Dual regulatory switch through interactions of Tcf7l2/Tcf4 with stage-specific partners propels oligodendroglial maturation

Constitutive activation of Wnt/β-catenin inhibits oligodendrocyte myelination. Tcf7l2/Tcf4, a β-catenin transcriptional partner, is required for oligodendrocyte differentiation. How Tcf7l2 modifies β-catenin signalling and controls myelination remains elusive. Here we define a stage-specific Tcf7l2-regulated transcriptional circuitry in initiating and sustaining oligodendrocyte differentiation. Multistage genome occupancy analyses reveal that Tcf7l2 serially cooperates with distinct co-regulators to control oligodendrocyte lineage progression. At the differentiation onset, Tcf7l2 interacts with a transcriptional co-repressor Kaiso/Zbtb33 to block β-catenin signalling. During oligodendrocyte maturation, Tcf7l2 recruits and cooperates with Sox10 to promote myelination. In that context, Tcf7l2 directly activates cholesterol biosynthesis genes and cholesterol supplementation partially rescues oligodendrocyte differentiation defects in Tcf712 mutants. Together, we identify stage-specific co-regulators Kaiso and Sox10 that sequentially interact with Tcf7l2 to coordinate the switch at the transitions of differentiation initiation and maturation during oligodendrocyte development, and point to a previously unrecognized role of Tcf7l2 in control of cholesterol biosynthesis for CNS myelinogenesis.

O ligodendrocyte (OL) myelination permits saltatory propagation of nerve signals and is critical for cognitive and motor functions in the vertebrate central nervous system (CNS) [1][2][3][4][5] . During myelination, OLs pass through multiple developmental stages, including OL precursor cell (OPC), immature premyelinating OL and mature myelinating OL stages. A series of signalling pathways including Wnt/b-catenin, BMP/Id and Notch/Hes signalling have been shown to negatively regulate OL differentiation 6,7 . Hyperactivation of canonical Wnt signalling leads to the inhibition of OL differentiation and myelination through constitutively activated b-catenin 8-10 , Wnt3a ligand treatment [11][12][13] or the loss of signalling inhibitors, as observed in Apc Min (ref. 14) or Apc knockout mice 15 . In addition to these signalling pathways that sense the presence of extrinsic factors in the environment, intrinsic factors such as transcription factors including Olig1/2, Sox10, Zeb2/Sip1, Yy1, Zfp191 and Myrf/Gm98 positively regulate OL development 6,7 . Despite the critical roles of these signalling pathways and transcriptional regulators in OL differentiation, the means by which different signalling pathways and transcriptional regulatory circuitries are integrated to control OL differentiation remains poorly understood.
Activation of Wnt signalling on Wnt ligand binding results in the stabilization and subsequent nuclear translocation of b-catenin 16 . Nuclear b-catenin binds to T-cell factor (TCF)/ lymphoid enhancer-binding factors including Tcf7l2 (a.k.a. Tcf4). The TCF complex with nuclear b-catenin activates Wnt target genes 16 and b-catenin-mediated transcriptional activation is principally through TCFs 17 . Tcf7l2 is a major transducer of b-catenin activity 17 and is highly expressed by OL lineage cells 8,9 . Although hyperactive Wnt signalling inhibits OL differentiation, intriguingly, the loss of the b-catenin effector Tcf7l2 leads to a block of OL differentiation in Tcf7l2-null animals 9,18 . At present, the mechanisms underlying Tcf7l2 regulation of CNS myelination and remyelination remain elusive. Importantly, direct transcriptional targets of Tcf7l2 have not been identified in OLs, a particular concern, as Tcf7l2 is a known partner of b-catenin, raising the paradox of how Tcf7l2 might exert functions opposing inhibitory functions of b-catenin.
By generating mice lacking the Tcf7l2 DNA-binding transcription-activating domain, we showed that Tcf7l2 transcriptional activity is crucial for OL myelination and remyelination. We further conducted genome-wide chromatin immunoprecipitation sequencing (ChIP-seq) profiling to comprehensively map the Tcf7l2 direct targets at different stages of OL development and find that Tcf7l2 engages the OL genome through its sequential interactions with stage-specific co-regulators, including the non-canonical Wnt signalling repressor Kaiso/Zbtb33 at the early phase of OPC differentiation and a differentiation-promoting factor, Sox10, later in OL differentiation. Our data further suggest that Tcf7l2 interacts with Kaiso to antagonize Wnt signalling activity at the differentiation onset, while coordinating with Sox10 to promote myelin gene expression during OL maturation. Furthermore, we find that Tcf7l2 and Sox10 interaction controls the cholesterol biosynthesis pathway for myelinogenesis. Thus, our studies define stage-dependent functions of Tcf7l2 during OL lineage development mediated through switching binding partners and provide a molecular framework for understanding the context-specific control of CNS myelination.

Results
Tcf7l2 transcriptional activity is vital for OL myelination. Tcf7l2 consists of several functional domains including the b-catenin-binding domain, Groucho/TLE-binding domain and HMG (high mobility group) DNA-binding domain (Fig. 1a) 19 . To assess the role of Tcf7l2 transcriptional activity during OL development, we generated mutant mice carrying a transcriptionally inactive Tcf7l2, in which the floxed Tcf7l2 exon 11 encoding the DNA-binding HMG box 20,21 was excised by an OL lineage-expressing Olig1-Cre 9,22 , to generate a Tcf7l2 in-frame mutant without the HMG domain. This yielded control (Tcf7l2 fl/ þ :Olig1-Cre þ / À ) and mutant (Tcf7l2 fl/fl : Olig1-Cre þ / À ) mice (designated as Tcf7l2DHMG; Fig. 1a). We confirmed the excision of the exon 11 in complementary DNAs of OPCs isolated from Tcf7l2DHMG neonates by quantitative reverse transcriptase-PCR (qRT-PCR), using primers spanning the deleted exon 11 (Fig. 1b). Expression of Tcf7l2DHMG, but not full-length Tcf7l2 protein, was detected in Tcf7l2-mutant spinal cords, although its level is lower when compared with that of control animals by western blot analysis (Fig. 1b).
To investigate OL differentiation in Tcf7l2 mutants, we first examined myelin gene expression in the brain. In the Tcf7l2DHMG cortex, we found that in contrast to their robust expression in the control, Mbp and Plp1 (proteolipid protein 1) expression was remarkably reduced at postnatal stages (Fig. 1c). The dysmyelinating phenotype persisted to adulthood (Fig. 1c). Myelin basic protein (MBP) expression or green fluorescent protein (GFP) signals from the CNP-mGFP transgenic line 23 was also reduced at P14 (Fig. 1d,e), consistent with the sustained decrease in the number of Plp1 þ OLs at P60 (Fig. 1f).
Impaired terminal OL differentiation in Tcf7l2 mutants might be due to a shortage in OPCs. We then assessed the formation of cortical OPCs using platelet-derived growth factor receptor-a (PDGFRa) immunolabelling and a PDGFRa-GFP reporter 24 . In the Tcf7l2DHMG cortex at P7, the number of PDGFRa þ OPCs was comparable to that in the control (Fig. 1g-j). The rate of OPC proliferation was also unaltered, as shown by 5-bromodeoxyuridine (BrdU) incorporation (Fig. 1i,j). Furthermore, inactivation of Tcf7l2 in the OL lineage did not affect the generation of neurons, astrocytes or microglia, as immunolabelled by NeuN, glial fibrillary acidic protein and Iba1, respectively ( Supplementary Fig. 1). These results indicated that Tcf7l2 is not essential for OPC formation.
In light of OL differentiation deficits noted in the Tcf7l2 mutants, we used electron microscopy to analyse myelin sheath morphologies in the corpus callosum and in optic nerves. Consistent with the decrease in myelin gene expression in the brain, the optic nerves of mutant mice exhibited a significantly reduced proportion of myelinated axons at early postnatal stages such as P14, this relative hypomyelination persisted into adulthood (Fig. 1k). Similarly, hypomyelination was also observed in the corpus callosum at P60 (Fig. 1l), in which the percentage of myelinated axons remained significantly lower than controls (Fig. 1m). These data suggest that the loss of Tcf7l2 transcriptional activity impairs myelination in both developing and adult brains.
To further verify the stage-specific function of Tcf7l2 in OL differentiation, we inactivated Tcf7l2 using other OL-lineage expressing Cre drivers, including Olig2-Cre, the expression of which begins in early OL progenitors 8 . Similar to Olig1-Cre-mediated mutagenesis, ablation of Tcf7l2 by Olig2-Cre resulted in a substantial reduction in the expression of the myelin genes Mbp and Plp1 in the mutant cortex at P14 (Supplementary Fig. 2a). In addition, in mice with Tcf7l2 ablated by a CNP-Cre line 25 , wherein Cre expression commences at early postmitotic OPC stages, there was a significant decrease in the number of CC1 þ or MAG þ myelinating OLs ( Supplementary  Fig. 2b,c). Together, these observations suggest that Tcf7l2 regulates myelination-associated gene expression and is critical for OPC differentiation.
Tcf7l2 activity is required for OL remyelination. In the developing spinal cord of Tcf7l2DHMG mice, expression of Mbp was substantially reduced at P0 and P7 (Fig. 2a), whereas the number of PDGFRa þ OPCs was comparable to controls at P7 and P14 ( Supplementary Fig. 3). In adulthood at P60, however, Mbp expression was similar to control (Fig. 2b). Consistently, myelin ultrastructure, the percentage of myelinated axons and their g-ratios were all comparable between control and adult Tcf7l2DHMG spinal cord at P60 (Fig. 2c-e), indicating a delayed myelination process in the Tcf7l2-mutant spinal cord.  As myelination in the spinal cord fully caught up when the Tcf7l2 mutants reached adulthood, we then capitalized on this phenotype to assess the function of Tcf7l2 in remyelination by employing the lysolecithin (LPC)-induced demyelination 26 . Local injection of LPC in the white matter induces rapid myelin breakdown and removal of myelin from adult CNS; myelin regenerates through an OPC recruitment phase at 7 days post lesion (dpl) and a remyelinating phase at 14 dpl 26   control mice, Tcf7l2 was drastically upregulated at 7 and 14 dpl within the LPC lesions (Fig. 2f), consistent with the previous findings 8 . To determine whether Tcf7l2DHMG is required for myelin repair, we analysed myelin gene expression in the lesions at 14 dpl, a phase of OL regeneration and remyelination. Compared with controls, we detected substantially lower levels of Mbp and Plp1 during remyelination in Tcf7l2DHMG mice (Fig. 2g,i), although PDGFRa þ OPCs were generated normally (Fig. 2g,h). Importantly, many fewer myelinated axons were detected in the lesions of Tcf7l2DHMG mice than in controls (Fig. 2j). The percentage and thickness of newly generated myelin sheaths around axons were significantly reduced in the Tcf7l2 mutants (Fig. 2k,l). These observations indicate that Tcf7l2 activity is critical for remyelination after demyelinating injury.
Downregulation of myelin genes in Tcf7l2DHMG mutants. In light of our data demonstrating impaired re/myelination capacity in the absence of Tcf7l2, we sought to identify the Tcf7l2-regulated genes. We carried out RNA-sequencing (RNA-seq) analysis using the OL-enriched optic nerves from control and Tcf7l2DHMG mice at P12, to identify differentially expressed genes. In accordance with dysmyelinating phenotypes, expression of myelin genes such as Cnp, Mbp, Ugt8a and Mog, and myelination-regulatory genes such as Sox10, Olig1 and Myrf, was significantly reduced in Tcf7l2 mutants ( Fig. 3a,b). In contrast, we observed an upregulation of OL differentiation inhibitors, including Id2, Notch2, Tgfb1/Tgfbr3 and Bmp6, as well as Wnt signalling pathway components including Wnt4/6, Axin2, Ctnnb1, Ccnd1 and Sp5 (Fig. 3a,b). Gene ontology analysis of downregulated genes identified cholesterol biosynthesis, axon ensheathment, OL differentiation and myelination (Fig. 3c), congruent with impaired OL differentiation in Tcf7l2 mutants. We further confirmed the downregulation of these myelinationassociated genes and differentiation regulators by qRT-PCR analysis (Fig. 3d). Moreover, overexpression of Tcf7l2 in OPCs enhanced expression of myelin genes such as Mbp, Cnp, Plp1 and Mag, while repressing Id2 expression (Fig. 3e). These observations suggest that Tcf7l2 transcriptional activity is both necessary and sufficient for OPC maturation.
Stage-specific Tcf7l2 targeting for OL lineage progression. To determine the expression pattern of Tcf7l2 during OL lineage progression, we treated rat OPCs with triiodothyronine (T3) for different durations. Tcf7l2 messenger RNA upregulated in differentiating OPCs after 1 and 3 days of T3-induced differentiation (Fig. 3f), but fell to a lower level in terminally differentiated OLs 5 days after T3 treatment. Similarly, Tcf7l2 immunoreactivity was weakly detected in PDGFRa þ OPCs but increased substantially in CNP þ differentiating OLs and then decreased in terminally differentiated MBP þ OLs (Fig. 3g).
To gain insights into the direct targets regulated by Tcf7l2, we carried out ChIP-seq analysis for Tcf7l2-chromatin occupancy in OPCs, immature OLs (iOL, OPC exposure to T3 for 1 day) and maturing OLs (mOLs, OPC exposure to T3 for 3 days) 27 (Fig. 3h). The closest annotated gene to each Tcf7l2-binding site was identified as a presumed target. The number of Tcf7l2-targeted sites was B1,125 in iOLs and 14,541 in mOLs, respectively (Fig. 3i). The majority of Tcf7l2-binding peaks in iOLs overlapped with those identified in mOLs, but were of lower intensity (Fig. 3h). The increase in targeted sites and signal intensity from OPC to mOL correlated with the progression of OPC differentiation. In contrast, Tcf7l2-targeted sites in iOL and mOLs, respectively, overlapped by only 10 and 3% with those in other cell types such as H4IIE liver cells (Fig. 3i), suggesting that Tcf7l2 targets unique sets of genes in the OL lineage and possesses a distinct role in the control of OL differentiation.
The Tcf7l2-binding sites were co-localized with evolutionarily conserved enhancer elements marked by an activating histone mark H3K27ac 28 in iOL and mOL cells (Fig. 3j,k). To evaluate the global distribution of Tcf7l2-binding loci, we plotted the number of Tcf7l2 sites against their distance to the nearest transcription start site (TSS). We detected a strong enrichment for Tcf7l2 binding around TSSs within 5-kb promoter regions of genes marked by the histone mark H3k4me3 (ref. 29), in particular in mOLs (Fig. 3l). These observations indicate that Tcf7l2 targets primarily to enhancer/promoter regions to regulate target gene expression.
Wnt inhibitor Kaiso is a Tcf7l2 co-factor in iOLs. To investigate whether certain DNA motifs were enriched in Tcf7l2-binding sites, we applied a motif-discovery algorithm, HOMER 30 . The sequence motif A(C/G)(A/T)TCAAAG identified in iOLs matches the consensus-binding motif for Tcf7l2 in previous ChIP-seq data sets ( Fig. 4a) 31,32 . Enhancer regions typically have binding sites for co-factors, which bind within B100 bp of the Tcf7l2 peak summit. We found that a substantial proportion (B28%) of Tcf7l2-binding sites were significantly overrepresented with the binding-motif of Zbtb33/Kaiso (Fig. 4a). Kaiso is a transcriptional repressor of Wnt signalling, it interacts with TCF factors to inhibit b-catenin-dependent activation of transcription [33][34][35] . Kaiso expression increased as OPC differentiated into iOL, but downregulated in mOL by qRT-PCR and western blot analyses (Fig. 4b,c). Consistently, by immunostaining, in contrast to weak expression in PDGFRa þ OPCs, Kaiso was upregulated and co-localized with Tcf7l2 in iOLs (Fig. 4d), while downregulated in MBP þ OLs, suggesting a potential role of Kaiso at the onset of OPC differentiation.
Gene ontology analysis indicated that Tcf7l2-targeted genes in iOL were significantly enriched in components of the Wnt signalling pathway (Fig. 4e). Tcf7l2 targeted to a number of the prototypical Wnt-responsive genes including Axin2, Sp5, Lef1, Ctnnb1 and Ccnd1 on the transition of OPCs to iOLs (Fig. 4f) and this was maintained in mOLs. We further confirmed that these loci were also co-occupied by Kaiso using ChIP-qPCR in iOLs (Fig. 4g). To examine the effects of Kaiso on Wnt signalling activity, we transfected Kaiso-expressing vectors together with constitutively active b-catenin (DN89 b-catenin) 36 and a luciferase reporter for b-catenin/TCF activation, Topflash 37 , into HEK293T cells. Expression of DN89 b-catenin activated the Topflash reporter significantly; however, Kaiso expression suppressed the Topflash activity induced by the activated bcatenin (Fig. 4h). This suggests that Kaiso expression attenuates Wnt signalling activation, which negatively regulates OL differentiation 8,9 .
Co-immunoprecipitation revealed that endogenous Tcf7l2 and Kaiso were co-associated in the same complex in iOLs (Fig. 4i). To further investigate the functional interactions between Tcf7l2 and b-catenin or Kaiso, we performed co-immunoprecipitation in 293T cells transfected with expression constructs carrying Tcf7l2, b-catenin or Kaiso alone or in combination. Consistent with previous studies 17 , Tcf7l2 and b-catenin were detected in the same complex; however, in the presence of Kaiso, the interaction between Tcf7l2 and b-catenin was abolished (Fig. 4j). Similarly, overexpression of Kaiso substantially attenuated the interaction between endogenous Tcf7l2 and b-catenin in Oli-neu cells, an oligodendroglial cell line 38 (Supplementary Fig. 4a). These observations suggest that Kaiso competes with b-catenin for Tcf7l2 binding and thereby inhibits Wnt/b-catenin signalling. Furthermore, overexpression of Kaiso in rat OPCs enhanced the expression of myelin-associated genes such as Cnp, Mbp and Plp1 (Fig. 4k).
To determine the effect of Kaiso on the genome occupancy of Tcf7l2 on different target genes, we performed Tcf7l2 ChIP-qPCR in primary rat OPCs transfected with a Kaiso-expressing vector. We found that Tcf7l2 occupancy on the promoter of Wnt target genes (that is, Sp5, Ctnnb1, Ccnd1, Wnt11 and Wnt10a) was reduced in Kaiso-overexpressing cells (Fig. 4l). In contrast, Tcf7l2_ChlP-seq   ARTICLE Tcf7l2 occupancy on the promoter of a myelination-promoting factor Myrf was enhanced (Fig. 4l), suggesting that Kaiso expression levels modulate the genome occupancy of Tcf7l2 in OPCs. To further examine the effect of Kaiso knockdown on the OL differentiation programme, small interfering RNA (siRNA) targeting Kaiso was transfected into rat OPCs and cultured under differentiation conditions for 72 h. Kaiso knockdown resulted in the downregulation of myelination-associated genes that included Mbp, Plp1, Cnp, Mag and Myrf, while upregulating the Wnt signalling effectors Sp5, Wnt10a, Wnt4 and Wnt11 (Fig. 4m), all as compared with scrambled siRNA-transfected OPCs. Furthermore, we found that Kaiso overexpression enhanced the expression levels of myelin-associated genes, while Tcf7l2 knockdown blocked the ability of Kaiso to promote their expression (Fig. 4n,o), suggesting that the effect of Kaiso on enhancing OL differentiation gene expression depends on Tcf7l2.
In addition, to determine the in vivo role of Kaiso in OL development, we analysed the phenotype of Kaiso-null mice 39 and found that the number of CC1 þ differentiating OLs and MBP signal intensity were substantially reduced in the corpus callosum of Kaiso mutants at P7 (Fig. 4p,q) compared with controls, despite normal PDGFRa þ OPC formation (Fig. 4q).
Collectively, these data suggest that Kaiso promotes OPC differentiation programmes, while repressing Wnt signalling activity and is required for normal OL differentiation.
Sox10 is a Tcf7l2 co-regulator for OL maturation. As in iOLs, we found that Tcf7l2-binding peaks in mOLs match its consensus DNA-binding motif (Fig. 5a); however, B38% of Tcf7l2-binding sites present in mOLs were predominantly enriched with the Sox-consensus binding motif C(T/A)TTG(T/A)(T/A), which matches most significantly with an OL differentiation-promoting factor Sox10 motif 40,41 in the HOMER-motif discovery programme (Fig. 5a), but to a lesser extent with the Kaiso-binding motif. To determine whether Sox10 and Tcf7l2 co-target to the same regulatory elements in mOL, we then performed genome-wide occupancy of Sox10 in mOL using ChIP-seq. We found that B44% of Tcf7l2 peak sets overlapped with those of Sox10 occupancy ( Fig. 5b and Supplementary  Fig. 5). Sox10 motifs were enriched in Tcf7l2-binding sites in mOLs (Fig. 5b,c), suggesting that Tcf7l2 and Sox10 co-target the same elements in mOLs. To further determine whether Tcf7l2 may co-associate with Sox10, we performed coimmunoprecipitation assays and showed that Tcf7l2 and Sox10 were present in the same complex in mOLs (Fig. 5d), suggesting that Sox10 is a co-factor of Tcf7l2 during OL maturation. A large proportion of Tcf7l2/Sox10 co-occupancy was within the regulatory elements of myelination-associated genes including Mbp, Myrf, Olig1, Utg8 and Zfp191 in mOLs, but not in OPCs or iOLs (Fig. 5e). In addition, although Sox10 or Tcf7l2 expression in rat OPCs elevated the expression level of myelin-associated genes such as Mbp and Cnp, co-expression of Sox10 and Tcf7l2 further increased their levels (Fig. 5f), suggesting that Sox10 and Tcf7l2 cooperate to promote myelin gene expression.
How does Tcf7l2 coordinate distinct co-factors to control OL lineage progression? Kaiso expression increases in iOL, but downregulates in mOL, while the Sox10 expression level is maintained throughout the OL lineage. To determine whether the temporal sequence of Tcf7l2 recruitment of Kaiso and Sox10 depends on their expression levels, we co-transfected Tcf7l2 with a varied amount of Kaiso, while keeping the Sox10 level constant in 293T cells. We found that Tcf7l2 was associated preferentially with Kaiso when Kaiso expression levels were high (Fig. 5g). As the Kaiso level fell, Tcf7l2 was found to interact with Sox10 (Fig. 5g). Furthermore, we found that overexpression of Sox10 substantially reduced association of Tcf7l2 with Kaiso in Oli-neu cells ( Supplementary Fig. 4b), suggesting dosage-dependent competitive binding of Tcf7l2 with Kaiso and Sox10. Thus, there appears to be a two-step recruitment process in which Tcf7l2 association with 'early-' and 'late-binding' transcription factors, Kaiso and Sox10, respectively, is responsible for OL lineage progression. Together, these observations suggest that Tcf7l2 coordinates the stepwise OL differentiation process through interacting with Kaiso to suppress inhibitory Wnt signalling in OPCs, while associating with Sox10 to promote myelination-associated programmes during OL maturation (Fig. 5h).
Tcf7l2 activates cholesterol biosynthesis for OL maturation. Superimposing Tcf7l2 ChIP-seq from mOLs and RNA-seq data revealed that B174 targeted genes with substantial changes of expression in Tcf7l2DHMG mutants also exhibited strong Tcf7l2 binding to their proximal promoter regions (Fig. 6a). These genes were overrepresented in the functional categories of steroid biosynthesis and cholesterol metabolism (Fig. 6b). Among these Tcf7l2 targets in mOLs, we identified a cohort of genes that encode enzymes involved in de novo cholesterol biosynthesis, including hmgcs1, Fdps, Fdft1, Lss, Cyp51, Hsd17b7 and Dhcr24 (Fig. 6c,d). Although Tcf7l2 did not appear to directly target the gene locus of Hmgcr ( Supplementary Fig. 6a), it was highly enriched on the promoter region of Srebf2 ( Supplementary Fig. 6b), a key upstream regulator of Hmgcr expression 42,43 .
When comparing Tcf7l2 occupancy on the promoters of cholesterol pathway genes with that of Olig2, an OL specification factor, we found that Tcf7l2 targeting was substantially enriched in parallel with OL maturation, and that these promoters were marked with the activating histone marks H3K4me3 and H3K27Ac 28,29 (Fig. 6d). In contrast, Olig2 binding on these promoter regions was barely detectable in mOLs (Fig. 6d), suggesting a distinct role between Olig2 and Tcf7l2 in the differentiation of mOLs.
We next asked whether Tcf7l2 acted with partners to support oligodendrocytic cholesterol synthesis and found that in mOLs, Tcf7l2 co-occupied the promoters of cholesterol pathway genes with Sox10 (Fig. 6d); ChIP-qPCR further confirmed the enrichment of Tcf7l2 and Sox10 binding on the promoter elements of these cholesterol synthetic genes (Fig. 6e). The co-occupancy of the gene loci by Tcf7l2 and Sox10 is consistent with observations that Tcf7l2 and Sox10 form a complex in mOL.
The presence of H3K4me3 and H3K27Ac in the targeted promoters, both of which are indicators of active transcription state 28,29 , suggests that Tcf7l2 positively regulates expression of cholesterol biosynthesis genes (Fig. 6d). Accordingly, the mRNA expression levels of these Tcf7l2-targeted cholesterol biosynthesis genes was significantly reduced in Tcf7l2DHMG mutant optic nerves (Fig. 6f).
To compare the activity of Tcf7l2 with a key regulator for the cholesterol biosynthesis Srebf2/Srebp2 on target gene expression, we cloned the promoter regions carrying Tcf7l2-binding sites of cholesterol biosynthesis genes including Srebf2, Hmgcs1, Hmgcr, Fdps, Lss, Cyp51, Hsd17b7 and Dhcr24 into a luciferase reporter system. We found that similar to Srebf2, Tcf7l2 stimulated the luciferase reporter activity driven by these regulatory elements/enhancers (Fig. 6g). In contrast, these effects could not be observed with Tcf7l2DHMG (Fig. 6g).
To further explore the relative contribution of Tcf7l2 to expression of cholesterol biosynthesis genes in OLs, we next transfected purified OPCs with control and expression vectors carrying Tcf7l2 and found that overexpression of Tcf7l2 induced the expression of a number of genes encoding essential enzymes for cholesterol synthesis, including Lss, Cyp51, Hsd17b7 and Dhcr24 (Fig. 6h), as well as Axin2, a generic Wnt/Tcf7l2 target gene, and the myelination-promoting gene Myrf, suggesting that activation of Tcf7l2 promotes expression of cholesterol biosynthesis genes for OL differentiation.

Tcf7l2-induced cholesterol synthesis for OL differentiation.
Cholesterol biosynthesis is required for myelin sheath outgrowth as shown in Hmgcs1 mutant zebrafish 44 and Fdft1 mutant mice 45 .
To determine the role of other Tcf7l2-regulated cholesterol biosynthesis genes in OL differentiation, we carried out siRNA knockdown during OPC differentiation. Knockdown of Cyp51,  Dhcr24, Hsd17b7 and Lss (Fig. 7a) reduced expression of myelin genes including Mbp, Plp1, Cnp and Mag, although Id2 and Id4 expression was not significantly altered (Fig. 7b).
As Tcf7l2 activated the expression of cholesterol biosynthesis genes, we hypothesized that dysregulation of cholesterol synthesis might account for the impairment of OL maturation in Tcf7l2DHMG mice. Previous studies reported that cholesterol, a rate-limiting lipid component for myelin sheath growth, is not imported into the brain from the circulation but rather synthesized locally by myelin-forming OLs 46,47 . On that basis, we next asked whether exogenous cholesterol would restore the differentiation capacity of OPCs isolated from Tcf7l2 mutants. Under differentiation conditions, control OPCs readily differentiated into CNP þ and MBP þ mature OLs, whereas the majority of Tcf7l2-mutant OPCs failed to mature into OLs with elaborate membrane processes (Fig. 7c). In addition, the proportion of CNP þ or MBP þ OLs was significantly reduced in Tcf7l2 mutants compared with controls (Fig. 7d,e). Strikingly, exogenous cholesterol partially restored the percentage of CNP þ and MBP þ OLs in mutant cultures (Fig. 7c-e). Consistently, expression of myelin-associated genes increased in Tcf7l2-mutant OPCs treated with cholesterol (Fig. 7f). To further determine whether cholesterol supplementation could rescue the OL differentiation defect in Tcf7l2-mutant animals, we fed the pregnant mice carrying control and Tcf7l2DHMG mutants with cholesterol-enriched diet during gestation, to allow cholesterol intake into the CNS. The pups were harvested at P14 for analysis. Although cholesterol supplementation did not alter the number of OLs in the spinal cords of control animals, it significantly increased the number of CC1 þ OLs and expression of MBP in Tcf7l2 mutants (Fig. 7g,h). These observations suggest that cholesterol is responsible, at least in part, for Tcf7l2-dependent control of OL differentiation, and that Tcf7l2 directly activates the genes encoding cholesterol biosynthesis enzymes to promote OL differentiation and maturation.

Discussion
Wnt/b-catenin/Tcf7l2 signalling has been suggested to positively or negatively regulate OL development, which may depend on the context 8,9,[48][49][50][51] . Identification of Tcf7l2 direct targets and its co-regulators is of vital importance to understand the Tcf7l2mediated process for OL development. Using integrative unbiased genomics analyses of multiple stages in the OL lineage, we present the first genome occupancy mapping of stage-specific targets of Tcf7l2 during OL lineage progression and further identify Tcf7l2-engaging partners that control transcriptional switches at the transitions of differentiation initiation and OL maturation. Strikingly, we uncover a non-canonical Wnt signalling corepressor Kaiso as a Tcf7l2 partner at the OL differentiation onset to inhibit b-catenin activity. During OL maturation, Tcf7l2 further recruits another partner Sox10, a differentiationpromoting factor, to activate the myelinogenic transcriptional programme. Thus, Tcf712 executes a dual regulatory control by recruiting stage-specific partners to switch its functions to propel stepwise OL differentiation. Importantly, our studies further provide evidence that Tcf7l2 directly targets and activates cholesterol metabolism-associated genes and thereby regulates de novo cholesterol biosynthesis necessary for myelinogenesis, pointing to a previously unappreciated role of Tcf7l2 for cholesterol homeostatic control of CNS myelination. Constitutive activation of Wnt/b-catenin signalling inhibits OL differentiation, suggesting that Wnt signalling may need to be attenuated for developmental OL differentiation. We found that the association of Kaiso with Tcf7l2 reverses the role of the b-catenin/Tcf7l2-mediated transcriptional complex from a differentiation repressor to an activator, which then promotes the transition of OPCs from an undifferentiated to differentiating state. At the onset of OPC differentiation, Kaiso is upregulated and appears to act as a competitive inhibitor of b-catenin, to oppose Wnt signalling activation by disrupting the b-catenin/ Tcf7l2 complex. Kaiso has been shown to interact with the NCoR/ Hdac1 complex to repress target gene expression 35,52 . It is possible that Tcf7l2 forms a potent repressive complex with Kaiso and Hdac1/2, and perhaps Hdac-associated Groucho/TLE repressors as well 9,53 , so as to inhibit Wnt/b-catenin signalling.
Our data indicate that Tcf7l2 targets a number of prototypical Wnt-responsive genes, including Axin2, Ccnd1, Ctnnb1, Lef1 and Sp5 on OPC differentiation into iOLs and persists with maturation (Fig. 4f). This suggests that Tcf7l2 may continue to modulate b-catenin activation and downstream target expression during OL differentiation. Persistent control of Wnt target gene expression and signalling by Tcf7l2 and its co-factors probably generates a high degree of regulatory complexity in response to different Wnt signalling inputs over the course of OPC differentiation.
We found that the transcriptional activity of Tcf7l2 is critical for OL differentiation and myelin repair. Although we could not rule out a potential dominant-negative effect of the Tcf7l2 mutant, the Tcf7l2DHMG mutant mice yielded a dysmyelinating phenotype similar to that of full-length Tcf7l2-deletion mutants 18 . Of note, we did not observe any discernable phenotype in Tcf7l2DHMG heterozygous mice. It thus seems unlikely that any dominant-negative effect would significantly contribute to the dysmyelinating phenotype. Although Tcf7l2DHMG was able to interact with Kaiso and Sox10 when overexpressed in 293T cells ( Supplementary Fig. 7a,b), it could not activate myelin gene expression as wild-type Tcf7l2 (Supplementary Fig. 7c), but rather blocks Kaiso and Sox10 activity for myelin gene activation ( Supplementary Fig. 7d,e). This suggests that Tcf7l2 activity is required for OL differentiation-promoting machinery. Tcf7l2DHMG mutant protein appears at a lower level in the Tcf7l2-mutant spinal cord compared with its wild-type counterpart (Fig. 1b). This is probably due to the severe reduction of differentiating OLs, where Tcf7l2 is highly expressed, in the mutants.
During OL maturation, we detect stronger signal intensity and an increase in Tcf7l2-targeted binding sites, which largely overlap with those identified in rat spinal cord 54 (Supplementary Fig. 8). We find that Sox10 acts a co-regulator with Tcf7l2 in mOLs, and that Tcf7l2 and Sox10 cooperate to promote myelinogenic programmes, consistent with the notion that Tcf7l2 has a b-catenin independent function during OL differentiation 18 . These findings suggest that Tcf7l2 functions through the sequential operation of two interlaced gene regulatory networks: one blocking inhibitory Wnt signalling activity at iOL stages through recruiting transcriptional repressors such as Kaiso, and the other promoting a 'terminal' differentiation process in committed OLs in cooperation with Sox10. The switch of Tcf7l2 co-regulators from Kaiso to Sox10 while transitioning from iOL to mOLs may lead to the further induction of myelin-specific genes, thus facilitating OL maturation (Fig. 5h). Cooperation with stage-specific, transiently expressed transcriptional partners may also explain the distinct stagespecific role of Tcf7l2 in the initiation and maintenance of OL differentiation.
Besides the temporally dynamic nature of Tcf7l2 interactions with other transcriptional modulators, we have also observed a distinct role for Tcf7l2 in regulating the extent of myelination. In contrast to persistent myelination defects in the brain of Tcf7l2 mutants, Tcf7l2 appears dispensable for myelination in the adult spinal cord, where Tcf7l2 is hardly detectable. At present, the underlying mechanisms for this regional specificity remain undetermined. The efficient Tcf7l2 allele recombination in the spinal cord (Fig. 1b) indicates that a region-specific function of Tcf7l2 rather than incomplete recombination accounts for the phenotypic difference.
A novel finding of our present study is that Tcf7l2 directly regulates de novo cholesterol biosynthesis and metabolism. We find that Tcf7l2 target genes are selectively enriched on the promoter regions of genes involved in cholesterol biosynthesis in mOLs. In contrast, in liver cells, Tcf7l2 targets are enriched on the promoter/enhancers of classical lipid synthesis-related genes such as Fabp1, Apod, Scd1, Mttp, Hnf4a and Fas, which are associated primarily with lipid metabolic pathways, such as the synthesis of ketone bodies, transport of fatty acids and gluconeogenesis 32,55 ( Supplementary Fig. 9). This suggests a conserved yet distinct role for Tcf7l2 in the control of lipid biosynthesis and metabolism, by regulating different sets of genes in a cell-type-specific manner.
Cholesterol is a rate-limiting lipid component for OL myelination and myelin sheath growth 46,47 . Defects in myelination has been observed in mutants in the genes responsible for cholesterol biosynthesis such as in Fdft1 mutant mice 45 and in Hmgcs1 mutant zebrafish 44 . We find that cholesterol supplementation could, at least in part, rescue OL differentiation defects in Tcf7l2DHMG mutants. Interestingly, a similar observation was recently reported that cholesterol injection could rescue the axon wrapping defect in hmgcs1 mutant zebrafish, suggesting that cholesterol synthesis is necessary for OL maturation and axon wrapping 44 . A cell sensing insufficient cholesterol could block OL differentiation and expression of myelination-associated genes. It is possible that a negative feedback or 'check point' mechanism exists, whereby sufficient cholesterol biosynthesis for myelin biosynthesis must be verified before transcription of myelination-promoting factors. Given that Tcf7l2 mutations lead to downregulation in the expression of cholesterol biosynthesis enzymes, the requirement of Tcf7l2 for OL myelination can be mediated, at least partially, through its role as a regulator of cholesterol biosynthesis during OL maturation. Collectively, Tcf7l2 may have a parallel role in regulating the expression of cholesterol biosynthesis genes in addition to directly activating the myelinogenic transcriptional programme.
Previous studies show that TCF7L2 expression rose during active remyelination in both human patients and rodent models 8,56 , and yet is completely absent in the lesions of chronic multiple sclerosis 57 . This observation suggested the potential requirement of TCF7L2 for active myelin repair. Here we provide evidence that Tcf7l2 transcriptional activity is crucial for remyelination after demyelination, as well as to normal myelin formation during brain development. Indeed, as the impact of the Tcf7l2 knockout on myelination is more severe in the brain than in the spinal cord, it would seem likely to be that the remyelination defects observed in the Tcf7l2DHMG spinal cord should manifest in the brain as well.
Recent studies indicate that TCF7L2 is a key regulator of the metabolic gene programme that controls transcriptional responses to metabolic challenge in the liver, muscle or white adipose tissues 32 . For instance, TCF7L2 single-nucleotide polymorphism variants and mutations contribute to type 2 diabetes mellitus [58][59][60] . Yet, type 2 diabetes impairs OL regeneration in the white matter lesions after ischaemic injury 61 . Our findings suggest then that TCF7L2 mutations or single-nucleotide polymorphism variants might modify both the degree of ischaemic myelin damage in patients with type 2 diabetes and their ability to remyelinate. Together, integrative analyses of Tcf7l2 occupancy, gene expression profiling and binding motifs at multiple OL stages reveal that Tcf7l2 appears to exercise its functions by cooperating with stage-specific co-regulators to control OL differentiation programmes. More broadly, the present genome-wide Tcf7l2 target identification should provide us a framework to identify novel therapeutic avenues for myelin repair in the CNS.

Methods
Generation of Tcf7l2 conditional knockout mice. Tcf7l2 floxed mice 20 were crossed with Olig1-Cre 22 , Olig2-Cre 62 (Olig2otm2(TVA,cre)Rth4/J from Jackson Laboratory) and CNP-Cre 25 (gift of Dr Klaus-Armin Nave) mice to generate Tcf7l2DHMG (Tcf7l2 fl/fl :CNP-Cre þ / À ) and heterozygous control (Tcf7l2 fl/ þ : CNP-Cre þ / À ) mice. As LoxP sites flank exon 11 encoding the highly conserved DNA-binding domain of Tcf7l2, Cre-mediated recombination results in Tcf7l2 mutant protein lacking DNA-binding transcriptional activity, while the mutant mRNA and protein may still be expressed. The control mice developed and behaved the same as wild type. PDGFRa-GFP (Jackson Laboratory Stock Number 024098) and CNP-mGFP reporter mice 23 , and Zbtb33/Kaiso mutant mice 39 were used for phenotype analysis. We used both male and female mice for the study. The mouse strains used in this study were generated and maintained on a mixed C57Bl/6;129Sv;CD-1 background. All animal use and studies were approved by ethical committees of our institutions and by the Institutional Animal Care and Use Committee of Cincinnati Children's Hospital Medical Center, USA.
RNA in situ hybridization was performed using digoxigenin-labelled riboprobes as described previously 63 . The probes used were as follows: murine PDGFRa, Plp1/Dm-20 and Mbp.
Oligodendroglial cell culture and transfection. Isolation of primary rat OPCs from cortices of P2 pups was performed as previously described 64 . OPCs were differentiated in OL differentiation medium (Sato medium supplemented with 15 nM T3 and 10 ng ml À 1 ciliary neurotrophic factor) for 24 and 72 h to become iOL and mOL as the initiation and maturing phases of OLs, respectively, as previously described 27 . Mouse OPCs were isolated from P5 to P7 cortices of control and Tcf7l2DHMG mutants by immunopanning with antibodies Ran-2, GalC and PDGFRa sequentially as previously described 65 . The mouse OL cell line Oli-neu cells were maintained in proliferation medium consisting of DME/F12 medium supplemented with Sato, NT3, CNTF, B27, 0.5 mM T3, 0.5 mM T4 and 1% horse serum. The cells were induced to differentiate with 1 mM cyclic AMP supplementation 38 .
Co-immunoprecipitation and luciferase assays. Co-immunoprecipitation and western blotting were performed as described previously 27 . Briefly, for coimmunoprecipitation in iOLs and mOLs, 600 mg of cell lysate proteins were incubated with 3 mg anti-Tcf7l2, anti-Kaiso or anti-Sox10 for immunoprecipitation assay. For Tcf7l2/b-catenin complex competition assay, in Oli-neu cells, cells were transfected with 3 mg each pCS2-Myc-Kaiso or pcDNA3-HA-Sox10; in HEK293T cells, cells were transfected with pcDNA3 Flag-tag Tcf7l2, b-catenin, Myc-tag Kaiso or pcDNA3-HA-Sox10. For luciferase assays, HEK293T cells were transiently transfected with pCS2-Myc-tag Kaiso and/or b-catenin 9 , together with a Topflash reporter, or transfected with pcDNA3-Flag-Tcf7l2 or Srebf2 with individual luciferase reporters driven by the enhancers of cholesterol biosynthesis genes. After western blotting, proteins were detected with appreciate secondary antibodies by using chemiluminescence with the ECL kit (Pierce) according to the instructions of the manufacturer. The primers used for cloning of cholesterol biosynthesis gene promoters are listed in Supplementary  Supplementary Figs 11 and 12, respectively.
Quantitative real-time PCR analysis. RNAs were isolated with Trizol (Invitrogen Inc.) from cells or snap-frozen tissues. Reverse transcription was performed with the cDNA Reverse Transcription Kit (Bio-Rad) with iQ SYBR Green Supermix (170-8880). qRT-PCR was carried out using the ABI Prism 7900 Sequence Detector System (Perkin-Elmer Applied Biosystems) using Gapdh as an internal control. Each analysis was performed in triplicates and the results were normalized to Gapdh for each sample. The qRT-PCR primer sequences are listed in Supplementary Table 1. RNA-seq and data analysis. We isolated RNAs from the optic nerves of control mice and Tcf7l2DHMG mutants at P12 and subjected samples to RNA deep sequencing and data analysis as previously described 27 . RNA-seq libraries were prepared using Illumina RNA-seq Preparation Kit (Illumina) and sequenced on a HiSeq 2000 sequencer. RNA-seq reads were mapped using TopHat with default settings (http://tophat.cbcb.umd.edu). TopHat output data were then analysed by Cufflinks to (1)  changes of gene expression between Tcf7l2DHMG and control. Heatmap of gene differential expression was generated using R language (http://www.r-project.org).
ChIP-sequencing and ChIP-qPCR. ChIP assays were performed as previously described, with minor modifications 27 . Briefly, OPCs, iOLs and mOLs (B20 million cells) were fixed for 10 min at room temperature with 1% formaldehyde-containing medium. Nuclei were isolated and sonicated in sonication buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 0.5 mM EGTA and protease inhibitor cocktail). Sonicated chromatin (B300 mg) was used for immunoprecipitation by incubation with appropriate antibodies (4 mg) overnight at 4°C. Ten per cent of chromatin used for each ChIP reaction was kept as input DNA. Prerinsed magnetic protein A/G beads (50 ml) were added to each ChIP reaction and reactions were incubated for 1 h at 4°C. The beads were then incubated in 200 ml elution buffer at 65°C for 20 min to elute immunoprecipitated materials. We performed duplicate ChIP-seq assays using chromatin from at least two different cell cultures. The ChIP-seq libraries were prepared using NEBNext ChIP-seq Library Prep Master Mix Set for Illumina (NEB catalogue number E6240L) and then run on the Illumina sequencer HS2000. The antibodies used were as follows: Kaiso (Abcam, ab12723), TCF7L2 (Santa Cruz, sc-8631) and Sox10 (Abcam, ab155279). The primers used for ChIP-PCR are listed in Supplementary  Table 1.
ChIP-seq peak-calling and data analysis. All sequencing data were mapped to rat genome assembly rn5 and peak calling was performed as previously described 27 using MACS (Model-based Analysis of ChIP-seq) version 1.4.2 (http:// liulab.dfci.harvard.edu/MACS) with default parameters, to get primary binding regions. To ensure that our data were of high quality and reproducibility, we called peaks with enrichment Z10-fold over control (Pr10 À 9 ) and compared the peak sets using the ENCODE overlap rules 66 . These identified primary regions were further filtered using the following criteria, to define a more stringent protein-DNA interactome: (1) the P-value cutoff was set to o10 À 9 ; (2) an enrichment of 5-fold and tag number 420, and subtracted with background regions.
The genome-wide distribution of protein binding regions was determined by CisGenome2.0 (http://www.biostat.jhsph.edu/Bhji/cisgenome) in reference to Ensembl RGSC3.4.61 release. This information was also used to group binding regions by the distance between peak summits and TSS. Venn diagrams were constructed using 3Venn Applet. De novo motif discovery was performed using Homer software and novel motifs were compared with JASPAR database (http://jaspar.genereg.net). For all ChIP-seq data sets, WIG files were generated with MACS, which were subsequently visualized using Mochiview v1.46. Tcf7l2 ChIP-seq heatmaps were ordered by strength of binding. The heatmaps were drawn using the Heatmap tools provided by Cistrome (http://cistrome.org/ap).
Gene ontology and enriched motif identification. For ChIP-seq data, binding peaks in rn5 were annotated with MACS 1.4.2. Functional classification of annotated binding genes from ChIP-seq and differentially expressed genes in RNA-seq data was performed using ToppGene (https://toppgene.cchmc.org/). For RNA-seq data, functional classifications were performed using DAVID (http://david.abcc.ncifcrf.gov). Enriched motifs of Tcf7l2-binding peaks in mOLs and iOLs were identified by HOMER (http://homer.salk.edu/homer/ngs/peakMotifs.html). The script 'findMotifsGenome.pl' was run for 'Homer Known Motif Enrichment Results' with default parameters.
Cholesterol supplementation in OPC culture and in vivo. Primary mouse OPCs from P5 to P6 cortices of control and Tcf7l2DHMG pups were prepared by immunopanning with antibodies Ran-2, GalC and PDGFRa sequentially as previously described 65 . Cholesterol (Sigma, C4951) were prepared in ddH2O and added into the culture medium at 10 mg ml À 1 . Cells in growth medium without Platelet-derived growth factor-AA were treated with cholesterol or vehicle control for 4 days and fixed in 4% PFA for 10 min, and stained as described.
For the cholesterol-enriched diet feeding experiment, pregnant female mice were separated into two groups and fed with 2% cholesterol diets (Harlen, TD.01383) or normal diet, respectively, from the time of gestation when plugs were detected after initial mating. The control and Tcf7l2DHMG pups were kept together with their mothers and fed through lactation until harvested at the indicated time point.
LPC-induced demyelinating injury in the spinal cord. LPC-induced demyelination was carried out in the ventrolateral spinal white matter of B8-week-old mice. Anaesthesia was induced and maintained by peritoneal injection of a mixture of ketamine (90 mg kg À 1 ) and xylazine (10 mg kg À 1 ). After exposing the spinal vertebrae at the level of T9-T12, meningeal tissue in the intervertebral space was cleared and the dura was pierced with a dental needle. One per cent LPC (L-a-lysophosphatidylcholine; 0.5 ml) via a Hamilton syringe attached to a glass micropipette was injected into the ventrolateral white matter via a stereotactic apparatus. Spinal cord tissues carrying the lesions were collected at time points as follows: 7 dpl, representing peak OPC recruitment, and 14 dpl, representing OL differentiation and new myelin sheath formation (at least 5 mice per control and mutant groups were used for each time point analysis).
Electron microscopy. Electron microscopy was performed essentially as previously described 22 . Anaesthetized mice were perfused briefly with 0.1 M cacodylate and followed by 2.5% glutaraldehyde/2.5 PFA in 0.1 M cacodylate (pH 7.2). The optic nerves and spinal cord were removed and fixed in fresh fixative overnight at 4°C. Tissues were rinsed in PBS, postfixed in 1% OsO 4 in PBS for 1 h, dehydrated in a graded ethanol series, infiltrated with propylene oxide and embedded in Epon. Semi-thin sections were stained with toluidine blue and thin sections were stained with lead citrate.