TCF7L2 regulates postmitotic differentiation programmes and excitability patterns in the thalamus

ABSTRACT Neuronal phenotypes are controlled by terminal selector transcription factors in invertebrates, but only a few examples of such regulators have been provided in vertebrates. We hypothesised that TCF7L2 regulates different stages of postmitotic differentiation in the thalamus, and functions as a thalamic terminal selector. To investigate this hypothesis, we used complete and conditional knockouts of Tcf7l2 in mice. The connectivity and clustering of neurons were disrupted in the thalamo-habenular region in Tcf7l2−/− embryos. The expression of subregional thalamic and habenular transcription factors was lost and region-specific cell migration and axon guidance genes were downregulated. In mice with a postnatal Tcf7l2 knockout, the induction of genes that confer thalamic terminal electrophysiological features was impaired. Many of these genes proved to be direct targets of TCF7L2. The role of TCF7L2 in terminal selection was functionally confirmed by impaired firing modes in thalamic neurons in the mutant mice. These data corroborate the existence of master regulators in the vertebrate brain that control stage-specific genetic programmes and regional subroutines, maintain regional transcriptional network during embryonic development, and induce terminal selection postnatally.


Introduction
Studies of invertebrates have shown that terminal differentiation gene batteries in individual classes of neurons are induced and maintained by specific transcription factors called terminal selectors, that are expressed throughout the life (Hobert and Kratsios, 2019).
However, regulatory strategies of postmitotic maturation and terminal selection in vertebrates are unclear. Till date, only a few terminal selectors have been identified in vertebrates (Cho et al., 2014;Flames and Hobert, 2009;Kadkhodaei et al., 2009;Liu et al., 2010;Lodato et al., 2014;Wyler et al., 2016). Neurons of the thalamus and habenula are derived from a single progenitor domain (prosomere 2) and are glutamatergic (Watson et al., 2012), except for GABAergic interneurons in the rostral thalamus (Evangelio et al., 2018). The thalamus is a sensory relay centre and part of cortico-subcortical loops that process sensorimotor information and produce goal-directed behaviours (Sherman, 2017). The habenula controls reward-and aversion-driven behaviours by connecting cortical and subcortical regions with the monoamine system in the brainstem (Benekareddy et al., 2018;Hikosaka, 2010). During postmitotic differentiation, thalamic and habenular neurons segregate into discrete nuclei (Shi et al., 2017;Wong et al., 2018), develop a variety of subregional identities (Guo and Li, 2019;Nakagawa, 2019;Phillips et al., 2019), extend axons toward their targets (Hikosaka et al., 2008;López-Bendito, 2018), and acquire electrophysiological characteristics postnatally (Yuge et al., 2011). The knowledge of the mechanisms that control postmitotic development in this region is important, because its functional dysconnectivity, which possibly originates from the period of postmitotic maturation, is implicated in schizophrenia, autism and other mental disorders (Browne et al., 2018;Steullet, 2019;Whiting et al., 2018;Woodward et al., 2017).
The network of postmitotically induced transcription factors that regulate the maturation of prosomere 2 neurons has only begun to be deciphered. Gbx2 and Pou4f1 are early postmitotic markers of thalamic and habenular neurons, respectively. GBX2 plays a transient regulatory role in the initial acquisition of thalamic molecular identities and thalamocortical axon guidance, and then its expression is downregulated in the majority of the thalamus (Chatterjee et al., 2012;Chen et al., 2009;Li et al., 2012;Mallika et al., 2015;Miyashita-Lin et al., 1999). In contrast, POU4F1 (alias BRN3A) is not essential for the growth of habenular axons, but it maintains the expression of the glutamate transporter gene Vglut1/Slc17a7 and other habenula-abundant genes in adults, though its impact on electrophysiological responses of habenular neurons was not tested (Quina et al., 2009;Serrano-Saiz et al., 2018). Subregional transcription factors RORA and FOXP2 regulate some aspects of postmitotic differentiation in thalamic subregions during embryogenesis Development • Accepted manuscript (Ebisu et al., 2016;Quina et al., 2009;Vitalis et al., 2017), but their role in terminal differentiation in this region was not investigated.
Tcf7l2, a risk gene for schizophrenia and autism (Bem et al., 2019) that encodes a member of the LEF1/TCF transcription factor family (Cadigan and Waterman, 2012), is the only shared marker of prosomere 2 neurons (Nagalski et al., 2013;Nagalski et al., 2016). Its function was tested only during early postmitotic period in zebrafish (Beretta et al., 2013;Husken et al., 2014) and mice (Lee et al., 2017;Tran et al., 2020). In Tcf7l2 -/mouse embryos, some markers were misexpressed in the region of prosomere 2, and the formation of axonal tracts was disrupted. TCF7L2 expression is maintained throughout life and the TCF7L2 motif is overrepresented in putative enhancers of adult thalamus-enriched genes (Nagalski et al., 2016;Wisniewska et al., 2012), suggesting that this factor can play a role of prosomere 2 terminal selector.
The present study used complete and conditional knockout mice to explore the role of Tcf7l2 in postmitotic anatomical maturation, maintenance of molecular diversification, adoption of neurotransmitter identity, and postnatal acquisition of electrophysiological features in the thalamus and habenula. We show that TCF7L2 orchestrates the overall morphological differentiation process in this region by regulating stage-specific gene expression directly or via subregional transcription factors. We also report that TCF7L2 functions as a terminal selector of postnatally-induced thalamic electrophysiological characteristics but not glutamatergic (VGLUT2) identity.

Generation of mice with the complete and conditional knockouts of Tcf7l2
Tcf7l2 expression in prosomere 2 is induced in mice during neurogenesis (Cho and Dressler, 1998) and maintained in the thalamus and medial habenula throughout life (Nagalski et al., 2013). In wild type (WT) embryos on E12.5, we observed high levels of TCF7L2 protein in the superficial portion of the thalamus and habenula, which is populated by postmitotic neurons (Fig. 1A). We also observed several TCF7L2-positive cells in the prethalamus. Possibly, these cells migrate from the GABAergic rostral thalamus to take part in the formation of the intergeniculate leaflet and ventral lateral geniculate nuclei that derive from prethalamic and rostral thalamic progenitors (Jeong et al., 2011). At late gestation, TCF7L2 was observed in the entire caudal thalamus (hereinafter referred to as the thalamus; a glutamatergic domain) and medial habenula (Fig. 1B). Relatively lower levels of TCF7L2 were present in the ventrobasal complex (VB), nucleus reuniens, and recently identified perihabenula (Fernandez et al., 2018). Low levels of TCF7L2 were also present in the lateral habenula and the derivatives of the rostral thalamus.
To investigate the role of TCF7L2 as a selector of morphological and electrophysiological characteristics of prosomere 2, we used two knockout models in mice.
The complete knockout of Tcf7l2 was generated by insertion of the tm1a(KOMP)Wtsi allele with lacZ cassette upstream of the critical 6 th exon of the Tcf7l2 gene ( Fig. 2A). Tcf7l2 tm1a allele led to the lack of TCF7L2 protein, confirmed by immunostaining and Western blot, and ectopic expression of β-galactosidase from the lacZ locus ( Fig. 2B-C). Tcf7l2 -/mice die after birth. To create a thalamus-specific postnatal knockout of Tcf7l2, we first crossed Tcf7l2 +/tm1a mice with a flippase-expressed strain and then with mice that expressed CRE recombinase from the Cholecystokinin (Cck) gene promoter (Fig. S1). Cck is upregulated postnatally and its expression overlaps with the expression of Tcf7l2 in the thalamus (Allen Brain Atlas, 2011;Nishimura et al., 2015). The expression from the Cck Cre locus, visualised in Cck Cre :tdTomato fl/+ reporter line, was high in lateral parts of the thalamus, and lower in thalamic medial and midline parts, including anterodorsal (AD), paraventricular (PV) and parafasicular (PF) nuclei (Fig. S2). Cck-driven knockout of Tcf7l2 was induced postnatally and completed in the thalamus by P14 (Fig. 2G-H). In the resulting Cck Cre :Tcf7l2 fl/fl mice, TCF7L2 was absent in most thalamic nuclei in adults, except for the PV and PF (Fig. S3). In the AD and midline nuclei, Tcf7l2 was partially knocked out.

Normal neurogenesis but disrupted anatomy and connectivity of prosomere 2
To confirm that proliferation and neurogenesis occurred normally within prosomere 2 in Tcf7l2 -/mice, we stained E12.5 brain sections with antibodies specific for the KI-67 antigen and TUJ1, markers of proliferating progenitors and young postmitotic neurons, respectively.
Prosomere 2 was identified with Tcf7l2 probe both in WT and knockout (KO) embryos, taking advantage of preserved expression from the targeted Tcf7l2 locus (Fig. 3A). The complete knockout of Tcf7l2 did not cause any apparent defects in proliferation or neurogenesis in prosomere 2 on E12.5 (Fig. 3B), consistent with previous results in another Tcf7l2 knockout strain (Lee et al., 2017).
To investigate whether TCF7L2 is required for the initial acquisition of postmitotic molecular identities in the thalamus and habenula, we examined the expression of the Gbx2 gene and POU4F1 protein (the earliest markers of postmitotic neurons in the thalamus and habenula, respectively) during neurogenesis. Both Gbx2 mRNA and POU4F1 were highly expressed in the prosomere 2 area in WT and Tcf7l2 -/embryos on E12.5, indicating that their expression is not induced by TCF7L2 (Fig. 3C). The number of POU4F1-positive cells visibly increased, suggesting that prosomere 2 cells more readily adopted habenular fate in Tcf7l2 -/embryos at this stage. Gbx2and POU4F1-positive areas expanded extensively into each other's territory, implying a defect in thalamo-habenular boundary formation.
To further investigate if TCF7L2 regulates structural maturation of prosomere 2, we focused on late gestation (E18.5), when nucleogenesis is already concluded and axonal connections are well developed in this region. The anatomy of the area was analysed by Nissl staining . The boundaries between prosomere 2 and neighbouring structures, i.e., the prethalamus (rostral boundary) and pretectum (caudal boundary), were not morphologically detected in sagittal sections from Tcf7l2 -/embryos. On coronal sections, the habenula was fused with the thalamus, and nuclear groups within prosomere 2 were not well demarcated. The whole region was reduced in the radial dimension and elongated dorsoventrally, resulting in an oval-like shape. Then, we analysed circuit formation in the thalamo-habenular region by tracing thalamocortical axons with DiI and immunostaining brain sections with an antibody specific for L1 cell adhesion molecule (L1CAM) that marks growing axons. Consistent with a previous report (Lee et al., 2017), thalamocortical axons were not detected in KO brains (Fig. 3F). The bundles of stria medullaris, which include afferent fibres from the basal forebrain and lateral hypothalamus to the habenula, and were previously reported to be normal in Tcf7l2 -/embryos on E16.5 (Lee et al., 2017), were disorganised and less compact on E18.5 (Fig. 3G). The major habenular efferent tract, i.e., the fasciculus retroflexus was split into thinner fascicles. The L1CAM staining also showed major disruption in the general topography of axonal connections in the whole thalamic area. This demonstrated that TCF7L2 is critically involved in the development of habenular and thalamic anatomy and connectivity.

Impaired cell sorting in the diencephalon
Because anatomical boundaries of prosomere 2 were blurred in Tcf7l2 -/embryos, we hypothesized that cells in this region do not segregate properly. To identify borders of prosomere 2 and its main subdivisions at molecular levels, we analysed the expression pattern of several diencephalic markers by in situ hybridisation. We used a Tcf7l2 probe to stain the pretectum (prosomere 1) and prosomere 2, Gbx2 probe to stain the glutamatergic thalamus, Nkx2-2 and Sox14 probes to stain the rostral thalamus and Pax6 probe to stain the prethalamus (prosomere 3) and pretectum. These stainings confirmed that the thalamo-habenular area was fused and malformed in Tcf7l2 -/embryos ( Fig. 4A-B). Also, we noticed that Gbx2 staining was still present in the periventricular area, where the youngest neurons are located, but absent in the intermediate and superficial portions of the thalamus (Fig. 4B), suggesting premature downregulation of Gbx2. The rostral thalamus area was elongated laterally (Fig. 4C). The boundary between prosomere 2 and 3, which was demarcated by the expression of Pax6, was disrupted ( Fig. 4D).
To examine the boundaries at the cellular level, we stained brain sections with anti-PAX6, anti-SIX3 (marking the prethalamus), anti-NKX2-2 and anti-POU4F1 antibodies.
Prosomere 2 cells were identified with anti-TCF7L2 (WT mice) or anti-β-galactosidase (KO mice) antibodies. In control embryos, the thalamic area was delineated rostro-ventrally by a narrow strip of PAX6-positive cells in a prethalamic subdomain (Fig. 4E), which separated the thalamus from SIX3-positive prethalamic area (Fig. 4F), and the NKX2-2-positive area of the intergeniculate leaflet and ventral lateral geniculate nucleus did not overlap with TCF7L2-high area of the caudal thalamus (Fig. 4G). Habenular cells were easily identified by POU4F1 staining, and the differences in cell densities distinguished the lateral from medial part (Fig.   4H). In contrast, in Tcf7l2 KO embryos, many PAX6-, SIX3-, NKX2-2-and POU4F1-positive cells were intermingled into the neighbouring thalamic territories (Fig. 4E-H). Sparse distribution of these cells, in particular POU4F1-positive cells, pointed to their unusual migration rather than identity switch. Consequently, the border between prosomere 2 and 3 was devoid of its sharpness, and thalamo-habenular border did not exist. This indicated that TCF7L2 plays a critical role in the segregation of cells in subregional clusters in prosomere 2.

Disrupted prosomere 2-specific regulatory network and altered expression of morphogenesis effector genes
To investigate the possible role of TCF7L2 in the regulation of a genetic program of region-specific maturation in prosomere 2, we analysed global gene expression by RNA-seq in the thalamo-habenular region in WT and Tcf7l2 -/embryos on E18.5. 210 genes were significantly downregulated and 113 were upregulated in KO embryos by ≥ 0.4 or ≤ -0.4 log 2 fold-change (FC) (Fig. 5A, Table S1). Gene ontology (GO) term analysis of the differentially expressed genes (DEGs) revealed an overrepresentation of genes that are involved in transcription factor activity, anatomical structure development, neuron differentiation, axon guidance, cell adhesion, regulation of cell migration, regulation of transcription and synaptic signalling ( Fig. 5A, Table S2). To determine whether the E18.5 DEGs from these groups are specific for prosomere 2, we inspected the corresponding in situ hybridisation images of brain sections in the Allen Brain Atlas. 100% of the downregulated genes in the selected groups were enriched in the thalamus, habenula, or both (Fig. 5A). Among them were prosomere 2specific transcription factor genes, including known regulators of thalamic or habenular development -Rora, Foxp2, Etv1 and Nr4a2 (Ebisu et al., 2016;Quina et al., 2009;Vitalis et al., 2017), cell adhesion molecules Cdh6, Cdh8 and Cntn6 (Bibollet-Bahena et al., 2017) and axon guidance genes such as Epha4, Ntng1, and Robo3 that encode important regulators of the guidance of thalamic or habenular efferent connections and the segregation of neurons in this region (Belle et al., 2014;Braisted et al., 2000;Dufour et al., 2003;Lehigh et al., 2013). Also excitability genes, such as thalamus-enriched serotonin transporter gene Slc6a4 that is expressed only during embryogenesis to regulate arborisation of thalamocortical axons (Chen et al., 2015) were downregulated in Tcf7l2 KO embryos. In contrast, the list of the upregulated genes was dominated by the ones that were specifically expressed along thalamic borders or depleted from the thalamus, such as Reln that is involved in neuronal migration and positioning (Hirota and Nakajima, 2017). An increased level of the rostral thalamus markers Nkx2-2, Sox14 and Lhx5 was also observed, and was likely caused by the expansion of this domain into the caudal thalamic area (Fig. 4C).
To confirm that the knockout of Tcf7l2 caused misexpression of prosomere 2-enriched or depleted genes, we validated several of the identified E18.5 DEGs by in situ hybridisation.

Normal acquisition of glutamatergic identity but an impaired expression of postnatally induced synaptic and excitability genes in the thalamus
We then asked if TCF7L2 acts also as a terminal selector of thalamic phenotype, which is underlined by the expression of region-specific genes that determine neurotransmitter identity and electrophysiological characteristics. Habenular and thalamic neurons in rodents (except for GABAergic interneurons that are derived from the rostral thalamus (Evangelio et al., 2018)) are glutamatergic and express high levels of a vesicular glutamate transporter VGLUT2 (Fremeau et al., 2001;Herzog et al., 2001). To determine whether TCF7L2 is involved in the adoption of glutamatergic fate in the thalamus and habenula, we examined the expression patterns of Vglut2/Slc17a6 and Gad67/Gad1 (a marker of GABAergic neurons) in the diencephalon in Tcf7l2 -/embryos and Cck Cre :Tcf7l2 fl/fl P60 adult mice. Both knockout strains exhibited a pattern of GABAergic and glutamatergic cell distribution that was similar to the wild type condition, with predominant Vglut2/Slc17a6 expression in prosomere 2 ( Fig. 6A and S5A). Thus, TCF7L2 is not involved in the specification and maintenance of VGLUT2identity in prosomere 2.
To investigate the hypothesis that TCF7L2 regulates terminal gene batteries, we compared global gene expression profiles in the thalamus between Cck Cre :Tcf7l2 fl/fl and WT mice on P60 by RNA-seq. 310 genes were significantly downregulated and 227 were upregulated in KO mice by ≥ 0.4 or ≤ -0.4 log 2 FC (Fig. 6B, Table S3). GO term enrichment analysis of the P60 DEGs revealed significant enrichment with terms that clustered into groups of synaptic proteins and regulators of membrane conductance: regulation of ion transport, voltage-gated channel activity, regulation of membrane potential, G-protein coupled receptor signalling pathway, regulation of trans-synaptic signalling and regulation of synapse organisation (Fig. 6B, Table S4). Transcription factor genes were not overrepresented in the P60 DEGs. However, 6 thalamus-enriched subregional transcription factor genes were significantly downregulated or upregulated in Cck Cre :Tcf7l2 fl/fl mice on P60 (Fig. 6C), including Rreb1 that is expressed in the thalamus postnatally, Lef1 and Rora, the latter confirmed by in situ hybridisation ( Fig. 6D and S5B).

Development • Accepted manuscript
To investigate if TCF7L2 regulates thalamus-specific or generic neuronal features, we examined spatial expression profiles of the identified excitability/synaptic genes in this cluster in the Allen Brain Atlas. The vast majority of the downregulated genes are expressed specifically in the thalamus (Fig. 6B), such as Kcnc2 and Cacna1g, which encode subunits of K v3.2 voltage-gated potassium channels and Ca v3.1 voltage-gated calcium channels, respectively (Kasten et al., 2007;Kim et al., 2001). The downregulation of Ca v3.1 was further confirmed by immunohistochemistry ( Fig. 6E and S5C). Also, the habenular and thalamic glutamate transporter gene Vglut1/Slc17a7 was downregulated, but not Vglut2/Slc17a6 that encodes the main thalamic glutamate transporter, consistently with the in situ hybridisation results ( Fig. 6A and S5A). Conversely, only a few of the upregulated excitability/synaptic genes were thalamus-enriched.
To investigate the hypothesis that TCF7L2 regulates a genetic program of terminal selection that is activated postnatally, we crossed the selected group of genes with a list of genes that were differentially expressed between E18.5 and P60 in WT mice. Almost 90% of the synaptic/excitability P60 DEGs that were thalamus-enriched were induced after embryogenesis, confirming that TCF7L2 functions as a terminal selector during postnatal development. Thus, postnatally, TCF7L2 only partly regulates the expression of thalamic transcription factors but controls a battery of genes that are induced postnatally and shape terminal electrophysiological identities of thalamic neurons.

Direct regulation of thalamic terminal effector genes by TCF7L2
To understand how TCF7L2 regulates terminal effector genes in the thalamus, we performed a ChIP-seq analysis on the thalami isolated from adult WT mice on P60. We used the same antibody that we used for Western blot and immunofluorescence/immunohistochemistry in this study, and which was validated with samples from the mutant animals ( Fig. 2B-G). This antibody was previously used by other authors on different cell types (Frietze et al., 2012;Geoghegan et al., 2019;Norton et al., 2011). Analysis resulted with 4625 peaks in the anti-TCF7L2 precipitated samples with fold enrichment (FE) ≥ 10 over input sample, which annotated to 3496 unique genes (Table S5).
Analysis of motif enrichment (the AME algorithm from the MEME suite) showed significant overrepresentation of the consensus motif for TCF7L2 in the sequences bound by the anti-TCF7L2 antibody. This motif was detected in almost 85% of the TCF7L2 ChIP-seq peaks ( Fig. 7A), validating the experiment. In addition, we used a thalamic sample from Cck Cre :Tcf7l2 fl/fl mice. 94,3% of the peaks identified in the wild type condition were not detected in this sample, proving specific target recognition in our assay.

Development • Accepted manuscript
To find the most frequent motifs in the TCF7L2 ChIP-seq peaks, we performed denovo motif discovery with MEME-ChIP. 17 motifs were identified. The most significantly overrepresented motif was identical to the TCF7L2/TCF7L1 consensus binding site (E value = 4.3e -425 ; Fig. 7B). The motifs of GCR (NR3C1), RREB1 and RORA were also overrepresented (E value = 1.3 -32 , 2.3 -30 , 2.2 -23 , respectively; Fig. S6A). These transcription factors are enriched in the thalamus, their expression was altered in Cck Cre :Tcf7l2 fl/fl mice, and their genes were identified by the ChIP-seq, suggesting that not only are they downstream targets of TCF7L2 but also cooperate with TCF7L2 in gene expression regulation. The peaks annotated to genes that were expressed in the thalamus on P60 were most frequently localised in intronic regions that may act as intragenic enhancers ( Fig. S6B and Table S5). The remaining peaks (i.e., annotated to non-expressed genes) were mainly annotated to predicted genes and pseudogenes, and were located in distal intergenic regions, suggesting that they represent distal enhancers of unidentified genes.
GO term enrichment analysis performed on genes significantly bound by TCF7L2 AND expressed at P60 revealed an overrepresentation of genes related to the regulation of membrane potential and synaptic signalling, calcium and potassium ions transmembrane transport, also the regulation of neuron projection development and cell adhesion (Table S6).
The ChIP-seq peaks were detected in 31% genes that were upregulated, and in 45% genes that were downregulated, included many thalamus-enriched genes that are involved in synaptic signalling or membrane excitability ( Fig. 7D-E and Table S5). These gens were either broadly expressed in the thalamus (such as Cacna1g, Gabrd,Kcnc2,Syt7,Gabra4,Grm1,Grid2ip or Synpo2) or restricted to thalamic subregions (such as Grm1 to the anterodorsal and mediodorsal nuclei, Cacng3 to the PV and midline nuclei, Kcnab2 to the PF, AD and ventral nuclei, and Kcnd2 to the AD, PV and habenula). This confirmed that TCF7L2 is directly involved in the activation of genes that define pan-thalamic terminal identity and subregional identities in the thalamus. The DEGs with no annotated ChIP-seq peaks can be indirect targets of TCF7L2 or are regulated by TCF7L2-dependent distal enhancers.

Severe impairments in excitability of thalamic neurons in adult mice
To functionally test the role of TCF7l2 in the regulation of thalamic neurons electrophysiological properties, we used whole-cell patch-clamp recordings in Cck Cre :Tcf7l2 fl/fl mice from targeted thalamocortical neurons in the VB. We first tested the basic properties of VB neurons. Resting potential and capacitance were similar in WT and KO mice, but input Development • Accepted manuscript resistance significantly decreased in KO mice (Fig. 8A). We then examined whether the targeted neurons were able to evoke action potentials (APs; e.g., tonic, burst, and rebound burst modes) that are typical for thalamocortical cells. We used depolarising current steps to evoke both sustained trains of APs (tonic) and burst firing at the beginning of a train (Fig. 8B-C). Thalamic cells produced fewer APs in Cck Cre :Tcf7l2 fl/fl mice in both tonic and burst firing modes (Fig. 8D). Rebound bursts, which are crucial for the response of thalamocortical neurons to inhibitory input, were evoked by steps of hyperpolarising current (Fig. 8E). Most neurons from Tcf7l2 KO mice did not show any rebound bursts at the hyperpolarising membrane potential (-65 mV; Fig. 8F). Hyperpolarising steps that were applied at the resting membrane potential (~ -57 mV) evoked rebound burst spiking in VB neurons in Cck Cre :Tcf7l2 fl/fl mice, but the number of spikes was approximately twice as low as those of WT mice (Fig. 8G). These dramatic impairments in electrophysiological responses demonstrated that TCF7L2 is essential for the establishment of unique excitability and firing patterns in thalamocortical neurons.

Discussion
Little is understood about how the lengthy process of postmitotic differentiation is regulated in the vertebrate brain. The present study identifies TCF7L2 as a master regulator of regional transcription factors in the thalamus and habenula, and a selector of stage-specific developmental programs that switch postnatally from morphological to electrophysiological maturation.

TCF7L2 and a network of transcription factors regulate morphological maturation of the thalamus and habenula
TCF7L2 is the only developmentally regulated transcription factor that is expressed

Development • Accepted manuscript
Anatomical impairments were much more severe in Tcf7l2 -/embryos, but much of this phenotype may be attributed to secondary effects that result from the spread of POU4F1positive cells throughout lateral part of prosomere 2 in Tcf7l2 -/embryos.
Presumably, cell non-autonomous and secondary mechanisms contribute to morphological malformation of the thalamo-habenular region. Considering that Pax6-positive prethalamic cells do not express Tcf7l2 in wild type embryos, abnormal intermingling of these cells into thalamic territory must be cell non-autonomous. The same may apply to the impaired segregation of rostral thalamic and habenular cells. Mechanisms that regulate cell migration and nucleogenesis in the diencephalon are poorly understood. We speculate that misexpression of cell adhesion genes in the thalamus, such as ectopic expression of Reln and decreased expression of thalamus-specific genes Cdh6, Cdh8, and Cntn6, could turn the thalamus into a permissive environment for cells migrating from the neighboring Relnpositive structures, i.e., prethalamus, rostral thalamus, habenula and, possibly, pretectum.
Considering that topographic axonal connections can create physical boundaries in the developing brain, disorganised stria medullaris or afferent connections from the retina, pretectum and midbrain, where Tcf7l2 is expressed at high levels (Nagalski et al., 2013;Vacik et al., 2011), may also play a role.
A previous research showed that the aberrant growth of thalamocortical axons toward the hypothalamus instead of the ventral telencephalon in Tcf7l2 -/embryos resulted from unresponsiveness of thalamic cells to Slit repulsive ligands, due to decreased expression of genes that encode Slit receptors Robo1 and Robo2 (Lee et al., 2017). We did not observe any changes in the levels of Robo1 and Robo2 mRNA. The expression of these genes is specific for prosomere 2 only at earlier stages (Allen Brain Atlas: Developing Mouse Brain, 2008); hence it may not depend on TCF7L2 at late gestation. Instead, we observed decreased expression of genes that encode habenular axon-navigating molecules Robo3 and Rgma and thalamic axon-navigating molecules that are later induced and subregion-specific, e.g., Ntng1, EPHA1, 3, 4, 8. Eph receptor A4 (EPHA4) regulates topographical sorting of VB axons in the ventral telencephalon at late gestation (Dufour et al., 2003). This implicates TCF7L2 in controlling the sequential steps of thalamocortical axon navigation and subregional sorting.

TCF7L2 controls the acquisition of characteristic excitability patterns in the thalamus
Many genes that were downregulated in mice with the postnatal knockout of Tcf7l2  (Kasten et al., 2007) in thalamocortical neurons, similar to Tcf7l2 knockout. These results implicate TCF7L2 in the regulation of postnatal genes that control thalamic terminal excitability patterns.
The thalamus is molecularly distinguishable from other brain structures, but many thalamus-enriched genes are differentially expressed between thalamic nuclei or groups of nuclei (Nagalski et al., 2016;Phillips et al., 2019). TCF7L2 was proved to regulate genes that are broadly expressed in the thalamus and those that are specifically expressed in groups of thalamic nuclei. Although Tcf7l2 was not knocked out in PV and PF, and was less efficiently knocked out in the AD or midline nuclei, ChIP-seq analysis identified TCF7L2 peaks in excitability/synaptic genes whose expression is enriched specifically in these regions, implicating TCF7L2 in the direct control of subregional as well as pan-regional terminal selection in the thalamus. Cooperation with subregional thalamic transcription factors, such as RORA, NR3C1 and RREB1, as suggested by the overrepresentation of the corresponding binding motifs in the TCF7L2 ChIP-seq peaks, could contribute to TCF7L2-dependent regulation of differentially expression thalamic genes, but this question needs further investigation.  (Hendricks et al., 2003;Liu et al., 2010;Wyler et al., 2016), and neurotransmitter metabolism genes together with ion channel genes constitute a co-varying module in midbrain dopaminergic neurons (Tapia et al., 2018). Furthermore, studies on glutamatergic neurons in mice concluded that POU4F1 and FEZF2 are selectors of VGLUT1 identity in the medial habenula and corticospinal neurons, respectively, as well as regulators of many other neuron subtype-specific genes (Chen et al., 2008;Lodato et al., 2014;Serrano-Saiz et al., 2018 (Pereira et al., 2015), and ttx-3 controls the terminal differentiation of neurosecretory-motor neurons, but not their serotoninergic identity (Zhang et al., 2014). More research is needed to build models of the postmitotic regulation of neurons in vertebrates and compare regulatory strategies between vertebrates and invertebrates.

Conclusion
The present study sheds new light on vertebrate regulatory strategies in the postmitotic differentiation of molecularly diverse neurons that share a glutamatergic identity.
We found that temporarily separated developmental events and molecular diversification of neurons within a region can be controlled by a single regional transcription factor, as exemplified by TCF7L2 and prosomere 2. Finally, we showed that electrophysiological maturation can be uncoupled from the selection of neurotransmitter identity. Considering that Tcf7l2 is associated with mental disorders, our findings also provide a new insight into the aetiology of thalamic and habenular dysfunction that are observed in these disorders.

Development • Accepted manuscript
Brain fixation and brain slice preparation. Embryos were collected on embryonic day 12.5 (E12.5) or E18.5. Noon on the day of appearance of the vaginal plug was considered E0.5.
Timed-pregnant dams were sacrificed by cervical dislocation, the embryos were removed and decapitated. E18.5 brains were dissected out and fixed overnight in 4% paraformaldehyde Tomomi Shimogori from the RIKEN Center for Brain Science, RIKEN, Saitama, Japan (Reln) (Chiara et al., 2012); David Price from the Centre for Integrative Physiology, University of Edinburgh, UK (Pax6) (Walther and Gruss, 1991

Development • Accepted manuscript
DiI axon tracing. PFA-fixed brains were separated into hemispheres and small DiI crystals ThermoFisher Scientific) were placed in the exposed thalamic surface.
Tissue was then incubated in 4% PFA at 37 o C for 18-21 days. The hemispheres were then embedded in 5% low-melting-point agarose and cut into 100 μm thick coronal sections in a vibratome. The sections were counterstained with Hoechst, mounted onto glass slides, and secured under a coverslip with Vectashield Antifade Mounting Medium.
Western blot analysis. Protein extracts were obtained from 6 animals per genotype from at least two litters. The thalamo-habenular regions were dissected from the brains (Fig. S7A-B) and homogenised in ice-cold RIPA buffer. Protein concentrations were determined using Bio- A3854, Sigma-Aldrich) and anti-GAPDH (1:1000; Cat. no. SC-25778, SantaCruz) antibodies.
The staining was visualized with peroxidase substrate for enhanced chemiluminescence (ECL) and 200 μM coumaric acid. Images were captured using Amersham Imager 600 RGB (General Electric).

Quantification of Ki-61.
Ki-67-positive cells were counted manually in the prosomere 2 region in E12.5 brain sections from control and Tcf7l2 -/animals from three different litters (3 mice, 4 sections each per genotype). Areas of each prosomere 2 section were measured in ImageJ and the number of Ki-67-positive cells by 1 mm 2 was calculated. Two-tailed Student's t-test was used to test for the difference between two groups.
RNA isolation and RNA-seq analysis. Mice were collected on E18.5 and P60. The thalamohabenular regions were dissected-out immediately (Fig. S7A-B), and the RNA was extracted using QIAzol (Cat. no. 79306,Qiagen) and the RNeasyMini Kit (Cat. no. 74106,Qiagen). The quality of RNA was verified with Bioanalyzer (Agilent). RNA samples from three animals (two litters) for each genotype were sequenced on the same run of Illumina HiSeq2500. The reads were aligned to the mouse genome mm10 assembly from UCSC, using HISAT (Kim et al., 2015) and their counts were generated using HTSeq (Anders et al., 2015). Differential gene expression analysis was performed with DeSeq2 (Love et al., 2014). Genes with log 2 (FC) ≥ 0.4 and log 2 (FC) ≤ -0.4 and FDR adjusted p value (q value) ≤ 0.05 were considered to be the differentially expressed up-and downregulated genes. options. For de novo motif discovery, MEME-ChIP (comprising MEME (Bailey and Elkan, 1994), DREME (Bailey, 2011) and CentriMO (Bailey and Machanick, 2012)) was used with DNA HOCOMOCO Mouse (v11 CORE) database (Kulakovskiy et al., 2013;Ma et al., 2014).

Development • Accepted manuscript
Statistical significance was tested with Fisher's exact test, and p values were corrected for multiple testing (E value).
In vitro slice electrophysiology. Brain slices (300 μm thick) from control and Cck Cre :Tcf7l2 fl/fl mice (4 animals per genotype) of both sexes on P21-23 were prepared by an "along-row" protocol in which the anterior end of the brain was cut along a 45° plane toward the midline (Ying and Goldstein, 2005 (Molecular Devices) amplifier and Digidata 1550A digitizer and pClamp10.6 (Molecular Devices). Recordings were sampled and filtered at 10 kHz. Analysis of action potentials was performed in Clampfit 10.6. Intensity to Voltage (I-V) plots were constructed from a series of current steps in 40 pA increments from -200 to 600 pA from a holding potential of -65 mV or at the resting membrane potential (around -57 mV). Two-tailed Mann-Whitney test was used to test for the difference in resting membrane voltage, membrane capacitance, numbers of action potentials and spiking frequency. Two-tailed Student's t-test was used to test for the difference in series resistance (after confirming the normal distribution of the data).

Data availability
The RNA-seq raw FASTQ files are available at the EMBL-EBI data repository -ArrayExpress, under E-MTAB-8755 number. The ChIP-seq files are in the process of uploading at the NCBI Gene Expression Omnibus data repository. The number will be provided as soon as the datasets are accepted. (g-h) The hypothalamus is removed with a straight cut below the thalamus. Cx, cortex; Hb, habenula; HTh, hypothalamus; Mb, midbrain; OB, olfactory bulb; Th, thalamus. Development: doi:10.1242/dev.190181: Supplementary information Table S1. RNA-seq data for E18.5 wild type and Tcf7l2 -/mice (DeSeq2).

Figures
all_expressed_WT_KO_E18.5 -lists of all genes with non 0 normalised values or non NA p values; E18.5_sign._log2FC<-0.4_or_>0.4 -differentially expressed genes (q value ≤ 0.05) with arbitrary log2 fold change cutoffs ≥ 0.4 or ≤ -0.4. n=3 (independent biological replicates). Table S2. Gene ontology enrichment analysis of the E18.5 DEGs. Reported GO terms (q value ≤ 0.01) are divided into 3 ontologies: biological process, molecular function and cellular component.  replicates, for each replicates; chromatin from 6 mice was pooled for each replicate. Table S6. Gene ontology enrichment analysis of the genes identified by the TCFL2 ChIPseq and expressed in the thalamo-habenular region in wild type mice on P60. Reported GO terms (q value ≤ 0.01) are divided into 3 ontologies: biological process, molecular function and cellular component.
Click here to Download Table S4 Click here to Download Table S5 Click here to Download Table S6 Development : doi:10.1242/dev.190181: Supplementary information