In a previous study, we determined the brain precursor map in the anterior epiblast of chicken embryos marked by Sox2 N2 enhancer activation (Yoshihi et al., 2022a). In this study, we extended the analysis to anterior spinal cord (SC) precursors, also included in the N2 enhancer-marked anterior epiblast (Figures 1AA’). The fundamental technique was to randomly label the chicken embryo epiblast at st. 5 (∼20 h of incubation) by electroporating it with a Supernova vector cocktail at st. 4 (∼18 h of incubation) and to trace cells that contributed to the individual CNS portions at st. 8 (over ∼18 h). The previous analysis indicated that the epiblast regions for the developmental potential of the forebrain (FB), midbrain (MB), and hindbrain (HB) are arranged from anterior to posterior in the anterior epiblast. As shown in Figure 1A’, the SC precursors were mapped posterior to the HB precursors in the anterior epiblast.

The SC precursors migrated toward the midline and then posteriorly during early st. 5 to late st. 6, as shown by the trajectory of randomly labeled epiblast cells shown in Figure 1B, where the representative migration paths of the SC precursors are drawn in orange arrows. This migration path contrasts with the anterior migration of the brain precursors after convergence to the midline (green arrows). The vectors in Figure 1C show the average migration directions and velocities around the grid points at the end time point of the trajectories. Fast posterior migration of SC precursors along the midline (138 ± 39 (SD) µm/h) was observed.

The posterior migration of the anterior epiblast-derived SC precursors was reflected by the distribution of cells labeled by N2-activated mCherry, as shown in Figure 1D. The most posteriorly located SC precursors migrated over an axial distance of 2 mm; no more posterior labeling was observed (Figure 1D), indicating that the anterior epiblast-derived cells resided only in the anterior spinal cord. N2 enhancer activity depends on Otx2 activity (Iwafuchi-Doi et al., 2012; Uchikawa et al., 2017), but Otx2 expression is lost along the posterior migration route of SC precursors (Chapman et al., 2002); hence, the N2-indicating fluorescence declined but allowed tracing of the initially N2-positive cells in the spinal cord (Figure 1D, broken orange line).

In mouse embryos, the N2 enhancer-active region of the epiblast and forming neural tissues change analogously to the chicken embryos after E7 (Iwafuchi-Doi et al., 2011). In addition, analysis of stochastically labeled cell clones demonstrated clones comprising the brain and spinal cord subclones (Tzouanacou et al., 2009). Therefore, the mouse embryo anterior epiblast also serves as the origin of the CNS region spanning from the FB to the anterior SC.

EpiSCs are maintained in culture media containing FGF2 and activin; the latter is an in vitro Nodal replacement that inhibits neural development (Camus et al., 2006). EpiSCs enter the neurogenetic process after removing activin/Nodal signaling and develop into neurons and glial cells (Brons et al., 2007; Tesar et al., 2007; Iwafuchi-Doi et al., 2012). A detailed analysis of the change in the gene expression profiles after activin removal indicated that the first day after activin removal from the culture gives rise to the transitory state between the epiblast state and the neurogenic state: the high Pou5f1 expression level is still maintained, yet neural primordium-characteristic Zeb2 (Sip1) expression (Chng et al., 2010) has started (Iwafuchi-Doi et al., 2012). Therefore, EpiSCs on the first day of activin removal mimic the transition state from the anterior epiblast to the neural primordium in embryos. To capture the neural primordial state immediately after the epiblast-to-neural-primordium transition, we placed activin-free EpiSCs in medium containing EGF and FGF2, which arrests neural development but maintain the cells as early-stage neural primordial cells. As discussed in the Introduction, establishing a stem cell line stabilizes a cell state that is transient in the normal developmental process. As the Wnt signal was a strong candidate agent specifying the anteroposterior regionality in the CNS primordia, we also manipulated Wnt signal levels on the NSC-generating EpiSCs in the following ways: 1) with the addition of Wnt signal antagonist XAV939 (XAV) at 10 µM to the culture medium (low Wnt condition) to obtain cell lines numbered with the prefix XN; 2) without the addition of Wnt antagonist/agonist allowing unconstrained action of cell-endogenous Wnt signaling (medium Wnt condition) for the cell lines with the prefix KN; and 3) with the addition of Wnt signal agonist CHIR at 2 µM (high Wnt condition) for the cell lines with the prefix CN. The cultures were passaged at an interval of 2–3 days on gelatin-coated dishes.

In the first 4–5 days, cells developed as cell clumps loosely attached to the culture dish, and then cells adherent to the gelatin-coated dish started to migrate out of the cell clumps in a week; these cells proliferated well and prevailed in the cultures after serial passages (Figure 2A). The resultant cultures at 4 weeks appeared homogeneous in cell morphology in culture (Figure 2B). All these cell lines exhibited the characteristics of neural stem cell (NSC) lines (Supplementary Table S1), as detailed below.

We conducted NSC-producing cultures using different Wnt signal input levels in five series, as summarized in Supplementary Table S1, indicating the reproducibility of the NSC line derivation protocol. The six cell lines XN4, XN1, KN3, KN1, CN1, and CN4 were characterized in detail below, concerning gene expression profiles. The NSC lines were generated mostly using a mouse 129 line-derived EpiSC (Tesar et al., 2007). However, a BDF1-derived EpiSC line (Y62) used in series 2 similarly produced NSC lines, indicating that the derivation of an NSC line from EpiSCs using our protocol does not depend on the mouse origin of EpiSCs.

The cells in the NSC-generating culture multiplied similarly in the EGF- and FGF2-containing culture medium regardless of the addition of Wnt antagonist or agonist for ∼4 weeks. The resultant cell populations were morphologically homogeneous in culture but morphologically discernible between the NSC lines (Figure 2B). Notably, after 4 weeks of culture, the NSC lines started to display cell line-specific growth requirements, as summarized in Table 1. The cell lines XN1, CN1, and CN4 exhibited decreased cell growth under standard culture conditions, but the addition of the Rho kinase inhibitor Y27632 restored cell proliferation. In contrast, the addition of Y27632 inhibited the growth of XN4. The NSC lines also started to show different preferences concerning the culture substrate (gelatin or laminin) for their growth. Furthermore, the KN3 line was found to be sensitive to Accutase treatment.

In summary, the EpiSC-derived NSC lines acquired each unique trait after 4 weeks of culture: responses to the Rho kinase inhibitor, the culture substrate, and Accutase treatment. This observation may indicate that the cell populations reach the state of mature and stable cell lines with specific growth requirements. A feature shared by all six NSC lines was the dependence of their cell growth on the local cell density; in a rough estimation, a local cell density (not the average of a culture) of 400 cells/mm² was required for their growth.

These cell lines were proven to be NSC lines that could develop into neurons expressing ßIII tubulin (Tuj1) and glial cells (astrocytes expressing glial fibrillary acidic protein (GFAP) and oligodendrocytes expressing the O4 antigen) under their differentiation conditions (Figure 2C, Supplementary Table S1).

The transcriptome profiles of the NSC lines derived from EpiSCs under different Wnt signal inputs were characterized using microarrays. The expression of genes was also compared with the FB, MB, HB, and cervical SC (anterior to the forelimb buds) of E10.5 mouse embryos concerning the patterns and expression levels (Figure 3).

Analysis of landmark gene transcripts characteristic of EpiSCs and NSCs indicated overt differences between these cells. EpiSC-characteristic genes, including Pou5f1, Nanog, and Fgf5, were all downregulated in all six NSC lines (Figure 3A). In contrast, the NSC-characteristic genes itemized below were activated to levels comparable to those in the E10.5 CNS (Figure 3B). Neural progenitor-characteristic TF transcripts Zeb2 (Sip1) (Chng et al., 2010; Dang et al., 2012; Hegarty et al., 2015) and Sox2, the primary B1 Sox gene (Uchikawa et al., 2011; Kamachi and Kondoh, 2013), were expressed at significant levels. At least one of the subsidiary B1 Sox genes, Sox3 and Sox1, was strongly expressed; Sox3 expression was low in the XN4 cells, while Sox1 expression was low in the XN1 cells. Pax6 expression was also high in the NSC lines except for the XN4 cells. The variations in Sox3/1 and Pax6 expression presumably reflect the anteroposterior regional specification of individual NSC lines, which is presented below.

To allocate the EpiSC-derived NSC lines along the anteroposterior axis of the embryonic CNS, the expression levels of brain region-characteristic TF genes (Foxg1, the forebrain (FB); Otx2, the FB and midbrain (MB); and En2, the MB and hindbrain (HB)), as well as the Hox genes expressed posterior from the HB, were analyzed. To graphically represent the data, the highest expression level among the ten sample groups (six NSC lines and the four brain portions) measured in the microarray analysis was referred to as 1.0 (Figure 4). Overall, the TF genes mentioned above were expressed in the NSC lines at the level ranges observed in the E10.5 mouse CNS.

The expression profiles of the brain-characteristic genes (Figure 4Aa) showed that XN4 expressed Foxg1 at a level comparable to that of the embryonic FB and low levels of Otx2 and En2 (MB/HB) but no Hox genes (posterior from HB), indicating that the XN4 line primarily showed FB regionality.

Other NSC lines expressed various Hox genes. The regionality codes of Hox gene expression from the hindbrain to the SC around E11 were determined first based on the data by Carpenter (2002), the whole-mount in situ hybridization data of E10.5 mouse embryos in the EMAGE database, and the CSC data here, as summarized in Figure 4Ba. The data for the expression of Hoxa, Hoxb, and Hoxc genes in the NSC lines are shown in Figures 4Ab–d, which indicated that Hoxa1, a2, a3, and a4, as well as Hoxb1, b2, b3, and b4, collectively called the HB class in this study, have the anterior expression boundary in the HB domain. Likewise, Hoxa5, Hoxb5, and Hoxb6 (the CSC class) have their anterior expression boundary in the cervical spinal cord (CSC, anterior to the forelimb bud). Hoxa6, a7, a9, and a10 (the TSC class) appeared to be expressed posterior to the CSC in the thoracic spinal cord (TSC). The anterior expression limits for Hoxb7, b8, and b9 (also the TSC class) are presumably around the CSC and TSC junction, as a low level of expression was measured in the CSC specimens (Figure 4Ba). Hoxa13 or Hoxb13 characteristic of the caudal SC were not expressed in any of the specimens. Hoxa13 or Hoxb13, characteristic of the caudal SC, were not expressed in any samples. Important information concerning the thoracic and lumber levels is that Hoxc9 expressions present in the TSC, while Hoxc10 is expressed posterior to the lumbar SC (LSC).

Based on the above criteria, anteroposterior regional coverages of NSC lines expressing Hox genes were tentatively assessed (Figure 4Bb).

1) The XN1 line expressed Otx2, En2, HB class Hox genes, and Hoxa5 but not Hoxb5, indicating its regional coverage from MB to the anterior CSC. (The observation places the anterior boundary of Hoxa5 expression anterior to that of Hoxb5.)

Significantly, the domain coverages by the NSC lines were ordered from anterior to posterior according to the Wnt signal inputs. Very similar TF expression profiles indicative of posterior positioning of CN1 and CN4 that developed in the presence of the Wnt signal agonist CHIR at 2 μM confirm this notion. The lines KN3 and KN1 developed solely under cell-endogenous Wnt signaling but differed slightly in regional coverage. This difference can be accounted for by the fluctuation of endogenous Wnt signal levels in the starting EpiSCs, which was reported in human ESCs resembling EpiSCs (Moya et al., 2014).

Whereas the XN4 line expressed Foxg1, a hallmark of the forebrain precursor, XN1 line did not express Foxg1 but expressed anterior Hox genes and midbrain-centered characteristics. The XN4 line was produced using 10 µM XAV from supplier B (Supplementary Table S1), but the XN1 line was generated using 10 µM XAV from supplier A, which developed crystalline precipitates during the culture incubation (Supplementary Figure S1; Supplementary Table S1). This crystal precipitation must have decreased the effective concentration of XAV from supplier A, as discussed in Supplementary Figure S1 legend. This difference in the effective XAV concentrations accounts for the different anteroposterior regionalities between the XN4 and XN1 lines.

In summary, all six NSC lines derived from the EpiSCs under different levels of Wnt signal input displayed distinct anteroposterior regionalities following the Wnt signal strengths, from the FB (the lowest Wnt signal input) to the TSC (the highest Wnt signal input). However, their regionalities were broad, covering more than one CNS domain (Figure 4Bb), which was investigated using single-cell transcriptome analysis presented below.

In the mouse neural tube, dorsoventral regionalization also occurs from E8 to E11 during embryogenesis under the influence of dorsal-character-inducing Wnt/BMP signals and ventral-character-inducing sonic hedgehog (Shh) signals (Andrews et al., 2019; Sagner and Briscoe, 2019). Dorsoventrally arranged neural tube subregions are marked by differential TF expression (Figure 5A). In contrast to the anteroposterior regionalities given to the NSC lines, these lines did not develop clear dorsoventral identities (Figure 5B). The only commonly expressed TF genes were members of an intermediate class (Irx3 and Irx5), covering the wide dorsal region, but they were generally unaccompanied by dorsal neural progenitor-characteristic Pax3/Pax7 expression (Gard et al., 2017). The ventral p3 domain-characteristic Nkx2-2 and Nkx6-1 are activated in response to Shh signaling in spinal cord neural progenitors (Andrews et al., 2019; Sagner and Briscoe, 2019), but the NSC lines, e.g., XN1, expressed Nkx2-2 without concurrent Nkx6-1 or correlation with endogenous Shh expression levels, ruling out the model that Nkx2-2 expression represents ventral CNS specification. Most notably, the TF gene expression analysis showed no dorsal or ventral bias in response to the Wnt signaling levels during the cell line derivation, although the Wnt signal is one of the primary cues to activate TF genes of the dorsal class (Muroyama et al., 2002). This observation confirmed that the NSC lines from activin-free EpiSCs represent an earlier developmental period than the dorsoventral specification of neural tubes.

The changes in TF gene expression were evaluated using RT‒qPCR at 1-week intervals during the development of the XN4, KN3, and CN4 lines under the following conditions: with the addition of 10 mM XAV, without any additives, or with the addition of 2 mM CHIR. The aim was to investigate how the NSC characteristics converge after 4 weeks of NSC line derivation (Figure 6).

The EpiSC-characteristic genes Pou5f1 and Nanog were quickly inactivated in a week in all three NSC culture conditions (Figure 6Aa). Although Bra is strongly activated in EpiSCs in the activin-FGF2 medium with 2 mM CHIR, Bra was not activated in the NSC-developing culture conditions, even during the CN4 line derivation with 2 mM CHIR (Figure 6Aa). NMPs characterized by the coexpression of Bra and Sox2 around the time of dichotomic fate choice (Figure 6Aa) (Kondoh and Takemoto, 2012), although the possibility of a transient activation Bra during the sampling intervals was not formally excluded. This notion is consistent with the observation that the contribution of anterior epiblast-derived cells to the spinal cord was limited to the anterior part of the spinal cord (Figure 1D; Figure 4Ad).

The characteristic neural genes, Sox1, Zeb2, and Pax6, which were silent in the EpiSCs, were activated in 1 week in all three culture conditions (Figure 6Ab). The low-Pax6 expression of the XN4 line was observed from the earliest time point.

Hoxb6 and Hoxa7 genes representing the thoracic levels of the spinal cord were quickly activated by 1 week under the high Wnt condition, and the expression level was maintained up to the 4-week point when the CN4 line was established (Figure 6Ac). The level of Otx2 expression was sustained under the low Wnt condition (XAV addition) that gave rise to the XN4 line. In contrast, under the high Wnt condition (CHIR addition), Otx2 expression was turned off from the beginning despite being expressed in the EpiSCs. The case of the KN3 line was intermediate.

During the culture process to produce the NSC lines, the cells were passaged and maintained under the same culture conditions with or without the addition of a Wnt antagonist (XAV) or Wnt agonist (CHIR). However, the anteroposterior regional specificities were created during the development of the NSC lines rather than in response to the culture conditions after the NSC line establishment.

After 3 weeks of culture in the presence of Wnt antagonist, Wnt agonist, or no addition, subcultures were prepared to compare the effects of different culture conditions on Wnt signal levels. 1) Four subcultures of cells developed in the XAV-containing medium were prepared; a pair was maintained in the medium with XAV (X=>X) and the second pair without XAV (X=>O). The subcultures were separately maintained over a week with two to three passages. 2) Similarly, six subcultures of cells that had developed in the absence of Wnt antagonist/agonist were cultured pairwise in media with XAV (O=>X), without addition (O=>O), or with CHIR (O=>C). 3) Four subcultures of cells developed in the CHIR-containing medium were placed pairwise in the agonist-free medium (C=>O) or the CHIR-containing medium (C=>C). After a week of subculturing (i.e., the same chronological point as the establishment of the XN4, KN3, and CN4 lines), the sublines were examined for the expression of an anterior-characteristic gene Otx2 and for a posterior characteristic gene Hoxb6.

These cell subcultures under conditions of altered Wnt signal inputs did not affect the anteroposterior characteristics of the genes. All X=>X and X=>O cultures maintained a level of Otx2 expression, whereas Otx2 expression remained very low in O=>X and O=>O cultures (Figure 6Ba). Similarly, all C=>C and C=>O cultures displayed high levels of Hoxb6 expression, while Hoxb6 expression in O=>C and O=>O cultures remained low (Figure 6Bb).

These observations indicate that the anterior and posterior regional characteristics imposed by the Wnt antagonist and Wnt agonist in the anterior epiblast-derived neural primordia had already developed during a week of culture. The stabilized anteroposterior and other characteristics of the NSC lines developed during the 4-week culture period.

The microarray data for the cell groups of the NSC lines presented in Figure 4 indicated that each NSC line has a relatively broad anteroposterior regional coverage. For instance, the XN1 line expressed the Otx2, En2, and Hoxa genes of HB regionality, covering the MB and HB regions. It is critical to determine whether the broad anteroposterior specificity is intrinsic to individual cells of the cell line or whether the cell lines are composites of separable subpopulations with narrower anteroposterior specificities. To address this issue, we performed single-cell transcriptome analysis (single-cell RNA-seq) of the four cell lines XN4, XN1, KN1, and CN4 (Figure 7). The uniform manifold approximation and projection (UMAP) of the single-cell transcriptome data indicated that individual cell lines formed tight clusters. The arrangement of the four clusters representing the NSC lines, showing the relative distances among the cell states (Figure 7Aa), was consistent with the primary component analysis (PC1 and PC2) of the microarray data of the bulk transcripts (Figure 7Ab), indicating that the general features of the bulk transcriptomes are reflected in the single-cell transcriptomes.

Each transcriptome cluster representing an NSC line was divided into two (or three at most) subclusters, indicated by the broken lines separating the subclusters in Figure 7Aa. However, these subclusters did not represent separable cell populations but the cells in different cell cycle phases. As shown in Figure 7Ba, one of the subclusters was filled with cells expressing Hmgb2, Mki67, and Top2a, which are characteristic of cells in the G2 to M phases (Pallier et al., 2003; Daniloski et al., 2019; Uxa et al., 2021).

Concerning the brain region characteristic genes (Figure 7Bb), the XN4 cluster contained Foxg1-expressing cells, and the XN1 cluster had cells with Otx2 expression and En2 expression. The cells where these transcripts were detected were distributed more or less randomly in the NSC clusters, indicating that the anteroposterior regionality does not divide the NSC clusters. Similarly, Hoxa5-expressing cells were abundant in the XN1, KN1, and CN4 clusters, Hoxa7-expressing cells in the KN1 and CN4 clusters, and Hoxa9 specifically in the CN4 cluster, following the more posterior activation of Hoxa genes with larger numbers. Again, the cells with the detection of these Hox transcripts were distributed randomly in an NSC cluster.

Similarly, the cells with genes normally involved in the dorsoventral regionalization of the neural tube were distributed with the NSC clusters as units. Irx3 transcript-captured cells were abundant in the XN4, XN1, and CN4 clusters, Pax6-captured cells in the XN1, KN1, and CN4 clusters, and Dbx2-captured cells in the CN4 cluster.

The random distribution of the cells with expression of TF genes for different anteroposterior characters in an NSC line cluster opposed the model that the NSC lines were composites of distinct subpopulations bearing narrow regionalities.

To clarify whether single cells simultaneously express genes representing different regional specificities, transcriptomes in individual cells were analyzed. The results are presented in Figure 8 using area-proportional Venn diagrams. The numerals indicate the percent of cells where the transcript was detected.

Figure 8A illustrates the expression and coexpression rates of Otx2, En2, and Hoxa5 in the XN1 cell population. A substantial fraction of the cell population expressed two or three genes together without specific indication of coexpression preferences. Particularly noteworthy is the presence of cells expressing both Otx2 and Hoxa5 (3.2% among 45.3% of cells with transcript capturing of one of these TFs), with a cell population size larger than those with Otx2-En2 coexpression (2.3%). Cells expressing Otx2, En2, and Hoxa5 (0.8%) were also identified. The situation was analogous concerning the expression of TF genes, which are normally expressed according to dorsoventral regionalities (Figure 8B), and among the genes for anteroposterior regionalities (En2 and Hoxa5) and dorsoventral regionality gene Pax6 (Figure 8C). As the coexpression rates of two TF genes appeared to reflect the fraction of cells expressing individual TFgenes, we tested the model in which individual TF genes were expressed stochastically. Namely, if TF genes P and Q are expressed in the fractions p and q of the cell population, the coexpression frequency for P and Q is predicted to be p multiplied by q. Figure 8D compares predicted and observed TF gene coexpression frequency using data shown in Figures 8A–C. The data were on the regression Line y = 1.178x through the axis origin with R ² = 0.996. The nature of single-cell transcriptome analysis accounts for the 17.8% increase in the rate of observed coexpression. The sequence read depths are highly variable among the cells. As the detection rate of coexpressed gene transcripts is higher in more deeply read cells, the observed coexpression rate becomes higher than the predicted rate based on the averaged detection rates of single TF genes. The same scenario also applies to the TF genes representing anteroposterior regionalities in the XN4, KN1, and CN4 NSC lines (Supplementary Figure S2).

Taken together, the transcriptome analysis at the single-cell level (Figures 7, 8) indicated that each NSC line consisted of cells bearing a range of anteroposterior specificities, which accounted for the microarray expression profile of the NSC lines, as displayed in Figure 4. The analysis identified cells expressing two or three such genes simultaneously and cells expressing single genes representing different regional characteristics. The statistical analysis indicated that the coexpression of multiple genes occurred stochastically, without a preference for anteromedial or medioposterior regional preferences.

A study using chicken embryos for brain portion development (Yoshihi et al., 2022a; Yoshihi et al., 2022b) and spinal cord development (Figure 1) indicated that anteroposterior regional specificities are grossly specified at the epiblast stage, before neural plate development. In chicken embryogenesis, the CNS precursors in the Otx2-expressing, and hence N2-enhancer active, anterior epiblast first converge to the midline, as do other cells involved in gastrulation through the primitive streak. In contrast, in mouse embryos, the cell convergence process is missing; anterior epiblast cells develop in situ into brain tissues (Cajal et al., 2012), and only primitive streak-abutting cells ingress into the axial mesoderm compartment, at least during the early stages of gastrulation (Williams et al., 2012). Nevertheless Metzis et al. (2018), using an ESC-derived neurogenesis in vitro model, showed that the chromatin accessibility pattern of anteroposterior regional characteristics of the CNS primordia is established before neurogenesis initiates, indicating that early-stage anteroposterior specification is common between chicken and mouse embryogenesis. The model of the brain portions and the anterior spinal cord arising from a consecutive epiblast domain in mouse embryos without the involvement of NMPs, as in chicken embryos, has been supported by the presence of randomly labeled clones encompassing the brain and spinal cord without including somites (Tzouanacou et al., 2009). Additionally, the anterior region of the spinal cord must develop as a part of the anterior CNS without involving NMPs, as Tbx6 ^−/− mutant embryos display NMP-dependent conversion of the paraxial mesoderm to extra neural tubes only at axial levels posterior to the seventh somite (Chapman and Papaioannou, 1998; Takemoto et al., 2011). Thus, anterior CNS domain specification appears to proceed similarly between chicken and mouse embryos, except that the mouse CNS precursors bypass the step of cell convergence toward the midline. Thus, mouse EpiSCs provide an excellent model for the early-stage specification of the anteroposterior regionality of the brain domains.

EpiSCs enter the neurogenetic process after removal of activin/Nodal signaling (Brons et al., 2007; Tesar et al., 2007; Iwafuchi-Doi et al., 2012). On the first day after activin removal, EpiSCs enter the transitory state between the epiblast and neurogenic states (Iwafuchi-Doi et al., 2012). We thus placed activin-free EpiSCs in medium containing EGF and FGF2, which may have resulted in the establishment of NSC lines representing an early stage of CNS development. We also manipulated the Wnt signal levels delivered to the NSC-generating EpiSCs using a Wnt antagonist and a Wnt agonist. We also manipulated the Wnt signal levels delivered to the NSC-generating EpiSCs using a Wnt antagonist and a Wnt agonist.

The activin-free, EGF and FGF2-supplemented culture conditions quickly changed the EpiSCs into neural cell populations in 1 week (Figures 2, 6), from which NSCs developed as cells adherent to gelatin-coated culture dishes (Figure 2). NSC lines with different morphological characteristics and culture requirements developed in 4 weeks (Figure 2B; Table 1). These were, in fact, NSCs, as they gave rise to neurons and glial cells under appropriate culture conditions (Figure 2C).

Transcriptome analysis of the six NSC lines established under different Wnt signal inputs demonstrated that the characteristics of the NSC lines were ordered from anterior to posterior according to the Wnt signal input levels during the derivation of the cell lines. In the XN4 NSC line, where the Wnt signal input was the lowest with 10 µM XAV, Foxg1, characteristic of the anterior forebrain (Shimamura et al., 1997; Martynoga et al., 2002), was the prominent TF gene associated with low-level Otx2 and En2 expression (Figure 4A, Supplementary Figure S2). The XN1 line that developed under a lower concentration of XAV and hence a milder Wnt signal inhibition expressed Otx2, En2, Hoxa5 (and other Hoxa genes with the anterior expression limit in the hindbrain), indicating that the regional specificity of the XN1 line covers the mid-hindbrain region.

KN3 and KN1 lines were established under culture conditions without a Wnt antagonist or agonist. This condition provided a unique situation in which cells responded to endogenous Wnt3 signaling. EpiSC-endogenous Wnt signaling was moderate and failed to activate Bra expression, whereas high Bra expression was induced by exogenous CHIR at 2 µM (Figure 6A). In addition, the EpiSC-endogenous Wnt signaling level likely fluctuates, as demonstrated for human ESCs (Moya et al., 2014), which resemble EpiSCs in many respects (Brons et al., 2007; Matsuda et al., 2017). KN3 cells expressed En2 and cervical level-starting Hoxa genes more strongly than thoracic level genes and no Hoxb genes. In contrast, KN1 cells did not express En2 but expressed high levels of cervical level Hoxa5, 6, and 7 and anterior thoracic level-starting Hoxa7 and Hoxb genes (Figure 4A). However, the KN1 cells did not express posterior thoracic level-starting Hoxa9. These observations indicated that the regional specificity of the KN3 cells covers the hindbrain to the cervical spinal cord, whereas that of the KN1 cells covers the cervical to the anterior thoracic level spinal cord. The differences in the regional coverages, although overlapping, are accounted for by the fluctuation of cell-endogenous Wnt signaling among EpiSCs.

The CN1 and CN4 NSC lines were established in the presence of 2 µM CHIR, which overrides endogenous Wnt signaling. The Hox gene expression patterns of the CN1 and CN4 lines were similar, presumably reflecting an equivalent level of exogenous Wnt signaling (Figure 4A). These NSC lines expressed thoracic level-characteristic Hoxa9, Hoxa10, and Hoxc9 but did not express lumbar level-characteristic Hoxc10 or Hoxc11. These observations indicate that the CN1 and CN4 NSC lines share regional specificities covering mainly the thoracic spinal cord.

During the derivation of CN4 or any other NSC lines, activation of the Bra gene was not detected (Figure 6A), indicating that NMPs are not involved in the derivation of these lines. However, the possibility of a transient activation Bra during the sampling intervals was not formally excluded. Presumably, the thoracic level spinal cord is the posterior limit of the CNS portion coverable by the NSCs derived from the Otx2-expressing (N2 enhancer-active) state of the epiblast; by increasing the CHIR concentration, no NSC outgrowth was observed. Spinal cord cells at the lumber and more posterior levels may develop fully from the NMPs. Presumably, in the range of thoracic levels, the origin of the spinal cord tissues gradually changes from the Otx2-expressing epiblast to NMPs. Concerning the chicken embryo case shown in Figure 1D, the posterior limit of the Otx2-expressing epiblast-derived spinal cord precursors is marked by the distribution of N2 enhancer-labeled cells.

The anteroposterior regionalities of the above NSC lines already manifested after a week of culturing under the respective conditions (Figure 6Ac). These regionalities, once established, were maintained stably under culture conditions with different Wnt signal input levels, as shown in Figure 6B.

In contrast to the TF genes involved in the anteroposterior regional specification, those engaged in the dorsoventral specification in the neural tubes were expressed in a dysregulated fashion, presumably without any constraints (Figure 4), indicating that the NSCs were fixed at a stage earlier than when the neural plate begins to fold and the dorsoventral specification of the CNS cells starts. Interestingly, the expression of dorsal character TF genes, such as Olig3 and Pax3, was generally very low. As Wnt signaling promotes dorsal side regional specification (Muroyama et al., 2002), the Wnt agonist CHIR would have promoted the expression of such genes if NSCs represented postneural plate developmental stages.

These observations were strongly in support of the following model: 1) the EpiSC-derived NSC lines produced by immediate exposure to the NSC-generating condition represent the neural precursor cell stage shortly after leaving the epiblast state; 2) at that stage, the gross anteroposterior regionalities are specified; and 3) the Wnt signal input levels in the cells are crucial to specifying the anteroposterior regionality. Thus, the early-stage anteroposterior regionality specification before neural plate development, as observed in chicken embryos (Figure 2A), was successfully modeled in mouse EpiSC-derived NSCs.

The bulk transcriptome analysis of the NSCs summarized in Figure 4 indicated that individual cell lines were endowed with anteroposterior regional specificities reflecting the Wnt signal strength imposed during NSC derivation. However, the regional specificities had ranges. A critical question was whether these ranges reflect the composite of narrower specificity or reflect the character of individual cells.

The single-cell transcriptome analysis shown in Figures 7, 8 indicated that the NSC lines consisted of cells expressing TF genes of a given anteroposterior range. The study identified cells expressing one of the genes representing different regional characteristics within the regionality range and those expressing two or three such genes simultaneously. The statistical analysis indicated that the coexpression of multiple genes resulted from stochastic expression of individual genes without anteromedial or medioposterior regional correlations. A possible model to account for this observation is that individual cells have the potential to express all TF genes of the anteroposterior range characteristic of the cell line but express these genes transiently and stochastically, which may deserve further investigation.

Presumably, the characteristics of these NSC cells represent the initial phase of anteroposterior regionality specification in the neural primordium, although equivalent analyses have yet to be conducted using embryonic tissues. This study investigated six NSC lines that developed under discrete-level changes in Wnt signaling. In embryos, the Wnt signal levels will continuously change depending on the anteroposterior direction, and an array of neural primordia (possibly transient NSCs) bearing overlapping yet shifting regional specificities will be produced. These neural precursors may narrow their regional coverages in the cell population with slightly different regional specificities and during the course of neural plate formation.

The basic strategy in this study was to stabilize cells as NSC lines in a stage of neural development. We demonstrated that early-stage anteroposterior regionality is specified according to Wnt signal levels. In the mouse embryo CNS, various finer subdivisions are established by E11.5 (Bulfone et al., 1993; Shimamura et al., 1995; Shimamura et al., 1997; Gray et al., 2004; Berg et al., 2019; La Manno et al., 2021), which develop into functional CNS subdomains in adults (Thompson et al., 2014; Zeisel et al., 2018; Berg et al., 2019). These subdivisions will be generated under the intersections of various signaling pathways at different time points. The present approach may be widely useful in determining the elements of intersecting signaling and event chronology by changing the timing of exposure to the NSC-generating condition.

Anteroposterior regional specification reflecting differential Wnt signal input levels is not unique to neural plate precursors. Using human pluripotent stem cells, it has been shown that neural crest (NC) precursors are efficiently produced by 2 days of exposure to the Wnt antagonist CHIR, a condition for neural plate border cells (Thawani and Groves, 2020), and that NC cells develop after a 3-day incubation period (Leung et al., 2016; Gomez et al., 2019b; Heckland et al., 2019). Notably, lower-CHIR conditions enable the development of anterior characteristics, including Otx2 expression in NC cells, while higher CHIR conditions elicit trunk-characteristic Hox gene-expressing NC development (Gomez et al., 2019a; Heckland et al., 2019). In the latter case, NC cells develop via NMPs coexpressing Sox2 and Bra (Gomez et al., 2019a; Heckland et al., 2019), and combination with Fgf signaling enables the development of the posteriormost characters expressing Hox13 paralogs (Heckland et al., 2019). The combination of high Wnt and FGF signaling is the condition needed to activate the Sox2 N1 enhancer that regulates NMP cells (Takemoto et al., 2006; Takemoto et al., 2011).

The approach taken in our study was to stabilize the early stage of CNS precursors as NSC lines rather than promoting their progression into neural development. The time course of Bra and the Hox gene expression (Figure 6) indicated that the NSC lines likely did not pass through the NMP stage, possibly because these cells were incompatible with the EGF-FGF cosignaling condition. Overall, the anteroposterior regional specification processes appear to proceed analogously between the CNS and NC cells that develop in parallel in embryos, reflecting the Wnt signal input levels.

An important message derived from the study on human pluripotent stem cells by Leung et al. (2016) is that the Wnt signal has an additional role earlier than the anteroposterior regional specification of NC cells, i.e., to generate the NC cell lineage separable from the neural plate cell lineages in the presence of a sufficient level of the endogenous BMP signal; otherwise, human pluripotent stem cells develop into CNS cells (Leung et al., 2016). BMP signaling inhibits Sox2 expression and neural plate development (e.g., Takemoto et al., 2006). In contrast, the entire population of mouse EpiSCs was found to develop into CNS cells two days after activin removal (Iwafuchi-Doi et al., 2012), indicating that EpiSC-derived cells exist under low-BMP conditions. This possible difference in the endogenous BMP level accounts at least partially for the lack of NC traits in the EpiSC-derived NSC lines investigated in this study.

St. 4 chicken embryos (Hamburger and Hamilton, 1951) isolated on a ring-shaped filter paper support were electroporated using the electrode setting described by Uchikawa et al. (2017) with 1 µg/µl N2-tkmCherry [a modified version of N2-tkmRFP1 (Uchikawa et al., 2017)] or a Supernova vector cocktail consisting of 1 µg/µL pK031 (TRE-CRE-pA), 0.5 µg/µL pK038 (CAG-LoxP-Stop-LoxP-GFP-IRES-tTA-pA) (Mizuno et al., 2014), and 0.4 µg/µL pTK-CAG-tTA (provided by Y. Takahashi) with four pulses of 8 V with 50 ms duration at 1 s intervals using CUY21SC (NEPAGENE, Ichikawa, Japan). The embryos were cultured on agar-albumen medium with the ventral side upward (Uchikawa et al., 2003; Yoshihi et al., 2022b). Time-lapse embryo images were recorded at 10-min intervals for 18 h under an AF6000 inverted microscope (Leica, Wetzlar, Germany). The trajectories of the Supernova-labeled cells were analyzed according to the method detailed by Yoshihi et al. (2022a).

EpiSC lines provided by Paul Tesar (Tesar et al., 2007; Iwafuchi-Doi et al., 2012) and Y62 line established from DBF1 epiblasts according to Sumi et al. (2013) were cultured under feeder-free conditions (Iwafuchi-Doi et al., 2012) and used to establish the NSC lines (Table 1). The basic culture medium was 50% DMEM (Sigma‒Aldrich, D6421)/50% Neurobasal Medium (Invitrogen, 21103-049), 0.004% BSA, with 2 mM L-glutamine, 0.5× N-2 (R&D, AR003) and 0.5 × B-27 (Invitrogen, 17504-044) supplements. The EpiSCs were cultured in basic culture medium with the addition of 20 ng/mL Activin A (PEPROTECH, 120-14), 10 ng/mL FGF2 (PEPROTECH, 100-18B), and 2 µM XAV939 and on a NunclonΔ surface coated with 10 µg/mL fibronectin (BD, 354008). NSCs were cultured in medium containing 10 ng/mL EGF (PEPROTECH, 100-15) and 10 ng/mL FGF2 in dishes coated with 0.2% gelatin (Sigma). When needed, 10 µM XAV939, 2 µM CHIR99021 (STEMGENT, 100-15), or 10 µM Y-27632 (Calbiochem, 146986-50-7) was added to the culture medium diluting the stock solutions prepared in dimethylsulfoxide (Table 1 and Supplementary Table S1). XAV939 was obtained from two different sources, Sigma‒Aldrich (X-3004) and Selleck (S1180). To transfer the cells, Accutase (Sigma or Nacalai) was used to dissociate cells except for the KN3 line, where the cell sheets were mechanically dissociated into cell clumps (Table 1). RNA was extracted from the cultures as described in Inamori et al. (2020). To differentiate the NSCs into neurons or glial cells, the dissociated cells were plated on uncoated dishes and cultured in basic culture medium for 3 days for neuronal development, or in medium containing 1% fetal bovine serum for 5 days for glial cell development. The cultures were fixed with 4% paraformaldehyde in PBS for 1 h and immunostained using the following combinations of primary and secondary antibodies: Tuj1 mouse monoclonal IgG (R&D MAB1195) and Alexa Fluor 488-donkey anti-mouse IgG (Abcam, ab150109); anti-O4 mouse monoclonal IgM (Sigma O7139) and Alexa Fluor 555-goat anti-mouse IgM (Invitrogen, A-21426); anti-GFAP rabbit polyclonal IgG (Santa Cruz SC-6170) and Alexa Fluor 568-donkey anti-goat IgG (Abcam, ab175704).

FB, MB, HB, and CSC portions of the E10.5 CNS were isolated for comparison with the TF gene expression patterns of the NSC lines. The E10.5 stage was chosen because it is the developmental stage when the posterior-dominant Wnt expression profile is lost (Takemoto et al., 2016) and when the basic TF expression pattern in the CNS domains is established yet neurogenesis is minimal (Gary et al., 2004; Thompson et al., 2014; Andrews et al., 2019; Sagner and Briscoe, 2019). Two litters of E10.5 embryos of C57BL/6J mouse, with 5 and 6 embryos of the right stage, were dissected to isolate the embryonic portions containing FB, MB, HB, and the CSC (anterior to the forelimb bud). In PBS containing 10 μg/mL Dispase II (Wako) and 0.1% BSA, eyes, otic vesicles, and attached non-CNS tissues were removed. The separate CNS domains were pooled, and RNAs were extracted using TRIzol™ Plus RNA Purification Kit (ThermoFisher).

mRNA from biological replicates of the six NSC lines and four E10.5 CNS domains/portions were analyzed using SurePrint G3 Mouse GE Microarray 8 × 60K Ver. 2.0 (Agilent). The data were deposited in GEO with the accession number GSE237042. Principal component analysis of the data was performed using Subio software (Subio Inc., Amami, Japan). RT‒qPCR analysis of individual mRNA levels was performed as described in Inamori et al. (2020) using the primers described by Iwafuchi-Doi et al. (2012), except for Hoxb6 (Forward, TTC​CTA​TTT​CGT​GAA​CTC​CAC​CTT; Reverse, CCG​CAT​AGC​CAG​ACG​AGT​AGA) and Hoxa7 (Forward, ACG​CGC​TTT​TTA​GCA​AAT​ATA​CG; Reverse, GGG​TGC​AAA​GGA​GCA​AGA​AG) (Dickson et al., 2009).

The NSC lines XN4, XN1, KN1, and CN4 in the growing phase were dissociated using Accutase, passed through a 20 µm cell strainer, suspended in 10x volumes of CellCover (Anacyte Laboratories) (Akiyama-Oda et al., 2022), and stored at 4°C for up to 2 weeks until the cells were processed for single-cell transcriptome library preparation. After washing in 50% DMEM/50% neurobasal medium with 0.004% BSA, the cells were loaded on a BD Rhapsody cartridge (BD Biosciences) to capture mRNAs of cells on isolated magnetic beads. Single-cell RNA sequencing libraries were prepared for each NSC line using a BD Rhapsody WTA Amplification kit (633801) according to the manufacturer-supplied protocol. The libraries were sequenced using Hi-SeqX (Illumina) for 150 b paired-end reads to gain Fastq data. The sequence data were processed using the BD Rhapsody WTA pipeline on the Seven Bridge Genomics platform (Tzani et al., 2021). R1 reads were used to gain transcript data in individual cells, and R2 reads were used to align them on the mouse mm10 (GRCm38) genome sequence. Then, spreadsheet data were produced for the captured transcript counts of each gene in each cell. The Seurat package in RStudio (v. 4.3.0) (Satija et al., 2015) was used to compare the four NSC lines using plots of uniform manifold approximation and projection (UMAP). After removing ribosomal protein and mitochondrial gene reads, the data for 4000 randomly selected cells from each NSC line library were normalized; then, the data were merged to compare the four cell lines. The transcriptome profiles in individual cells were analyzed using Modern CSV (https://www.moderncsv.com/) and Excel (Microsoft). The area-proportional Venn diagrams were drawn using eulerAPE (Micallef and Rodgers, 2014). The original and processed data were deposited to GEO with the accession number GSE239383.