ or Vhl^fl/fl) pups at P5. Most Sox10:VHL cKO pups died by the first postnatal week. B, Representative confocal images of Sox10 (red) and differentiated OL marker CC1 (green) in P5 spinal cord (Ctrl, n = 5; cKO, n = 3). Blue, DAPI nuclear counterstaining. C, Quantification of Sox10⁺CC1⁺ OLs and Sox10⁺PDGFRa⁺ OPCs (right) in P5 spinal cord (n = 5 Ctrl, n = 3 cKO). D, qRT-PCR assay for myelin-specific genes in the spinal cord at P5. Plp, Proteolipid protein; Cnp, 2′,3′-cyclic-nucleotide 3′-phosphodiesterase; Mobp, myelin oligodendrocyte basic protein; Mog, myelin oligodendrocyte glycoprotein. E, F, mRNA in situ hybridization (ISH) of Plp and Mbp in the spinal cord (E) and forebrain (F) of Cnp-Cre:Vhl^fl/fl (Cnp:VHL cKO) and littermate control (Cnp-Cre:Vhl^fl/+, Vhl^fl/+, or Vhl^fl/fl) mice at P14. G, Myelination in the adult spinal cord indicated by toluidine blue staining of semithin sections (left) and TEM of ultrathin section (right). Red asterisks in G denotes blood vessels, the density of which is elevated in Cnp:VHL cKO mice (Guo et al., 2015). H, Mbp mRNA levels quantified by qRT-PCR in Cnp:VHL cKO and Ctrl mice at different time points. Spc, Spinal cord; Fbr, forebrain. I, densities of Sox10⁺CC1⁺ OLs and Sox10⁺PDGFRα⁺ OPCs in P8 spinal cord of Cnp:VHL cKO and Ctrl mice. Scale bars: A, 100 µm; B, 20 µm; E, F, 200 µm; G, left, 10 µm; G, right, 1 µm. Data are presented as the mean ± SEM; gray bars are from Ctrl, white bars from cKO; two-tailed Student's t test, *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. The above data presentation and statistics are applied to all figures unless otherwise indicated).

To determine the effect of HIFα hyperactivity on subsequent oligodendrocyte maturation, we analyzed Cnp-Cre:Vhl^fl/fl (Cnp:VHL cKO) mutants. Previous studies have demonstrated that, compared with Sox10-Cre, Cnp-Cre drives gene disruption in later stages of oligodendrocyte development (Moyon et al., 2016; Zhang et al., 2018a,b). Cnp:VHL cKO mice were viable and indistinguishable from their littermate controls (Vhl^fl/fl, Vhl^fl/+, or Cnp-Cre:Vhl^fl/+) throughout the early postnatal and adult ages. In sharp contrast to Sox10:VHL cKO mice, Cnp:VHL cKO mice had normal levels of myelin gene expression (Fig. 2E,F,H); unperturbed myelination, evidenced by toluidine blue myelin staining (Fig. 2G, left) and TEM myelination assay (Fig. 2G, right); and normal numbers of OLs and OPCs (Fig. 2I) despite the elevated density of blood vessels in the CNS we recently reported (Guo et al., 2015; Zhang et al., 2020). These data indicate that HIFα hyperactivity in OLs does not affect their maturation. Together, our genetic evidence convincingly demonstrates that elevated HIFα activity perturbs developmental myelination by inhibiting OPC differentiation but not OL maturation, suggesting a lineage stage-dependent role of HIFα in developmental myelination.

We next used a time-conditional and OPC-specific genetic model to circumvent the early death of Sox10:VHL cKO mice and investigate whether time-conditional HIFα activation in postnatal OPCs impacts their differentiation. Toward this end, we generated Pdgfrα-CreER^T2:Vhl^fl/fl (PDGFRα:VHL cKO) mutants. Our data showed that OPC-specific HIFα activation during neonatal ages (Fig. 3A,B) significantly reduced the expression of genes encoding major myelin proteins (Fig. 3C) and decreased the densities of mature OLs, but not OPCs, in the spinal cord (Fig. 3D). IHC of myelin protein PLP demonstrated a marked hypomyelination (Fig. 3E). Further analysis of myelination by transmission electron microscopy (Fig. 3F) demonstrated that the density of myelinated axons was reduced by ∼42% in the ventral spinal white matter of PDGFRα:VHL cKO mice compared with littermate controls (Fig. 3G), which is consistent with histologic myelin protein assays (Fig. 3F). To assess the long-term effect of HIFα activation, we induced HIFα activation by tamoxifen injections to neonatal mice (Fig. 3H,I) and analyzed oligodendroglial phenotypes in the adult CNS (P60–P70). We found that the number of OLs, but not OPCs, was persistently reduced in the corpus callosum of PDGFRα:VHL cKO mice compared with control mice (Fig. 3J), suggesting that neonatal HIFα activation inhibits OPC differentiation in the adult CNS. Given that virtually all axons in the optic nerve are myelinated in the adult rodents (Bercury and Macklin, 2015), we then analyzed myelinated axons visualized by toluidine blue myelin staining of semithin resin sections (Fig. 3K1–K4). Our quantification demonstrated an ∼44% reduction in the density of myelinated axons in the optic nerve of adult PDGFRα:VHL cKO mutants compared with age-matched fully myelinated control optic nerves (Fig. 3K5). Together, our data suggest that HIFα activation inhibits long-term OPC differentiation and results in hypomyelination throughout postnatal CNS development.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Inducible HIFα stabilization in postnatal OPCs inhibits OPC differentiation. A, experimental design for B–F. PDGFRα:VHL cKO (Pdgfrα-CreER^T2:Vhl^fl/fl) and littermate control (Vhl^fl/fl) mice were treated with tamoxifen intraperitoneally at P1, P2, and P3, and analyzed at P8. B, qRT-PCR quantification of HIFα target gene hexokinase 2 Hk2 in the spinal cord. C, Expression of major myelin proteins in the spinal cord quantified by qRT-PCR. D, Representative confocal images and the density of Sox10⁺/CC1⁺ OLs and Sox10⁺/PDGFRα⁺ OPCs. E, IHC of PLP showing myelination in the spinal cord ventral white matter. F, G, Representative TEM images (F) and the density of myelinated axons (G) in the spinal cord ventral white matter. H, Experimental design for I–K. PDGFRα:VHL cKO and littermate control mice were treated with tamoxifen intraperitoneally at P1, P2, and P3, and analyzed at P60–P70. I, Representative confocal images showing HIF1α stabilization in PDGFRα⁺ OPCs (arrowheads) of PDGFRα:VHL cKO mice. J, Densities of CC1⁺ mature OLs and PDGFRα⁺ OPCs in the corpus callosum. K1–K4, Toluidine blue staining for myelin on semithin (500 nm) resin sections showing hypomyelination in the optic nerve of adult PDGFRα:VHL cKO mice. Boxed areas in K1 and K3 were shown at higher-magnification images in K2 and K4, respectively. K5, Density of myelinated axons in the adult optic nerves. Scale bars: D, 25 µm; E, 100 µm; F, 0.5 µm; I–K5, 10 µm.

Together, our in vivo data derived from three different genetic mouse models suggest that HIFα hyperactivity disturbs developmental myelination by inhibiting OPC differentiation.

Disrupting HIFα in OPCs but not OLs delays developmental myelination

Transient HIFα stabilization (Fig. 1A–E) suggests that HIFα may regulate OPC differentiation and/or myelination. To define the role of physiological HIFα stabilization in developmental myelination, we analyzed oligodendroglial phenotypes in HIFα cKO genetic mice. Our extensive analysis of HIF1α or HIF2α conditional knock-out mice driven by Pdgfrα-CreER^T2 and Cnp-Cre revealed no phenotypic abnormalities of the oligodendroglial lineage, suggesting functional redundancy between HIF1α and HIF2α. We therefore generated HIF1α and HIF2α double conditional knockout (referred to as HIFα cKO).

To determine the role of HIFα in OPC differentiation, we used Pdgfrα-CreER^T2:Hif1α^fl/fl:Hif2α^fl/fl (PDGFRα:HIFα cKO) transgenic mice. Tamoxifen was administered to PDGFRα:HIFα cKO and littermate control (Hif1α^fl/fl:Hif2α^fl/fl) pups at P1, P2, and P3, and the CNS tissue was analyzed at P8. Our previous data show that this tamoxifen paradigm induces >85% of gene cKO efficiency in OPCs (Zhang et al., 2018a,b). The density of differentiated OLs (CC1⁺Olig2⁺), but not OPCs (PDGFRα⁺Olig2⁺), was significantly decreased in the spinal cord white matter of PDGFRα:HIFα cKO mice compared with HIFα controls (Fig. 4A,B), suggesting that HIFα regulates OPC differentiation and is dispensable for OPC population expansion. In agreement with impaired OPC differentiation, the mRNA levels of major myelin protein (Mbp and Plp) and mature OL-enriched protein [Opalin and QK (called CC1)] were significantly reduced in PDGFRα:HIFα cKO mice compared with littermate controls (Fig. 4C). In the forebrain, developmental myelination in the periventricular white matter was delayed in PDGFRα:HIFα cKO mice at P8 (Fig. 4D) and became indistinguishable from that in littermate control mice at P14 (Fig. 4E). These data suggest that HIFα regulates the timing of developmental myelination.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

HIFα inactivation transiently delays developmental myelination by controlling OPC differentiation but not subsequent OL maturation. A–E, PDGFRα:HIFα cKO and littermate control (Hif1α^fl/fl:Hif2α^fl/fl) mice were treated with tamoxifen at P1, P2, and P3, and analyzed at P8 (n = 4 each group) and P14 (n = 3 each group). A, B, IHC (A) and quantification (B) of Olig2⁺CC1⁺OLs (arrowheads) and Olig2⁺PDGFRα⁺ OPCs (arrows) in the spinal cord. C, Relative expression of myelin protein genes of Mbp and Plp and mature OL-specific gene of Opalin and quake [Qk (called CC1)] in the forebrain measured by qRT-PCR. D, IHC of myelin marker MBP and pan-axonal marker SMI312 in forebrain periventricular white matter (left, marked by dotted area) and quantification (right) at P8. EC, External capsule; IC, internal capsule; Ctx, cortex. E, IHC of MBP in forebrain periventricular white matter at P14 (tamoxifen at P1, P2, and P3). F–J, Cnp:HIFα cKO and littermate control (Hif1α^fl/fl:Hif2α^fl/fl) mice were analyzed at P2, P8, and P14. F, qRT-PCR quantification of HIFα target genes Ldha and Vegfa in P8 spinal cord. G, Myelin staining by MBP in the forebrain. CC, Corpus callosum; Hip, hippocampus. H, I, Densities (#/mm²) of Olig2⁺CC1⁺ OLs (H) and Olig2⁺PDGFRα⁺ OPCs (I) at P14. Spc-WM, Spinal cord white matter; Fbr-CC, forebrain corpus callosum. J, qRT-PCR assay of exon3b-containing Plp mRNA, which is specific to mature OLs, in the spinal cord at different time points. K, IHC of CC1 and HIF1α in P8 forebrain periventricular white matter (Fbr-PWM). Arrowheads point to double-positive cells. LV, Lateral ventricle. Scale bars: A, K, 10 μm; D, 20 μm; E, 100 μm; G, 250 μm.

To determine the role of HIFα in oligodendroglial maturation, we used Cnp-Cre:Hif1α^fl/fl:Hif2α^fl/fl (Cnp:HIFα cKO) and littermate control (Hif1α^fl/fl:Hif2α^fl/fl) mice. Cnp:HIFα cKO mice were born at expected Mendelian ratios and displayed no behavioral abnormalities during postnatal development and throughout adult ages. The expression of canonical HIFα target genes Ldha and Vegfa (Sharp and Bernaudin, 2004) was significantly reduced in Cnp:HIFα cKO mice (Fig. 4F), thus validating the disruption of HIFα function. Unexpectedly, we found no significant difference in developmental myelination (Fig. 4G), the number of CC1⁺Olig2⁺ mature OLs (Fig. 4H) and PDGFRα⁺Olig2⁺ OPCs (Fig. 4I), and myelin gene expression (Fig. 4J) between Cnp:HIFα cKO and littermate control mice at different ages (P2, P8, and P14). Double IHC showed that some CC1⁺ OLs were positive for HIF1α in the periventricular white matter of the brain at P8 (Fig. 4K, arrowheads) but not at P14 (data not shown).

These data indicate that HIFα plays a minor role in oligodendrocyte maturation. Collectively, the contrast phenotypes derived from PDGFRα:HIFα cKO versus Cnp:HIFα cKO mice support a working model that HIFα controls the timing of developmental myelination by regulating the differentiation of OPCs into OLs but not subsequent oligodendrocyte maturation.

Disrupting oligodendroglial HIFα does not affect axonal integrity or cell survival

HIFα is required for neuronal survival, as HIFα deficiency in Nestin-expressing neural cells results in massive neuronal death and neurodegeneration (Tomita et al., 2003). Recent data reported that HIFα deficiency in Olig1-expressing neural cells leads to neuronal death and widespread axonal injury in the corpus callosum at the perinatal ages (Yuen et al., 2014). To determine whether OPC- or OL-specific HIFα deficiency causes axonal damage and cell death, we used IHC of SMI32 monoclonal antibody and cleaved Caspase 3 (CC3) antibody. SMI32 labels functionally compromised axons (Soulika et al., 2009; Yuen et al., 2014) and a subset of pyramidal neuronal cell bodies and dendrites during normal development (Campbell and Morrison, 1989; Voelker et al., 2004; Haynes et al., 2005; Fig. 5A, Ctx). We did not observe SMI32-immunoreactive signals in the corpus callosum of Cnp:HIFα cKO (Fig. 5A) and PDGFRα:HIFα cKO (Fig. 5C) mice compared with respective littermate controls. In contrast, many tightly packed axons running through the corpus callosum were clearly labeled by the pan-axonal marker SMI312 (Fig. 5B,D). Furthermore, the densities of CC3⁺ apoptotic cells and CC3⁺/Sox10⁺ apoptotic oligodendroglial lineage cells were statistically indistinguishable between PDGFRα (or Cnp):HIFα cKO mutants and their respective littermate controls (Fig. 5E–G). We used TEM to assess the structural integrity of axons in the corpus callosum of PDGFRα:HIFα cKO mice at P8, a time point at which few axons are myelinated. The density of axons with diameter ≥400 nm, which will be myelinated by oligodendrocyte during postnatal development (Lee et al., 2012; Hines et al., 2015), was similar between PDGFRα:HIFα cKO mice (458,401 ± 59,957 axons/mm²) and littermate controls (485,558 ± 38,780 axons/mm²; t₍₄₎ = 0.4, p = 0.72), suggesting that axons in HIFα cKO mice are not subjected to degeneration. We did not find visible morphologic differences in axonal structure (Fig. 5H,I, Ax) and axonal mitochondria (Fig. 5H,I, Mt) between PDGFRα:HIFα cKO and littermate control mice. Thus, disrupting oligodendroglial HIFα does not affect neuronal cell survival and axonal integrity in the white matter in the early postnatal CNS.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Oligodendroglial HIFα is dispensable for axonal integrity or cell death. A, Immunostaining images of P14 forebrain using SMI32, a monoclonal antibody recognizing the nonphosphorylated neurofilament proteins of heavy chain (NFH), which has been reported labeling injured axons and a subset of pyramidal neuron bodies and dendrites. Boxed areas show the corpus callosum (CC) in higher-magnification images of SMI32 and MBP in the right panels. Ctx, Cortex; Hip, hippocampus. Note the absence of SMI32-positive signals in the CC of both Cnp:HIFα cKO and Ctrl mice. B, Immunostaining images of P14 CC using SMI312, a monoclonal antibody recognizing both phosphorylated and nonphosphorylated NFH (a pan-axonal marker). C, D, IHC of SMI32/MBP (C) and SMI312 (D) in the CC of P8 non-Cre Ctrl and PDGFRα:HIFα cKO and littermate control (Hif1α^fl/fl:Hif2α^fl/fl) mice (tamoxifen injection at P1, P2, and P3). E, F, Representative confocal images and quantification of cells positive for cleaved CC3 and/or pan-oligodendroglial lineage marker Sox10 in P8 CC of PDGFRα:HIFα cKO and control mice. Arrowheads point to CC3⁺/Sox10⁺ cells. G, Quantification of CC3⁺ and CC3⁺/Sox10⁺ cells in P14 CC of Cnp:HIFα cKO and control mice. H, I, Representative TEM images showing the cross-sections of axons in the corpus callosum of HIFα Ctrl (I) and cKO (H) mice at P8 (tamoxifen injections at P1, P2, and P3) when myelination is barely detectable at the TEM level. Note that the morphology of axons (Ax) and axonal mitochondria (Mt) were indistinguishable between HIFα Ctrl and cKO mice. Scale bars: A, low magnification, 100 µm; A, high magnification, 20 µm; B–E, 20 µm; H, I, 400 nm.

Oligodendroglial HIFα plays a minor role, if any, in regulating Wnt/β-catenin activity

A previous study reported that aberrant HIFα stabilization in OPCs activates Wnt7a/Wnt7b expression and canonical Wnt signaling (i.e., Wnt/β-catenin; Yuen et al., 2014). The unbiased RNA sequencing (RNAseq) analysis of Sox10-Cre:Vhl^fl/fl mice prompted us to revisit HIFα-Wnt regulation in oligodendroglial lineage cells (Fig. 6A–E). Sox10-Cre-mediated VHL cKO markedly elevated HIFα activity, as evidenced by the increased expression of canonical HIFα target genes Hk2, Ldha, Glut1, Bnip3, and Vegfa (Fig. 6A). In stark contrast to HIFα activation, there were no significant changes in the expression of Wnt genes Wnt7a and Wnt7b, and canonical Wnt signaling target genes Axin2 and Naked1 (Fig. 6B). The KEGG Pathway of HIF-1 signaling and the gene ontology (GO) term of response to hypoxia were significantly overrepresented among 255 upregulated genes (Fig. 6C,D), whereas the GO terms of oligodendrocyte differentiation and sterol biosynthesis were significantly overrepresented among 351 downregulated genes (Fig. 6E), which is in agreement with elevated HIFα signaling and diminished OPC differentiation in the CNS of Sox10:VHL cKO mice we documented in Figure 2A–D. However, we did not find overrepresentation of Wnt-related KEGG Pathways or GO terms among the differentially regulated genes. The unbiased analysis of gene regulation at the transcriptomic level strongly suggests that oligodendroglial HIFα stabilization plays a minor role in regulating canonical Wnt signaling.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Stabilizing HIFα activates HIFα-mediated signaling but does not perturb Wnt/β-catenin signaling in the oligodendroglial lineage and in the CNS. A, B, Relative expression of HIFα and Wnt target genes from unbiased RNAseq analysis. FPKM, Fragments per kilobase of transcript per million mapped reads, a normalized measurement of gene expression. The RNA was prepared from the spinal cord of P5 Sox10-Cre:Vhl^fl/fl mice (n = 3, white bars) and littermate control (Vhl^fl/+ or Vhl^fl/fl; n = 3, gray bars) mice. C–E, Top three significant enriched KEGG Pathway and GO terms among 255 upregulated and 351 downregulated genes derived from unbiased RNAseq. Note that HIFα signaling pathway (C) and response to hypoxia (D) were significantly upregulated, validating HIFα hyperactivation in Sox10-Cre:Vhl^fl/fl mice and that oligodendroglial cell differentiation was significantly downregulated (E), which is in agreement with the data of Figure 2A–D. F, Experimental design for G and H. PDGFRα:VHL cKO (n = 3, white bars) and littermate control (Vhl^fl/fl, n = 3, gray bars) mice were treated with TM at P5, and forebrain OPCs at P6 were acutely purified by MACS. G, H, qRT-PCR quantification of HIFα target gene (G) and Wnt target gene (H) expression in purified OPCs. I, Expression of Wnt target gene (Axin2 and Naked1) and Wnt7a/7b in P8 spinal cord (n = 3, Ctrl, gray bars; n = 5, PDGFRα cKO, white bars) quantified by qRT-PCR. J, Western blot images and relative expression (normalized to the internal loading control β-actin) of the active form of β-catenin (ABC) and Axin2 in P14 forebrain (n = 3 each group). K, Expression of Wnt target genes (Axin2 and Naked1) and Wnt7a/7b in P8 spinal cord measured by qRT-PCR (n = 4 each group).

To strengthen our conclusion derived from RNAseq, we analyzed OPCs isolated from PDGFRα:VHL cKO mice. Tamoxifen was injected to PDGFRα:VHL cKO and littermate control mice at P5, and brain OPCs were acutely purified at P6 (Fig. 6F) by MACS. As shown in Figure 6G, HIFα was indeed functionally stabilized in purified VHL-disrupted OPCs, as evidenced by the elevated expression of HIFα target genes Hk2, Ldha, and Glut1. However, there were no significant changes in the expression of canonical Wnt/β-catenin signaling target genes Axin2 and Naked1 and Wnt ligands Wnt7a and Wnt7b in VHL-disrupted OPCs compared with VHL-intact OPCs (Fig. 6H), suggesting that genetically stabilizing HIFα does not perturb the activity of autocrine Wnt/β-catenin or the expression of Wnt7a and Wnt7b in OPCs.

To corroborate our findings, stabilizing HIFα in PDGFRα⁺ OPCs did not affect the activity of Wnt/β-catenin signaling in the spinal cord at P8 (tamoxifen at P1, P2, and P3) and in the forebrain at P14 (tamoxifen at P6 and P7) of PDGFRα:VHL cKO mice compared with littermate controls, which was supported by the unaltered expression of canonical Wnt target genes at the mRNA and protein levels (Fig. 6I,J). We also found that the expression of Axin2, Naked1, Wnt7a, and Wnt7b was not perturbed in PDGFRα:HIFα cKO mice compared with littermate controls (Fig. 6K). Thus, HIFα does not regulate canonical Wnt signaling activity or Wnt7a/Wnt7b expression in the oligodendroglial lineage and in the CNS.

Blocking oligodendroglial-derived Wnt signaling does not affect HIFα hyperactivation-elicited inhibition of OPC differentiation and myelination

That nonperturbation of Wnt signaling in HIFα-stabilized OPCs led us to hypothesize that HIFα hyperactivation inhibits myelination in a manner independent of autocrine Wnt activation in OPCs. To test our hypothesis, we stabilized HIFα function and simultaneously blocked oligodendroglial-derived autocrine Wnt signaling.

Previous studies demonstrated that disrupting WLS blocks the secretion of Wnt ligands from Wnt-producing cells and inhibits intracellular Wnt signaling activity in Wnt-receiving cells (Bänziger et al., 2006; Bartscherer et al., 2006). To determine whether WLS deficiency blocks autocrine Wnt/β-catenin activity in OPCs, we knocked down WLS in Wnt3a-expressing primary OPCs (Fig. 7A–C). We chose Wnt3a because it is a typical Wnt ligand that activates the canonical Wnt signaling in Wnt-receiving cells. Wnt3a expression in OPCs significantly increased Wnt3a concentration and simultaneous WLS knockdown abrogated Wnt3a elevation in the culture medium of primary OPCs (Fig. 7D), suggesting that WLS is required for Wnt3a secretion from OPCs. We found that the activity of the canonical Wnt signaling, as assessed by Wnt target genes Axin2 and Sp5, was upregulated in Wnt3a-expressing OPCs but blocked in Wnt3a-expressing/WLS-deficient OPCs (Fig. 7E,F). These results demonstrate that disrupting WLS blocks OPC-derived autocrine Wnt/β-catenin signaling activity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

WLS deficiency inhibits Wnt3a secretion and blocks autocrine Wnt/β-catenin signaling activity in OPCs. A, Experimental design of cell transfection of Wls shRNA and Wnt3a-expressing plasmids to primary brain OPC maintained in the growth medium. B, C, qRT-PCR assay of Wls (B) and Wnt3a (C) expression in transfected primary OPCs. One-way ANOVA followed by Tukey's multiple-comparisons test: Wls, F_(3,8) = 15.39, p = 0.001; Wnt3a, F_(3,8) = 32.81, p < 0.0001. D, ELISA quantification of Wnt3a in the growth medium in the absence or presence of OPCs transfected by Wls shRNA and/or Wnt3a. One-way ANOVA followed by Tukey's multiple-comparisons test, ***p < 0.001; ns, not significant. F_(4,10) = 6.70, p < 0.0069. Note that Wnt3a concentration in the GM in the presence of primary OPCs is statistically indistinguishable from that in the GM alone, suggesting that intact primary OPCs do not secrete Wnt3a. E, F, qRT-PCR quantification of Wnt target genes Axin2 and Sp5 in OPCs transfected with Wls shRNA and/or Wnt3a. One-way ANOVA followed by Tukey's multiple-comparisons test: Axin2, F_(3,8) = 11.06, p = 0.0032; Sp5, F_(3,8) = 13.25, p = 0.0018.

Genetically disrupting WLS in PDGFRα⁺ OPCs had no rescuing effects on HIFα hyperactivation-elicited hypomyelination (Fig. 8A–C), impaired OPC differentiation (Fig. 8D,E), and diminished myelin gene expression (Fig. 8F) in PDGFRα:VHL/WLS cKO mice, suggesting that autocrine Wnt signaling is unlikely a downstream pathway in mediating HIFα stabilization-induced inhibition of OPC differentiation. To corroborate this conclusion, we used a different inducible genetic model, Sox10-CreER^T2:Vhl^fl/fl:Wls^fl/fl (iSox10:VHL/WLS cKO; Fig. 9A). Conditionally stabilizing HIFα in Sox10-expressing oligodendroglial lineage cells resulted in impaired OPC differentiation and hypomyelination in iSox10:VHL cKO mice; however, simultaneous WLS cKO did not affect the degree of inhibited OPC differentiation and hypomyelination in iSox10:VHL/WLS cKO mice (Fig. 9B–E). We further demonstrated that disrupting WLS alone in Sox10-expressing oligodendroglial lineage cells had no detectable effect on normal OPC differentiation and developmental myelination in Sox10-CreER^T2:Wls^fl/fl (iSox10:WLS cKO) mice compared with non-Cre littermate control mice (Fig. 10). Collectively, our results derived from cell-specific genetic approaches suggest that autocrine Wnt/β-catenin signaling plays a dispensable role in HIFα-regulated OPC differentiation and myelination.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Disrupting WLS in PDGFRα-expressing OPCs does not affect the degree of inhibited OPC differentiation and hypomyelination elicited by HIFα hyperactivation. A, Experimental design for B–F. PDGFRα:VHL cKO, Pdgfrα-CreER^T2:Vhl^fl/fl (n = 5); PDGFRα:VHL/WLS cKO, Pdgfrα-CreER^T2:Vhl^fl/fl:Wls^fl/fl (n = 4); non-Cre Ctrl carrying Vhl^fl/fl and/or Wls^fl/fl (n = 8). B, C, Representative confocal images (B) and quantification of myelination (C) by MBP staining. The signal of SMI312, a pan-axonal marker is indistinguishable among each group and used as an internal control of MBP quantification. One-way ANOVA followed by Tukey's multiple-comparisons test: F_{(2, 14)}=31.93 p < 0.0001. D, E, Representative confocal images (D) and quantification (E) of Sox10⁺/CC1⁺ differentiated OLs (D, arrowheads), Sox10⁺/PDGFRα⁺ OPCs (D, arrows), and Sox10⁺ oligodendroglial lineage cells in the spinal cord. One-way ANOVA followed by Tukey's multiple-comparisons test: Sox10⁺, F_(2,14) = 17.34, p = 0.0002; Sox10⁺/CC1⁺, F_(2,14) = 18.44, p = 0.0001; Sox10⁺/PDGFRα⁺, F_(2,14) = 2.967, p = 0.0843. F, qRT-PCR quantification of mature OL-enriched genes in the spinal cord. One-way ANOVA followed by Tukey's multiple-comparisons test: Mbp, F_(2,14) = 11.53, p = 0.0011; Mag, F_(2,14) = 39.88, p < 0.0001; Cnp, F_(2,14) = 38.78, p < 0.0001. Scale bars: B, D, 20 µm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Disrupting WLS in Sox10-expressing oligodendroglial lineage cells does not affect HIFα hyperactivation-elicited inhibition of OPC differentiation and hypomyelination. A, Experimental design for B–E. Tamoxifen-inducible Sox10-CreER^T2:Vhl^fl/fl (iSox10:VHL cKO, n = 3); Sox10-CreER^T2:Vhl^fl/fl:Wls^fl/fl (iSox10:VHL/WLS cKO, n = 3); non-Cre Ctrl carrying Vhl^fl/fl and/or Wls^fl/fl (n = 5).& B, C, Representative confocal images (B) and quantification (C) of myelination by MBP staining of the spinal cord. SMI312 signal was used as an internal control of MBP quantification. One-way ANOVA followed by Tukey's multiple-comparisons test: F_(2,8) = 8.484, p = 0.0105. D, E, Representative confocal images (D) and quantification (E) of Sox10⁺/CC1⁺ differentiated OLs (D, arrowheads) and Sox10⁺/PDGFRα⁺ OPCs (D, arrows) in the spinal cord. One-way ANOVA followed by Tukey's multiple-comparisons test: Sox10⁺/CC1⁺, F_(2,8) = 11.90, p = 0.004; Sox10⁺/PDGFRα⁺, F_{(2, 14)} = 2.880, p = 0.1143. Scale bars: B, D, 20 µm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Disrupting WLS in Sox10-expressing oligodendroglial lineage cells does not perturb normal OPC differentiation and myelination. A, Experimental design for B–G. Tamoxifen-inducible Sox10-CreER^T2:Wls^fl/fl (iSox10:WLS cKO, n = 4); littermate non-Cre control Wls^fl/fl mice (n = 3). B, C, qRT-PCR quantification of Wls (B) and differentiated OL-enriched gene Cnp and Ugt8a (C) in the spinal cord. D, E, Representative confocal images (D) and quantification (E) of Sox10⁺/CC1⁺ differentiated OLs (D, arrowheads), Sox10⁺/PDGFRα⁺ OPCs (D, arrows), and Sox10⁺ oligodendroglial lineage cells in the spinal cord. F, G, Representative confocal images (B) and quantification (C) of myelination by MBP staining of the spinal cord. SMI312 signal was used as an internal control of MBP quantification. Scale bars: D, 20 µm; E, 50 µm.

Sustained Sox9 activation mediates HIFα stabilization-elicited inhibition of OPC differentiation

The mRNA level of Sox9 was significantly upregulated in primary OPCs maintained in growth medium that had been treated with DMOG, a potent HIFα stabilizer (Yuen et al., 2014), and this upregulation was abrogated on simultaneous treatment of chetomin (Fig. 11A), an inhibitor of HIFα transcriptional activity (Kung et al., 2004; Viziteu et al., 2016). To support the functional regulation, our assay of ChIP followed by qPCR assay (ChIP-qPCR) showed that HIFα physically bound to the promoter region (−828 bp) of the rat Sox9 gene (Fig. 11B). We found that Sox9 mRNA expression was significantly upregulated in the spinal cord of PDGFRα:VHL cKO mice compared with littermate controls at P8 (Fig. 11C) and P14 (Fig. 11D). At the cellular level, sustained Sox9 expression was observed in HIFα-stabilized Sox10-expessing oligodendroglial lineage cells in Sox10:VHL cKO mice (Fig. 11E, right, arrowheads), in sharp contrast to control mice in which Sox9 was expressed at a much lower level or was barely detectable in Sox10-expressing cells (Fig. 11E, left, arrowheads). Our quantification showed that the percentage of Sox10⁺ cells expressing Sox9 (Fig. 11F) and the level of Sox9 mRNA (Fig. 11G) were significantly increased in Sox10:VHL cKO mice. Furthermore, Sox9 expression was significantly elevated in primary OPCs, which were purified from Sox10:VHL cKO neonatal brain (Fig. 11H). Sox9 is expressed in virtually all astrocytes in the postnatal CNS (Sun et al., 2017; Zhang et al., 2018b). We found that the density of Sox9⁺GFAP⁺ astrocytes was not significantly different in the spinal cords of Sox10:VHL cKO mutants from that in control mice (Fig. 11I). We also found that the percentage of Sox10⁺ oligodendroglial cells expressing Sox9 (Fig. 11J1,J2) was unchanged in PDGFRα:HIFα cKO mice (Fig. 11J3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 11.

HIFα stabilization activates Sox9 expression. A, Sox9 expression in the purified primary OPCs assessed by qRT-PCR. Purified cortical OPCs from neonatal rat brains were treated with HIFα stabilizer DMOG (1 mm) in the absence and presence of HIFα signaling blocker chetomin (100 nm) for 7 h. n = 4 each group. One-way ANOVA followed by Tukey's multiple-comparisons test: F_(2,9) = 62.53, p < 0.0001. B, Potential HIFα binding sites (CGTG) in the upstream sequence of rat Sox9 genes [+1 bp is defined as the transcription start site (TSS)] and ChIP-qPCR verification of physical binding of HIFα to the promoter region of −828 bp in primary OPCs treated with DMOG. One-way ANOVA with Dunnett post hoc test comparing each group with negative control: F_(2,12) = 69.21 p < 0.0001. C, qRT-PCR quantification of Sox9 expression in P8 spinal cord of PDGFRα:VHL cKO (n = 5) and littermate control (Vhl^fl/fl, n = 3) mice that had been treated with tamoxifen at P1, P2, and P3. D, qRT-PCR quantification of Sox9 expression in P14 spinal cords of PDGFRα:VHL cKO (n = 3) and littermate control (Vhl^fl/fl, n = 3) mice that had received tamoxifen at P6 and P7. E, Confocal images showing that Sox9 is barely expressed in Sox10⁺ oligodendroglial lineage cells in the control spinal cord but upregulated in Sox10:VHL cKO mice at P5. WM, White matter; GM, gray matter. The boxed areas of #1 and #2 were shown at higher magnification at the right. F, Percentage of Sox10⁺ oligodendroglial lineage cells expressing Sox9 in the spinal cord (n = 3 each group). G, qRT-PCR quantification of Sox9 mRNA in the spinal cord of Sox10:VHL cKO (n = 5) and littermate control (Vhl^fl/+ and/or Vhl^fl/fl, n = 8) mice at P5. H, qRT-PCR assay of Sox9 expression in primary OPCs isolated from the neonatal brain of Sox10-Cre:Vhl^fl/fl (n = 4) and littermate control (n = 6) pups. I, Representative confocal images and quantification of Sox9⁺GFAP⁺ astrocytes in the ventral spinal cord of P5 Sox10-Cre:Vhl^fl/fl (n = 3) and littermate control (n = 6) mice. J1–J3, Representative confocal images and quantification of Sox10 and Sox9 in the ventral spinal cord of P8 PDGFRα:HIFα cKO (n = 4) and littermate control (n = 4) mice (tamoxifen at P1, P2, and P3). Arrowheads point to Sox10⁺Sox9⁺ cells. Scale bars: E, 50 μm; I, 10 μm; E, boxed areas, 10 μm; J1, J2, 20 μm.

Recent data demonstrate that Sox9 is rapidly downregulated on OPC differentiation (Reiprich et al., 2017), prompting us to hypothesize that sustained Sox9 expression may inhibits OPC differentiation. To test this hypothesis, we overexpressed Sox9 in primary OPCs in growth medium and assessed OPC differentiation in differentiating medium (Fig. 12A). Sox9 mRNA and protein were significantly increased in primary OPCs transfected with Sox9-expressing plasmid compared with those transfected with empty vector (EV; Fig. 12B–D). Sox9 overexpression reduced the mRNA expression of the most abundant myelin protein Plp and the potent prodifferentiation factor Myrf (Emery et al., 2009) assessed by qRT-PCR (Fig. 12E), the number of ramified MBP⁺/Sox10⁺ oligodendrocytes assessed by immunocytochemistry (Fig. 12F), and the protein level of MBP assessed by Western blot (Fig. 12G). To define the role of Sox9 activation in HIFα-regulated OPC differentiation, we stabilized HIFα in OPCs by DMOG treatment and simultaneously knocked down Sox9 expression by siRNA (Fig. 12H–J). Our data showed that Sox9 downregulation rescued the degree of myelin gene inhibition elicited by DMOG treatment, as evidenced by the density of MBP⁺/Sox10⁺ differentiated OLs with ramified morphology (Fig. 12K,L) and the expression of myelin gene Mbp (Fig. 12M). Collectively, our results suggest that sustained Sox9 expression is one of the downstream molecular pathways in mediating HIFα-regulated OPC differentiation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 12.

Sustained Sox9 expression inhibits OPC differentiation. A, Experimental design for B–G. Cortical primary OPCs in growth medium were transfected with Sox9-expressing plasmid (Sox9, n = 3) or EV (n = 3), then differentiated for 24 h in serum-free differentiating medium. B, qRT-PCR quantification of Sox9 mRNA expression. C, Representative immunocytochemical images of Sox9/Sox10. Arrowheads point to Sox9⁺/Sox10⁺ oligodendrocytes and arrows point to Sox9⁺/Sox10^– cells, which are presumably astrocytes. Scale bar, 20 μm. D, Percentage of Sox10⁺ oligodendroglial lineage cells that express Sox9 (top) and Sox9-immunoreactive intensity normalized to Sox10 (bottom). E, qRT-PCR assay of exon3b-containing Plp (Plp-E3b), which specifically expresses in differentiated OLs, and myelin regulatory factor (Myrf), a potent prodifferentiation gene. F, Representative immunocytochemical images of MBP/Sox10 and density (#/mm²) of Sox10⁺ oligodendroglial cells of complex ramified morphology indicated by MBP staining (F, arrowheads), Scale bar, 10 μm. G, Western blot images and quantification of MBP and internal loading control β-actin. H, Experimental designs for I–M, primary OPCs were treated with HIFα stabilizer DMOG and transfected with Sox9 siRNA twice with 24 h apart. I, J, Expression of Sox9 (I) and HIFα target gene Bnip3 (J) quantified by qRT-PCR. One-way ANOVA followed by Tukey's test: Sox9, F_(3,8) = 10.33, p = 0.004; Bnip3: F_(3,8) = 20.83, p = 0.0004. K, Representative immunostaining images of differentiated OLs labeled by MBP and Sox10. Scale bar, 10 µm. L, M, Density (#/mm²) of MBP⁺/Sox10⁺ differentiated OLs (L) and qRT-PCR assay of Mbp mRNA expression (M). One-way ANOVA followed by Tukey's test: MBP⁺/Sox10⁺, F_(3,8) = 33.29, p < 0.0001; Mbp, F_(3,8) = 97.96, p < 0.0001. N, Previously proposed and our alternative working model explaining the role of HIFα in OPC differentiation. Endogenous HIFα regulates CNS myelination by controlling upstream OPC differentiation but not downstream OL maturation in perinatal and early postnatal CNS, whereas aberrant HIFα activation (via VHL mutation of hypoxia-ischemia insult) inhibits OPC differentiation through the Wnt-independent and Sox9-dependent pathway.