Generation of thalamic neurons from mouse embryonic stem cells

The thalamus is a brain region in the diencephalon, which is located between the telencephalon and the midbrain. The thalamus receives and processes sensory signals from various subcortical systems including somatosensory, auditory, visual and motor systems. It sends the processed signals to the cerebral cortex via thalamocortical projections and receives regulatory feedback signals from the cortex via corticothalamic projections. The major function of the thalamus is to relay, gate and modulate sensory information to the cerebral cortex (Mitchell et al., 2014). During development, the thalamic primordium arises at a specific position of the neural tube. It is located anteroposteriorly in the caudal part of the forebrain and dorsoventrally in the alar plate or the dorsal part of the neural tube. Based on the concept that brain regions are specified by positional signals along the anteroposterior and dorsoventral axes of the neural tube (Stern, 2001), it seems likely that the thalamus could be specified by a set of patterning signals that are spatiotemporally expressed. However, the molecular mechanisms that mediate the development of the thalamic primordium remain largely unknown.

To reveal the molecular mechanisms and the minimum patterning signals necessary for thalamic development, we used an in vitro differentiation culture of embryonic stem cells (ESCs) (Muguruma and Sasai, 2012). This is a powerful tool for determining the minimum requirements for differentiation in a self-organizing tissue. We previously established an efficient culture system for selective neural differentiation of mouse and human ESCs, namely, serum-free culture of embryoid body-like aggregates with quick reaggregation (SFEBq) (Watanabe et al., 2005; Wataya et al., 2008). When cultured in a growth factor-free chemically defined medium (gfCDM), ESCs differentiate into hypothalamus, the anterior-most region of the neural tube, as a default fate. Insulin in the SFEBq culture functions as a weak caudalizing factor and induces structures of the midbrain-hindbrain boundary (MHB), including the isthmus (Wataya et al., 2008; Muguruma et al., 2010). The isthmus functions as an organizer that releases a key molecule, Fgf8, which further regulates the differentiation of the midbrain and the hindbrain (Wurst and Bally-Cuif, 2001). It is thought that insulin and Fgf8 are key molecules for the rostral-caudal specification of early neural development.

In the present study, we searched for an efficient method to differentiate mESCs into thalamic neurons by modifying the SFEBq culture. We found that a low concentration of insulin combined with MAPK/ERK kinase (MEK) inhibitor induced the expression of Pax6 and Otx2, markers for the caudal forebrain, while suppressing the midbrain marker En2. We further found that BMP7, which is expressed in thalamic primordium, promoted the induction of thalamic precursors such as Tcf7l2⁺/Gbx2⁺ and Tcf7l2⁺/Olig3⁺ cells. These mESC-derived thalamic precursors expressed the glutamate transporter vGlut2 (Slc17a6 – Mouse Genome Informatics). They selectively extended their neurites to explants of rat cerebral cortex in organotypic co-cultures. Furthermore, thalamic precursors transplanted in the subcortical regions of neonatal rat extended their axons through the white matter to the neocortex and innervated the upper layers. Thus, we have identified the minimum elements sufficient for in vitro generation of thalamic neurons. These findings expand our knowledge on the development of the thalamus and further open the way for regenerative medicine to address thalamic diseases.

The rostral forebrain region is induced without any patterning signals in SFEBq culture. To induce caudal forebrain, we first attempted to caudalize the mESC-derived neural progenitors by adding insulin to the SFEBq culture (Fig. 1A,B). In qPCR analysis of regional marker genes (Fig. S1A-D), the titrated insulin (1 µg/ml; note that 7 µg/ml was used by Wataya et al., 2008; Muguruma et al., 2010) significantly decreased the expression of Six3 and Rax (rostral forebrain) in day 7 mESC-differentiated cells (Fig. 1C,D). Otx2 (caudal forebrain) and Irx3 (caudal diencephalon and brain tissues caudal to it) were expressed without obvious changes in insulin-treated mESCs (Fig. 1E,F). Pax6 (caudal forebrain) expression significantly decreased, whereas En2 (around the isthmic organizer) expression substantially increased (Fig. 1G,H) in insulin-treated mESCs. Thus, titrated insulin certainly caudalized the rostral forebrain as expected; however, the induced region became midbrain rather than caudal forebrain. Actually, Fgf8 expression was significantly increased by the insulin (Fig. 1I). These results raise the possibility that Fgf8 generates midbrain by acting as an inducing factor like the isthmic organizer and prevents the induction of caudal forebrain. To test this hypothesis, we tried to suppress the activity of FGF signaling.

FGF signaling is mediated by interaction with the FGF receptor (FGFR). It modulates downstream pathways including PI3K/AKT and MEK. To examine whether suppression of Fgf8 activity could induce the development of caudal forebrain, we applied downstream inhibitors of FGF signaling in SFEBq culture. Addition of FGFR inhibitor (SU5402, 10 nM) or MEK inhibitor (MEKi; PD0325901, 1 µM) on day 4 led to decreased expression of En2 and Fgf8 (Fig. 1J; Fig. S1E), whereas these genes were expressed without obvious changes after treatment with PI3K inhibitor (ZSTK474, 250 nM) (Fig. S1F,G, lane 3). In addition, treatment with PD0325901 (Fig. 1K, lane 3), compared with SU5402 (Fig. 1K, lane 2), significantly increased Pax6 expression. Immunohistochemical analysis showed that En2⁺ cells substantially decreased (Fig. 1L-N), whereas Pax6⁺/Otx2⁺ cells increased in PD0325901-treated mESCs on day 7 (Fig. 1O-Q). As expected, phosphorylation of ERK in the insulin-treated differentiating mESCs was markedly blocked by PD0325901 (Fig. 1R,S).

These findings suggest that suppression of Fgf8 activity via the MEK/ERK pathway induces the generation of Pax6⁺/Otx2⁺ caudal forebrain in SFEBq-cultured mESCs (Fig. S1H).

It is thought that the zona limitans intrathalamica (ZLI) is contributes to the development of caudal forebrain via secreted patterning signals, such as sonic hedgehog (Shh) and Wnts (Ishibashi and McMahon, 2002; Braun et al., 2003; Kiecker and Lumsden, 2004; Vue et al., 2009). We tested the effects of hedgehog agonist (SAG, 10 nM) or GSK3β inhibitor (CHIR99021, 1 μM) on the induction of caudal forebrain, but these compounds did not lead to any obvious changes in the expression of En2, Fgf8 or Pax6 on day 7 (Fig. S1I-K).

Because the transcription factor Tcf7l2 starts to be expressed in the thalamus (embryonic day 9 onwards in mice) (Fig. 2A-F; Fig. S2A), we used Tcf7l2::Venus mESCs (Tcf7l2-Venus) to visualize thalamic precursors. However, Tcf7l2 is also expressed in regions other than the thalamus, including the midbrain and hindbrain (Fig. S2B,C). The transcription factors Gbx2 and Olig3, specific markers for thalamic primordium, are expressed in the marginal zone (MZ) and the ventricular zone (VZ) in thalamic primordium, respectively (Fig. 2A-F″). Thus, we used Tcf7l2 in combination with Gbx2 or Olig3 for identification of thalamic precursors.

As described above, PD0325901 induced generation of Pax6⁺/Otx2⁺ caudal forebrain; however, it only slightly increased Tcf7l2-Venus⁺ cells on day 12 (Fig. S2D-K). We searched for further culture conditions to promote the generation of thalamic progenitors by modifying patterning signals in SFEBq culture. Wnt1, Wnt3a and BMP4 are expressed around the thalamic region (Furuta et al., 1997; Suda et al., 2001; Braun et al., 2003; Ishibashi and McMahon, 2002). When these patterning molecules were added in SFEBq culture, qPCR analysis did not show any obvious changes in Tcf7l2, Gbx2 and Olig3 expression (Fig. S2L-N).

Bmp7 is expressed in the MZ of thalamic primordium (Suzuki-Hirano et al., 2011), and loss of Bmp7 function causes caudal forebrain defects (Dudley and Robertson, 1997; Solloway and Robertson, 1999). When BMP7 (30 ng/ml) was added on day 4 with insulin and PD0325901 (IP) treatment in SFEBq culture (SFEBq/IP) (Fig. 2G), we observed Tcf7l2-Venus⁺ mESC-derived cell aggregates in the culture on day 12 (Fig. 2H). The increase in Tcf7l2⁺ cells was confirmed by fluorescence-activated cell sorting (FACS) analysis (Fig. 2I,J). Immunohistochemical analysis also showed that most of the cells in BMP7-treated mESC-aggregates were positive for Tcf7l2-Venus and co-expressed with Gbx2 or Olig3 (Fig. 2K,L). qPCR analysis showed that BMP7 in addition to IP treatment also significantly increased Gbx2 and Olig3 expression (Fig. 2M,N). The frequency of Venus⁺ aggregates increased by treatment with PD0325901 plus BMP7. Within the Venus⁺ aggregates, the number of Tcf7l2-Venus⁺ co-expressing Olig3⁺ or Gbx2⁺ cells was significantly increased by the treatment (Fig. 2O).

BMP7 is well known as a dorsalizing factor in neural development, but we observed little change in expression of Nkx6.1 (Nkx6-1), a specific marker for the ventral region (Fig. S2O). BMP7 also had little effect on the expression of regional marker genes along the anteroposterior axis (Fig. S2P-R).

Taken together, our findings indicate that BMP7 and PD0325901 cooperatively induce the differentiation of caudal forebrain, especially thalamus, in mESC-SFEBq culture (SFEBq/IPB, hereafter).

The caudal forebrain can be divided into several subdomains according to the expression of molecular markers (Fig. 3A; Fig. S3A-H). For example, Ascl1 and neurogenin 2 (Ngn2; Neurog2), which is expressed by precursors of glutamatergic projection neurons (Vue et al., 2007), show a complementary expression pattern (Fig. 3A-B″; Fig. S3A,A′). SFEBq/IPB cell aggregates formed a Tcf7l2-Venus⁺ neuroepithelial-sphere structure with the apical side to the interior and labeled by strong N-cadherin (cadherin 2, Cdh2) and atypical protein kinase C (aPKC) expression (Fig. 3C,D). Ki67 (Mki67)⁺ or phospho-histone H3⁺ mitotic cells were found at or near the apical area (Fig. 3C,D). The precursors of caudal thalamus that were identified by Tcf7l2-Venus⁺ and co-expression of Olig3, Ngn2, Olig2 or Dbx1 were localized in the inner zone close to the apical cavity, whereas those with Tcf7l2-Venus⁺/Gbx2⁺ cells were located in the outer zone (Fig. 3E-H; Fig. S3B,F), recapitulating the stage of maturation along the apical-basal axis of the thalamic primordium. Cells expressing pretectal markers, Tcf7l2/Ascl1/Pax7, appeared in the region adjacent to the caudal thalamic primordium (Fig. 3A,F,I,J; Fig. S3B,D). A prethalamic marker, Isl1, or a basal plate marker, Nkx6.1, was expressed in the Tcf7l2-Venus^– regions (Fig. S3G,H). It is known by studies using mouse genetics that excess Shh signals enlarge the Ascl1-positive rostral thalamic subdomain (Vue et al., 2009). We tested the effect of excess Shh signals on the differentiation of mESC-derived cells. When SAG, a hedgehog agonist, was added in the cultures on day 7 (Fig. 3K), Tcf7l2-Venus⁺/Ascl1⁺/Pax7⁻ cells were greatly increased (Fig. 3L-O). This result indicates that rostral thalamic precursors were increased, consistent with the in vivo findings. Taken together, the results of the detailed analysis on marker expression patterns indicate that several subdomains of caudal forebrain, including thalamic primordium, differentiated in the SFEBq/IPB cell aggregates in a spatially distinct manner.

Tcf7l2 is expressed in the thalamus and the pretectum but not in the prethalamus in embryonic day (E)15 mice. Gbx2 and Sox2 are expressed in restricted regions, such that the thalamus is divided into a Gbx2⁺/Tcf7l2⁺ caudo-medial portion and a Sox2⁺/Tcf7l2⁺ rostral portion (Fig. 4A; Fig. S4A-F) (Nakagawa and O'Leary, 2001; Hashimoto-Torii et al., 2003). It is known that Sox2, a well-known marker for proliferative cells during embryogenesis, is also expressed by postmitotic cells in the thalamus (Vue et al., 2007). In SFEBq/IPB-derived thalamic tissue, Tcf7l2-Venus was intense in the basal region and mitotic Sox2⁺/TuJ1 (Tubb3)⁻ cells were found in the apical region, similar to the VZ (Fig. 4B,B′). Tcf7l2-Venus⁺/Gbx2⁺ and Gbx2⁺/Sox2⁺ cells were observed in SFEBq/IPB-derived thalamic tissue (Fig. 4C,C′). Sox2⁺ cells in the apical region expressed Ki67, a mitotic cell marker (Fig. 4D-D‴), whereas those in the basal region expressed TuJ1, a postmitotic neuronal marker (Fig. 4E-E‴). Immunohistochemical analysis showed that Gbx2⁺, Sox2⁺/TuJ1⁺ and Gbx2⁺/Sox2⁺ cells were 7.3%, 23.1% and 2.8% of cells in Tcf7l2-Venus⁺ cells, respectively (Fig. 4F).

Collectively, we found that mESCs differentiated into thalamic primordium with a rostral-caudal spatial pattern in SFEBq/IPB culture. Furthermore, mESCs differentiated into a diencephalic region including the thalamus, that is, Foxa2 (specific marker for the basal plate of embryonic brain) was expressed outside of the Tcf7l2-Venus⁺ region (Fig. S4G,H). Foxa2⁺ and Tcf7l2⁺ cells were clearly divided in neuroepithelial structures in SFEBq/IPB culture. Pax6 was expressed in prethalamus and co-expressed with Tcf7l2 in a caudal part of pretectum (Suda et al., 2001), but not in the thalamus in E15 mice (Fig. S4I). Pax6 was expressed both in Tcf7l2-Venus⁻ and Tcf7l2-Venus⁺ regions in mESC-derived progenitors (Fig. 4G; Fig. S4J). Isl1, a marker for the prethalamus, was expressed in the part of Tcf7l2⁻ regions that were distinct from those expressing basal plate markers, Nkx2.1 (Nkx2-1) or Nkx6.1 (Fig. S4K-M). These results suggest that mESCs differentiated into prethalamus, thalamus and pretectum with basal plate; in other words, our SFEBq/IPB culture can steer the differentiating region into diencephalon, including thalamus.

The transcription factors Lhx2 (Rétaux et al., 1999), Prox1 (Lavado and Oliver, 2007) and Foxp2 (Ferland et al., 2003) are co-expressed with Tcf7l2 in subsets of presumptive thalamic nuclei (Fig. S4N-P). These transcription factors were also expressed in Tcf7l2-Venus⁺ cells cultured in SFEBq/IPB (Fig. 4H-J).

It is well known that hedgehog and/or Wnt/β-catenin signaling are involved in identification of thalamic progenitors (Ishibashi and McMahon, 2002; Braun et al., 2003; Bluske et al., 2012; Robertshaw et al., 2013; Kiecker and Lumsden, 2004; Vieira et al., 2005; Vue et al., 2009; Jeong et al., 2011). We tested the effects of a GSK3β inhibitor (CHIR99021, 1 μM) or a hedgehog agonist (SAG, 10 nM) on the SFEBq/IPB culture on day 7 (Fig. S4Q-T). The number of Tcf7l2-Venus⁺ cells on day 14 in SFEBq/IPB culture did not change (Fig. S4U). However, addition of CHIR99021 caused an increase of Gbx2⁺/Tcf7l2⁺ cells (Fig. S4V, middle column), whereas the number of Sox2⁺/TuJ1⁺/Tcf7l2⁺ cells did not change (Fig. S4W, middle column). Addition of SAG did not cause changes in the number of Gbx2⁺/Tcf7l2⁺ cells (Fig. S4V, right column) or Sox2⁺/TuJ1⁺/Tcf7l2⁺ cells (Fig. S4W, right column). These results suggest that activation of Wnt signal by the GSK3β inhibitor CHIR99021 was involved in the induction of a caudal region of thalamus, but hedgehog signals had no substantial effect on regionalization of thalamus.

The thalamus contains vGlut2-expressing glutamatergic neurons that project their axons into the neocortex (Varoqui et al., 2002). The neuropeptide VGF is expressed in embryonic sensory thalamic nuclei and transported to the axon terminal (Fig. 5A-C; Fig. S5A,B′) (Sato et al., 2012). To examine the characteristics of SFEBq/IPB-cultured cells, we used 3D culture in collagen gel (Fig. 5D). The explants dissected from E15 mouse thalamus extended their axons radially in the collagen gel and the axons expressed VGF and vGlut2 (Fig. 5E-E″). The Tcf7l2-Venus⁺ portion was dissected from SFEBq/IPB mESC-derived tissues on day 12 and cultured in collagen gel. Tcf7l2-Venus⁺ neurites extended radially, similar to in vivo explants (Fig. S5C) and these neurites expressed VGF and vGlut2 (Fig. 5F-G‴, arrows; Fig. S5D-D‴′). We also observed VGF⁺/vGlut2⁺/Venus⁻ neurites (Fig. 5F,F′, arrowheads), which were presumed to be neurites of the ventral posterior nucleus (Fig. S5A-B′). Most of the neurites emerging in the collagen gel cultures were Tcf7l2-Venus⁺ (Fig. S5E-E‴). vGlut2⁺ glutamatergic fibers were frequently observed, whereas GABAergic fibers were rarely observed (data not shown).

Next, we sorted Tcf7l2-Venus⁺ and Tcf7l2-Venus^– cells on day 14 in SFEBq/IPB culture by FACS (Fig. 5H), and quantitative PCR analysis showed significantly greater Tcf7l2 expression in the Venus⁺ versus Venus^– population (Fig. 5I). Gbx2 and Vgf expression were also significantly higher in the Venus⁺ fraction compared with the Venus^– fraction (Fig. 5J,K). These data suggest that Tcf7l2-Venus⁺ neurites show similar characteristics to thalamic cells in vivo.

To examine whether the SFEBq/IPB-cultured thalamic axons project to corresponding regions of neocortex, we performed organotypic culture in mESC-derived thalamic tissues with rat neonatal cortex (Fig. 6A). When rat thalamic explants (E15) were placed at the ventricular side of a rat cortical slice [postnatal day (P)1 to day 2] (Fig. 6B), thalamic fibers extended radially through the ventricular surface of the cortical slice as shown in previous reports (Yamamoto et al., 1989; Bolz et al., 1990; Molnár and Blakemore, 1991). When the dissected mESC-derived Tcf7l2-Venus⁺ portion on day 15 was co-cultured with a neonatal (P2) rat cortical slice for 6 days, Venus⁺ neurites extended into the cortical slice (Fig. 6C). Thalamocortical axons formed a branch in layer 4 of the cortex. RAR-related orphan receptor alpha (RORα) is expressed in layer 4 of the postnatal mouse cortex (Nakagawa and O'Leary, 2003). As shown in previous studies, Venus⁺ neurites do not terminate in layer 4 labeled with RORα, but rather reach the marginal zone and then grow laterally along the pial surface. In our culture, immunostained Venus⁺ neurons showed high intensity around the RORα⁺ layer (Fig. 6C-F; Fig. S6A-C). RORα is known to be expressed in subsets of thalamic nuclei, and was actually expressed in the mESC-derived aggregates (Fig. S6C). Thus, RORα cells in the host cortical tissues could be migrated from the aggregates. However, this possibility was excluded by the observation that RORα⁺ cell nuclei were not surrounded by GFP⁺ perikarya (Fig. S6B). When we excluded data from the pial surface where axons grew laterally, the GFP intensity was correlated with that of RORα (correlation r=0.5776; Fig. 6G). To examine the preference of mESC-derived neurons for another target tissue, we also performed co-culture with a slice of cerebellum. In the cerebellar slice, Tcf7l2-Venus⁺ neurites did not extend radially but showed a preference for fasciculate extension into the white matter as is the case with E15 thalamic explants co-cultured with a P2 rat cerebellar slice (Fig. S6D-H).

Finally, we performed the transplantation study to confirm the maturation and integration of mESC-derived thalamic neurons in P5 rat brain. To label the transplanted neurons with extended axon, we generated ROSA26::MT/MG knocked-in Tcf7l2-Venus mESCs. The dissociated mESC-derived thalamic precursors on day 15 were injected into caudate putamen (Fig. 6H; 20,000 cells per neonate). The grafted cells were observed in subcortical regions such as caudate putamen, white matter and cortical plate. They had characteristic properties observed in thalamus in vivo (e.g. cruciate dendrites) (Fig. 6I-L; Fig. S6I-J″) (Friedlander et al., 1981). This morphology could resemble that of oligodendrocytes, but these cells were not labeled with an oligodendrocyte marker, oligodendrocyte specific protein (OSP) (Fig. S6K-M″). The membrane Tomato⁺ transplanted cells extended their axons through the white matter towards the neocortex into the upper cortical layers (Fig. 6M,N). Such axonal projection patterns are similar to those of thalamic neurons transplanted into the postnatal rat (Kurotani et al., 1993). These findings show that we established an in vitro differentiation method for thalamus.

Based on accumulated knowledge of differentiation of other brain regions (Wataya et al., 2008; Muguruma et al., 2010), we successfully generated thalamic tissues in mESC culture using a 3D self-organizing system. In our SFEBq/IPB-cultured ESC aggregates, thalamic tissues co-existed with thalamus-neighboring tissues including basal plate, prethalamus and pretectum (Figs 3 and 4). This phenomenon may be consistent with the hypothesis that formation of ZLI from basal plate and regionalization of caudal forebrain should result from interaction between transcription factors expressed in prethalamus and thalamus (Suda et al., 2001; Braun et al., 2003; Hirata et al., 2006; Scholpp et al., 2007).

The spatial pattern of induced thalamic neurons in the neural rosettes resembled the following in vivo spatial expression pattern: Tcf7l2-Venus⁺/Olig3⁺ VZ cells were localized in the inner zone close to the apical cavity, and Tcf7l2-Venus⁺/Gbx2⁺ MZ cells were away from the apical side (Fig. 3E). Once caudal forebrain was induced in the cell aggregates on day 7, thalamic tissues self-organized without exogenous patterning signals (Fig. 4). Treatment with GSK3β inhibitor showed no obvious changes but tended to increase the Tcf7l2-Venus⁺/Gbx2⁺ population. Somatosensory or visual nuclei emerge from the rostral part of the thalamus, whereas auditory and some cognitive nuclei emerge from the caudal part of the thalamus. The mechanism of thalamic nuclear organization, however, remains largely elusive. Our mESC-derived 3D-cultured thalamus could be a new tool for addressing this question.

In this study, we offer new insight into patterning signals for early thalamic development. The efficient induction of thalamic tissues was triggered by transient treatment with the MEK inhibitor PD0325901 and BMP7. FGFR inhibitor or MEK inhibitor induced caudal forebrain tissues in mESC aggregates by suppressing midbrain induction (Fig. 1), which is consistent with a previous study of Fgfr1 mutant mice (Lahti et al., 2012). These data suggest that robustness of the caudal forebrain and MHB regions depends on the FGF signaling pathway. Fgfr1 is weakly expressed throughout neural tube and Fgfr2 and Fgfr3 are expressed in caudal forebrain (Ishibashi and McMahon, 2002; Walshe and Mason, 2000). Actually, when a piece of isthmic tissue or FGF8-soaked beads are implanted in the caudal forebrain of chick embryos, they induce ectopic midbrain-hindbrain regions (Crossley et al., 1996; Martinez et al., 1999). Therefore, the MEK/ERK pathway is thought to remain inactive because the caudal forebrain is somewhat distant from the FGF8 signaling center (isthmic organizer).

Compared with FGF, Wnt and Shh, BMP signals have been rarely discussed in terms of thalamic early development. In the process of establishing this thalamic differentiation method, we found that the addition of BMP7 along with MEK inhibitor promoted the induction of thalamic primordium (Fig. 2). However, the mechanism for thalamic induction by BMP signals remains unclear. The findings obtained from an in vitro system can inform future in vivo analysis in order to increase our understanding of early thalamic development.

The mESC-derived neurons contained neuropeptides, such as VGF, similar to those of thalamic projection neurons (Fig. 5) and projected their axons into the cortex (Fig. 6). The thalamus relays sensory inputs to the neocortex and modulates cognitive processes including memory, attention and decision-making (Mitchell et al., 2014). Recently, several studies have supported a correlation between neurodevelopmental/psychiatric disorders and defects in the thalamocortical neural network (Parnaudeau et al., 2013; Chun et al., 2014; Wells et al., 2016). Thalamocortical axons also have been shown to influence the survival and identity of postmitotic cortical neurons via thalamus-derived molecules including VGF (Sato et al., 2012; Li et al., 2013; Gerstmann and Zimmer, 2015). Understanding the pathogenesis of neurodevelopmental disorders, however, remains challenging. In a preliminary study, we are applying the differentiation culture to human induced pluripotent stem cells. In the future, analysis of thalamic development using our ESC-derived thalamus might be useful for elucidating the mechanisms underlying human neurodevelopmental disorders.

mESCs (EB5, Tcf7l2::Venus, and ROSA26::MT/MG knocked-in Tcf7l2::Venus) were maintained as described (Watanabe et al., 2005). For SFEBq culture, mESCs were dissociated into single cells in 0.25% trypsin-EDTA and quickly re-aggregated in differentiation medium (3000 cells per 150 μl well) in 96-well low-cell-adhesion plates (Thermo Fisher Scientific, 174925). The differentiation medium used during days 0-7 was gfCDM (Johansson and Wiles, 1995; Wataya et al., 2008), consisting of Iscove's modified Dulbecco's medium/Ham's F-12 1:1, chemically defined lipid concentrate, penicillin/streptomycin, monothioglycerol (450 μM), apo-transferrin (15 μg/ml) and crystallization-purified bovine serum albumin (5 mg/ml, Sigma). The medium was supplemented with insulin (1 μg/ml) and SB431542 (10 μM, Tocris) on day 0 (the starting day of differentiation culture).

To generate the targeting construct, we used the rapid two-recombination (EG construction) method (Ikeya et al., 2005). Briefly, the 5′ arm (6.2 kbp) and 3′ arm (4.2 kbp) were excised in vitro from the BAC clones (129J mouse) using the Red/ET cloning. The cDNA of Venus, encoding a yellow variant of GFP (Nagai et al., 2002), was fused in-frame into the first exon of the Tcf7l2 gene at the initial ATG. A Pgk promoter-driven neomycin-resistance selection cassette flanked by loxP sites was inserted downstream of Venus. After electroporation of the linearized vector into EB5 ESCs, homologous recombinant ESCs were selected with neomycin and then screened. Three targeted clones (#1-2, #1-11, #2-34) were confirmed by Southern blot analysis with the 3′, 5′ and neo probes. The floxed PGK-neo cassette was removed from #1-11 by transient transfection with Cre-expressing plasmid using Amaxa Nucleofector. The resultant subclones (clone #1-11-10, #1-11-11 and #1-11-18) were confirmed by genotyping PCR with the Tcf7l2 and neo probes, and they exhibited indistinguishable abilities to differentiate into Tcf7l2⁺ thalamic progenitors; subclone #1-11-18 was mainly used for the experiments shown in this report. The ROSA26::MT/MG knocked-in Tcf7l2::Venus mESC line was produced in a similar fashion. The resultant subclones (clone #1-11-11-1 and #1-11-11-2) were confirmed by Tomato signal and genotyping PCR with the ROSA probes, and exhibited indistinguishable abilities to differentiate into Tcf7l2⁺ thalamic progenitors; subclone #1-11-11-2 was used for the experiments shown in this report.

For the neural differentiation culture, PD0325901 (1 μM, Stemgent) and BMP7 (30 ng/ml, R&D Systems) were added to gfCDM containing insulin on day 4. On day 7, the cell aggregates were transferred to a Petri dish in DMEM/F12 GlutaMAXI/N2/10% fetal bovine serum.

Immunohistochemistry (IHC) was performed as described (Watanabe et al., 2005). Images were analyzed with vHCS Discovery Toolbox software (Thermo Fisher Scientific) and ImageJ (National Institutes of Health; http://rsb.info.nih.gov/ij/rsb.info.nih.gov/ij/). Values shown in graphs represent mean±s.d. Antibodies to the following proteins were used at the indicated dilutions: Pax6 (1:250, PRB278P, BAbCO; 1:200, sc-7750, Santa Cruz Biotechnology), Otx2 (1:100, ab21990, abcam; 1:100, sc-30659, Santa Cruz Biotechnology), En2 (1:100, sc-8111, Santa Cruz Biotechnology), Foxg1 (1:5000, AS3514, custom, Watanabe et al., 2005), Rax (1:3000, MS8407-3, custom, Ikeda et al., 2005), Six3 (1:1000, AS3984, custom, Wataya et al., 2008), Irx3 (1:1000, M230, Takara), phospho-Erk1/2 (1:1000, 4370, Cell Signaling), Tcf7l2 (1:1000, 2569, Cell Signaling; 1:1000, 05-511, Millipore), Gbx2 (1:10,000, no.1-1, custom, Muguruma et al., 2015; 1:500, AF4638, R&D Systems), Olig3 (1:500, MAB2456, R&D Systems), GFP (1:1000, 04404-84, Nacalai Tesque; 1:1000, 598, MBL), neurogenin 2 (shown as Ngn2; 1:1000, sc-19233, Santa Cruz Biotechnology), Mash1 (shown as Ascl1; 1:200, 556604, BD Pharmingen), Ki67 (1:1000, NCL-Ki67p, Novocastra; 1:200, 556003, BD Pharmingen), N-cadherin (1:1000, 610920, BD Transduction), PKCζ (shown as aPKC; 1:100, sc-216, Santa Cruz Biotechnology), phospho-histone H3 (1:500, 9706, Cell Signaling), Olig2 (1:1000, AB9610, Millipore), Dbx1 (1:2000, Ms-1513, custom synthetic peptide antigen conjugated to KLH; antibodies to C+DEDEEGEEDEEITVS were raised in guinea pig and antiserum was affinity purified), Pax7 (1:100, MAB1675, R&D Systems), TuJ1 (1:300, MMS-435P, Covance), Sox2 (1:250, sc-17320, Santa Cruz Biotechnology), Lhx2 (1:100, sc-19344, Santa Cruz Biotechnology), Prox1 (1:200, MAB5654, Millipore), Foxp2 (1:1000, ab16046, abcam), VGF (1:5000, ab69989, abcam), vGlut2 (1:1000, AB5907, Chemicon), RORα (1:100, sc-6062, Santa Cruz Biotechnology), DsRed (1:500, 632496, Clontech), Isl1 (1:100, 40.3A4, Developmental Studies Hybridoma Bank), Nkx6.1 (1:200, HPA036774, Sigma), Foxa2 (1:400, 8186S, Cell Signaling; 1:300, AF2400, R&D Systems), TTF-1 (shown as Nkx2.1; 1:100, 12373S, Cell Signaling), oligodendrocyte specific protein (shown as OSP; 1:500, ab53041, abcam), MAP2 (1:500, M9942, Sigma). Fluorescence-tagged secondary antibodies were as follows: Cy2-, Cy3-, Cy5-, Alexa Fluor 488- or Alexa Fluor 647-conjugated (Jackson Immuno Research). The signal for En2 was detected using biotinylated anti-goat IgG with Cy2- or Cy3-conjugated streptavidin (Vector Laboratories). DAPI was used for counterstaining the nuclei (Molecular Probes). Stained sections were analyzed with LSM710 and LSM780 confocal microscopes (Carl Zeiss). Images were assembled using Adobe Photoshop CS5.

Quantitative PCR was performed using the 7500 Fast Real Time PCR System (Applied Biosystems) and data were normalized to Gapdh expression. PCR primers are listed in Table S1. The values shown on graphs represent mean±s.d. For quantitative analysis, 8-16 aggregates were examined for each experiment, which was repeated at least three times.

For FACS analysis, cells were counted with FACSAria (Becton Dickinson), and the data were analyzed with the FACSDiva software (Becton Dickinson). The cells were dispersed to single cells using Nerve-Cell Culture System/Dissociation Solutions (Sumitomo Bakelite) and analyzed at 4°C. The cells were dispersed on day 12 or day 14, and filtered through a Cell Strainer (BD Biosciences).

The procedure for collagen gel culture was modified as described previously (Shirasaki et al., 1995). ESC aggregates were dissected and embedded in collagen gel matrix. ESC aggregates were cultured in DMEM/F12/GlutaMAXI/N2 medium at 37°C in 5% CO₂ for 3 days.

Organotypic co-culture was carried out using the modified procedure in Yamamoto et al. (1992). Briefly, fetal mice (E15) were removed from pregnant females (ICR), and the whole brain was dissected from each fetal mouse and kept in cold DMEM medium. Thalamic explants were dissected in the medium under a binocular microscope. A Tcf7l2-Venus⁺ portion of mESC aggregate (day 15) was dissected in medium for neural differentiation culture. The cortical coronal slices and cerebellar sagittal slices (300-500 μm thickness) were dissected from newborn rats (P1-2). These neural tissues were kept in cold DMEM medium.

Rat slices (cortex or cerebellum), and thalamic explant or dissected cell aggregates were plated on the collagen-coated membrane (Coster, Transwell-COL #3942), separated at a distance of less than 0.5 mm. Thalamic explant or dissected cell aggregate was plated so that it faced the ventral side of the cortical or cerebellar slice. Serum-free, hormone-supplemented medium (DMEM/F12/GlutaMAXI/N2) was used to enhance neuronal survival and to suppress glial proliferation. In addition, the culture medium contained 5% fetal bovine serum for the first 2 days.

For the distribution histogram of normalized intensity of GFP and RORα, ten regions of interest (0.1 mm×0.4 mm) per slice were manipulated from pia to 1 mm depth. Images were analyzed with ImageJ.

The transplanted cells were Tcf7l2::Venus/ROSA26::MT/MG mESCs dispersed by Nerve-Cell Culture System/Dissociation Solutions on day 15. To perform the transplantation, P5 neonatal rats were deeply anesthetized by inhalation of sodium pentobarbital. Two microliters of cell suspension (20,000 cells per μl) was injected into the white matter of right hemisphere of each neonate using a 26-gauge needle attached to a 10-μl Hamilton syringe (Hamilton). The brain was dissected 1 week after transplantation, sectioned at a 50-μm thickness, and the sections were immunostained.

ICR mice and SD rats were obtained from Nihon-SLC. The animals were cared for in accordance with the Regulations on Animal Experiments of RIKEN.

All data are shown as mean±s.d. Statistical analyses were performed with PRISM software (GraphPad, version 6). Statistical significance was tested by unpaired Mann–Whitney test (nonparametric) for two-group comparisons or Kruskal–Wallis with Dunn's test for multiple-group comparisons.

We thank F. Matsuzaki, M. Eiraku and M. Ohgushi for discussion; M. Kawada for technical advice for generation of knocked-in mESCs; and M. Shiraishi, C. Shiraishi, K. Shiraishi, K. Fujiwara, M. Takeichi, T. Hirano, and all members of the Sasai group and Matsuzaki group for support and encouragement. A.S. offers special thanks to Yoshiki Sasai for enthusiastic mentoring, and sincere respect for his integrity to science and precious time he spent on the Sasai group.