Cerebellar granular neuron progenitors exit their germinative niche via BarH-like1 activity mediated partly by inhibition of T-cell factor Free

The Wnt/β-catenin cell-to-cell signaling pathway coordinates development and is one of the most conserved in the animal kingdom. The vast majority of Wnt/β-catenin transcriptional targets are regulated by T-cell factor/lymphoid enhancer-binding factor 1 (TCF/LEF1) transcription factors (TFs) (reviewed by Bou-Rouphael and Durand, 2021; Hoppler and Waterman, 2014). Investigation of the developmental fate of Wnt/β-catenin-responsive cells in embryonic and postnatal mouse brains reveals that long-lived neural stem cells (NSCs) retain Wnt/β-catenin responsiveness throughout development (Bowman et al., 2013), and persistent TCF transcriptional activity appears to be a hallmark of long-lived NSCs (reviewed by Ding et al., 2020; Nusse, 2008; Sokol, 2011; Urbán et al., 2019). Besides Wnt and TCF, the Notch pathway and its downstream effectors, hairy and enhancer of split (Hes/Hey), that repress proneural gene expression are required for maintenance of NSCs, and a proper control of neurogenesis in both embryonic and adult brain (reviewed by Ables et al., 2011; Alunni and Bally-Cuif, 2016; Lampada and Taylor, 2023). Whereas TCF activation and Notch dependance appear to be hallmarks of both stem cell and cancer stem cell, transcriptional cross-regulations between the two pathways are poorly understood (Acar et al., 2021; Bayerl et al., 2021; Espinosa-Sánchez et al., 2020; Fendler et al., 2020).

A crucial component of the central nervous system (CNS) in all jawed vertebrates is the cerebellum, which is involved in executing motor functions as well as participating in higher cognitive processes such as decision-making, emotional and social behavior, and expectation of reward (reviewed by Reeber et al., 2013; Deverett et al., 2018; Carta et al., 2019; Haldipur et al., 2022). The cerebellum has two major stem cell niches: the ventricular zone (VZ) adjacent to the fourth ventricle, which produces all cerebellar GABAergic inhibitory neurons (Hoshino et al., 2005; Pascual et al., 2007; Yamada et al., 2014); and the upper rhombic lip (URL) which is the origin of glutamatergic excitatory neurons, derived from atonal homologue 1 (Atoh1)-expressing progenitors. In amniotes, the URL gives birth first to the deep cerebellar nuclei (DCN), followed by the unipolar brush cells (UBCs) and the granular neuron progenitors (GNPs), which, in turn, produce granule neurons (GNs), the predominant neuronal population in the entire CNS (Ben-Arie et al., 1997; Wang et al., 2005; reviewed by Leto et al., 2016; Lackey et al., 2018; Joyner and Bayin, 2022). Defects in the human URL developmental program lead to serious diseases (Hendrikse et al., 2022; Smith et al., 2022; reviewed by Haldipur et al., 2022). Once they are born, GNPs undergo a tangential migration on the surface of the cerebellar plate, giving rise to the external granule layer (EGL).

Amphibian cerebellum displays morphological features resembling those found in higher vertebrates (Gona, 1972; Herrick, 1914; reviewed by Hibi et al., 2017; Miyashita and Hoshino, 2022). Studies performed in the amphibian Xenopus at pre-metamorphosis stages reveal the presence of a non-proliferative EGL-like structure, which is unique compared with other anamniotes (Butts et al., 2014). These studies also indicate that the developmental processes that lead to the formation of GNs, specifically the presence of an Atoh1-expressing URL and EGL, and the expression of basic helix loop helix (bHLH) neurogenic differentiation factor 1 (Neurod1), which is a marker of GN differentiation, are close to those described in higher vertebrates (Butts et al., 2014; D'Amico et al., 2013). Moreover, the amphibian URL maintains itself until post-metamorphic stages (Butts et al., 2014; Gona, 1972).

In the cerebellar primordium, although the VZ appears to be TCF inactive, positive TCF transcriptional activity has been documented in the URL of mouse, human and Xenopus species (Borday et al., 2018; Garbe and Ring, 2012; Selvadurai and Mason, 2011; Wizeman et al., 2019). In the rodent EGL, Notch signaling maintains sonic hedgehog (SHH)-dependent GNP proliferation (Solecki et al., 2001; Machold et al., 2007; Adachi et al., 2021; reviewed by Miyashita and Hoshino, 2022). Although levels of Notch activity have been demonstrated to pattern the VZ in discrete subdomains (Khouri-Farah et al., 2022; Zhang et al., 2021), contributions of TCF and Notch to URL and early EGL biology, and their regulatory modes, have not been clearly identified.

ATOH1 directly induces its own expression as well as the expression of the two homeodomain (HD)-containing TFs, Barhl1 and Barhl2, which are mammalian homologues of the Drosophila Bar-class HD, BarH1 and BarH2 (reviewed by Bou-Rouphael and Durand, 2021; Reig et al., 2007). In the Xenopus organizer, Barhl2 binds to TCF, and enhances TCF repressor activity, thereby preventing β-catenin-driven activation of TCF target genes. Barhl2 maintains expression of TCF target genes repressed by a mechanism that depends on histone deacetylase 1 (Juraver-Geslin et al., 2011; Sena et al., 2019). In mice, Barhl1 and Barhl2 transcripts are detected in the outer URL and in the posterior EGL from E12.5 onwards (Aldinger et al., 2021; Bulfone, 2000; Kawauchi and Saito, 2008; Li et al., 2004; Mo et al., 2004). Barhl1 participates in the generation of the EGL (Kawauchi and Saito, 2008) and is one of the major TFs that regulate the radial migration of GNP in a mechanism involving neurotrophin 3 (NT3). Furthermore, an impairment in GN survival and an attenuated cerebellar foliation are observed in Barhl1^−/− mice (Li et al., 2004). Although Barhl2 is expressed in the amniote EGL (Mo et al., 2004), a potential role of Barhl2 in cerebellar development has not been investigated, and whether BARHL1 interacts with, and regulates, TCF transcriptional activity is unknown.

Here, using Xenopus as a model system, we investigated the role of Tcf gene activity, and Barhl genes in early GNP development. We establish that markers of GNP commitment and differentiation are conserved in Xenopus compared with amniotes, and confirmed the presence of atoh1-expressing EGL (Butts et al., 2014). We also establish that the URL is proliferative, Tcf active and expresses the hairy and enhancer of split genes hes4 and hes5. Whereas the EGL is non-proliferative and displays low Tcf gene activity, it expresses high levels of hes4 and hes5. Using gain- and loss-of-function approaches, immunoprecipitation, and a X. tropicalis Wnt reporter transgenic line, we demonstrate that Tcf-mediated transcriptional activation is strictly necessary for URL emergence and maintenance. We did not detect barhl2 in the Xenopus cerebellar anlage and focus our study on barhl1. We show physical interactions between Barhl1, Groucho (Gro), transducin-like enhancer of split (TLE) and Tcf isoforms. Barhl1 overexpression phenocopies premature inhibition of Tcf, whereas Barhl1 loss of function dramatically increases Tcf activity in the URL, leading to a major enlargement of the URL and to significant delays in GNP differentiation. Using a transcriptomic approach, we confirm that GNP depleted of Barhl1 stay in a proliferative, Tcf active and hes5-expressing state. We identify direct and indirect Barhl1 target genes in the cerebellar URL, and perform in silico analysis of Barhl1 target genes regulatory regions. Among the most upregulated target genes we identified were markers of Tcf activity and of neural stemness, and Hes and Hey genes. Downregulated genes were markers of neuronal differentiation. Our study establishes a Barhl TF as an inhibitor of Tcf activity and a repressor of some Hes and Hey genes that limits the size of a long-lived germinative niche.

In amniotes, the GNP developmental path is marked by expression of specific TFs, including Atoh1, the expression of which initiates in the rhombic lip (RL), is maintained in the EGL during GNP proliferation, and is lost in differentiated GN that start expressing Neurod1 (Flora et al., 2009; reviewed by Leto et al., 2016). In addition, expression of paired box protein 6 (Pax6) and Barhl1 occurs, both of which are markers of GNP commitment (Aldinger et al., 2021; Carter et al., 2018; Hanzel et al., 2019; Machold and Fishell, 2005; Miyata et al., 1999; Wizeman et al., 2019).

We performed in situ hybridization on X. laevis tadpoles through pre-metamorphic froglets, and assessed the expression of genes involved in the development of the Atoh1 lineage in rodents, focusing on GNs. We used the initiation of pax6 (Fig. S1C) and barhl1 (Fig. 1F) expression (stage 38) as a landmark of GNP induction. From stage 38 onwards, atoh1 is expressed in the URL and in a layer of three or four cells bordering the URL, which we considered part of the EGL (Fig. 1B; Fig. S1A). nmyc (mycn) is similarly expressed in the URL, but is also detected in the VZ (Fig. 1Ab,C; Fig. S1B). At both stages 42 and 48, hes5 and hes4 (El Yakoubi et al., 2012; Grbavec et al., 1998) strongly label part of the EGL (Fig. 1D). The dynamics of pax6, barhl1 and neurod1 expression within R1 from stage 38 to stage 48 reveals that they are first detected in the caudal region of the URL and EGL, and then in the cerebellar plate (Fig. 1E-G; Fig. S1C-E). At later stages, their expression spread within the inner cerebellar tissue, where they undergo their final differentiation (Fig. 1E-G; Fig. S1D,E). Although, in amniotes, orthodenticle homeobox 2 (otx2) expression is limited to the posterior lobes of the EGL, at stage 40 we detected otx2 expression in the caudal EGL, which was subsequently extended to the cerebellar plate (Fig. 1H). In rhombomere 1 (R1), we detected tcf7l1 transcripts mostly in the rostral URL and in part of the EGL, and, to a lesser extent, we also detected tcf7 transcripts (Fig. S1G). We did not observe expression of barhl2 in the cerebellar anlage at all developmental stages investigated (Fig. S1H).

These observations established a GNP development map that outlines the developmental progression of atoh1 lineage cells within R1 (Fig. 1Ac). It reveals strong similarities between Xenopus and amniotes in the expression of genes involved in induction of the URL and specification/differentiation of GNP. As previously reported (Butts et al., 2014), we detect an EGL along the R1 antero-posterior axis, which is marked by atoh1, hes4 and hes5.1. Our observations also reveal a gradient in GNP differentiation, initiated in the caudal R1 at stage 38 and progressing to the rostral part up to stage 50.

We focused our study on the role of Tcf and Barhl1 in URL establishment and maintenance, during the time window when GNPs are produced. Development of the URL and GNPs was investigated using atoh1, pax6, barhl1 and neurod1 as commitment and/or differentiation markers for the URL (atoh1), EGL (atoh1) and GNPs (pax6, barhl1 and neurod1). We used tcf7l1-Δβcat-GR, an inducible form of tcf7l1 that lacks its β-catenin-binding domain and thus acts as a inducible constitutive inhibitor of Tcf transcriptional activity (Molenaar et al., 1996) (Fig. S2).

At a high dose, tcf7l1-Δβcat-GR overexpression induced a dramatic reduction in the size of the URL, associated with the disappearance of the expression of its key marker: atoh1. This effect was restricted to R1 (Fig. 2Aa-a″,c). At a lower dose, tcf7l1-Δβcat-GR overexpression induced a decrease in atoh1 expression (Fig. 2Ab,c). This decrease was associated with both an increase of expression and a rostral shift, indicated by expression of the three commitment and/or differentiation markers pax6, barhl1 and neurod1 within R1 (Fig. 2Ba-d).

Similarly, barhl1 overexpression phenocopies Tcf inhibition (Fig. S2). barhl1 overexpression induced a decrease in atoh1 transcripts levels within the URL (Fig. 2Ca-a″,d), which was associated with an increase in pax6 and neurod1 expression, and with a rostral shift in the expression of both markers within R1 (Fig. 2Cb-d). In both cases, the loss of atoh1 expression was limited to the R1 territory with no effect on isthmic nuclei or atoh1 expression in R2 (Fig. 2Aa-a″,c,Ca-a″,c).

These data indicate, first, that Tcf transcriptional activity is necessary for the expression of atoh1 within the URL and, second, that, once GNPs are within the URL, premature inhibition of Tcf activity leads to an accelerated GNP differentiation. Similarly, overexpression of barhl1 in the cerebellar primordium results in URL induction defects that are associated with premature GNP differentiation.

To decrease Barhl1 activity within the cerebellar anlage, we designed and validated two morpholinos (MO), MObarhl1-1 and MObarhl1-2, specifically targeting Xenopus barhl1 mRNA (Fig. S2; Fig. S3; see Materials and Methods). We investigated whether MO-mediated Barhl1 knockdown (Barhl1-KD) alters development of the URL, the EGL and/or GNPs.

At all stage analyzed, depletion of Barhl1 induced an increase in atoh1 expression, with atoh1-expressing cells being expressed in the URL and spreading across the surface of the cerebellar plate (Fig. 3Aa-a″,Ba-a″; Fig. S3B-D). This expansion of the atoh1 expression territory is associated with an increase in nmyc expression (Fig. S3C), and a major decrease in both pax6 and neurod1 expression (Fig. 3Ab,Bb,C,Da-d; Fig. S3B,D). Furthermore, barhl1 overexpression rescued the decrease in neurod1 expression induced by Barhl1-KD (Fig. 3D).

We next asked whether inhibition of Tcf activity compensates for Barhl1-KD, using pax6 as a marker of GNP commitment. As previously observed, Barhl1-KD delayed GNP differentiation process, while tcf7l1-Δβcat-GR overexpression accelerated it (Fig. 3Ea-c). Indeed MObarhl1-1 co-injected with two different doses of tcf7l1-Δβcat-GR mRNA rescued the pax6 expression defects (Fig. 3Ed-f).

These results indicate that Barhl1 depletion maintains GNPs in an early progenitor state, and delays their differentiation. The Barhl1-KD phenotype is rescued by inhibition of Tcf activity, revealing that Barhl1 and Tcf act in opposing ways within the URL and the EGL.

We next asked whether Barhl1 directly controls Tcf transcriptional activity within the cerebellar URL. We investigated interactions between Barhl1, Groucho (Gro) 4 and Tcf isoforms, by performing co-immunoprecipitation experiments on protein extracts from HEK293 T cells transfected with tagged constructs of Tcf7l1, Tcf7l2, Tcf7, Gro4 and Barhl1 (Fig. 4A). In agreement with Barhl1 containing two engrailed homology 1 (EH1) motifs known to interact with the WD-repeat domain of Gro, Barhl1 co-immunoprecipitated with Gro4 (Fig. 4Aa). Using Tcf7l1 as bait, we observed that Tcf7l1 immunoprecipitated Barhl1, in the presence and absence of Gro4 (Fig. 4Ab). Finally, whereas Tcf7l1, Tcf7l2 and Tcf7 immunoprecipitated Barhl1, we observed that Tcf7l1 was the most efficient (Fig. 4Ac).

Concatemers of the consensus TCF-binding motif have been used to generate Wnt/TCF reporter lines, such as the Xenopus tropicalis (X. tropicalis) transgenic pbin7LefdGFP line (Borday et al., 2018; Tran and Vleminckx, 2014; Tran et al., 2010), which contains one copy of a wnt reporter gene. Using this reporter line, we assessed TCF activity from stage 42 up to stage 50, and observed positive TCF activity in the URL. Contrastingly, we did not detect any TCF activity in the VZ and the EGL at similar developmental stages. Noteworthy, TCF activity is stronger at the rostral end of the URL up to stage 48. This asymmetry of activity is lost at stage 50 (Fig. 4Ba-d). Similarly, atoh1 is mostly detected in the rostral URL up to stage 50, where it starts to be expressed throughout the URL (Fig. 4Ba′-d′). In addition, TCF activity partly overlaps the barhl1 expression domain (Fig. 4Bc″,c‴,d″).

We next assessed the impact of Barhl1 gain and loss of function on TCF activity (Fig. 4C). Whereas mouse Barhl1 overexpression decreased TCF activity (Fig. 4Ca), we observed a threefold increase in TCF activity upon Barhl1 downregulation with MObarhl1-1 (Fig. 4Cb,c). In contrast, MOct had no effect on TCF activity (Fig. S4A). We further inhibited Barhl1 by selective knockout (KO) of the Xenopus barhl1 gene in the pbin7LefdGFP line (F0 generation) using Crispr/Cas9 genome editing technology (Fig. S4B). We observed that Barhl1 KO induced, on average, a twofold increase in TCF transcriptional activity (Fig. 4Da-c). We assessed phenotypic penetrance in Crispr/Cas9-injected embryos. We observed different levels of phenotypic severities in over 70% of injected embryos, ranging from a slight increase in TCF activity observed in ∼20% of injected embryos to over 40% of injected embryos exhibiting strong to complete penetrance, as observed by a significant increase in TCF activity in R1 (Fig. 4d).

To determine whether the effects of Barhl1 were mediated through its interaction with Gro, we used an inducible form of Barhl2-EHsGR, which contains the two EH1 domains of Barhl2 (Fig. S2) and has been demonstrated to act as a dominant-negative for TCF repressive activity by competing for Gro binding (Sena et al., 2019). Overexpression of Barhl2EHsGR induced a phenotype similar to that of Barhl1-KD, an increase of the URL/EGL size at the expense of GNP commitment/differentiation (Fig. 4Ea,b; Fig. S4). In conclusion, Barhl1 directly interacts with TCF and Gro, and normally limits TCF transcriptional activity in the cerebellar primordium.

The URL germinative zone is characterized by its proliferative state and by its bordering of the roof plate. We asked whether the enlargement of the URL/EGL territories observed in Barhl1-KD tadpoles was corroborated by increased proliferation in the URL and/or within the EGL. Using immunofluorescence staining for phosphorylated-histone H3 (PHH3), a marker of cells undergoing mitosis, we followed proliferation in tadpoles injected with either MOct or MObarhl1-1 at stages 45 and 48.

In MOct injected tadpoles, PHH3⁺ cells were solely detected within the URL (Fig. 5Aa) (Butts et al., 2014). In contrast, we observed changes in the PHH3⁺ cell pattern in Barhl1-KD embryos, including a 1.2-fold lengthening of the URL on the injected side relative to the control side at both stages 45 and 48 that was associated with the presence of PHH3⁺ cells within the cerebellar plate (Fig. 5Ac-e). Because it was morphologically easier to distinguish the URL from the VZ at stage 48, we performed subsequent analysis at this later developmental stage (Fig. 5Ab). We measured, on average, a twofold increase in the number of PHH3⁺ cells on the injected side compared with the control side (Fig. 5Ac,d,f), and quantified the presence of ectopic proliferating cells in the cerebellar plate (Fig. 5Ac,d,g).

We next investigated whether TCF inhibition counteracted the Barhl1-KD effect on URL lengthening using in situ hybridization for nmyc, a marker of both the URL and the R1 caudal boundary (Fig. 5B). As previously observed, Barhl1-KD increased the URL length, whereas tcf7l1-Δβcat-GR overexpression reduced it (Fig. 5Ba-c,f). However, co-injection of MObarhl1-1 and tcf7l1-Δβcat-GR mRNA returned the size of the URL to normal (Fig. 5Bd-f). In conclusion, Barhl1-KD cells are compromised both in their ability to leave the URL niche and to become postmitotic. Barhl1-KD defects are at least partly due to over-activation of TCF.

To further document Barhl1 URL activity, we designed an RNA-sequencing experiment allowing the identification of Barhl1 direct and indirect target genes in the early Xenopus cerebellum. We isolated and sequenced RNA from dissected stage 42 R1 tadpoles injected with MObarhl1-1, MObarhl1-2 or MOct. Samples were compared through differential expression (DE) analysis (Fig. S5) (Table S1-S3). Principal component analysis of these R1 samples demonstrated that they clustered by Barhl1-KD status (Fig. S5B), indicating that changes in gene transcription were consistent across different clutches. We identified 1622 and 830 differentially expressed genes between, respectively, MObarhl1-1 and MObarhl1-2 injected R1, compared with MOct-injected R1, with 575 DE genes common to both MO-injected samples (Fig. S5C,D). A heatmap (Fig. 6A) and Volcano plots (Fig. 6B) representing upregulated and downregulated DEGs for both MOs are shown. Using the clusterProfiler algorithm (Wu et al., 2021), we performed gene ontology (GO) analysis, and compared altered biological functions between both Barhl1-KD conditions (Fig. 6C,D). This GO analysis reveals that the most significantly upregulated genes acted as transcriptional activators (Fig. 6C), whereas the downregulated DEGs are involved in neuronal differentiation (Fig. 6D).

We next investigated the presence of Barhl1 cis regulatory motifs (CRMs), defined as CAATTAC/G and its mirror motif (Chellappa et al., 2008), within the regulatory sequences – 5 kb or 30 kb upstream or downstream of the transcription start site (TSS) – of previously identified DEGs common to MObarhl1-1 and MObarhl1-2 conditions. For 5 kb, we observed that 71% of DEG regulatory regions contain at least one Barhl1 CRM, whereas 34% contain two or more Barhl1 CRMs (Fig. 6E; Fig. S5, Table S4). For 30 kb, all DEG regulatory regions contain at least two Barhl1 CRMs, 87.5% contain five or more Barhl1 CRMs and 33% contain 10 or more Barhl1 CRMs (Fig. 6E; Table S5). To investigate which Barhl1 target genes are also regulated by TCF we similarly searched for a TCF CRM defined as CTTTGAA/CTTTGAT within the regulatory sequences of previously identified DEGs common to both MObarhl1-1 and MObarhl1-2 (Kjolby et al., 2019; Nakamura et al., 2016). We observed that 76% of Barhl1-depleted DEG regulatory regions contain at least one Tcf CRM: 26% contain one CRM and 49% contain at least two Tcf CRMs (Fig. 6E; Fig. S5, Table S6). In agreement with our functional data, among the Barhl1-KD DEG we observed an upregulation of sp5, a bona fide direct target gene of TCF (Fig. 6A,B; Table S1-S3) (Wu et al., 2012). We also observed an upregulation of Wnt ligand secretion mediator (wls), a URL marker (Yeung et al., 2014) (Fig. 6A,B; Table S1-S3). Although both wnt2b and wnt8b are upregulated by Barhl1 KD their expression levels are very low in the cerebellar anlage (Fig. S5E; Table S1-S3). Barhl1 depletion thus activates TCF activity throughout the cerebellar anlage, specifically within the URL and the cerebellar plate.

Among the DEGs, we also observed an upregulation of Hes5 family genes (hes5.1, hes5.2, hes5.3 and hes5.4), and HES/HEY-like TF (helt) (also known as Heslike and Megane), which are downstream effectors in the Notch pathway (reviewed by Kobayashi and Kageyama, 2014). We also observed a downregulation of the Delta/Notch-like epidermal growth factor (EGF)-related receptor (dner), which has been suggested to be a neuron-specific Notch ligand (Eiraku et al., 2005) that inhibits neural proliferation and induces neural and glial differentiation (Hsieh et al., 2013). Using in situ hybridization, we validated cerebellar upregulated expression for hes5.1 in the EGL (Fig. 6Ga,d). We further asked whether hes5.1 upregulation observed in Barhl1 morphants depended on Notch signaling. MOBarhl1-injected tadpoles were grown either in carrier or in LY411575, a potent inhibitor of γ-secretase that is strictly necessary for intracellular transmission of Notch signals (Belmonte-Mateos et al., 2023). As expected, LY411575 induced a downregulation of hes5.1 expression levels within the cerebellum (Fig. 6Ge) (Jacobs and Huang, 2019; Myers et al., 2014). However, despite γ-secretase inhibition, Barhl1 depletion upregulated hes5.1 expression (Fig. 6Gb-d).

Our differential expression dataset also reveals that genes that are the most upregulated in the Barhl1-KD conditions are involved in adult neural stem cell (NSC) maintenance. For example, dmrta2 (which encodes doublesex and mab-3-related TF A2; also known as dmrt5), the orphan nuclear receptor subfamily 2 group E member 1 (nr2e1) [commonly known as Tailless (Tlx)] and zic3 [a member of the zinc finger of the cerebellum (Zic) family known to be involved in regulation of neuronal progenitor proliferation versus differentiation, and cerebellar patterning; reviewed by Aruga, 2004; Aruga and Millen, 2018; Houtmeyers et al., 2013] are all upregulated (Fig. 6A,B; Table S1-S3). Using in situ hybridization, we validated the upregulated cerebellar expression of zic3, transcripts of which are present in the URL, and of otx2, which is detected in a subset of GNPs at stage 41-42 (Fig. 6Gd; Fig. S5F). Finally, in agreement with our functional data, basic helix-loop-helix family member E22 (bhlhe22), a downstream target of Neuro D1, is downregulated in Barhl1-depleted R1 (Ma et al., 2022) (Fig. 6A; Fig. S5E; Table S1-S3).

Taken together, our transcriptomic analysis identifies direct and indirect Barhl1 target genes, and confirms our functional data. Our in silico search of Barhl1 target gene regulatory regions further validates the theory that Barhl1 partly acts by inhibiting TCF transcriptional activity. Our data also argue that Barhl1 directly or indirectly inhibits expression of Hes5 genes independently of Notch signaling.

This study, conducted in amphibian, establishes that: (1) TCF transcriptional activity appears necessary for inducing both atoh1 expression and the cerebellar URL; (2) there is a role for Hes4 and Hes5 genes in the EGL that is independent of proliferation; 3) Barhl1 is necessary for URL cells to exit their niche, partly through direct repression of TCF transcriptional activity; and 4) Barhl1 KD maintains GNP in a non-proliferative immature state, while increasing both TCF activity and Hes5 gene expression. Similarly, in amniotes, TCF activity is detected in the URL and switched off in the EGL. Notch target genes and Barhl1 and Barhl2 transcripts are detected in the amniote URL and EGL. Taken together, these observations argue for a conserved role of a Barhl protein in maintenance of the URL and/or EGL germinative niche via TCF and potentially Hes4 and Hes5 gene regulation.

Our data provide a developmental map of GNP development in Xenopus, revealing that the processes leading to the emergence of URL derivatives and maturation of GN are similar to those seen in higher vertebrate (Gona, 1972; Herrick, 1914) (reviewed by Hibi et al., 2017; Miyashita and Hoshino, 2022). Our analysis confirms the absence of proliferating cells within the EGL (Butts et al., 2014), and shows expression of hes5 and hes4, markers of stem/progenitor cells, in the EGL. Our data argue that, at least in amphibians, Hes genes are involved in maintaining GNPs in an immature state, independently of proliferation.

Our data indicate that TCF transcriptional activity is necessary for induction of atoh1 expression, and of the URL territory. Studies performed in mouse neuroblastoma and in neural progenitor cells in culture identified two TCF/LEF-binding sites in the 3′ enhancer region of Atoh1 that are required for Atoh1 activation (Shi et al., 2010). In these cells, the concomitant inhibition of Notch signaling and activation of WNT/TCF are required for Atoh1 expression (Shi et al., 2010). In mouse, low levels of Notch activity are necessary to induce a URL fate (Khouri-Farah et al., 2022; Zhang et al., 2021). These data argue that concomitant TCF activation and Notch inhibition are responsible for Atoh1 URL induction. Once activated, Atoh1 directly induces its own expression, which maintains the URL territory (Klisch et al., 2011).

Both tcf7l1 and tcf7 transcripts are detected in the amphibian cerebellar anlage and could mediate Wnt signaling in this germinative niche. Of note, Wnt canonical ligands present in the MHB, such as Wnt1, and in the roof plate, specifically Wnt3 and Wnt3a, may participate in TCF activation within the URL (Fig. S5E) (Roelink and Nusse, 1991). Whether TCF is activated in a Wnt-dependent manner in the URL at the analyzed stage has not been proven.

Three out of the four Tcf isoforms (Tcf7l2, Tcf7 and Lef1) mostly act as transcriptional activators, whereas the fourth (Tcf7l1) mostly acts as a transcriptional repressor (Liu et al., 2005) (reviewed by Arce et al., 2006). Transcriptomic analyses of human cerebellar development reveal that the transcriptional activator TCF7 is active in the human URL (Aldinger et al., 2021), whereas Tcf7l2 is detected in the mouse URL (Carter et al., 2018). In both species, Tcf7l1 is associated with differentiated GNs (Aldinger et al., 2021; Wizeman et al., 2019). Therefore, our findings in amphibian are relevant for amniote URL biology.

Barhl1 depletion and KO dramatically increases TCF transcriptional activity in the URL and/or EGL. Both the increase in URL length, and the delay in GNP commitment and/or differentiation induced by Barhl1 depletion, are fully compensated for by co-expression of a constitutive inhibitory form of Tcf7l1. Our R1 transcriptomic analysis of Barhl1-KD embryos reveals an increase of TCF activity. Seventy-five percent of Barhl1-depleted DEG regulatory regions contain at least one TCF CRM. These include markers of the URL and EGL, and wls, which in mice has been described to orchestrate cerebellum development (Yeung and Goldowitz, 2017; Yeung et al., 2014). Barhl1 KD significantly upregulated genes act by regulating the fine equilibrium between the proliferative state and commitment, and/or by maintaining the stem/progenitor features. In rodents, dmrta2 (dmrt5) expression maintains NSC self-renewing ability and is transcriptionally regulated by TCF (Konno et al., 2012; Young et al., 2017). The primary function of the orphan nuclear receptor Nr2e1 (also known as Tlx) is to maintain NSC pools in an undifferentiated self-renewing state (Kandel et al., 2022) (reviewed by Islam and Zhang, 2015; Wang et al., 2013). otx2 expression is associated with a high GNP proliferation rate (El Nagar et al., 2018; Fossat et al., 2006), and zic3 is involved in maintaining pluripotency in both embryonic stem cells (ESCs) (Lim et al., 2007, 2010) and neural progenitor cells (Inoue et al., 2007). Finally, most downregulated DEGs, including Bhlhe2, are involved in terminal neuronal differentiation, which, in mice, is a regulator of post-mitotic GN radial migration towards the IGL (Ma et al., 2022; Ramirez et al., 2021 preprint).

The current model of Wnt canonical pathway activation is that β-catenin promotes transcription of Wnt-responsive gene by either displacing Gro (Daniels and Weis, 2005; Roose et al., 1998) or displacing the whole Tcf7l1-Gro repressor complex and replacing it with an activator complex containing β-catenin in association with Tcf7 (Hikasa et al., 2010; Roël et al., 2002). Importantly, this switch from repression to activation of transcription for canonical Wnt pathway target genes depends on the stability of the Tcf7l1-Gro repressor complex, which promotes compaction of chromatin when the Wnt/β-catenin pathway is switched off (Shy et al., 2013). We propose that, in the amphibian cerebellar URL, activation of Tcf7 and/or Tcf7l1 participates in atoh1 expression and initiates the GNP developmental program. A gradient of GNP differentiation initiated in the caudal R1 and progressing to the R1 rostral region is established. Once activated, Atoh1 directly induces barhl1 transcription in a caudal R1 URL subpopulation (Kawauchi and Saito, 2008; Klisch et al., 2011). There, Barhl1 binds to Tcf7l1 and/or Tcf7, thereby preventing their further activation and switching off the GNP ‘URL’ program. GNPs then move out of the URL and progress towards commitment and differentiation. The role of Barhl1 in GNP development, through its inhibition of Tcf activity, is similar to Barhl2 activity during Spemann organizer formation (Sena et al., 2019). In silico analysis performed on active medulloblastoma enhancers, together with our previous study on Barhl2, indicate that Barhl1 and Barhl2 have long-range activity via their specific binding on DNA, perhaps on super-enhancers and via chromatin modifications (Lin et al., 2016; Sena et al., 2019). Indeed, all DEGs containing TCF CRMs also contain at least two Barhl1 CRMs 30 kb upstream or downstream of their TSS. A more-thorough study is necessary to understand how TCF and Barhl1 interact in vivo and regulate transcription within the cerebellar primordium. Our study does not differentiate gene regulation by Barhl1 alone or via its interaction with Tcf.

Depletion of Barhl1 also leads to a significant upregulation of hes5 isoforms and helt, which all directly repress proneural genes expression (Ables et al., 2011; Alunni and Bally-Cuif, 2016; Giachino and Taylor, 2014; Imayoshi et al., 2010; Kageyama et al., 2019; Lampada and Taylor, 2023). Importantly γ-secretase inhibition in Barhl1 morphants does not prevents hes5 upregulation. Gro/TLE, the main co-repressor partner of TCF, also participates in Notch signaling (reviewed by Bou-Rouphael and Durand, 2021; Cinnamon et al., 2008). Taken together, our results argue that, within the cerebellar anlage, Barhl1 transcriptionally inhibits hes5 and/or helt independently of Notch signaling, possibly via the recruitment of a protein complex containing Gro and/or Tle. Of note Hes5 is not expressed in rodent EGL (Khouri-Farah et al., 2022), and Hes4, a marker of stemness in Xenopus (El Yakoubi et al., 2012), is not expressed in mice, although its expression is found in human. Whether, in amniotes, a HES-related protein maintains some GNPs in an immature state remains to be investigated.

In rodents, ATOH1 directly induces the expression of both Barhl1 and Barhl2 (Kawauchi and Saito, 2008; Klisch et al., 2011). scRNA-sequencing analysis of mouse cerebellar Atoh1 lineage cells reveals that Barhl1 is associated with early stages of GNP differentiation, whereas Barhl2 expression is uniquely associated with early fate commitment (Carter et al., 2018). Barhl1 and Barhl2 are highly conserved through evolution (reviewed by Bou-Rouphael and Durand, 2021). Their functional conservation is evidenced through studies in various species, including mouse, C. elegans and the acorn worm Saccoglossus kowalevskii (Schwartz and Horvitz, 2007; Yao et al., 2016). Mice carrying a KO insertion within the mouse Barhl2-coding frame die before P24 after, among other signs, defective weight gain and impaired motor coordination that reflects cerebellar deficiency (Ding et al., 2009). Although a role for one of the Barhl TFs in the emergence and maintenance of the URL germinative niche in amniotes is predictable, Barhl2 appears to be a more suitable candidate than Barhl1 for this function.

X. laevis embryos were obtained by conventional methods of hormone-induced egg laying and in vitro fertilization, and were staged according to Nieuwkoop and Faber (1994). X. tropicalis transgenic Wnt reporter pbin7LefdGFP was generated as previously described (Borday et al., 2018; Tran and Vleminckx, 2014; Tran et al., 2010). Briefly, the synthetic Wnt-responsive promoter consists of seven copies of TCF- and/or LEF1-binding sites, and a TATA box driving destabilized green fluorescent protein (eGFP) and a polyA sequence. gfp expression reveals Wnt and/or TCF activity. X. tropicalis embryos were obtained by in vitro fertilization. Experimental procedures were specifically approved by the ethics committee of the Institut de Biologie Paris Seine (IBPS) (Authorization 2020-22727 given by CEEA 005) and have been carried out in strict accordance with the European community directive guidelines (2010/63/UE). B.C.D. carries the authorization for vertebrate experimental use (75-1548).

mbarhl1-HA-GR contains the full-length mbarhl1 sequence [two engrailed-homology (EH1) motifs, nuclear localization signal (NLS); homeodomain (HD); and the C-terminal region], followed by a HA tag in the C-terminal domain. This construct is inducible as it contains a glucocorticoid receptor that can be activated by dexamethasone (10 μM). Dexamethasone-inducible mbarhl2-EHs-GR contains the first 182 amino acids of mouse Barhl2 full-length cDNA, which correspond to the N-terminal EHs Gro-binding domains, and has been shown to act as a dominant negative (Sena et al., 2019). The full-length mbarhl1-HA-GR and truncated mbarhl2-EHs-GR constructs were generated in pCS2+ by Vector Builder. Non-inducible mbarhl1-Myc and xbarhl1-Flag were generated by GeneScript. Peptide sequences of the tags used are as follows: HA (YPYDVPDYA); FLAG (DYKDDDDK) and MYC (EQKLISEEDL). The constitutive repressor pCS2-Tcf7l1-Δβcat-GR was a gift from H. Clevers (Molenaar et al., 1996) and consists of the full-length Tcf7l1 lacking the β-catenin-binding domain (BCBD), which reinforces its repressive activity. Constructs used for immunoprecipitation assay are pCS2+ mbarhl1-3xFlag-HA, which was generated by Vector Builder. It contains the full-length mbarhl1 sequence followed by three Flag tags and one HA tag in the C-terminal region. pCS2+ Myc-Tcf7l1, pCS2+ Myc-Tcf7l2, pCS2+ Myc-Tcf7, pCS2+ Flag-Gro4 and pCS2+GroHA have been previously described (Liu et al., 2005; Sena et al., 2019). All necessary sequences were obtained from the NCBI database. Constructs were validated by western blot on extracts from injected embryos or cell lysates.

Capped messenger RNAs (mRNAs) were synthesized using the mMessage mMachine kit (Invitrogen) and resuspended in RNAse-free water. Antisense morpholino oligonucleotides (MOs) were generated by Gene Tools. ATG start-site MObarhl1-1 and MObarhl1-2 were designed to block initiation of xBarhl1 protein translation. The MOs were designed in a region overlapping the translation initiation site, so that they did not recognize mouse Barhl1 or xbarhl2 mRNA (Fig. S3). To establish the specificity of the MO effect, we tested the ability of MObarhl1-1 and MObarhl1-2 to specifically inhibit translation of xbarhl1 mRNA. Flag-tagged xbarhl1 (xbarhl1-flag) or myc-tagged mBarhl1 (mBarhl1-myc) were co-injected with MObarhl1-1, MObarhl1-2 or a control MO (MOct) (Fig. S2). Western blot analysis on extracts from injected embryos confirmed a MObarhl1-mediated dramatic decrease in Xenopus Barhl1 protein levels, whereas MOct had no effect. We also observed that MObarhl1-1 did not decrease mBarhl1-myc protein levels (Fig. S2; Fig. S3). MObarhl1-1 was used for both X. laevis and X. tropicalis as the mRNA sequence of barhl1 is highly conserved between both species, more specifically in the region in which MObarhl1 is hybridized. A standard control MO from Gene Tools was used in this study. MO sequences and doses are summarized in Table S7.

Xenopus embryos were injected unilaterally in one dorsal blastomere at the four- and eight-cell stage together with gfp as a tracer for phenotype analysis by in situ hybridization, except for CRISPR/Cas9 genome editing and RNAseq analysis (see corresponding sections in the Materials and Methods). MOs were heated for 10 min at 65°C before use. Injected embryos were transferred into 3% Ficoll in 0.3×Marc's Modified Ringer's (MMR) buffer (stock solution: 1 M NaCl, 20 mM KCl, 20 mM CaCl₂, 10 mM MgCl₂ and 50 mM HEPES at pH 7.4). 10 nl of mRNA or MO solution was injected together with a tracer in X. laevis, while 5 nl were injected in X. tropicalis. In X. laevis, MOs or mRNAs were co-injected with gfp mRNA (100 pg). MOs or mRNAs were co-injected with mcherry (100 pg) in X. tropicalis. Concentrations of injected mRNA and MOs per embryo were optimized in preliminary experiments. The minimal mRNA or MO quantity that induced the specific phenotype without showing toxicity effects was used. For embryos injected with inducible constructs, half of the injected embryos were treated with 10 μM dexamethasone at stage 35/36 while the other half were left untreated and served as controls. All necessary Xenopus sequences were obtained from Xenbase.

Embryos were staged according to Nieuwkoop and Faber (1994) and collected at the desired stage, then fixed in 4% PFA for 1-2 h at room temperature and dehydrated in 100% methanol. In situ hybridization was carried out using digoxigenin (DIG)-labeled probes. Antisense RNA probes were generated for the following genes: atoh1, barhl1, hes4, hes5.1, neurod1, pax6, nmyc, otx2, tcf7l1, tcf7l2, tcf7, lef1, zic3, wnt2b, wnt8b and gfp according to the manufacturer's instructions (RNA Labeling Mix, Roche). pCS2-Gfp was a gift from David Turner (University of Michigan, Ann Arbor, MI, USA; Addgene #200393). pBSK+xBarhl1 was a gift from Roberto Vignali (Unità di Biologia Cellulare e dello Sviluppo, Pisa, Italy). pCS2-Atoh1 was a gift from G. Schlosser (University of Galway, Ireland). pBSK+Wnt2b was a gift from S. Sokol (Icahn School of Medicine at Mount Sinai, NY, USA). pBSK+Wnt8b was a gift from J. Christian (University of Utah, USA; Addgene #16862). Tcf isoforms in pCS2 were provided by S. Hoppler (Institute of Medical Sciences, University of Aberdeen, Aberdeen, UK). In situ hybridization was processed following the protocols described by El Yakoubi et al. (2012) and Sena et al. (2019). Double in situ hybridization was processed as described by Juraver-Geslin et al. (2014). For X. laevis embryos, after rehydration, the eyes and ectoderm overlying the anterior neural tube were removed, which allows the omission of further proteinase K (PK) treatment. Dissections were not performed on X. tropicalis embryos that were treated with PK. In both cases, bleaching was carried out, and samples were incubated with the probes overnight. Alkaline phosphatase-conjugated anti-DIG or anti-FLUO antibodies (Roche) were incubated for 3 h at room temperature. Enzymatic activity was revealed using NBT/BCIP (blue staining) and INT/BCIP (red staining) substrates (Roche). After in situ hybridization, post-fixation was carried out in PFA 4% and the neural tubes of control and injected X. laevis embryos were dissected in PBS-0.1% Tween and stored in 90% glycerol. X. tropicalis embryos were stored in 4% PFA. Dissected neural tubes or embryos were photographed on a Leica M165 FC microscope equipped with Leica DFC320 camera using the same settings to allow direct comparison. Dorsal and lateral views of the dissected neural tubes were photographed.

Embryos were treated either with 25 μM of the γ-secretase inhibitor LY411575 (Sigma-Aldrich) or with DMSO for controls and added to Xenopus culture environment for 24 h.

Immunofluorescence was carried out as previously described (Juraver-Geslin et al., 2011). The entire brains of wild-type and MO-injected X. laevis embryos were carefully dissected and transferred into a tube containing PBS-0.1% Tween, where they were progressively permeabilized. Samples were incubated with primary antibody (anti-phospho-histone H3; Upstate Biotechnology, 06–570; 1:500) at 4°C overnight. Cellular nuclei were stained with BisBenzimide (BB) (Sigma), which was added to the solution containing diluted secondary antibody (Alexa Fluor 488 donkey anti-rabbit IgG; Invitrogen, A-21206; 1:500) and incubated at 4°C overnight. Neural tubes were captured on a Zeiss Axio Observer.Z1 microscope equipped with apotome. Acquisitions were taken using the z stack tool from the most superficial layer to deeper layers.

HEK293T cells were originally obtained from ATCC (CRL-3216), verified to be free of mycoplasm and cultured in supplemented Dulbecco's modified Eagle's medium (DMEM) (Gibco). Cells were transfected with expression vectors for pCS2-mbarhl1-3xFlag-HA, pCS2-mbarhl1-Myc, pCS2-Tcf7l1-Myc, pCS2-Tcf7l2-Myc, pCS2-Tcf7-Myc pCS2-Gro-Flag and pCS2-Gro-HA (gifts from S. Hoppler) encoding tagged proteins using the phosphate calcium method. Plasmids encoding pCS2+ or pSK+ were used as a supplement to ensure that cells in different dishes were transfected with the same quantity of expression vectors and plasmids (a total of 2 μg). Thirty-six hours post-transfection, cells were harvested and lysed in ice-cold lysis buffer (20 mM Tris at pH7.6, 150 mM NaCl, 1% Triton and 1 mM EDTA) supplemented with completeTM protease inhibitor (Roche). Cell lysates were centrifuged for 15 min at 16,900 g. Protein complexes were precipitated from the cell lysates with anti-c-Myc antibody (clone 9E10). Protein complexes were then precipitated with protein A-sepharose beads (Sigma) pre-washed with lysis buffer. Immunoprecipitated proteins were eluted from protein A beads by heating beads in Laemmli sample loading buffer (BioRad).

Western blot (WB) analysis was performed on protein extracts from injected and/or wild-type Xenopus embryos, and on extracts from transfected HEK923 T cells. Xenopus embryos were injected with mbarhl1HAGR, xbarhl1Flag, mBarhl1Myc or mBarhl2EHsGR mRNA at the two-cell stage, targeting both blastomeres. Proteins were extracted at stage 10 with lysis buffer (10 mM Tris-HCl at pH 7.5, 100 mM NaCl, 0.5%NP40 and 5 mM EDTA supplemented with a cocktail of protein inhibitors). Western blotting was carried out using conventional methods. Proteins were separated by 10% SDS-polyacrylamide gel and transferred to nitrocellulose membrane. Membranes were blocked using 5% milk and incubated with the corresponding primary and secondary antibodies diluted in 5% milk (summarized in Tables S8 and S9). Proteins were detected with Western Lightning Plus-ECL (Perkin Elmer Life Sciences). Membrane stripping was carried out between two staining steps using stripping buffer (Thermo Scientific) for the removal of primary and secondary antibodies from the membranes. ChemiDoc MP Imaging System (BioRad) was used for imaging the blots.

Three CRISPR target sites (barhl1-1, GAGTCGGACGAGGCCATGGAAGG; barhl1-2, ACCAGCTCTGTGCGACAGAATGG; and barhl1-3, AGAGTTGGACTCCGGGCTGGAGG) cutting respectively at 2, 37 and 230 bp from the beginning of the coding sequence were selected for their high predicted specificity and efficiency using CRISPOR online tool (http://crispor.tefor.net/). Alt-R crRNA and tracrRNA were purchased from Integrated DNA Technologies (IDT, Coralville, IA, USA) and dissolved in duplex buffer (IDT) at 100 µM each. cr:tracrRNA duplexes were obtained by mixing equal amounts of crRNA and tracrRNA, heating at 95°C for 5 min and allowing the mixture to cool down to room temperature. The gRNA:Cas9 RNP complex was obtained by incubating 1 µl of 30 µM Cas9 protein (kindly provided by TACGENE, Paris, France) with 2 µl cr:tracrRNA duplex in a final volume of 10 µl of 20 mM Hepes-NaOH (pH 7.5) and 150 mM KCl for 10 min at 28°C. X. tropicalis one-cell stage embryos were injected with 2 nl of gRNA:Cas9 RNP complex solution and were cultured to the desired stage. For co-injection, the three complexes were mixed in equal quantities.

Single embryo genomic DNA was obtained by digesting for 1 h at 55°C in 100 µl lysis buffer (100 mM Tris-HCl at pH 7.5, 1 mM EDTA, 250 mM NaCl, 0.2% SDS and 0.1 µg/µl Proteinase K), precipitating with 1 volume of isopropanol and resuspending in 100 µl PCR-grade water. The region surrounding the sgRNA binding sites was amplified by PCR using X. tropicalis Xt_barhl1_F (CAGCTCCTCCGACTTTTGTG) as a forward primer and Xt_barhl1_R (GTTGCCCGTTGCTGGAATAA) as reverse primer. CRISPR efficiency was assessed using a T7E1 test (Mashal et al., 1995) on mono-injected embryos and by detecting deleted fragments on co-injected embryos.

X. laevis embryos were injected with three different conditions: MObarhl1-1, MObarhl1-2 and MOct in the two dorsal blastomeres at the four-cell stage. At stage 42, neural tubes were extracted in RNAse-free conditions, and rhombomere 1, which includes the URL was carefully dissected. For each condition, three biological replicates were collected. Each replicate contains three rhombomeres, which was the optimal number needed to achieve the minimal RNA concentration required for this experiment (total RNA concentration was ∼30 ng per sample). Briefly, total RNA was extracted using the TRIzol reagent (Ambion) according to the manufacturer's instructions. The overall RNA quality was assessed using Agilent High Sensitivity RNA ScreenTape System. Samples with an RNA Integrity Number (RIN) greater than 9 were used for subsequent analysis. Poly A RNA was purified. Sequencing was performed using anIllumina NovaSeq (paired-end sequencing) by Next Generation Sequencing Platform (NGS) (Institut Curie). RNAseq data processing was performed using the Galaxy server of ARTBio platform (IBPS). The raw RNA-sequencing data and a brief description of the method are available at https://www.ebi.ac.uk/ena/browser/view/PRJEB64149, secondary accession ERP149284 (Role of barhl1 in the amphibian cerebellum).

Datasets were aligned against the X. laevis v10.1 genome assembly downloaded with its corresponding annotation file from Xenbase (Fortriede et al., 2020). Alignment was carried out using two read mapping programs: STAR v2.7.8a (Dobin et al., 2013) and HISAT2 v2.2.1 (Kim et al., 2015). Quality control checks were assessed using FastQC v0.73 (Andrews, 2010) and are summarized in a single report generated by MultiQC v1.9 (Ewels et al., 2016). As both alignment programs provided comparable results, we proceeded with STAR alignment tool. The number of aligned reads was counted by featurecounts tool v2.0.1 (Liao et al., 2014). Finally, we used the DESeq2 v2.11.40.6 package (Love et al., 2014) to determine differentially expressed genes (DEGs) from count tables. In the present study, genes with adjusted P value pAdj<0.001 were selected as significant DEGs. Venn diagrams were produced with JVenn v2021.05.12 (Bardou et al., 2014). Volcano Plots v0.0.5 were generated to show significantly upregulated and downregulated genes, only a selection of DEG names are represented.

Further analysis and data visualization were performed using R v4.2.1package. A heatmap was generated to visualize gene expression across the samples. To overcome the lack of Xenopus gene ontology (GO) annotation, we replaced X. laevis gene symbols with the human orthologs. Functional enrichment analysis was performed using the compareCluster function of ClusterProfiler v4.8.1 (Yu et al., 2012) to identify GO-term enrichment among DEGs with pAdj<0.001 as threshold. It provides the biological processes, cellular components and molecular functions of DEGs, and compares each of the three subgroups between both knockdown conditions.

For in situ hybridization performed on embryos injected unilaterally, comparison of the expression levels between injected and control sides was assessed using a specific macro from ImageJ v2.1.0/1.53c (Abràmoff et al., 2004; Schneider et al., 2012). The macro functions were based on the RGB color mode. RGB images are split into three channels (red, green and blue) and pixel values corresponding only to the blue channel are recorded, excluding the red and green channels, since the signal recorded on the blue channel represents the expression levels. For each image, the region of interest (ROI) was specified and its dimensions were fixed, such that the same ROI is placed on the control and injected side of the embryo, which prevents any subjectivity in ROI determination. The following were measured: the area corresponding to the blue signal; the mean or average value of signal within the selected ROI; and the integrated density, which is the equivalent of the product of area and mean, as it sums the values of pixels in the selection. In this study, the ratio of integrated density measured in the injected versus control side was assessed. The macro used is available at https://zenodo.org/records/12205788 and https://github.com/BeaDurand/Macro_quantificationISH.

The same macro was used for the analysis of CRISPR/Cas9-injected embryos, except that the ROI was placed completely in rhombomere 1 as the entire embryo was targeted. The mean integrated density values of control embryos were compared with each individual integrated density value of control and injected embryo. Phenotype penetrance was evaluated by counting and classifying embryos based on the intensity of gfp expression increase.

For immunofluorescence, z-stack images were reconstructed and processed using ImageJ v2.1.0/1.53c. PHH3-positive cells were counted and the length of the RL was measured on the control and injected side. Ratio of PHH3-positive cells and RL length in the injected versus control side were measured.

For the same experiment, all images were acquired using the same magnification and camera settings. In this way, all images were processed in a standardized manner, such that results are objectively analyzed. Final images were processed with Adobe Photoshop (v24.00).

Three independent experiments were performed for each condition analyzed. Dissected neural tubes and embryos were analyzed individually, and the results were pooled for data representation. Statistical analyses were implemented with R. Normality in the variable distributions was assessed using the Shapiro-Wilk test. Furthermore, the Levene test was performed to probe homogeneity of variances across groups. Variables that failed the Shapiro-Wilk or the Levene test were analyzed with non-parametric statistics using the one-way Kruskal–Wallis analysis of variance on ranks followed by Nemenyi test post-hoc and Mann–Whitney rank sum tests for pairwise multiple comparisons. Variables that passed the normality test were analyzed by means of one-way ANOVA followed by Tukey's post-hoc test for multiple comparisons or by unpaired Student's t-test for comparing two groups. P<0.05 was considered to be statistically significant. Results are presented as the mean± s.e.m. The statistical tests used are described in each figure legend.

The authors gratefully acknowledge M. Perron, D. Turner, P. Kreig, J. Christian, R. Vignali and G. Schlosser, P. Walentek, S. Hoppler, and M. Salonna for gifts of materials. We thank the IBPS Aquatic platform supported by Sorbonne Université, the CNRS, the IBISA and the Conseil Régional Ile-de-France, specifically S. Authier and M. Abdelkrim, for Xenopus care. We thank C. Antoniewski (ArtBio platform, FR3631 IBPS) for tutoring in RNAseq data analysis and the team of G. Almouzni (Institut Curie) for providing a high-sensitivity RNA screentape system. We acknowledge Ann Lohof and Rachel Sherrard for editing work on the manuscript. We thank A. El Helou for technical help.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202234.reviewer-comments.pdf