The Foxi3 transcription factor is necessary for the fate restriction of placodal lineages at the neural plate border

As vertebrate embryos undergo gastrulation and lay down their body plan, their anterior ectodermal layer becomes patterned along the medial-lateral axis to give rise to four lineages: neural plate, which will form the entire central nervous system; neural crest cells, which will form most of the peripheral nervous system, in addition to much of the bone and cartilage of the head and some pigment cells; cranial placodes, which form some paired sensory organs and sensory ganglia of the head; and epidermis. The segregation of these four lineages results from the interaction of several signaling families, such as FGFs, BMPs and WNTs, and the establishment of domains of lineage-specific transcription factors (Albazerchi and Stern, 2007; Grocott et al., 2012; Groves and LaBonne, 2014; Pla and Monsoro-Burq, 2018; Prasad et al., 2020; Rogers et al., 2011; Schlosser, 2021; Streit et al., 2000; Thawani and Groves, 2020; Thiery et al., 2020).

These secreted signals initially partition embryonic ectoderm into broad pre-neural and non-neural domains marked by transcription factors such as Sox2/3, Otx2 and Geminin (neural; Bally-Cuif et al., 1995; Kroll et al., 1998; Papanayotou et al., 2008; Rex et al., 1997; Streit et al., 1997; Uchikawa et al., 2003), and Dlx5, Gata2/3, Msx1 and TFAP2a (non-neural; Hoffman et al., 2007; Li and Cornell, 2007; McLarren et al., 2003; Pera et al., 1999; Phillips et al., 2006; Sheng and Stern, 1999; Woda et al., 2003). The boundary between these two domains is termed the neural plate border. Subsequent signaling leads to the induction of two additional derivatives: (1) placodal progenitors, which emerge from the ectoderm closer to the future epidermis and express ‘pre-placodal’ transcriptional regulators such as Six and Eya family members (Christophorou et al., 2009; Grocott et al., 2012; Kwon et al., 2010; Sato et al., 2010; Schlosser, 2007; Zou et al., 2004); and (2) the neural crest, which forms adjacent to the neural plate and expresses transcription factor families such as Zic, Pax3/7, FoxD3 and, later, Sox9/10 (Cheung et al., 2005; Hong and Saint-Jeannet, 2007; Jaurena et al., 2015; Keuls et al., 2023; Lukoseviciute et al., 2018; McKeown et al., 2005; Milet et al., 2013; Murdoch et al., 2012; Pieper et al., 2012; Roellig et al., 2017). How these four domains become distinct is beginning to be understood, with mechanisms that include both positive feedback between transcription factors within a given domain, and cross-repression between transcription factors of neighboring domains in response to varying WNT, BMP and FGF signals (Grocott et al., 2012; Groves and LaBonne, 2014; Pla and Monsoro-Burq, 2018).

The advent of new technologies, such as high-throughput in situ hybridization and single cell RNA-seq, has revealed a far more dynamic and heterogeneous expression of genes affiliated with the different ectodermal lineages than previously thought (Lukoseviciute et al., 2018; Maharana and Schlosser, 2018; Pla and Monsoro-Burq, 2018; Rothstein and Simoes-Costa, 2020; Thiery et al., 2020). Yet it remains unclear whether the expression of genes from multiple lineages in a single cell means that it can still give rise to several different border lineages, or whether that cell has already become restricted to a particular lineage. Many models have been suggested to describe the fate competence and differentiation landscape of the developing neural plate border. The ‘binary competence model’ suggests that cells in the neural plate border region acquire either non-neural or neural competence early, and placodes and neural crest, respectively, emerge from a subset of those cell populations. An alternative ‘neural plate border model’ proposes that neural plate border cells remain multipotent until they become epidermis, neural plate or a unique progenitor population that subsequently gives rise to neural crest and cranial placodes (Schlosser, 2006, 2014; Schlosser et al., 2014). More recently, a ‘probabilistic model’ suggests a transcriptionally unstable and dynamic state of the neural plate border cells that are gradually restricted to one of the four ectodermal lineages based on their spatiotemporal positions (Thiery et al., 2023). Reconciling these models has been challenging, as this requires knowledge of gene expression in a cell and tracking its subsequent fate. In this study, we attempt to address these questions by studying a transcription factor expressed at the neural plate border, Foxi3, and assessing the fate of the Foxi3 cell lineage.

Previous studies from our lab indicate that Foxi3 is one of the earliest transcription factors expressed in neural plate border ectoderm in mouse and chick embryos (Birol et al., 2016; Khatri et al., 2014; Ohyama and Groves, 2004). Foxi3-null mice show a dramatic phenotype with absence or severe reduction of the otic and epibranchial placodes (Birol et al., 2016); however, the fate of Foxi3 mutant cells in these embryos is not clear. In this study, we report and use two novel knock-in mouse lines: (1) Foxi3-GFP mice, to detect the expression of Foxi3 in developing ectoderm at single cell resolution; and (2) Foxi3-CreER mice, to follow the fate of Foxi3-expressing neural plate border cells in normal and functionally null embryos, and to study the contribution of Foxi3 lineage to placodes. We show that Foxi3 marks the cells at the neural plate border adjacent to and overlapping the pre-neural domain; these Foxi3-expressing neural plate border cells mostly give rise to placodes and some epidermis from the onset of its expression. This suggests that cells expressing Foxi3 have already become fate-specified and segregated from neural crest and neural plate lineages by the end of gastrulation. We show that this behavior is in marked contrast to genes of the neural crest lineage such as Zic5 (Inoue et al., 2004), using a third newly generated mouse line: Zic5-CreER. The Zic5-positive progenitors contribute initially to all germ layers, but gradually become restricted to neural crest and dorsal neural tube derivatives by the start of neurulation. We also show that, in the absence of Foxi3, Foxi3-null neural plate border cells undergo a fate change and can now also give rise to the other three ectodermal lineages. Our data suggest that Foxi3 not only marks cells that have committed to a placode lineage, but that the action of Foxi3 is necessary to drive this lineage restriction and segregation.

To observe the spatiotemporal expression of the Foxi3 gene, we made a knock-in mouse line (Foxi3-GFP) that expresses functional Foxi3 and mVenus fluorescent protein as a fusion protein that is cleaved during translation at the intervening P2A sequence (Fig. S1). We harvested Foxi3-GFP embryos from pre-gastrulation stages [embryonic day (E) 6.5] to organogenesis after onset of neurulation (E9.0) (Fig. 1A) and immunostained them with anti-GFP antibodies as a proxy for Foxi3 expression (Fig. 1B). We observed the earliest expression of GFP at around E6.75, where a few weakly positive cells were restricted to extra-embryonic ectoderm lining the amnion (Fig. 1C, asterisk; also labeled with asterisks in Fig. 1B). As gastrulation progressed, GFP expression increased and spread to the lateral edges of the embryonic ectoderm (Fig. 1C). We observed a strong band of GFP staining at E7.5 in the non-neural ectoderm adjacent to the SOX2-positive neural plate with some overlap between the two domains (Fig. 1C). E7.5 is the first stage when the neural plate border is morphologically apparent: the boundaries around the neural plate are more opaque and the neural folds become distinct by E8.0 (Fig. 1A,B). Using Lightsheet microscopy, we were able to visualize whole Foxi3-GFP embryos stained for GFP and SOX2 to reveal their unique expression patterns and the co-expression of these markers at the neural plate border along the anterior-posterior axis of the embryo at several stages between gastrulation and neurulation: E7.0, E7.5, E8.0 and E8.75 (Fig. S2; Movies 1-8). We verified the overlap of Foxi3-GFP and SOX2 at the neural plate border at E7.5 with a FOXI3 monoclonal antibody (Fig. 1D-E′). Co-staining with antibodies to AP2A (Fig. 1F,F′), a non-neural ectoderm marker, also showed an overlap with SOX2 at the neural plate border and extensive co-expression with FOXI3 in the non-neural domain lateral to the neural plate border (Fig. 1E′,F′).

Foxi3 mRNA is rapidly downregulated from the neural plate border as individual placodes differentiate (Birol et al., 2016). We confirmed this downregulation in our Foxi3-GFP mice; as the neural folds became more pronounced, Foxi3-GFP expression receded from the anterior-most neural plate border (Fig. 1B,C,,Fig. S2) and Foxi3-GFP- and SOX2-expressing domains separated around E8.5 (Fig. 1C, Fig. S2C,D). GFP expression persisted through E8.5, after which it was restricted to the pharyngeal clefts (E9.0 image in Fig. 1B, arrowheads) as previously described (Edlund et al., 2014; Ohyama and Groves, 2004). Together, our data indicate that Foxi3 is one of the earliest genes whose expression can be used to define the neural plate border, and that Foxi3 expression disappears rapidly from the border region as neurulation proceeds.

Our data show that Foxi3 is expressed adjacent to the developing neural plate, and that some Foxi3-expressing cells at the neural plate border also express SOX2 protein (Fig. 1, Fig. S2). This raised the question of whether Foxi3-expressing cells can produce one or more of the four derivatives of the neural plate border: neural tube, neural crest, cranial placodes and epidermis. To determine this, we mated our newly generated Foxi3-CreER mice with Rosa26-tdT reporter mice (Fig. 2A) and gave tamoxifen to pregnant females at E7.5 (Fig. 2B) to map the fates of Foxi3-expressing cells. We chose E7.5 for CreER activation because our Foxi3-GFP expression data show that the peak of Foxi3 expression at the neural plate border occurs between E7.0 and E8.0 (Fig. 1). We harvested embryos at E11.5 (Fig. 2B) when all cranial placodes are morphologically distinct and easily identified in sections (placodes in the schematic in Fig. 2C; cryo-sectioning plane in Fig. 2D). We identified cells derived from Foxi3 progenitors by endogenous tdTomato fluorescence or anti-RFP immunostaining.

We saw robust tdTomato signal in only one of the ectodermal derivatives – the cranial placodes (Fig. 2C,E). All cranial placodes– the lens, olfactory pit, adenohypophysis, inner ear, and the trigeminal and epibranchial cranial ganglia – showed significant tdTomato labeling. In addition to the labeling of all cranial placodes, we saw some tdTomato signal in the epidermis (Fig. 2E, arrows). Notably, little to no tdTomato labeling was found in the cranial mesenchyme around placodal derivatives or in the neural tube (Fig. 2E). Only epidermal tdTomato labeling was observed in the trunk that is devoid of sensory placodes (Fig. S3).

To determine whether the Foxi3-expressing population had a broader range of fates at earlier stages, we expanded our lineage tracing to compare the fates of Foxi3 cells at several ages between gastrulation and neurulation. We delivered tamoxifen to pregnant dams at five different ages – E5.5, E6.5, E7.5, E8.5 and E9.5, and each experimental group was analyzed by tdTomato immunostaining at E11.5 (Fig. 3A). We observed very little tdTomato labeling anywhere in the embryonic head after tamoxifen delivery at E5.5 (Fig. 3B-E). Animals receiving tamoxifen at E6.5, shortly before Foxi3 first begins to be expressed (Fig. 1, Fig. S2) showed similar results to embryos treated at E7.5 (Fig. 3B-E). This indicates that the gene is activated in tandem with neural plate border formation.

Foxi3 is downregulated from the neural plate border region in an anterior-posterior direction from E8.5 onwards (Birol et al., 2016; Ohyama and Groves, 2004). Accordingly, when we administered tamoxifen at E8.5, tdTomato expression was almost absent in the derivatives of the anterior placodes (olfactory, lens and trigeminal ganglion) (Fig. 3B,C), and was greatly reduced in the otocyst derived from the otic placode (Fig. 3D), with only the posterior epibranchial ganglia (IX, X) showing significant labeling (Fig. 3E). Labeling of the epibranchial ganglion was further reduced when tamoxifen was delivered at E9.5. These observations are summarized in Figs S4 and Fig. 3F. Together, these results show that Foxi3-expressing cells contribute exclusively to placodal derivatives and epidermis from the onset of Foxi3 expression until its cessation at the neural plate border.

To confirm the exclusive contribution of Foxi3-expressing progenitors to cranial placodes and epidermis, we harvested embryos at E9.5, an earlier age when neural crest cells are migrating from the neural folds (Fig. 4A). We immunostained cryosections (Fig. 4B) and used ectodermal derivative-specific markers: SOX9 to identify neural crest cells; SOX2 to label sensory placodes, the neural tube and pharyngeal endoderm; TUBB3 to label differentiating neurons that help identify the periphery of the neural tube; and ECAD to label epidermis and endoderm (Fig. 4C-F). Confirming our previous analysis at E11.5, we found the Foxi3-lineage contributes only to placodes and epidermis when tamoxifen was given at E6.5, E7.5 or E8.5. No tdTomato-positive cells were found in the neural tube, dorsal-most neural tube/neural folds or SOX9-positive neural crest cells that delaminate and migrate from the neural folds at this stage of development (Fig. 4C-F). These data confirm that Foxi3-expressing neural plate border cells contribute to the placodal and epidermal lineages from the earliest age of its expression, but not neural crest cells or neural plate.

The restriction of Foxi3-expressing progenitors to exclusively placodal and epidermal fates was surprising, as genes associated with neural plate and neural crest derivatives, such as Sox2 and Foxd3, initially label pluripotent cells in the epiblast before gastrulation and gradually become restricted to neural and neural crest fates (Liu and Labosky, 2008; Roellig et al., 2017). Zic5, a zinc-finger transcription factor, is one of the earliest markers of neural crest fate reported in vertebrates (Inoue et al., 2004). Unlike Foxi3, Zic5 expression extends uniformly across the epiblast before gastrulation, restricting itself to only the neural folds and dorsal-most neural tube during neural tube closure (Inoue et al., 2004), thus labeling the pre-migratory neural crest cells. We created a Zic5-CreER knock-in mouse line to compare how the range of fates adopted by the epiblast expressing this transcription factor differs from those adopted by Foxi3-expressing progenitors (Fig. S5).

We administered tamoxifen to pregnant dams from Zic5-CreER mice mated with ROSA-tdT reporter mice at different times between E6.5 and 8.5; embryos were harvested at E9.5 (Fig. 5A,B). Unlike Foxi3 progenitors, Zic5 progenitors labeled at E6.5 contributed to all three germ layers (Fig. 5C-F), but after tamoxifen induction at E7.5, Zic5 descendants were limited to neural plate border derivatives: epidermis, placodes, neural crest and the neural tube (absence of tdTomato signal in endoderm; solid arrow in Fig. 5C-F). Cre activation at E8.25 and later ages showed further restriction of tdTomato labeling to the dorsal neural tube and migrating neural crest cells (Fig. 5C-F). Analysis of cranial placodal labeling in E11.5 embryos showed no tdTomato signal in placodes or epidermis when labeled at or after E8.5 (Fig. S6).

Our data show that the population of progenitors expressing Zic5 act very differently from the population that expresses Foxi3. At the onset of expression, the Zic5 population is pluripotent and is then constrained to the four derivatives of the neural plate border lineages, finally giving rise only to cells derived from the dorsal neural tube, including the neural crest cells along the anterior-posterior axis. The Zic5 lineage thus appears to align with a model established for other pluripotency genes expressed in neural crest cells, comprising pluripotent cells that gradually become restricted to a neural crest progenitor cell (Buitrago-Delgado et al., 2015; Geary and LaBonne, 2018; LaBonne and Bronner-Fraser, 1999; Liu and Labosky, 2008).

Previous studies from our lab reported that posterior placodes, such as the inner ear and some cranial ganglia, are either absent or significantly underdeveloped in Foxi3 mutant embryos (Birol et al., 2016). Although a small increase in apoptosis was observed in the mutant tissue at the level of the hindbrain, the fate of the remaining Foxi3 mutant cells was unclear. As our Foxi3-CreER knock-in mice replace the entire Foxi3-coding region, we were able to follow the fate of the Foxi3 lineage in the absence of Foxi3 protein. We crossed male and female Foxi3-CreER mice that were also homozygous for the ROSA-tdT allele. The resulting litters contained embryos homozygous for the Foxi3-CreER allele at an expected Mendelian proportion of 25% being functionally null (Fig. 6A). We again administered tamoxifen at E7.5 to follow the fates of the CreER-expressing Foxi3 mutant cells (Fig. 6B). As in our previously published studies of Foxi3-null mice (Birol et al., 2016), the functionally null Foxi3-CreER homozygous embryos were identified by absence of the otocyst and malformed pharyngeal arches that could be easily identified as early as E9.5 (example in Fig. 6C, arrowheads). We compared the fate of Foxi3 mutant cells with Foxi3-CreER heterozygous littermates that showed similar lineage labeling as in the previous experiments (Figs 2 and 3).

The lens and olfactory placodes were unaffected in homozygous Foxi3-CreER null embryos (Fig. 6D,E; Birol et al., 2016) but the posterior placodes were defective or absent. Most strikingly, the otocyst and statoacoustic (VIII) and geniculate (VII) ganglia were absent (Fig. 6G). In addition, the trigeminal ganglion (V) was smaller and segmented, with the placodal-derived ventral (or distal) segment (D'amico-Martel and Noden, 1983) showing high tdTomato labeling compared with a more even spread of tdTomato labeling in Foxi3-CreER heterozygous controls (Fig. 6F). The adenohypophyseal placode was present but smaller or mis-shapen (data not shown). The disruption of posterior placodes was indistinguishable from that seen in our previously published conventional Foxi3-null embryos (Birol et al., 2016), reinforcing our use of the homozygous Foxi3-CreER embryos to further analyze the null phenotype.

The most significant difference we observed between the derivatives of Foxi3-CreER homozygous (null) and heterozygote embryos was significantly higher tdTomato labeling in the other three neural plate border lineages. We saw far higher numbers of tdTomato-labeled cells in the epidermis of null mutant embryos compared with controls (arrows in Fig. 6D-G,J); this difference was also obvious in embryos harvested early at E9.5 in the cranial region (Fig. S7) and the trunk (Fig. S8). Moreover, we saw significant numbers of tdTomato-labeled cells in the craniofacial mesenchyme around the forebrain and epibranchial region in all Foxi3 mutant embryos examined at E11.5 (asterisks in Fig. 6H); many of these tdTomato-labeled cells stained for SOX9, a marker of migrating neural crest cells and cartilage progenitors (asterisks in Fig. 6H′). The tdTomato-labeled neural crest cells were also found in E9.5 embryos, but the number of positive cells was far fewer (Fig. S7 shows some of the clearest examples). We also found tdTomato-labeled cells in many regions of the brain neuroepithelium (four out of five mutant embryos) at E11.5 (Fig. 6I) and, as expected, we saw fewer labeled cells in E9.5 embryos (Fig. S7). Taken together, our data show that loss of Foxi3 causes cells that would normally be restricted to cranial placodes and epidermal fates to contribute to all four neural plate border lineages.

The Foxi3-CreER null mutant embryos in these experiments had two copies of the CreER gene compared with one copy in the control (heterozygous) embryos. To ensure the increased lineage contribution to epidermis, neural crest and neural tube was due to loss of Foxi3 and not to greater efficiency of Cre recombination due to twice the amount of Cre protein, we generated single Foxi3-CreER:Foxi3-null embryos by mating our new Foxi3-CreER::ROSA-tdT mice with previously reported Foxi3-null mice (Birol et al., 2016; Edlund et al., 2014). The Foxi3-CreER:Foxi3-null embryos in this mating scheme also had greater tdTomato labeling in the epidermis and crest cells compared with the heterozygous Foxi3-CreER controls (Fig. S9), confirming our conclusion that cells lacking Foxi3 are able to contribute to a wider range of neural plate border lineages.

Placodes and neural crest cells arising from the neural plate border give rise to sensory and skeletal structures in the vertebrate head that are considered to be key innovations for the evolutionary success of vertebrates. They are believed to have arisen independently in the vertebrate ancestors, and living invertebrate chordates deploy elements of neural crest and placode gene regulatory networks in different cell populations during embryonic and larval development (Abitua et al., 2012; Gans and Northcutt, 1983; Martik and Bronner, 2021). It is a continuing subject of debate whether the independent evolution of these two populations is reflected in an embryonic segregation of the crest and placode lineages in vertebrates, or whether neural plate border cells of vertebrate embryos pass through a transient period where a multipotent cell population simultaneously expresses gene regulatory networks of several – or all – border lineages (Schlosser, 2014; Schlosser et al., 2014; Thawani and Groves, 2020).

Our analysis of the Foxi3-expressing cell lineage in this study, together with our fate mapping of Foxi3 mutant cells suggest that cells expressing Foxi3 are restricted to the placodal and epidermal lineages from the onset of Foxi3 expression. We base this conclusion on three lines of evidence. First, our Foxi3-GFP mice show that Foxi3 expression first arises in non-neural ectoderm and spreads to the edge of the neural plate. Foxi3 is never expressed broadly throughout the epiblast and its expression overlaps significantly with other markers of non-neural ectoderm, such as AP2A (Fig. 1E-F′; Movies 1-8). Second, even though some Foxi3-expressing cells also express Sox2 at the neural plate border (Fig. 1C,E-E′), our lineage tracing data from Foxi3-CreER mice shows that most Foxi3-expressing cells give rise to placodes and a small amount of epidermis. By performing lineage tracing at a range of ages (E5.5-9.5), we show that this restriction to placodes and epidermis is apparent from the earliest times of Foxi3 expression (Figs 3 and 4). We contrast this fate commitment with cells expressing Zic5 – a gene that marks pre-migratory neural crest specifically by E8.5, but which is expressed broadly in the epiblast at early stages (Fig. 5). In contrast to Foxi3, lineage tracing of Zic5-expressing cells show that this population undergoes gradual restriction from pluripotent progenitors at E6.5 to progenitors capable of generating all four neural plate border lineages at E7.5, to restriction to neural crest and dorsal neural tube by E8.5 (Fig. 5, Fig. S6). Finally, our analysis of Foxi3-CreER null mutant cells shows that the fate restriction of Foxi3-expressing progenitors to placodes and epidermis is dependent on the function of Foxi3 itself. In the absence of Foxi3, these progenitors now adopt a wider range of fates at the neural plate border, giving rise to neural crest and neural tube cells (Fig. 6, Fig. S7).

To interpret our data in light of the different neural plate border patterning models, our lineage tracing results showing that Foxi3-expressing neural plate border cells normally give rise to only placodes and epidermis are congruent with the ‘binary competence’ model of neural plate border formation, where non-neural (placodal and epidermal) lineages segregate first from neural (neural crest and CNS) lineages, and the same population of progenitors gives rise to both epidermis and placodes. As Foxi3-expressing progenitors do not normally contribute to the neural lineages, our data do not support the ‘neural plate border state model’, in which placodes and crest arise from a common progenitor population. However, it should be noted that our Foxi3 lineage tracing labels only a subset of cells in each placode, likely due to both the level of Cre expression driven by the Foxi3 locus and the moderate dose of tamoxifen used to ensure embryo survival. Although we assume this is likely a technical constraint of our Foxi3-CreER mice, it is formally possible that other, Foxi3-negative, cell populations from the neural plate border also contribute to placodes, and that these populations retain the potential to contribute to neural crest and neural tube lineages. For the future experiments, it could be instructive to use a combination of gene-specific lineage tracing and clonal analysis [e.g. CARLIN (Bowling et al., 2020), MORF (Veldman et al., 2020) and Brainbow (Weissman and Pan, 2015)]. With these techniques, we can also test whether the multipotent Zic5-expressing cell population contains subpopulations that are restricted to a subset of border lineages or whether the clonal derivatives are spread across the ectoderm.

A recent scRNA-seq study of the chick neural plate border identifies a transient population of cells referred to as BLUPs (Border-Located Undecided Progenitors; Thiery et al., 2023) that simultaneously express gene regulatory networks from multiple neural plate border lineages. Under the probabilistic model proposed by these authors, transcriptional programs from different border lineages compete in these ‘undecided’ progenitors until external influences, likely the spatiotemporal position of a progenitor, cause one lineage program to be dominant in that cell and its progeny. Viewed in the light of this model, our data suggest that Foxi3 acts early to repress neural crest and neural tube lineage gene networks. In contrast, our Zic5 lineage tracing data suggests the possibility that individual Zic5 cells are still undecided – ‘BLUPs’ in the terminology of Streit and colleagues – when Foxi3⁺ progenitors have already committed to the placode lineage. Alternatively, it is formally possible that individual Zic5⁺ progenitors are already restricted to individual border lineages; evaluating these possibilities will require lineage tracing at single cell resolution.

At present, we do not know how Foxi3 acts to restrict cells to the placodal and epidermal lineages at the neural plate border. Foxi3 belongs to the Forkhead family of transcription factors, the Forkhead domain of which can engage the outer face of DNA through its winged-helix motif (Kaufmann and Knöchel, 1996; Kaufmann et al., 1994; Singh et al., 2018). This allows the Forkhead proteins to act as pioneer factors and to bind to their DNA targets even if they are in a transcriptionally inaccessible state associated with heterochromatin (reviewed by Zaret and Carroll, 2011). It is possible that Foxi3 binds to elements of the placode gene regulatory network at early stages and acts as a positive epigenetic mark to prepare the network for transcription. Alternatively, or perhaps additionally, Foxi3 might also target elements of the neural crest and CNS gene regulatory networks and either repress them directly or prevent neural crest or neural transcription factors from accessing their targets. The region of co-expression of Foxi3 and Sox2 in the neural plate border (Fig. 1) could be a site of possible mutual cross-repression of two competing gene regulatory networks. Accordingly, our data suggest that Foxi3 drives cells away from a neural lineage and towards placodal fates. We also note that Foxi3-CreER null mutant mice show the greatest malformations in posterior placode derivatives – the inner ear, and VIIth and VIIIth ganglia – but show no phenotypes in the most anterior placodal derivatives, such as the olfactory and lens placodes. It is possible that there are multiple transcription factors that work together to sub-divide the pre-placodal domain along the anterior-posterior axis once Foxi3 restricts the neural plate border to the pre-placodal fate. Future experiments to resolve this mechanism will require examining transcriptional heterogeneity along the anterior-posterior axis of the Foxi3-expressing neural plate border and its derivatives, and identifying the direct targets of Foxi3 transcription factors through epigenomic analyses using the transgenic mouse lines we describe here; however, this is currently technically challenging given the very small numbers of Foxi3-expressing cells present in the mid-gastrulation mouse embryo. Nonetheless, the fact that Foxi3 mutant cells can adopt neural crest and neural tube fates does indeed suggest that at least a partial function of Foxi3 is to repress the gene regulatory networks of non-placodal ectodermal lineages directly or indirectly.

Our study has characterized and provided novel transgenic tools to interrogate the fate of neural plate border cells at their earliest stages of development. In future studies, our new mouse models will allow us to specifically target only the placodal or neural crest progenitors to perform signaling pathway or fate modifications. Similar future transgenic tools will enable the isolation of additional populations of the neural plate border to identify transcriptomic and epigenomic restrictions influenced by border transcription factors and to establish a more detailed gene regulatory network of ectodermal patterning.

A GSG-P2A-mVenus construct was targeted downstream of the Foxi3-coding region as described previously (Ankamreddy et al., 2023). Foxi3-GFP mice were genotyped using primers on either side of the inserted mVenus gene (Foxi3-GFP_F, 5′-CCT TCA GCA GCC CTT TCT AC-3′; Foxi3-GFP_R, 5′-TAA GTC CCC TTT CCA AGA CG-3′) that amplify a 974 bp band in mutants and 191 bp from the wild-type allele.

Functionally null mice lacking exon 2 of Foxi3 were generated as described previously (Edlund et al., 2014). Foxi3-null mice were genotyped using the following primers: f3G1, 5′-GGC CTT GTC TCA ACC AAC AG-3′; f3G2, 5′-GTT TCC TGT ATC CCT GGC TG-3′; and f3G3, 5′-CTT GGA ATG GGT TGA CTG AG-3′. f3G1and f3G2 produce a 350 bp band corresponding to the wild-type allele, and f3G1 and f3G3 yield a 600 bp band corresponding to the Foxi3 null allele.

A CreER-P2A-EGFP construct was targeted to the Foxi3 locus to replace the entire Foxi3-coding region as described previously (Ankamreddy et al., 2023). Mice were genotyped for the Foxi3-promoter and Cre junction using the following primers: Foxi3-promoter_F, 5′-AAA GCC GCT GCC GCT CTG CA-3′; and CreER_R, 5′-TTG GTC GTG GCA GCC CGG AC-3′. These detect a 368 bp band. The EGFP construct in this mouse model does not produce detectable levels of fluorescence, as described by Ankamreddy et al. (2023).

B6.Cg-Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J mice (Ai9; Madisen et al., 2010), henceforth referred to as ROSA-tdT mice, were generated at the Allen Institute and obtained from Jackson Laboratory (stock 007909). Mice were genotyped using the following primers recommended by the Jackson Laboratory: oIMR9020, 5′-AAG GGA GCT GCA GTG GAG TA-3′; and oIMR9021, 5′-CCG AAA ATC TGT GGG AAG TC-3′ (which detect the 297 bp band in wild-type embryos); and oIMR9103, 5′-GGC ATT AAA GCA GCG TAT CC-3′; and oIMR9105, 5′-CTG TTC CTG TAC GGC ATG G-3′ (which detect 196 bp mutant band).

A construct carrying a P2A-GFP-CreER fusion protein, followed by a neomycin resistance cassette flanked by FRT sites was inserted between two 3.8 kb homology arms corresponding to sequences upstream and downstream of the start codon of Zic5 in exon 2. A DTA sequence for negative selection was inserted upstream of the 5′ targeting arm. The construct was electroporated into ES cells and the neomycin-resistant colonies were subsequently selected and injected into blastocysts. Germline founders were generated and mated with ROSA-FLP mice [B6.129S4-Gt(ROSA)26Sor^tm2(FLP*)Sor/J; Jackson Laboratory stock 012930] to remove the neomycin-resistance cassette. Mice were generated by the University of Rochester Mouse Genome Editing Facility as a fee-for service. Zic5-CreER mice were genotyped using GFP2ACRE F (5′-CAC GGC ATG GAT GAG CTG TA-3′) and GFP2ACRE R (5′-GGT CTT GGT CCT GCC AAT GT-3′) to amplify a 684 bp band from the P2A-GFP-CreER fusion construct, Neo5mer2 (5′-CAT GCT CCA GAC TGC CTT G-3′) and Zic5 3′ LRR2 (5′-GGA TAC AGC CAG GAG GCA GCA TCT GA-3′) to detect the neomycin cassette with 1.2 kbp band, AcGFPF (5′-GGC ATC GTG CCC ATC CTG ATC GAG CTG A-3′) and AcGFPR (5′-GCC AGC TGC ACG CTG CCA TCC TCG ATG T-3′) to obtain 508 bp amplicon in GFP, and Zic5WT F (5′-GTT CAG ACC ACA GGG AAT GT-3′) and Zic5WT R (5′-CTG CCA CTT ACT ACC TCA CTT ATT-3′) targeting the wild-type Zic5 sequence across exon 2 to amplify an 820 bp band. Unlike the Foxi3-CreER transgenic line, Zic5^CreER/CreER mice are viable into adulthood and can reproduce.

All genotyping was performed using ear punch tissue digestion buffer with 10 mM Tris-HCl (pH 9.5), 50 mM KCl, 1.5 mM MgCl₂ (pH 8.5), 0.45% NP40, 0.45% Tween-20 and 0.1 mg/ml proteinase K, followed by thermal inactivation and proteinase K, and a PCR reaction using 2×PCR Pre-Mix (MB067-EQ2G) from SydLabs, except Zic5-neomycin PCR, which required a TaKaRa Ex Taq DNA polymerase kit (RR001A).

Homozygous ROSA-tdT females were mated with Foxi3-CreER or Zic5-CreER heterozygous males. To activate Cre recombinase, a single dose of ∼0.08-0.2 mg/g of tamoxifen with equal parts of progesterone was given to pregnant females by oral gavage at the desired gestational day post-fertilization between day 5.5 and 9.5. Additional progesterone doses were given to pregnant females gavaged before day 6 to prevent litter miscarriage. To trace the lineage of Foxi3 mutant cells, homozygous ROSA-tdT females mice carrying one Foxi3-CreER were sib-mated or bred with male mice carrying one Foxi3-null allele. Dams were euthanized by CO₂ asphyxiation and uterine horns were harvested to access the embryos. Control embryos that did not receive tamoxifen were harvested for both Foxi3-CreER and Zic5-CreER mouse lines to show absence of Cre leakiness. All manipulations of mouse lines were performed according to NIH/AVMA guidelines, and the policies established by the Institutional Animal Care and Use Committee at Baylor College of Medicine.

Embryo heads were harvested at E9.5, E10.5 or E11.5 and fixed in 4% paraformaldehyde (PFA) followed by dehydration in 15% sucrose, embedded in 7.5% gelatin in 15% sucrose (300 Bloom) and stored at −80°C. Heads were cryosectioned at 15-18 µm thickness and immunostained with antibodies for RFP (to detect or amplify tdTomato signal; 1:500; Abcam, AB62341), SOX2 (1:100; R&D Systems, AF2018), TUBB3 (1:1000; Covance, 801213), SOX9 (1:1000; Abcam, AB5535), GFP (1:500; ThermoFisher, A11122), E-cadherin (1:500; R&D Systems, AF748), FOXI3 (1:100; concentrated supernatant from NeuroMab clone, N359/28; RRID:MMRRC_066258-UCD) and AP2a (1:50; DSHB supernatant, 3B5) overnight at room temperature, followed by relevant Alexa Fluor-conjugated secondary antibodies (ThermoFisher) for 2-3 h with DAPI counterstaining. Antibodies were diluted in a solution of 5% serum, 1% bovine serum albumin (BSA), 0.1% Triton-X 100 and 0.05% sodium azide, with the preceding blocking solution with higher concentrations of serum (10%) and detergent (1%). FOXI3 and AP2a immunostaining required an additional heat-induced epitope retrieval step before blocking solution treatment to unmask the nuclear proteins by immersing slides in 0.01 M citric acid (pH 6) and microwaving at ∼900 W power for 5 min, followed by 30 min at ∼100 W (protocol modified from Galli et al., 2014). The immunostaining solutions used were made with Tris-buffered saline instead of phosphate-buffered saline with serum, BSA and detergent concentrations the same as above.

For whole-mount staining, embryos from age E6.5 to E8.0 were dissected using the protocol described by Shea and Geijsen (2007), and fixed in 4% PFA overnight at 4°C. Samples were incubated with primary antibodies for 2-3 days at room temperature followed by overnight secondary antibody incubation. For light sheet imaging, immunostained embryos were mounted in 1% agarose in a 1 ml syringe with the tip cut off and cleared in ∼80% Nycodenz (VWR, 100356-726) in 0.02 M phosphate buffer (McCreedy et al., 2021) for 3 days or until imaged. The tissue-clearing protocol using Nycodenz was modified from Hsu et al. (2022) and McCreedy et al. (2021).

For sections, areas of interest were imaged using a Zeiss Axio Imager with Apotome 2 structured illumination. Whole-mount embryos were imaged using Zeiss SteREO Discovery V20 for bright-field images and Zeiss Lightsheet I with 20× lens for fluorescent images. Post-processing of the images was carried out using ZEN software (Zeiss), FIJI (NIH ImageJ 2) and Inkscape (open-source vector graphics editor). Imaris Stitcher was used for compiling z-stack tiles and Imaris Microscopy Image Analysis software (Oxford Instruments) was used for post-processing, snapshots and video rendering. Any brightness and contrast adjustments performed for better signal visualization were applied to all corresponding images to prevent signal bias, wherever relevant.

Each sectioned embryo was evaluated to record the induced tdTomato signal. To account for intra-litter age variation, and residual effects of tamoxifen in the bloodstream or active Cre recombinase, we classified a positive record of lineage tracing to each placodal derivatives when we observed equal to or more than five RFP-positive cells in the most highly labeled section of the organ of interest in E9.5 harvests, and 15 or more RFP-positive cells in E11.5 harvests. Anything less than that was classified as unappreciable lineage contribution to the region of interest.

We thank Alyssa Crowder for outstanding technical support, Sue (Ziyang) Li for troubleshooting some of the antibody staining protocols, and Sunita Singh for concentration and purification of FOXI3 monoclonal antibody supernatant. Fig. S5 was provided by Lin Gan of the Transgenic & Genome Editing core of the Medical College of Georgia. We thank Jason Kirk and Chih-Wei Logan Hsu of Optical Imaging and Vital Microscopy core at Baylor College of Medicine for training and assistance, Joanna Jankowsky for Imaris software access, and Jason Heaney, Isabel Lorenzo, Denise Lanza and Lan Liao of the Genetically Engineered Rodent Model (GERM) core at Baylor College of Medicine for their help with generation of Foxi3-GFP and Foxi3-CreER mice. The GERM Core is funded in part by a National Institutes of Health Cancer Center grant (P30 CA125123).