To gain new insights into the transcriptional targets of Notch signalling, we manipulated Notch activity in embryonic chicken otocysts and analysed the resulting changes in gene expression levels with RNA-Seq. To stimulate Notch activity (gain-of-function, or GOF), we electroporated in ovo the right otic cup of E2 (stage HH12-14) chicken embryos with a plasmid encoding the intracellular domain of the chicken NOTCH1 receptor (NICD1) (Figure 1A). Control embryos were electroporated with a monomeric red fluorescent protein (mRFP1) expression construct. Next, RNA was isolated from both transfected (right) and untransfected (left) otocysts of 3 embryos for each of the following conditions: Notch GOF 6 h post-electroporation (post-EP), Notch GOF 24 h post-EP, and Control 24 h post-EP. For blocking Notch activity (loss-of-function, or LOF), we treated E2.5 otocysts in vitro with the γ-secretase inhibitor LY411575 (GSI) for 6 and 24 h (Figure 1B). Three embryos were used at each time point, with the right otocysts treated with 10 µM LY411575 whilst the left ones were kept in medium supplemented with DMSO at matching concentration as a control. At the end of the treatment, RNA was extracted from each cultured otocyst and processed for RNA-Seq, then analysed using the Kallisto and Sleuth packages for differential gene expression analysis.

To assess the effectiveness of the different treatments, we first generated plots for unsupervised principal component analysis (PCA) for each condition. The left and right (transfected) otocysts from the same embryo clustered together in the control electroporation (control EP) (Figure 1C), suggesting very little or no effects of the electroporation itself. The plots for Notch GOF 6 h and GOF 24 h showed respectively little and no obvious separation between samples electroporated with NICD1 and their left untransfected counterparts, suggesting that the overexpression of NICD1 did not trigger major and reproducible changes in the transcriptomes of transfected otocysts. In contrast, right ears treated with LY411575 and left control ears from the same embryo separated along the PCA2 in the plots for Notch LOF 6 and 24 h, suggesting an effect of the treatment.

To see whether any of the GOF or LOF treatments significantly affected the expression of components of the Notch pathway, we generated heatmaps representing the normalised transcripts abundance of TPM (Transcripts Per Million) for the direct Notch targets and effectors of the HES family (HES5.1, HES5.2, HES5.3, HES4, HEY1, HEY2), Notch receptors (NOTCH1, NOTCH2) and ligands (DLL1, JAG1, JAG2), and the transcriptional regulator RBPJ (Figure 1D, TPM and log TPM values in Supplementary Table S1). Surprisingly, the expression of NOTCH1 and JAG2 was significantly decreased after electroporation with the control construct (Figure 1D). In the Notch GOF 6 and 24 h datasets, we detected a significant increase in NOTCH1 expression (q-value < 0.05, Figure 1D). However, further analysis (Supplementary Table S1) in which we modified the reference chicken genome showed that the number of NOTCH1 intracellular domain transcripts was significantly increased (q-value < 0.001), most likely reflecting overexpression of the construct, while the expression levels of the extracellular and transmembrane domain of the endogenous NOTCH1 gene did not change. Of note, none of the HES genes were affected in Notch GOF 6 h group and only HES5.2 increased its expression in the Notch GOF 24 h condition. In contrast, blocking Notch activity with LY411575 for 6 h lead to a significant decrease in the expression of 5 HES genes (HES5.1, HES5.2, HES5.3, HES4, HEY1) and an increase in DLL1, JAG2 and RBPJ (raw TPM and log TPM values in Supplementary Table S1). A clear, but not as strong as with the 6 h treatment, reduction in the expression of HES5.1, HES5.2, HES5.3, HES4, NOTCH1, and NOTCH2 was also observed after 24 h treatment with LY411575 (Figure 1D). These results show that LY411575 efficiently inhibits Notch activity, and the short treatment (6 h) with LY411575 appears to induce stronger changes in the expression of direct Notch target genes than the 24-h treatment.

The differential gene expression analysis revealed that 28 genes were significantly (q-value < 0.05) regulated in the control EP samples (Figure 2; Supplementary Table S1); these were removed from the Notch GOF 6 and 24 h datasets in the subsequent analyses. A total of 53 genes were downregulated and 34 upregulated in Notch GOF 6 h condition (Figure 2; Supplementary Table S2), while 19 genes reduced their expression and 3 genes increased in Notch GOF 24 h dataset (Figure 2; Supplementary Table S3). Among genes significantly dysregulated in the Notch LOF 6 h, 105 genes showed a decrease and 181 genes an increase in expression levels (Figure 2; Supplementary Table S4). A comparable result was obtained in the Notch LOF 24 h dataset, with 127 genes downregulated and 128 genes upregulated (Figure 2; Supplementary Table S5). Surprisingly, SOX2 and JAG1, two genes implicated in prosensory specification and presumed to be positively regulated by Notch activity, were not among the significantly regulated genes in any of the datasets.

Altogether, the PCA results and the analysis of the expression of Notch components indicate that the pharmacological treatment with LY411575 leads to a stronger and more consistent perturbation of Notch activity than the in ovo electroporation of NICD1.

We next used the ToppGene bioinformatics platform to perform a functional annotation of the differentially expressed genes (q-value < 0.05) of each dataset. The Notch GOF 6 h dataset, with 90 differentially expressed (DE) genes (Supplementary Table S2), contained primarily genes associated with metabolic and general cell biological processes, indicating that this time point was too early to detect Notch-induced changes in expression of classic Notch target genes after electroporation. The GOF 24 h dataset (Supplementary Table S3) contained only 22 DE genes, which is not sufficient for a robust statistical analysis of biological functions or pathway enrichment. Nevertheless, several of the DE genes have been implicated in inner ear development (LMX1A, LMX1B, WNT2B, BMPER, SLITRK6) (Nichols et al., 2008; Katayama et al., 2009; Sienknecht and Fekete, 2009; Mann et al., 2017).

We present in detail the results obtained for the Notch LOF 6 h and Notch LOF 24 h datasets, which contained a larger number of DE genes, enabling a more robust bioinformatics analysis of their functions.

The Notch LOF 6h and 24 h datasets shared 17 genes only (Figure 3A), suggesting that the transcriptional response to Notch inhibition varies greatly over time. Reassuringly, these include HES5.1, HES5.2, HES5.3, and HES4, which together with TCF12 belong to the basic helix-loop-helix (bHLH) family of transcription factors regulating neurogenesis (Figure 3B). All the HES genes showed a reduced expression in both datasets (Figure 3C) suggesting a sustained inhibition of Notch activity. The remaining genes were: HSPA5, PODXL, CDK6, COLEC12, RRM2, MCM3, HAS2, PCDH8, LIPG, DSCC1, COL4A6, and two unknown genes (Figure 3B).

Next, we used the ToppGene functional annotation platform to analyse the two datasets. We found that 12 of the top 50 Gene Ontology (GO) functions of each dataset were shared; these include functions related to tissue morphogenesis, embryonic development, and sensory and central nervous system development (Figure 3D; Supplementary Tables S4, S5). In the Notch LOF 6 h dataset, 7 GO functions were associated with ear and sensory organ development and 14 related to the formation of the nervous system and neuronal differentiation (Figure 3D; Supplementary Table S4). In contrast, the Notch LOF 24 h dataset contained mainly genes associated with various cell biological processes and embryonic organ development (Figure 3E; Supplementary Table S5). Among the top 50 GO functions for this dataset, 5 were linked to eye development and only 2 with central nervous system development. Furthermore, the Notch LOF 6 h dataset included 11 genes from the bHLH family, which was the top affected gene family (Supplementary Table S4), including the proneural transcription factor NEUROD1, which is critical for the differentiation of otic neurons as well as other bHLH factors associated with neurogenesis (TCF3, TCF12, ID2, ID3, SIM1, and HES1, HES4, HES5.1, HES5.2, HES5.3, HEY1). In contrast, the Notch LOF 24 h dataset contained genes associated with general cell biological or metabolic processes (Supplementary Table S5). The Toppgene pathway enrichment analysis for Notch LOF 6 h dataset revealed several matches for Notch signalling among the dysregulated pathways across all cross-referenced databases (Supplementary Table S4): Biocyc, KEGG, Reactome, the Pathway Interaction Database, Gene Set Enrichment Analysis (GSEA), Panther Classification System, The Rat Genome Database (RGD). On the other hand, Notch LOF 24 h dataset had only two matches for Notch signalling among significantly dysregulated pathways in Reactome and Panther databases (Supplementary Table S5). The marked differences in the 6 and 24 h transcriptional responses suggest that Notch inhibition triggers a rapid commitment of neurosensory progenitors to a neurogenic fate; however, after 24 h, it is possible that the neurosensory progenitor pool is depleted, due to excess formation of otic neurons, and that Notch inhibition regulates another set of non-neurogenic genes in the remaining otic cells.

The analysis of the enriched transcription factors binding sites (TFBS) showed that only 8 out of the top 50 TFBS were present in both LOF datasets: LEF1 (CTTTGT, V$LEF1 Q2), FOXO4 (TTGTTT, V$FOXO4 01), E12 (CAGGTG, V$E12 Q6), NFAT (TGGAAA, V$NFAT Q4), MAZ (GGGAGGRR, V$MAZ Q6), KDM7A Target Genes, and two unknown transcription factors with AACTTT and CTGCAGY binding motifs (Supplementary Tables S4, S5). Interestingly, the motif V$LEF1 Q2 was one of the most significantly enriched TFBS site in both datasets (p = 4.45e-15 in LOF 6 h; p = 4.73e-07 in LOF 24 h), with 51 genes from Notch LOF 6 h (Supplementary Figure S1; Supplementary Table S4) and 34 genes in Notch LOF 24 h (Supplementary Table S5). LEF1 is a member of the TCF/LEF transcription factor family and an essential component of the Wnt signalling cascade. Although these analyses were conducted using the human orthologues of the chicken genes, they suggest that some of the early and late transcriptional targets of Notch activity may be co-regulated by the Wnt pathway.

Next, we investigated the potential transcriptional targets of Wnt signalling during prosensory specification using two different approaches to inhibit Wnt activity (Żak and Daudet, 2021). Firstly, we electroporated chicken otic cups with a truncated form of β-catenin lacking the N- and C-terminus responsible for transcriptional activity (DNBCAT), which acts like a dominant-negative protein and inhibits Wnt signalling (Wnt LOF EP 24 h) (Figure 4A) (Żak and Daudet, 2021). The control group was electroporated with a plasmid driving expression of a monomeric red fluorescent protein (mCherry, Control EP) (Figure 4A). At 24 h post-electroporation, we collected left (untransfected) and right (transfected) otocysts from 3 embryos for both conditions. Secondly, we used a pharmacological inhibition of Wnt activity using IWR-1 (Wnt LOF IWR1 24 h) (Figure 4B), a tankyrase inhibitor that stabilises AXIN2, a member of the β-catenin destruction complex (Chen et al., 2009). For three E2.5 embryos, each right otic cup was incubated for 24 h in medium containing 300 μM IWR-1, while its left counterpart was maintained in medium containing DMSO at a matching 0.6% concentration as a control.

The PCA plot for the control electroporation (Control EP) showed no separation between left (control) and right (transfected) otocysts of each embryo (Figure 4C). Similarly, there was no clear separation of the Wnt LOF EP 24 h otocysts according to condition, (Figure 4C), suggesting little impact on the overall gene expression profile of transfected otocysts. In contrast, the PCA plot for Wnt LOF IWR1 24 h dataset showed a good separation of IWR-1 treated samples from the control DMSO treated samples along the PC1 (48% variance) (Figure 4C).

To assess the potential changes in Wnt signalling in response to our treatments, we analysed the expression levels of specific Wnt components (receptor, ligands, effectors, and modulators). Heatmaps of the abundance of log TPM for selected transcripts were generated for control EP, Wnt LOF EP 24 h and Wnt LOF IWR1 24 h groups. Mapping to the reference chicken genome with Kallisto (using default settings) detected 27 genes associated with Wnt signalling (Figure 4D, raw and log TPM values in Supplementary Table S1). Among these were the Wnt signalling modulators DKK1, DKK3, LGR4, LGR5, AXIN1, AXIN2, SFRP1 (Figure 4D). All the Wnt pathway transcriptional effectors (LEF1, TCF7, TCF7L1 and TCF7L2) as well as several receptors from the Frizzled family (FZD1, FZD2, FZD3, FZD4, FZD6, FZD7, FZD9, and FZD10) and Wnt ligands (WNT2B, WNT3A, WNT4, WNT5A, WNT5B, WNT6, WNT7B, WNT9A, WNT11) (Figure 4D) were detected. Out of these genes, SFRP1, WNT2B and AXIN2 from the Wnt LOF EP 24 h and SFRP1, WNT4 and WNT11 from Wnt LOF IWR1 24 h were significantly regulated (q-value < 0.05) (Figure 4D).

The control electroporation dataset did not have any significantly DE gene (q-value < 0.05, Figure 5E). In contrast, the Wnt LOF EP 24 h dataset contained 30 upregulated and 69 downregulated genes (Figure 4E), whilst the Wnt LOF IWR1 24 h dataset contained 109 upregulated and 132 downregulated genes (Figure 4E). Among the upregulated genes were the NOTCH1 receptor as well as proneural transcription factors, which are known direct target genes and canonical effectors of the Notch pathway.

However, the two Wnt LOF datasets shared only 9 candidate target genes (Figures 5A-C). Next, we used the ToppGene platform to compare the two Wnt datasets. Among top 50 significantly enriched biological functions, 12 were shared by Wnt LOF EP 24 h and Wnt LOF IWR1 24 h groups and were implicated with developmental processes and organ morphogenesis. Wnt LOF EP 24 h included 16 functions associated with neurogenesis and formation of neuronal tissue, and 2 functions related to regionalisation and body axis specification (Figure 5D; Supplementary Table S6). Additionally, among the top 200 significantly enriched GO functions were also 2 related to the regulation of the Wnt signalling pathway involved in dorsal/ventral axis specification. These bioinformatics results are in line with our recent findings showing that Wnt signalling regulates neurosensory specification along the dorso-ventral axis of the inner ear (Żak and Daudet, 2021). In contrast, the top 50 significantly enriched biological functions for Wnt LOF IWR1 24 h group were primarily associated to embryonic development, organ morphogenesis, and extracellular matrix organization (Figure 5E; Supplementary Table S7). None of the top 50 biological function terms was specifically associated with neurogenesis, but 5 of them were associated with the development of eye and sensory organs. It is worth noting that Wnt signalling was not among the significantly enriched gene families or signalling pathways in Wnt LOF EP 24 h (Supplementary Table S6) or Wnt LOF IWR1 24 h datasets (Supplementary Table S7). Nevertheless, both datasets contained an over-representation of genes with TFBS for LEF1 (CTTTGT, V$LEF1 Q2 and CTTTGA V$Lef1 Q2), which could represent direct targets of Wnt signalling. Among the other over-represented TFBS present in both datasets as well as the Notch LOF datasets were, FOXO4 (TTGTTT, V$FOXO4 01), E12 (CAGGTG, V$E12 Q6), MAZ (GGGAGGRR, V$MAZ Q6), and unknown transcription factor with AACTTT binding motif (Supplementary Tables S6, S7). The two Wnt datasets also shared 3 more TFBS: AP4 (CAGCTG V$AP4 Q5), MAX (V$MAX 01), and USF (V$USF 02).

Our analyses identified a large number of genes associated with “neurogenesis” processes that could potentially be regulated by Notch and/or Wnt activities during prosensory specification. We decided to further test the function of some of these genes by overexpression in the chicken otocyst. We selected several transcription factors, which are prime candidate regulators of cell specification and differentiation events: NEUROD1, MEIS1A, MEIS2A, EGR1, LHX1, LHX5, DLX5, TCF3, MSX1, MSX2, PROX1, BAMBI, SOX9, SOX10, TBX2, and TFAP2C (Table 1). Additionally, we tested the effects of overexpressing the Wnt ligands WNT4, WNT5A, WNT11, and two critical negative regulators of Wnt signalling, AXIN2 and APC (Table 1).

To mark transfected regions, plasmids expressing our genes of interest were co-electroporated with a plasmid encoding GFP. Transfected ears were collected 2 days post-EP and stained for the otic prosensory domain marker SOX2. We looked for ectopic or loss of SOX2 staining within the prosensory domains and abnormalities in the formation of otic neurons. Control experiments were performed with a GFP expression plasmid. Control ears (n = 4/4) developed normally forming posterior and anterior prosensory domains (occasionally including a well-segregated anterior crista) with elevated SOX2 expression and a sensory-competent domain with lower SOX2 expression stretching between them (Figures 6A–A”). Otocysts transfected with NEUROD1 (n = 4/4) (Figures 6B–B”) were very small in size, devoid of GFP expression, but surrounded by a large number of GFP-positive neuronal processes (Figures 6B–B”). Inside these small otocysts remained a population of SOX2-positive cells that did not form distinct prosensory domains. After overexpression of AXIN2 (Figures 6C–C”), WNT11 (Figures 6F–F”), LHX1 (Figures 6G–G”), LHX5 (Figures 6H–H”), the otocysts appeared to be smaller compared to the control ears and their prosensory domains seemed to be underdeveloped (n = 4/4), but otic neurons formed normally. Otocysts transfected with PROX1 had a normal appearance with posterior and anterior prosensory domains and otic neurons (n = 4/4) (Figures 6D–D”). However, the overexpression of PROX1 in the tissue surrounding the otocyst triggered formation of ectopic SOX2 domains (n = 2/4) (Figures 6E–E”). Overexpression of APC, WNT4, WNT5A, SOX9, SOX10, TBX2, DLX5, MSX1, MSX2, BAMBI, TCF3, MEIS1A, MEIS2A, and EGR1 (Supplementary Figure S2) did not trigger any changes in the formation of the otocyst and SOX2-expressing domains.

The comparison of two durations of GSI treatment, Notch LOF 6 h and LOF 24 h, revealed striking changes in the set of differentially expressed genes, which were associated with distinct classes of biological processes. In fact, only 17 genes were shared between the 6 and 24 h LOF datasets; among these were 4 Notch effectors of the HES family (HES 5.1, HES 5.2, HES5.3, HES4), whose downregulation confirmed a persistent blockade of Notch activity by GSI. In accordance with the role of Notch lateral inhibition in preventing excessive neuroblast formation in the otocyst (Haddon et al., 1998; Abello et al., 2007; Daudet et al., 2007), the loss of Notch activity in the LOF 6 h dataset affected the expression of several genes implicated in otic neurogenesis. Among these were the genes encoding NEUROD1, a proneural bHLH transcription factor essential for audiovestibular neuronal differentiation (Liu et al., 2000; Filova et al., 2022) and the Notch ligand DLL1, mediating lateral inhibition in otic neuroblasts (Adam et al., 1998; Daudet et al., 2007; Daudet and Żak, 2020). As expected from a failure of lateral inhibition, the expression of both NEUROD1 and DLL1 was upregulated in the 6 h LOF dataset. Other dysregulated genes associated to the differentiation of otic neurons and sensory organs included PROX1 (upregulated) (Stone et al., 2003; Bermingham-McDonogh et al., 2006; Fritzsch et al., 2010; Nishimura et al., 2017), FGF10 and FGF20 (downregulated) (Pauley et al., 2003; Hayashi et al., 2008; Munnamalai et al., 2012; Urness et al., 2015), FGF3 (upregulated) (Zelarayan et al., 2007; Olaya-Sanchez et al., 2017), BMP4 (downregulated) (Wu and Oh, 1996; Morsli et al., 1998; Chang et al., 2008). The LIM-homeodomain transcription factor LMX1B, which plays a critical role in the segregation of sensory organs (Nichols et al., 2008; Koo et al., 2009), was upregulated in the Notch LOF 6 h dataset, which is consistent with previous studies showing that Notch activity represses LMX1B expression in the chicken otocyst. Finally, there was an overrepresentation of dysregulated genes with E-box motifs (CAGGTG; Supplementary Table S4), which are binding sites for the ubiquitous E12/E47 bHLH proteins associated to the tissue-specific proneural bHLH transcription factors (Jennings et al., 1999; de Martin et al., 2021). Altogether, these results show that the blockade of Notch activity induces a rapid transcriptional response associated to neurosensory differentiation.

In contrast, the genes affected by the prolonged loss of Notch activity in the LOF 24 h dataset were primarily associated with general cellular biology processes such as protein translation and trafficking as well as various catabolic processes. Furthermore, the proneural gene NEUROD1 was no longer differentially expressed in the LOF 24 h dataset and E-box motifs were not over-represented among the DE genes. The differences with the shorter treatment could result from the more advanced developmental stage of the otocyst as well as the fact that at 24 h post-blockade, most neurosensory precursors committed to a neuronal fate have presumably delaminated from the otocyst. This could explain the reduction in expression levels of neurogenesis-associated genes within the otic samples processed for RNA-Seq.

Our recent work showed that the Wnt signalling pathway forms an activity gradient along the dorso-ventral axis of the otocyst to control sensory organ formation and neurogenesis in a dose-dependent manner (Żak and Daudet, 2021). To identify targets of Wnt signalling in this context we used two distinct approaches to block Wnt activity in the otocyst. Pharmacological treatment with IWR-1 increases the intracellular levels of AXIN2 protein, thereby promoting formation of a destruction complex and degradation of beta-catenin (Chen et al., 2009), whereas overexpression of truncated form of beta-catenin lacking transcriptional activity and acting as a dominant-negative form prevents endogenous beta-catenin from activating its transcriptional targets (Żak and Daudet, 2021). We have shown that both approaches can reduce the activity of a fluorescent Wnt reporter (comprising 5 TCF/LEF binding sites) in the chicken inner ear (Żak and Daudet, 2021).

The Wnt signalling pathway does not have a set of context-independent core target genes and the ear-specific Wnt target genes are still unclear but, both Wnt LOF datasets contained an over-representation of DE genes with TFBS for LEF1, suggesting some alteration of Wnt activity. Among the 9 shared genes between the Wnt LOF IWR1 24 h and Wnt LOF EP 24 datasets were MSX1 and MSX2, which are expressed in the vestibular system and endolymphatic duct (Wu and Oh, 1996; Pujades et al., 2006) where high levels of Wnt activity are present. Interestingly, Wnt signalling positively regulates expression of MSX1 and MSX2 genes in other systems and both were downregulated in our LOF datasets (Willert et al., 2002; Szemes et al., 2020). Two other candidate targets of Wnt signalling are SFRP1 and AXIN2, which are negative regulators of Wnt activity (Finch et al., 1997; Behrens et al., 1998; Seidensticker and Behrens, 2000). AXIN2 is positively regulated by Wnt signalling in other tissues (Jho et al., 2002) and was significantly downregulated in the Wnt LOF EP24 h dataset. SFRP1, a soluble inhibitor of Wnt signalling that binds to Wnt ligands or Frizzled receptors in the extracellular compartment (Lin et al., 1997; Bafico et al., 1999), was downregulated in both LOF datasets and its paralogue SFRP2 is a known Wnt target gene (Lescher et al., 1998) in other tissues. These results suggests that both IWR-1 and DNBCAT overexpression affect Wnt signalling and that MSX1, MSX2, SFRP1, and AXIN2 may be Wnt targets in the inner ear.

Similarly to our Notch datasets, we identified more DE genes through pharmacological treatment than after overexpression of DNBCAT. The biological functions in both Wnt datasets were associated with morphogenesis, organ development and neurosensory tissue formation, but only 12 out of the top 50 functional annotations were common. This relatively low level of overlap may be due to the two very different approaches (in ovo and in vitro) targeting different elements of the Wnt signalling cascade. From the bioinformatics analysis, the Wnt LOF EP 24 h dataset appeared better aligned with our current understanding of the role of Wnt signalling during early inner ear development. Indeed, it highlighted several biological functions related to neurogenesis as well as antero-posterior (A-P) and dorso-ventral (D-V) axis specification. This is in line with previous studies demonstrating the roles of Wnt signalling in establishing the D-V patterning of the otocyst and its neurosensory territories (Stevens et al., 2003; Riccomagno et al., 2005; Noda et al., 2012; Rakowiecki and Epstein, 2013; Żak and Daudet, 2021). Our data suggest a number of Wnt signalling components could contribute to the establishment of the D-V gradient of Wnt activity. Wnt antagonists such as SFRP1, which is expressed in the ventral region of the inner ear (Sienknecht and Fekete, 2008; Sienknecht and Fekete, 2009), and AXIN2 could be part of a negative feedback loop progressively reducing the levels of Wnt activity along the D-V axis. Furthermore, three Wnt ligands (WNT2B, WNT4 and WNT11) were downregulated in at least one of the LOF conditions, suggesting some form of positive feedback dependent on Wnt activity could regulate their expression. The Wnt pathway is known to form such feedback mechanisms in other systems through Wnt-mediated expression of ligands and inhibitors (Lustig et al., 2002; Stuckemann et al., 2017) and it will be important to further investigate their contribution to inner ear patterning.

With the aim to identify new genes regulating neurosensory specification, we overexpressed in the chicken otocyst 21 of the genes identified in our Wnt and Notch datasets encoding either transcription factors, which are prime candidate regulators of cell differentiation, or components of the Wnt pathway. Disappointingly, only two transcription factors, NEUROD1 and PROX1, produced a noticeable phenotype 48 h after in ovo EP. NEUROD1-overexpressing cells formed neuron-like cells which delaminated from and surrounded the otocyst, confirming its importance for neuronal differentiation (Liu et al., 2000; Kim et al., 2001). PROX1-overexpressing cells formed clusters with ectopic SOX2 expression, but only outside of the otocyst. The significance of this result is unclear, although previous studies have shown that PROX1 overexpression can induce the formation of neurons in the CNS (Karalay et al., 2011; Iwano et al., 2012; Kaltezioti et al., 2021). AXIN2 overexpression drastically reduced the size of the otocyst, possibly reflecting an effect on cell proliferation or survival, but did not affect the spatial pattern of SOX2 expression. None of the other Wnt components induced a phenotype. These results suggest that these factors do not regulate SOX2 expression, or may do so in cooperation with other signals. They also indicate that the endogenous program of neurosensory differentiation is remarkably robust and only a small number of key transcriptional factors, such as proneural genes, SOX2 and LMX1A (Neves et al., 2011; Mann et al., 2017), might be able to override the signals driving sensory versus non-sensory cell differentiation.

Fertilised White Leghorn chicken (Gallus gallus) eggs were obtained from Henry Stewart UK and incubated at 37.8°C for the designated times. Embryonic stages refer to embryonic days (E), with E1 corresponding to 24 h of incubation or to Hamburger and Hamilton stages (Hamburger and Hamilton, 1951). All procedures were approved by University College London local Ethics Committee and by the UK Home Office.

Electroporation of the otic placode/cup of E2 chick embryos (stage HH 10–14) was performed using a BTX ECM 830 Electro Square Porator as previously described (Freeman et al., 2012). The total concentration of plasmid DNA ranged for each set of experiments between 0.5 and 1 μg/μl. Unless otherwise specified, for the gain of function experiments a minimum number of 4 well transfected samples were examined for each experimental condition (Figure 1A). For the bulk RNA-seq analysis, embryos were collected and examined with a fluorescent dissecting stereomicroscope. Those with the most intense and widespread fluorescence signal within the otocyst (Figure 1A) were selected for analysis. Both left (non-transfected) and right (transfected) otocysts were dissected from each embryo in ice-cold L-15 medium (Leibovitz), cleaned from surrounding mesenchyme and immediately processed for total RNA isolation. The expression of mesenchymal markers was assessed for all datasets and included in Supplementary Table S1.

The plasmids used in this study and their origin are described in the Table 1. New constructs were generated using the In-Fusion HD Cloning Kit (Takarabio).

Both left and right ears were dissected from chicken embryos aged E2.5 in ice-cold L-15 medium, cleaned from surrounding mesenchyme and individually incubated as free-floating cultures. Left ears were incubated in media enriched with DMSO (control conditions) and right ears were treated with γ-secretase inhibitor LY411575 (10 µM, Notch LOF datasets) or IWR-1 (300 µM, Wnt LOF datasets) for 6 or 24 h.

Total RNA was extracted from each individual otocysts using the RNAqueous™-Micro Total RNA Isolation Kit (Ambion) according to the manufacturers protocol. The quality of isolated RNA was tested using Agilent 2200 Tapestation and only samples with a value of RNA integrity number of at least 9 were used for library preparation by UCL Genomics. The SMART-Seq v4 Ultra Low Input RNA Kit (Clontech Laboratories, Inc.) was used to generate cDNA libraries using 10 cycles of PCR. cDNA was checked for integrity and quantity on the Agilent Bioanalyser using the High Sensitivity DNA kit, and 200 pg of cDNA was then converted to sequencing library using the Nextera XT DNA protocol (Illumina, San Diego, United States). Samples were sequenced 43 bp paired-end read and ∼16M reads per sample length on NextSeq 500 instrument (Illumina, San Diego, CA, United States). Run data were demultiplexed and converted into fastq files using Illumina’s bcl2fastq Conversion Software v2.19. Kallisto package (Bray et al., 2016) was used to map reads to a chicken reference genome (Gallus_gallus-6.0) and to quantify abundances of transcripts. Differentially expressed genes were identified using Sleuth package (Pimentel et al., 2017) by comparing treated samples with control ears from the same embryo. Genes with q-value (Wald test) lower than 0.05 were considered as significant. Functional annotations were downloaded from ENSEMBL using biomaRt and PCA plots were generated using DESeq2. The heatmaps were generated in R using logarithm (base 2) of TPM values. Signalling pathway enrichment, transcription factor binding site enrichment and Biological Function enrichment analysis were performed using Toppgene online tool with default settings and using the human orthologues of the chicken genes as input. Visualizations of gene networks were generated using Cytoscape. To distinguish between the expression of endogenous NOTCH1 and overexpressed NIDC1, we included in the reference genome two separate sequences for the chicken intracellular (NICD1) and extracellular domains of NOTCH1. Next, we compared the number of reads for intracellular (NICD1) and extracellular sequences of NOTCH1 gene in each sample and analysed changes in their expression levels between the treated and non-treated samples. The tpm results are included in Supplementary Table S1 tab “NICD1_EP_tpm” The original datasets are publicly available in GEO. For the Wnt LOF datasets please see https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149310 and https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?&acc=GSE230083; for all Notch datasets https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE196999.

Chicken embryos (E3-E4) were collected, fixed for 1-h in 4% paraformaldehyde (PFA) in 0.1 M phosphate buffered saline (PBS), and processed for whole-mount immunostaining. The head was dissected along the midline, the hindbrain was removed, and the region surrounding the otocyst was only partially trimmed to facilitate orientation. Next, the tissue was permeabilized in PBS containing 0.3% Triton and 10% goat serum for 30 min at room temperature. Specimens were incubated with primary antibodies diluted in 0.1% Triton in PBS at 4°C overnight. On the next day, tissues were rinsed with PBS at room temperature and incubated with secondary antibodies diluted in 0.1% Triton and 10% goat serum at 4°C overnight. Afterward, tissues were again rinsed with PBS and mounted in Vectashield Antifade Mounting Medium (Vector laboratories). The following antibodies were used: mouse IgG1 monoclonal anti-Sox2 (BD Biosciences, San Jose, CA; 561469, 1:500). Secondary goat antibody conjugated to Alexa dye (1:1000) were obtained from Thermo Fischer Scientific (United Kingdom). Confocal stacks were acquired using a Zeiss LSM880 inverted confocal microscope, further processed with ImageJ and arranged in Adobe Illustrator.