The alx gene family confers segmental identity to frontonasal cranial neural crest cells

Abstract

Cranial neural crest cells in the pharyngeal arches give rise to vertebrate head skeletal structures. Hox genes confer segmental identity to the different arches, and Hox mutations result in mirror-image homeotic duplications. Frontonasal cranial neural crest cells reside anterior to the first pharyngeal arch, give rise to the midface skeleton, and are not considered part of the segmental series. Here, we demonstrate that alx genes confer segmental identity to this anterior-most population in zebrafish; alx mutations result in mirror-image homeotic transformations. We define the boundary between the frontonasal and arch populations and demonstrate that Alx expression is restricted to the anterior-most cranial neural crest cells in zebrafish and lamprey, suggesting these expression patterns were present early in the vertebrate lineage. We propose that frontonasal neural crest cells represent the anterior-most segment of the pharyngeal arch meristic series, and that Alx genes function as homeotic selector genes conferring their segmental identity.

Here they show homeotic transformation of the frontonasal region to upper jaw identity in zebrafish alx mutants. These findings revive a debated, historical theory of vertebrate head segmentation where the anterior-most region is segmentally homologous with the pharyngeal arches.

Introduction

The vertebrate cranial skeleton is largely derived from the segmental pharyngeal arches^1–3. Each pharyngeal arch has a distinct population of cranial neural crest cells with a unique identity along the anterior-posterior axis defined by hox gene expression. Pharyngeal arch one is considered the most anterior segment in the pharyngeal arch meristic series and is Hox negative, unlike the subsequent arch segments with nested Hox gene expression. Each arch gives rise to specific skeletal elements. In zebrafish, arch one cranial neural crest cells give rise to the larval upper jaw skeletal elements, including the cartilaginous palatoquadrate and pterygoid process, as well as the entopterygoid and maxilla bones. These distinct structures are patterned along the dorsoventral axis by nested expression of the homeodomain transcription factor encoding dlx gene family within pharyngeal arch one to establish different regions⁴.

In addition to the pharyngeal arches, there are cranial neural crest cells residing anterior to pharyngeal arch one, known as frontonasal cranial neural crest cells^5–7. These preoral frontonasal neural crest cells do not express dlx or hox genes^7–12. It is not known what genes might make these hox-negative frontonasal cells different from the, also hox-negative, pharyngeal arch one cells. No homeobox gene family has yet been associated with establishing the identity of these anterior-most cranial neural crest cells. To uncover zebrafish homeobox-containing genes that might function to establish frontonasal neural crest cell identity, we used unbiased single-cell RNA sequencing (scRNA-seq) in a previous study⁷. This approach parsed frontonasal neural crest cells into a distinct subpopulation enriched for the alx homeobox gene family⁷. Expression of alx genes in the frontonasal domain appears to be an ancestral vertebrate condition. In situ gene expression data shows Alx transcripts are present anterior to the stomodeum, the developing mouth, in the basal vertebrate sea lamprey¹³. Further, these genes are strong candidates for functioning in frontonasal neural crest cells because mutations in this gene family cause frontonasal dysplasia (FND) in humans. Mutations in different ALX genes produce different classes of FND each resulting in dysmorphologies in midface skeletal elements, which are likely derived from frontonasal neural crest cells^14–16. In mice, Alx4 single mutants produce phenotypes in the midface¹⁷, and mice mutant for Alx1 present median orofacial clefting and eye phenotypes¹⁸. While initial reports indicated that mouse Alx3 mutants produce no overt craniofacial phenotype¹⁹, mouse Alx3/4 double mutants produce dramatic craniofacial defects, suggesting this gene family functions together in midface development^18,19. Further, fate mapping in mice demonstrates that the most anterior element of the mouse craniofacial skeleton, the premaxillary bone, is derived from Alx3-expressing frontonasal neural crest cells¹⁰. These data demonstrate that the Alx gene family functions in the anterior-most cranial neural crest cell population, which gives rise to the vertebrate midface skeleton.

In zebrafish larvae, the ethmoid plate is a prominent midface skeletal element²⁰. The ethmoid plate has separate regions: a medial zone flanked by lateral domains^5,6,21. These medial and lateral ethmoid plate cells have distinct developmental timing, cellular morphology, and developmental origin^7,21. The lateral domains are derived from neural crest cells residing atop the stomodeum which is considered pharyngeal arch one^6,21,22. The medial domain derives from cells residing more anteriorly, near the choroid fissure of the eye²². Here, we posit that these anterior cells giving rise to the medial ethmoid plate are frontonasal neural crest cells. Of note, the precise boundary between pharyngeal arch one and frontonasal neural crest cells has not yet been clearly defined in zebrafish.

Consistent with the mixed origin of the ethmoid plate, frontonasal and pharyngeal arch one, we previously found that only the medial ethmoid plate is affected when the frontonasal-restricted alx3 gene is mutagenized⁷. Specifically, frontonasal cells fated to give rise to the medial ethmoid plate in wild types resemble pharyngeal arch one-derived lateral ethmoid plate cells in alx3 mutants. One possible mechanism underlying this phenotype is identity transformation, although we did not directly test this hypothesis in our previous study⁷. In that work, we found that mutation of only alx3 produced an overt phenotype as a single mutant, while alx1, alx4a, and alx4b single mutants were indistinguishable from wild types. A recent study reported more severe ethmoid plate defects variably develop when alx3 was mutagenized in an alx1 homozygous mutant background²³. These data motivate a careful examination of zebrafish alx combinatorial mutants.

Here, we test the hypothesis that alx genes function to specify segmental identity of frontonasal cranial neural crest cells. We used gene expression studies to demonstrate that zebrafish alx gene expression is nested within frontonasal neural crest cells. Fate mapping determined the contributions of the frontonasal versus pharyngeal arch one cells to the larval craniofacial skeleton and surrounding mesenchyme. We used transgene and hybridization chain reaction (HCR) gene expression analyses to define the precise boundary between frontonasal and pharyngeal arch one cranial neural crest cells in zebrafish. We discovered striking alx combinatorial mutant phenotypes suggestive of identity transformations. Specifically, alx1;alx3 double and alx1;alx3;alx4a triple mutants develop ectopic cartilage and bone elements resembling the upper jaw in place of the midface. These remarkable phenotypes suggest a homeotic transformation from frontonasal to pharyngeal arch one identity when alx function is lost. This conclusion is supported by in situ, transgene, and scRNA-seq gene expression studies, all finding that alx mutant cranial neural crest cells gain pharyngeal arch one identity at the expense of frontonasal identity. Further, we found that the boundary between cranial neural crest cell populations is present in sea lamprey, a basal vertebrate, suggesting that the mechanisms distinguishing these two populations were present early in the vertebrate lineage. Our study raises the intriguing possibility that the anterior-most population of cranial neural crest cells is the first segment of a meristic series with the pharyngeal arches.

Results

alx genes are expressed in frontonasal cranial neural crest cells

Our previous 24 h post-fertilization (hpf) scRNA-seq indicated that alx paralog expression was enriched in frontonasal cranial neural crest cells⁷. However, expression of all the alx paralogs was not equivalent across the frontonasal population. Subsets of cells express one, two, or three alx paralogs. To test if alx expression might be nested, similar to the dlx gene family in the pharyngeal arches, we utilized HCR in situ probes recognizing alx1, alx3, and alx4a transcripts. Simultaneously labeling multiple alx gene expression patterns in the same animal showed that after cranial neural crest cell migration (24 hpf), alx3 expression is the broadest, surrounding the nares and extending to the most dorsal anterior cranial neural crest cells, as well as ventral to the eye and into the choroid fissure. Interestingly, alx3 expression ends near the choroid fissure; at this stage, we did not detect any alx transcripts in the pharyngeal arches (Fig. 1A). The alx4a expression pattern is slightly more restricted. Transcripts were again detected between the nare and the eye but did not extend as far dorsal, anterior, or ventral as alx3. Expression was detected ventral to the eye and into the choroid fissure. Eye iridophores also expressed alx4a, consistent with previous studies²⁴. Finally, alx1 expression was limited to the cells between the nasal placode and the eye, with a thin patch extending ventral to the eye and into the choroid fissure. Overall, these data demonstrate that alx1 is nested within the alx4a domain, which is nested within the alx3 domain in the anterior-most post migratory cranial neural crest. To quantify this nested pattern, we measured the amount of expression overlap by HCR between these paralogs, finding that 54% of alx3-positive cells co-express alx4a and that 22% of alx3-positive cells express all three paralogs (Supplementary Fig. 1). Our previously published 24 hpf scRNA-seq dataset¹² is consistent with this nested expression pattern (Fig. 1B), and we generated two-color UMAP feature plots to observe overlapping expression for each pair of alx paralogs (Supplementary Fig. 1). Thus, alx genes form a concentric pattern. alx triple-positive cells are at the center, the area between the nare and the eye, surrounded by double-positive cells, and then an outer ring of single-positive cells. There is minimal alx expression posterior to the choroid fissure at the pharyngula stage.

Fig. 1. alx-enriched frontonasal neural crest cells give rise to the zebrafish larval midface.

A Wild-type embryos were labeled with HCR probes detecting alx3, alx4a, and alx1 transcripts and imaged in whole mount. The same pattern was observed across eight replicates. B single cell RNA-sequencing on 24 hpf FACS isolated (fli1:EGFP and sox10:mRFP double-positive) cranial neural crest cells. Cells were clustered into four distinct populations annotated by gene enrichment¹². Individual feature plots for alx genes reveal enrichment in frontonasal (fn) cranial neural crest cells relative to the anterior (aa) and posterior (pa) pharyngeal arches. C Schematic synthesizing fate mapping experiments, including D–G and Fig. S4, with previous studies. At least three tracings of each region were observed. D An unconverted sox10:Kaede control animal shows no photoconverted cells at either 28 hpf or 4 dpf. E An animal with frontonasal cells converted was imaged at 28 hpf then reimaged at 4 dpf. F Converted cells around the eye and posterior to the choroid fissure including the anterior arches were imaged at 28 hpf and again at 4 dpf. G Restricted area of converted cells near the choroid fissure was imaged at 28 hpf and again at 4 dpf. In E–G, approximate areas of photoconversion are outlined with dotted lines and structures containing photoconverted cells are labeled with purple font. Structures without photoconverted cells are labeled with white font. Asterisks indicate labeling on the contralateral side from the laser passing through the animal at 28 hpf, which was not considered when building the overall fate map. In all panels, frontonasal (fn), nares (n), and choroid fissure (cf) of the eye are indicated. Anatomical landmarks include the nare (n) and stomodeum (s). The dashed line indicates the choroid fissure plane. Frontonasal (fn), anterior (aa), and posterior (pa) pharyngeal arch cranial neural crest cells are indicated. Larval structures including the medial (mep) and lateral (lep) ethmoid plate, pterygoid process (ptp), palatoquadrate cartilage (pq), maxilla (mx), midface mesenchyme (mm), and nasal mesenchyme (nm) are marked. Scale bars are 100 μm.

To test if alx expression in this region is restricted to cranial neural crest cells, we examined expression of the most broadly expressed paralog, alx3, in transgenic embryos where fli1:EGFP labels post migratory cranial neural crest cells. Most fli1:EGFP positive cranial neural crest cells in the frontonasal region expressed alx3 (Supplementary Fig. 2). Only minimal signal was detected beyond fli1-positive cells, likely due to non-specific staining. Thus, at this post migratory stage alx3 expression appears restricted to frontonasal cranial neural crest cells, consistent with previous conclusions²⁵.

To determine when alx expression initiates in cranial neural crest cells, we observed expression of alx1, alx3, and alx4a at 12, 15, and 18 hpf. These stages correspond to neural crest cell specification, early migration, and late migration, respectively. In these experiments, we used the sox10:EGFP transgene to mark cranial neural crest cells during these early stages. Of the alx paralogs we examined, alx1 is expressed earliest. alx1 transcripts were detected in the anterior-most cranial neural crest cells just after specification (Supplementary Fig. 3). None of the other paralogs were detectable in cranial neural crest cells at this stage. At this early stage, we also detected alx1 in the cephalic mesoderm, consistent with previous studies^25,26. At 15 hpf, during early migration stages, alx1 remains the only paralog detected in cranial neural crest (Supplementary Fig. 3), present in the anterior-most stream of migrating cells. At this stage alx1 expression remains in the cephalic mesoderm along with weak alx3 expression. We did not detect alx4a expression at this stage. During late migration stages, 18 hpf, alx4a and alx3 expression had initiated in migrating cranial neural crest, while alx1 remains the strongest paralog expressed (Supplementary Fig. 3). Expression of alx genes is restricted to cranial neural crest cells at this stage. Similar to the postmigratory stage (Fig. 1A), no alx genes were detected in the cephalic mesoderm. This study reveals interesting alx paralog expression dynamics, alx1 is the first to be expressed and is present in the anterior-most cranial neural crest cells just after specification. alx3 and alx4a are not expressed until late migration stages, then in post-migratory cells, become strongly expressed, whereas alx1 becomes the most restricted.

Fate mapping reveals frontonasal cranial neural crest cell contributions to the midface

Cells residing below the eye, just anterior to the choroid fissure, have been fate-mapped to give rise to the medial ethmoid plate, while those directly below, and posterior to, the choroid fissure give rise to the lateral ethmoid plate^21,22(Fig. 1C). However, to our knowledge, alx-expressing frontonasal cells anterior to these landmarks have not yet been fate-mapped. Thus, we set out to determine their contribution to the larval midface. Compared with unconverted controls (Fig. 1D), photoconversion of most of the frontonasal cranial neural crest cells at 28 hpf in sox10:kaede transgenic animals showed broad labeling of the midface at 4 dpf (Fig. 1E). We found converted larval mesenchymal cells surrounding the nares, including cells between the nares and eyes. This broad label also marked mesenchymal cells residing dorsal and lateral to the ethmoid plate. Finally, this pan-frontonasal conversion gave rise to cells of the medial ethmoid plate but not the lateral ethmoid plate, consistent with previous studies²² (Fig. 1C). Thus, most alx-expressing frontonasal cells remain in position and do not contribute to skeletal structures at 4 dpf except for the medial ethmoid plate. Photoconversion of the converse population, ventral and posterior to the choroid fissure extending toward the first pharyngeal arch, labeled the lateral ethmoid plate, pterygoid process, palatoquadrate, and Meckel’s cartilage (Fig. 1F), again in line with previous fate mapping of pharyngeal arch one^5,27,28. These broad fate mapping studies revealed the respective contributions of frontonasal neural crest cells versus pharyngeal arch one neural crest cells to the zebrafish larval skeleton and facial mesenchyme. We performed finer mapping, photoconverting smaller groups of cells near the region of the presumptive boundary between frontonasal and pharyngeal arch one (Fig. 1G, Supplementary Fig. 4). These experiments further demonstrate that cells ventral to the eye but anterior to the choroid fissure contribute to the medial ethmoid plate while more posterior cells contribute to pharyngeal arch one-derived skeletal structures. Our findings reproduce previous lineage tracing work on arch one derivatives and provide the first comprehensive fate map for zebrafish frontonasal cranial neural crest cells.

prrx1a expression delineates the extent of pharyngeal arch cranial neural crest cells

Our scRNA-seq and HCR gene expression analyses revealed that at the pharyngula stage, alx3 is a pan-frontonasal marker but is only minimally expressed in the pharyngeal arches (Fig. 1A, B). Lineage tracing indicates that frontonasal and pharyngeal arch one populations each give rise to distinct craniofacial populations at the larval stage. These findings suggest that there may be a boundary between the frontonasal and pharyngeal arch populations at embryonic stages, which is preserved during cranial skeletal development. To test this hypothesis, we searched our cranial neural crest scRNA-seq dataset for genes that were specifically expressed in the pharyngeal arches but restricted from the frontonasal population; the transcription factor encoding gene prrx1a is a good candidate fitting these requirements (Fig. 2A). In these data, prrx1a is strongly expressed in many cranial neural crest cells, including the anterior and posterior pharyngeal arches, but is largely restricted from the alx3-positive frontonasal cells. Although some frontonasal cells express prrx1a, we nevertheless reasoned that expression might be a useful indicator of the boundary between anterior arch and frontonasal populations in Uniform Manifold Approximation and Projection (UMAP) space. Consistently, there are few alx3;prrx1a double-positive cells in this dataset (Fig. 2A). To determine if this putative marker of the frontonasal-pharyngeal arch boundary might be present in intact embryos, we performed time-lapse imaging of a prrx1a:RFP knock-in reporter transgene crossed to fli1:EGFP, marking all post-migratory cranial neural crest cells. At early stages (24 hpf), strong RFP expression was found in all the pharyngeal arches, extending anteriorly to the cranial neural crest cells atop the stomodeum, including the fli1:EGFP positive cells residing below the eye. Expression continues until just below the choroid fissure, at which point, strong expression abruptly stops (Fig. 2B). Anterior to this boundary, there is only sparse prrx1a:RFP expression in fli1:EGFP-positive cells surrounding the nares. As development proceeds and morphogenesis occurs (32-40 hpf), the boundary between prrx1a-expressing and non-expressing cells remains; no new cells initiate prrx1a:RFP expression, but the boundary itself moves anterior and dives toward the midline (Fig. 2C, D, Supplementary Movie 1). We propose that this perduring transgene expression pattern reveals the in vivo position of the pharyngeal arch one-frontonasal cranial neural crest cell boundary, recapitulating the scRNA-seq analyses.

Fig. 2. The boundary between pharyngeal arch one and frontonasal cranial neural crest cell populations resides under the eye.

A Two-color UMAP feature plots indicate complementary expression of alx3 and prrx1a in the anterior arch and frontonasal cranial neural crest cell populations, respectively. Heat map indicates relative expression of each transcript. B–D Cranial neural crest cells were live labeled with fli1:EGFP and pharyngeal arches with prrx1a:RFP. Double transgenic animals were time-lapse imaged and single frames from the recording are presented. The same pattern was observed across six replicates. E–J Migrating cranial neural crest cells were live labeled with sox10:mRFP. Post-migratory cranial neural crest cells are double positive for sox10:mRFP and fli1:EGFP while vascular cells are fli1:EGFP single positive. Double transgenic animals were time-lapse imaged and single frames from the recording are presented with stages labeled. The same pattern was observed across six replicates. In all panels, frontonasal (fn), anterior arch (aa), and posterior arch (pa) cranial neural crest cell populations are indicated. Anatomical landmarks in embryos include the nares (n), choroid fissure (cf) of the eye, and stomodeum (s), dashed line indicates the choroid fissure plane. In E-J, the asterisk marks a fli1;sox10 double positive post-migratory cell, the arrowhead marks vascular cells, and pharyngeal arch one (a1) cells are indicated. Scale bar is 100 μm.

The fli:EGFP transgene is expressed in vascular cells in addition to cranial neural crest cells which may complicate interpretation of which cells in the boundary region express prrx1a. To confirm that the prrx1a:RFP expression boundary corresponds with a boundary between populations of cranial neural crest cells, we analyzed prrx1a:RFP expression in a sox10:EGFP background which marks cranial neural crest cells without labeling vascular cells (Supplementary Fig. 5). These studies revealed that prrx1a:RFP in this region is restricted to cranial neural crest cells and further support the model that prrx1a:RFP labels the extent of the maxillary region of pharyngeal arch one.

Distinct migratory cranial neural crest cells streams meet ventral to the eye

We next wished to determine if this boundary between frontonasal and pharyngeal arch one is established from different streams of migrating cranial neural crest meeting after migration near the choroid fissure plane. To track cranial neural crest cells during and after migration, we captured time-lapse recordings of wild-type animals carrying the sox10:mRFP and fli1:EGFP transgenes (Fig. 2E–J, Supplementary Movie 2). The sox10 reporter is expressed during early neural crest cell migration, while the fli1 reporter initiates after the cells reach their destination. By 15 hpf distinct streams of neural crest cells started their migration including those that will populate the frontonasal and pharyngeal arch one regions. By 16 hpf, the earliest arriving frontonasal neural crest cells initiated fli1:EGFP expression, while cell migration continued. The pharyngeal arch one stream began to migrate under the eye at this stage. By 20 hpf, the pharyngeal arch one and frontonasal populations have met each other under the eye as they initiate fli1:EGFP expression, indicating they are post-migratory. Importantly, fli1:EGFP single positive sox10:mRFP negative vasculature that invades the choroid fissure has begun to demarcate this boundary between the accumulating frontonasal and pharyngeal arch one cranial neural crest cell populations. By 22 hpf, the populations remained separate on each side of the developing choroid fissure plane marked by the forming vasculature. The pharyngeal arch one cells had further filled in under the eye, extending from the anterior boundary with the frontonasal population, posteriorly to the rest of the arch one population near the stomodeum, which was readily observable at this stage. By 24 hpf, fli1:EGFP expression was widespread among post-migratory cells as most of the migrating cells had arrived. The boundary between frontonasal and pharyngeal arch one cells remained near the choroid fissure. At 27 hpf, the choroid fissure boundary remained in place, as few cells trickled into the arches and the frontonasal domains. Overall, these recordings indicate that distinct streams of frontonasal and pharyngeal arch one cranial neural crest cells meet under the eye near the vasculature of the choroid fissure where they stop their migration.

alx3;alx4a double mutants develop dorsal ectopic cartilage

We next wished to examine phenotypes associated with mutating genes with expression restricted to the frontonasal side of the boundary. After alx3, alx4a has the next widest expression pattern (Fig. 1A, B), motivating us to examine alx4a function. Although alx4a single mutants do not have overt larval phenotypes⁷, we reasoned that double mutants between these two broadest alx paralogs might have more pronounced combinatorial phenotypes, as was found in mice¹⁹. Indeed, compared with wild-type siblings alx3;alx4a double homozygous larvae presented a “bulldog” phenotype^29,30. In these double mutants, the dorsal aspect of the anterior head was compressed compared with the lower jaw, resulting in apparent jaw protrusion. Meanwhile, overall development did not appear delayed (Fig. 3A). Dissection of the neurocranium revealed that these double mutants develop a reduced medial ethmoid plate and truncated parasphenoid bone, while patterning of the lateral ethmoid plate and the viscerocranium were overtly unaffected (Fig. 3B, Supplementary Fig. 6). Examination of whole-mount preps also revealed stacks of dorsally projecting Alcian Blue-positive structures extending from the lateral ethmoid plate, as well as free-floating Alcian Blue-positive nodules (Fig. 3A). These structures were present in 100% of double mutants and were never observed in wild types or single mutants, which retain mesenchyme at this position. Further analyses of double mutants using sox10:mRFP and sox9a:EGFP to label chondrocytes confirmed that these Alcian blue positive structures are ectopic dorsal cartilage elements (Fig. 3C). To closely examine if bone development is also affected in these mutants, we examined live animals with bone fluorescently labeled with Alizarin Red and the fli1:EGFP transgene labeling cranial neural crest-derived cells (Fig. 3D). In wild types, the pharyngeal arch one-derived maxilla was readily observed near the corner of the mouth. In double mutants, the maxilla appeared variably expanded dorso-medially; in addition, we observed free-floating ectopic bone that develops near the ectopic cartilage (Fig. 3D). These skeletal phenotypes are present at high penetrance in alx3;alx4a double homozygous mutants (Fig. 3E, Supplementary Table 1). Combining these data with our fate mapping, we conclude that cells fated to become midface mesenchyme in wild types instead develop into ectopic cartilage and bone in alx3;alx4a double mutants.

Fig. 3. Frontonasal cranial neural crest cells give rise to ectopic skeletal structures in alx3;alx4a double mutants.

A Wild-type and alx4a^-/-;alx3^-/- double homozygous mutant larvae were stained for cartilage and bone with Alcian Blue and Alizarin Red, respectively, and imaged in whole mount. Asterisk indicates an ectopic Alcian Blue-positive nubbin dorsal to the ethmoid plate. B Larvae stained as in A were dissected and the neurocranium was flat mounted and imaged. Guillemet indicates reduced medial ethmoid plate while asterisk marks ectopic Alcian Blue-positive column extending dorsally from the lateral ethmoid plate. C Double transgenic sox10:mRFP;sox9a:EGFP wild-type and double homozygous mutant larvae were imaged with a frontal view by confocal microscopy at 4 dpf. Enlargements are to the right for each genotype. Asterisks indicate ectopic sox9a:EGFP;sox10:mRFP double-positive cartilage structures near the ethmoid plate, arrowheads mark more dorsally residing ectopic cartilage nubbins. D Wild types and double mutants with cranial neural crest-derived cells labeled with fli1:EGFP and bone with Alizarin Red were live imaged with a frontal view by confocal microscopy at 6 dpf. Arrowheads denote ectopic Alizarin Red-stained bony elements, caret indicates expanded maxilla, and asterisk marks ectopic cartilage structures. E A penetrance table indicates phenotype frequency in wild type and double mutants, asterisks indicate significant difference (*0.05, **0.01, ***0.001, two-sided Fisher’s exact test) in penetrance between double mutants and wild types. The ethmoid plate (ep), maxilla (mx), palatoquadrate (pq), pterygoid process (ptp), entopterygoid (en), and parasphenoid (ps), medial (mep) and lateral (lep) ethmoid plate, nare (n), mouth (m) and Meckel’s cartilage (Mc) are indicated. Scale bars are 100 μm.

Ectopic structures resembling the pharyngeal arch one-derived skeleton develop in severe alx mutants

We next examined alx3;alx1 double homozygous mutants and alx3;alx4a;alx1 triple homozygous mutants, reasoning that the latter might represent complete loss of alx function phenotype in frontonasal cells. Imaging whole-mount skeletal preps laterally revealed a bulldog phenotype with a foreshortened midface, similar to alx3;alx4a double homozygous mutants, in both alx3;alx1 and alx3;alx4a;alx1 mutants (Fig. 4A). However, closer examination of dissected neurocrania revealed a different phenotype in these combinations compared with alx3;alx4a double homozygous mutants (Fig. 4B). Specifically, instead of the medial ethmoid plate, a bilaterally symmetrical plate-like structure forms near the midline, with paired processes extending ventro-laterally. These structures develop where the wild-type midface mesenchyme resides (Fig. 4A). Strikingly, these ectopic cartilages were shaped similarly to the palatoquadrate and the pterygoid process, pharyngeal arch one-derived skeletal elements. Characteristic of homeotic transformations, these structures exhibited mirror-image symmetry with the orthotopic skeletal elements. Specifically, the putative duplicated palatoquadrate cartilages extend dorsally with apparent ectopic pterygoid processes articulating with the lateral ethmoid plate near the location where the orthotopic pterygoid processes meet this structure. Moreover, dermal bones resembling ectopic entopterygoids develop along the putative duplicated pterygoid processes in both these mutant genotypes (Fig. 4B, D, Supplementary Table 1). A more careful analysis of the bones with Alizarin red-stained, fli1:EGFP transgenic fish from frontal views revealed that this bone resides near the midline, deeper in the pharynx (Fig. 4C). Further, in these fluorescent images, we observed an elongated maxilla projecting dorsally and medially in some individuals. Others developed separate bones shaped like the maxilla, positioned dorsal and medial to the orthotopic maxilla, which we interpret to be duplicated maxillae. These dramatic skeletal phenotypes, including the foreshortened midface in alx combinatorial mutants, suggest that alx genes provide frontonasal identity to the anterior-most population of cranial neural crest cells and that removing alx gene function may transform frontonasal cells into pharyngeal arch one identity.

Fig. 4. Frontonasal cells give rise to skeletal structures resembling pharyngeal arch one-derived structures in alx3;alx1 and alx3;alx1;alx4a combinatorial mutants.

A Wild-type, alx3^-/-;alx1^-/-, and alx3^-/-;alx1^-/-;alx4a ^-/- double and triple homozygous mutant larvae were stained for cartilage and bone with Alcian Blue and Alizarin Red and imaged in whole mount. B Larvae stained as in A were dissected and the neurocranium, together with the palatoquadrate and pterygoid process, were flat mounted and imaged. C Wild types, double, and triple mutants with cranial neural crest-derived cells labeled with fli1:EGFP and bone with Alizarin Red were live imaged with a frontal view by confocal microscopy at 6 dpf. D A penetrance table indicates the frequency of each phenotype, asterisks denote significant difference (*≤0.05, **≤0.01, ***≤0.001, ****≤0.0001, two-sided Fisher’s exact test, exact p-values in Supplementary Table 1) in penetrance between mutants and wild types. In all panels, the maxilla (mx), pterygoid process (ptp), entopterygoid (en), palatoquadrate (pq), lateral ethmoid plate (lep), medial ethmoid plate (mep), nares (n), and Meckel’s cartilage (Mc) are indicated wild types. In mutants, structures resembling duplicated, ectopic maxilla (mx’), pterygoid process (ptp’), entopterygoid (en’), and palatoquadrate (pq’) are marked. Scale bars are 100 μm.

Frontonasal neural crest cells ectopically express pharyngeal arch one-associated genes in alx mutants

Our interpretation that the alx mutant phenotypes are frontonasal-to-pharyngeal arch homeotic transformations motivates examining the boundary between these populations, which we discovered can be observed with prrx1a:RFP expression (Fig. 2). As we observed above, in wild types prrx1:RFP is strongly expressed in all the pharyngeal arches and extends under the eye to the choroid fissure plane, the putative boundary between pharyngeal arch one and frontonasal populations (Fig. 5A). In alx3;alx4a double mutants, which develop ectopic cartilages dorsal to the ethmoid plate at larval stages (Fig. 3), we observed fli1:EGFP positive frontonasal cranial neural crest cells ectopically expressing prrx1a:RFP including a few cells in both the midface mesenchyme and the nasal mesenchyme between the eye and the nares which reside in the position predicted to give rise to the ectopic cartilages in these mutants (Figs. 1C, 3). We also observed ectopic expression in cranial neural crest cells over the eye (Fig. 5A). In animals double homozygous mutant for alx1;alx3 which produce dramatic putative skeletal transformation phenotypes, we observed prominent ectopic prrx1a:RFP expression in frontonasal cells. Most of the cells anterior to the frontonasal-pharyngeal arch one boundary, marked by the choroid fissure plane, ectopically express prrx1a:RFP. There was similar broad ectopic expansion in frontonasal cranial neural crest cells in alx3;alx1;alx4a triple homozygous mutants. Of note, the choroid fissure angle does not appear changed in mutants compared to wild types at this stage. Thus, in all alx mutant combinations that produce ectopic skeletal structures the prrx1:RFP transgene, which is normally restricted to the pharyngeal arches at this stage, is ectopically expressed in frontonasal cells. We quantified the expression of the prrx1a:RFP transgene by measuring transgene volume (Supplementary Fig. 7). We find that the volume is significantly higher in mutant conditions compared with wild types. These results are consistent with our interpretation of frontonasal to pharyngeal arch one transformation (Fig. 4). In all mutant conditions, the fli1:EGFP transgene indicates that the frontonasal neural crest cells appear to reside in the correct position with similar numbers to wild types (Fig. 5A). To confirm this observation, we quantified the volume of the fli1:EGFP transgene and found no significant difference between wild types and mutants (Supplementary Fig. 7). Finally, we performed a TUNEL stain to assess differences in fragmented DNA representing cell death between wild types and mutants. Contrary to previous reports of apoptosis in Alx mutants^18,19, we found that cell death was unchanged between genotypes (Supplementary Fig. 7). These findings indicate that the mechanism underlying the phenotype is an identity change rather than changes in cell migration, proliferation, or survival similar to what we previously reported for alx3 mutants⁷.

Fig. 5. The alx gene family represses expression of pharyngeal arch-associated genes in frontonasal cranial neural crest cells.

A Double transgenic prrx1a:RFP;fli1:EGFP wild-type and alx mutant embryos were imaged laterally by confocal microscopy. B, C Wild-type and alx mutants were triple labeled with the fli1:EGFP transgene and HCR detecting prrx1a and dlx2a transcripts then imaged by confocal microscopy. The same individual embryos are presented in B (prrx1a and fli1:EGFP) and C (dlx2a and fli1:EGFP). B” and C” are enlargements from B and C demonstrating detected transcripts reside within frontonasal cranial neural crest cells. In all images, the frontonasal (fn) and anterior arch (aa) populations of cranial neural crest cells are labeled. Indicated anatomical landmarks in embryos include the nares (n), choroid fissure (cf) of the eye, forebrain (fb), and stomodeum (s). Dashed line indicates the choroid fissure plane. Asterisk indicates ectopic anterior arch gene expression in frontonasal cells. Caret in A indicates ectopic anterior arch gene expression in cranial neural crest cells over the eye. Numbers indicate the frequency of ectopic gene expression in frontonasal cells. The denominator in each ratio represents the number of replicates for that assay and genotype, the numerator indicates the frequency with which that pattern was observed. Scale bar is 100 μm.

To confirm that the prrx1a transgene expression, including the ectopic expression in mutants, faithfully recapitulates endogenous prrx1a gene expression, we used in situ HCR to monitor transcripts. We performed these experiments in the fli1:EGFP transgenic background, marking all cranial neural crest cells. Again, we observed no overt difference in the number or position of frontonasal cranial neural crest cells in any mutant combination compared with wild types (Fig. 5B, C). In 26 hpf wild-type embryos, we observed that prrx1a is expressed in the pharyngeal arches, as well as the cells residing atop the stomodeum, and under the eye. These data are consistent with our previous in situ hybridization results³¹. In wild-type embryos, prrx1a expression abruptly stops near the choroid fissure plane (Fig. 5B), like the prrx1a:RFP transgene (Fig. 2). Next, we examined expression in alx3^-/-;alx1^+/-;alx4a^+/- embryos. We found subtle ectopic prrx1a expression just anterior to the boundary. Finally, we examined alx mutant combinations that produce ectopic pharyngeal arch one-like structures in the larval skeleton. In double and triple mutants, we observed ectopic expression of prrx1a in the fli1:EGFP positive medial midface and nasal mesenchyme (Fig. 5B). We next looked for expression changes in other genes that are restricted to the pharyngeal arches. We examined the canonical pan-arch marker dlx2a, which is expressed in the forebrain and the pharyngeal arches but is excluded from frontonasal cells^1,4. In agreement with previous studies, we observed dlx2a expression in the pharyngeal arches and the forebrain at 26 hpf in wild-type larvae using in situ HCR (Fig. 5C). Strikingly, we found that alx double and triple mutants, which produce a homeotic transformation phenotype at larval stages, display ectopic dlx2a expression in frontonasal cells at the pharyngula stage (Fig. 5C). Enlarging the regions near the choroid fissure confirmed that the ectopic prrx1a and dlx2a expression was in fli1:EGFP expressing frontonasal cranial neural crest cells (Fig. 5B–C). Expression of prrx1a and dlx2a in the pharyngeal arches was unaffected (Fig. 5). Our finding that frontonasal cells express pharyngeal arch-associated genes further supports our hypothesis that frontonasal cells remain in the same position in alx mutant embryos but are transformed to pharyngeal arch identity.

Frontonasal cranial neural crest cell transcriptomic identity is transformed to pharyngeal arch identity in alx mutants

Single-cell transcriptomics is a highly sensitive method of detecting changes in cell type abundance^32,33. We used scRNA-seq to determine if the relative proportions of frontonasal versus pharyngeal arch cellular identity were different in alx mutants compared with wild types. To this end, we compared single-cell transcriptomes from FACS-isolated 48 hpf cranial neural crest cells (Fig. 6A) from wild types and the two mildest alx mutant conditions that produce a phenotype, alx3 single mutants and alx3;alx4a double homozygous mutants (Fig. 6B). We reasoned that a transcriptomic signature indicating cell identity changes should be present even in conditions where the hypothetical transformation is mild. At 48 hpf, wild-type cranial neural crest cells can be subdivided into different clusters, including the anterior arches, two frontonasal populations, and posterior arch populations (Fig. 6B). We also detected a cluster expressing landmark genes indicating differentiating skeletal cell identity. These results are consistent with previous transgenic and single-cell transcriptomic studies and our live transgenic imaging (Fig. 6A)^9,34. We also detected a few cells that clustered into a small population with neuronal identity. While all the same clusters were present in wild types and both mutant genotypes, the proportion of cells in the frontonasal and anterior arch clusters differed between mutants and wild types. Specifically, there were relatively fewer cells with frontonasal transcriptomic identity in mutants compared with wild types. Concurrently, we observed a higher proportion of cells with anterior arch identity in mutants (Fig. 6B). The relative proportion of cells in all other populations including posterior arches, the differentiating skeletal cells, and the neuronal population, was unchanged across genotypes (Supplementary Table 2), reflecting the skeletal and gene expression phenotypes we observed specifically affecting the frontonasal derived structures (Figs. 3–5). Consistent with the relatively mild skeletal phenotypes in alx3 mutants and moderately more severe phenotypes in alx3;alx4a double homozygotes, the proportional gain of anterior arch at the expense of frontonasal cells was gene dose dependent, becoming more pronounced as increasing numbers of alx genes were disabled (Fig. 6B). To investigate whether the frontonasal cells in our double mutant population are losing frontonasal identity and gaining anterior arch identity, we quantified overall direction of change for population specific genes. We used marker gene lists of frontonasal and anterior arch populations to identify changes in gene expression between wild types and double mutants (Fig. 6C). We found that 67% of frontonasal marker genes were lower in alx3;alx4a double mutants compared with wild types. We further found that 84.2% of anterior arch marker genes were higher in the frontonasal population in mutants compared to wild types. To confirm these trends were not true across all genes, we show that there is no clear bias in the direction of change between mutants and wild types when surveying all genes in the dataset. Taken together, these findings indicate that anterior arch identity is gained in the frontonasal population of alx3;alx4a double mutants at the expense of frontonasal identity. To more broadly look at changes to transcriptional programs in these mutants compared to wild types, we performed a GO enrichment analysis comparing the frontonasal populations of wild types and double mutants (Fig. 6D). We found decreases in transcriptomic programs associated with roof of the mouth development, neurocranium morphogenesis, skeletal morphogenesis, and neural crest cell differentiation. These changes are consistent with a loss of frontonasal identity and gain of pharyngeal arch identity. We see decreases in pathways associated with skeletal development of frontonasal-derived structures. Further, the frontonasal-derived structures become skeleton prior to pharyngeal arch-derived structures, explaining the decrease in skeletal morphogenesis pathways. Finally, differential expression analyses were used to identify candidate genes with increased or decreased expression in the frontonasal population (Fig. 6E). These changes were then visualized using violin plots. Expression of gata3, which is spatially restricted in wild types, increases in the frontonasal population in alx3;alx4a mutants. Expression of pax3a and pax9 is reduced in the frontonasal population in alx3;alx4a mutants. These analyses reveal new expression changes to explore in intact embryos to further test the transformation hypothesis at a molecular level. This transcriptomic evidence indicates that frontonasal identity is transformed to pharyngeal arch identity in alx mutants, even in a more mild mutant condition.

Fig. 6. The alx gene family functions to confer frontonasal versus anterior arch transcriptomic identity.

A Double transgenic sox10:mRFP;fli1:EGFP wild-type embryos, like those used for FACS and single cell transcriptomic analyses, were imaged from oblique lateral and lateral views. Scale bar is 100 μm, these patterns were observed across five replicates. B Double-positive sox10:mRFP;fli1:EGFP cells were isolated by FACS and analyzed by scRNA-seq. UMAP plots from wild type, single, and double mutants were clustered then annotated using landmark genes. The percentage of total cells with either frontonasal or anterior arch identity is indicated for each genotype. C Pie charts show changes in gene expression in the frontonasal population. From left to right, changes in frontonasal marker genes, anterior arch marker genes, and all genes in frontonasal cells are shown. D GO enrichment analysis was used to identify expression motifs that were decreased in the frontonasal population. Changes related to the frontonasal to pharyngeal arch identity transformation are labeled. E Violin plots of gene expression changes associated with a frontonasal to pharyngeal arch identity transformation are shown. The comparison used to run a two-sided Wilcoxon rank-sum test for differential gene expression analysis is indicated for each gene. The log2 fold change and p-values for the expression change of these genes in the indicated populations is shown. Frontonasal (fn), anterior (aa) and posterior (pa) pharyngeal arch populations of cranial neural crest cells are labeled. Nares (n) and differentiating skeletal cells (sk), and neuronal cells (nu) are indicated.

Frontonasal cells gain arch one gene expression and lose frontonasal gene expression in alx mutants

Our scRNA-seq findings prompted further exploration of markers of pharyngeal arch one identity and frontonasal identity in alx3;alx1 double mutants. Because this sequencing was not performed in the most severe genotype, we used fluorescent in situ hybridization to test whether markers of frontonasal and pharyngeal arch identity were changed in the severe alx3;alx1 genotype. We found that genes expressed in the maxillary portion of pharyngeal arch one in wild types were expanded in alx3;alx1 double mutants. The first candidate gene we investigated was gata3 (Fig. 6E). At 30 hpf and 48 hpf gata3 expression is restricted to the anterior-most region of pharyngeal arch one, right below the choroid fissure (Fig. 7A). At both stages, gata3 expression expands into frontonasal cranial neural crest cells, as visualized with fli1:EGFP, in alx3;alx1 double mutants. These findings align with the differential expression observed in the single-cell data. We next tested for expression changes of another maxillary pharyngeal arch one gene expressed near the stomodeum, barx1. We observed similar changes in barx1 expression (Fig. 7B). In wild types, at both 36 and 48 hpf, barx1 expression is restricted and strong in the anterior portion of pharyngeal arch one, which gives rise to the palatoquadrate. In alx3;alx1 double mutants, expression of barx1 expands into the frontonasal domain, supporting our hypothesis that a duplicated palatoquadrate is forming in these mutants. After looking at more restricted pharyngeal arch one genes, we investigated changes in a broader pharyngeal arch marker, lhx6a. Expression of the mouse ortholog, Lhx6, has been shown to expand in mouse Alx1 mutants¹⁸. We found broad expression of lhx6a in the pharyngeal arches in wild types at 30 and 48 hpf. However, this expression was restricted from the frontonasal cells (Fig. 7C). In alx3;alx1 double mutants, this expression expanded into the frontonasal region. These findings validate previous work done in the mouse system and demonstrate expansion of another pan-arch gene, similar to what we observe with prrx1a and dlx2a. We wondered whether genes with wild-type expression restricted to the ventral domain of arch one would be expanded in alx3;alx1 mutants. We therefore examined the expression of dlx5a, which is a ventrally restricted pan-arch marker (Fig. 7D). We found no change in dlx5a expression, further supporting that the ectopic skeletal structures represent duplication of the maxillary portion of pharyngeal arch one. We found evidence in our scRNA-seq data that pharyngeal arch one identity is gained in alx mutants at the expense of frontonasal identity (Fig. 7C, E). To further test this, we investigated frontonasal-restricted genes that were lost in the mutant single-cell dataset in intact animals. We performed in situ HCR and fluorescent in situ hybridization for pax9 and pax3a, respectively, in 48hpf embryos (Fig. 7E, F). We found that expression of both of these frontonasal genes was lost in the alx3;alx1 double mutants compared to wild types. Genes that show subtle changes in the scRNA-seq dataset show more dramatic spatial changes in intact embryos in the more severe mutant genotype. Of note, the choroid fissure appears to be improperly formed in these severe genotypes at 48 hpf, which was not observed at earlier stages. Taken together, these findings indicate that alx mutants lose frontonasal identity and gain pharyngeal arch identity in the anterior-most cranial neural crest.

Fig. 7. The frontonasal region gains anterior arch gene expression at the expense of frontonasal gene expression in alx mutants.

A Wild types and alx3;alx1 double mutants were double labeled with the fli1:EGFP transgene and fluorescent in situ hybridization for gata3 at 30hpf (A, A’) and 48hpf (A”,A”’). Expression of gata3 is shown in the presence (A, A”) and absence (A’, A”’) of the fli1:EGFP cranial neural crest marker. B Wild types and alx3;alx1 double mutants were labeled with fluorescent in situ hybridization for barx1 at 36hpf (B) and 48hpf (B’). C Wild types and alx3;alx1 double mutants were double labeled with the fli1:EGFP transgene and fluorescent in situ hybridization for lhx6a at 30hpf (C, C’) and 48hpf (C”, C”’). Expression of lhx6a is shown in the presence (C, C”) and absence (C’, C”’) of the fli1:EGFP cranial neural crest marker. D Wild types and alx3;alx1 double mutants were labeled with fluorescent in situ hybridization for dlx5a at 36hpf (D) and 48hpf (D’). E Wild types and alx3;alx1 double mutants were double labeled with the fli1:EGFP transgene and in situ HCR for pax9 at 48hpf. Expression of pax9 is shown in the presence (E) and absence (E’) of the fli1:EGFP cranial neural crest marker. F Wild types and alx3;alx1 double mutants were labeled with the fluorescent in situ hybridization for pax3a at 48hpf. G Cartoon summarizing wild-type gene expression at two stages and how those genes are regulated by alx. In all images, the frontonasal (fn) and anterior arch (aa) populations of cranial neural crest cells are labeled. Indicated anatomical landmarks in embryos include the nares (n) and choroid fissure (cf) of the eye. Dashed line indicates the choroid fissure plane. Asterisk indicates ectopic anterior arch gene expression in frontonasal cells. Caret indicates loss of frontonasal gene expression in frontonasal cells. Ratios in A–D indicate the frequency of animals with ectopic frontonasal gene expression. Ratios in E and F indicate the frequency of animals with frontonasal expression loss. The denominator in each ratio represents the number of replicates for that assay and genotype. Scale bar is 100 μm.

The boundary between frontonasal and pharyngeal arch cells is present in agnathans

Our hypothesis that alx and dlx genes delineate frontonasal versus pharyngeal arch identity in zebrafish motivated us to test if this configuration might be broadly present in other vertebrates. We elected to examine these frontonasal and pharyngeal arch markers in the basal vertebrate, sea lamprey. Previous work using colorimetric in situ hybridization on lamprey concluded that Alx and DlxB, respective orthologs of alx and dlx2 genes in zebrafish, overlap in the pre-oral or frontonasal domain of the developing lamprey head¹³. However, these colorimetric in situ analyses did not allow for high-resolution three-dimensional spatial information, or for simultaneous probing of multiple genes in the same embryo and larvae. We reexamined the spatial expression of these genes using HCR probes specific to the lamprey during relevant developmental stages allowing confocal three-dimensional analyses of these gene expression patterns (Fig. 8). Like in zebrafish, Alx expression in lamprey is restricted to the anterior-most cranial neural crest cells at stage T25.5, corresponding to approximately 24 hpf in the zebrafish embryo. In contrast to our previous report, this expression does not overlap with DlxB, which is expressed in the pharyngeal arches. To better observe three-dimensional gene expression, we prepared an animation revealing largely mutually exclusive Alx and DlxB expression in the frontonasal region and pharyngeal arches, respectively (Supplementary Movie 3). We observed the same restricted, mutually exclusive expression domains for both genes at stage T26.5, corresponding to 48 hpf zebrafish. We cannot be certain that this juxtaposition of gene expression patterns is in the same relative position as the one we observed near the zebrafish choroid fissure. Nevertheless, these results suggest that the genetic program delineating the frontonasal segment of cranial neural crest cells from the pharyngeal arches was present early in vertebrate evolution, predating gnathostomes and the evolution of a hinged jaw.

Fig. 8. Alx and DlxB expression in lamprey cranial neural crest cells is complementary.

A, B Alx and DlxB transcripts were fluorescently labeled by HCR in wild-type lamprey at stage T25.5 (A) and T26.5 (B) embryos. 3 of 3 embryos from each stage showed Alx and DlxB expression restricted to frontonasal (fn) and anterior pharyngeal arch (aa), respectively. Dashed line indicates the putative boundary between frontonasal and pharyngeal arch cranial neural crest cells delineated by Alx and DlxB gene expression. Scale bar is 20 μm.

Discussion

Several lines of evidence indicate that alx genes confer frontonasal versus pharyngeal arch identity to the anterior-most population of cranial neural crest cells. These include dramatic transformations of the medial ethmoid plate and surrounding mesenchyme to upper jaw skeletal elements, ectopic arch one transgene expression in frontonasal cranial neural crest cells, and cell proportion changes detected by transcriptomics in alx mutants. An extensive set of endogenous gene expression changes in alx mutants further support the transformation model. Although these in situ hybridization experiments are not quantitative in nature, they indicate that an array of arch one-associated genes expand into the frontonasal domain in alx mutants, while frontonasal associated genes are downregulated. Whether Alx proteins bind directly to these differentially expressed gene loci to control frontonasal versus pharyngeal arch identity is currently difficult to assess due to limitations in antibody availability in the zebrafish system.

As other homeobox gene families like dlx and hox subdivide cranial^35–37, ascribing homeotic selector function to the alx family aligns with our understanding of how positional identity is established during craniofacial development. Recent studies of mouse Alx1 mutants further support this model. Iyyanar et al. demonstrated that Alx1 mutants develop midface skeletal phenotypes and lose frontonasal mesenchyme identity markers like Pax7, concurrent with ectopic expression of jaw mesenchyme genes Lhx6 and Lhx8 in frontonasal cells. These skeletal and gene expression phenotypes were enhanced in Alx1;Alx4¹⁸. Although these authors did not explicitly propose that these were frontonasal to pharyngeal arch transformations, their work is consistent with our model. Additional work in mice demonstrates that altering Retinoic Acid signaling changes Alx gene expression, resulting in frontonasal to pharyngeal arch identity transformations reminiscent of those we describe here³⁸ These results lend further evidence for our model and suggest that the mechanisms conferring identity to the anterior-most population of cranial neural crest cells are conserved. In alx3;alx1 double and alx3;alx1;alx4a triple mutants, ectopic bone and cartilage skeletal structures resemble duplicated upper jaw elements in morphology and relative position to each other (Fig. 4). However, an alternative explanation of these phenotypes could be that precocious juvenile or adult skeletal structures are forming in these mutant larvae. For example, in the wild-type chondrocranium, the laminae orbitonasalis are dorsal projections of the anterolateral ethmoid plate that develop into plate-like cartilages by 5.6 mm or ~12 dpf²⁰. Yet, the ectopic structures observed in alx mutants do not develop in the same position as the laminae orbitonasalis, nor are they shaped like this structure. Further, the laminae orbitonasalis develop adjacent to the medial ethmoid plate, not in its place. Similarly, we considered that the ectopic bones in mutants might be precocious premaxillae, which are bilaterally paired dermal bones that form anterior to the maxillae and meet at the midline in wild types. However, the ectopic bones observed in alx3;alx1 double and alx1;alx3;alx4a triple mutants are not anterior to the maxillae. Given the gene expression changes that accompany the skeletal phenotypes in alx mutants, the best explanation of these phenotypes remains a homeotic transformation from frontonasal to pharyngeal arch one identity when alx genes are disabled.

That the anterior-most population of cranial neural crest cells is competent to acquire pharyngeal arch identity raises the intriguing possibility that this population is part of a meristic series with the pharyngeal arches. This interpretation is reminiscent of the “premandibular arch” hypothesis espoused by Huxley and DeBeer³⁹. Briefly, this theory proposes that the trabeculae, part of the neurocranium, were ancestrally gill-bearing structures arising from a premandibular pharyngeal arch segment, arch-0¹. We now know that the trabeculae are derived from the maxillary domain of pharyngeal arch one, discrediting the theory^1,40–42. However, this does not rule out that the anterior-most cranial neural crest cell population is part of the pharyngeal arch meristic series. In an essay countering the existence of a premandibular arch, Kimmel & Eberhart argued that understanding the geographical boundaries of the first arch is of central importance for considering the arch-0 problem. At the time, new data demonstrating that the mandibular portion of pharyngeal arch one gives rise to the trabeculae was a major argument against a premandibular arch. We agree with this interpretation and embrace their argument here by newly defining the anterior geographical boundary where pharyngeal arch one ends and frontonasal begins. They did not consider the frontonasal population, which largely remains as undifferentiated mesenchyme at larval stages, in their argument. Our new work here demonstrating that this population can behave like the first pharyngeal arch may warrant reconsidering the existence of a premandibular arch.

It is interesting that the observed transformation results in only dorsal derived maxillary structures developing from frontonasal cells. This might be partially explained by the absence of Endothelin signaling in the frontonasal region which is required for ventral mandibular skeletal structures to develop⁴³. Still, we have not ruled out the possibility that the phenotype we observe in our mutants is a transformation of only a maxillary-like dorsal subset of cells. Future work investigating different pharyngeal arch one-associated structures including nerves, muscles, and ligaments will help distinguish between a bona fide segmental homeosis and a partial transformation. Another alternative interpretation is that the frontonasal population is part of hox-negative pharyngeal arch 1 rather than an independent segment. Although this interpretation is not supported by our gene expression studies considering the geographical boundaries of the first arch.

One prediction of the hypothesis that frontonasal cells represent a premandibular arch is that alx genes function along with hox genes, conferring neural crest segmental identity along the anterior-posterior axis. There are several similarities between hox and alx genes in this context. Like the hox code, the “alx code” is established early in development; migrating cranial neural crest cells already express hox or alx genes based on their anterior-posterior position before they arrive in the arches of the frontonasal region, respectively^7,44,45. Our data here indicate that segmentally restricted alx expression was likely present in early vertebrates (Fig. 8), like hox genes^46,47. In fact, anterior-posterior regionalization by alx and hox may have already been in place in the common ancestor of amphioxus and vertebrates^48,49, prior to the evolution of cranial neural crest. Further, studies from Hydra indicate that alx may have functioned to confer anterior-most identity in early metazoans⁵⁰. Thus, it is intriguing to consider that neural crest segmentation could have been superimposed on preexisting regionalization along the anterior-posterior axis conferred by both hox and alx genes⁵¹.

Another similarity between hox and alx genes patterning the anterior-posterior axis is their interchangeability and redundancy. Some ‘hox-swap’ experiments suggest that hox paralogs are functionally equivalent for defining segmental identity^52–54. Nature has performed a similar experiment for the Alx genes. Alx3 has been independently lost in several vertebrate lineages including the chick. In mouse, Alx3, but not Alx1 or Alx4, expression encircles the eye. In the chick, where Alx3 was lost, this region expresses Alx1 and Alx4⁵⁵. Thus, one could ask what is important evolutionarily, the gene coding region, or just the circumocular expression by any Alx family member? Future alx-swap experiments could directly test an ‘alx-equivalency’ model during cranial neural crest segmental identity establishment.

Unlike the true pharyngeal arches, where there are endodermal pouches between segments, there is no described endodermal partition between the frontonasal and pharyngeal arch one domains in zebrafish. Here, we uncovered a previously cryptic boundary near the choroid fissure plane between these two populations, seen by gene expression and migration. However, evidence from other fishes suggests the possibility of an ancestral endodermal boundary separating these populations in the form of a pre-oral pharyngeal pouch in non-teleost fishes, such as gar and bichir^56,57. These results may provide additional evidence that the anterior neural crest cell population may be serially homologous with the pharyngeal arches and that an ancestral pre-mandibular endodermal pouch may have separated the pre-mandibular “pharyngeal arch-0” from the rest of the meristic series.

This work uncovers what genetic programs distinguish the anterior-most, frontonasal neural crest cell population from the more posterior cranial neural crest cell populations residing in the pharyngeal arches. While hox genes delineate the posterior pharyngeal arch segments from the hox-negative arch 1, in our model alx genes delineate the most anterior population from the alx-negative arch 1. Further, we propose that these frontonasal neural crest cells might be considered a pharyngeal arch segment, representing the first meristic structure in the pharyngeal arch series.

Methods

Zebrafish husbandry, strains, and mutagenesis

All zebrafish experiments used the AB strain

This zebrafish background lost the natural sex determinant⁵⁸, thus we cannot genotype for sex. All experiments are performed at stages before phenotypic sex is apparent, and phenotypes do not present with a sex bias. Animals were maintained and staged according to established protocols^59,60. All our work with zebrafish has been approved by the University of Colorado Institutional Animal Care and Use Committee (IACUC). Protocol number for animal research: 00188. Previously published alx3^co3003, alx4a^co3000, and alx1^co3002 mutant lines and genotyping protocols were used in this study⁷. The following transgenic lines have been previously reported: Tg(prrx1a:RFP)⁶¹, Tg(fli1:EGFP^y1)⁶², Tg(sox10:mRFP^vu234)⁶³, sox9a:EGFP (sox9a^zc81Tg)³⁴, and Tg(sox10:kaede) (zf393Tg)⁶⁴.

48 hpf FACS isolation and scRNA-seq

33 double transgenic fli1:EGFP;sox10:mRFP embryos per genotype were dissociated into single-cell suspension using cold-active protease from Bacillus licheniformis, DNase, EDTA, and trituration⁶⁵. Approximately 80,000 live, double-positive cells per genotype were sorted into tubes pre-coated with Fetal Bovine Serum with a MoFlo XDP100 and subjected to scRNA-seq. Sorted cells (at 3.27 × 10⁵ cells/ml in wild type, at 2.72 × 10⁵ cells/ml in alx3-/-, and at 2.86 ×10⁵ cells/ml in alx3-/-;alx4a-/- with ~85% viability) were loaded into the 10X Chromium Controller aiming to capture > 10,000 cells per sample. 2 × 150 paired-end sequencing Chromium 10X Genomics libraries were sequenced using the NovaSEQ6000. Recovered sequencing reads were mapped to a custom Cell Ranger reference based on the GRCz11/danRer11 genome assembly using the Cell Ranger pipeline (10X Genomics). This Cell Ranger reference contained the alx3 transcript from Ensembl release 91 (GRCz10) mapped to the GCRz11 genome in place of the release 101 (GRCZ11) version due to inaccurate truncation of the 3’ UTR in the most recent version⁷. The 10X Genomics Cell Ranger pipeline was used to identify cell-associate barcodes. Seurat v5.1.0⁶⁶ was used for quality control, filtering out cells with >10% mitochondrial counts, less than 500 features, and more than 6000 features. Cell cycle phase-associated genes were regressed out using a list of S- and G2M-phase genes built into Seurat, which was converted to zebrafish genes with BioMart⁶⁷. Dimensional reduction was performed using principal component analysis (PCA), where the first 10 PCs were used to generate clusters (resolution 0.3) which were visualized using a UMAP plot (dimensions 1:10). Major cranial neural crest cell derivatives were annotated based on published immunofluorescence and scRNA-seq profiling⁹. Gene expression was plotted on UMAP plots using FeaturePlot. Changes in marker genes in the frontonasal population were calculated by creating gene lists for markers of the anterior arch or frontonasal populations. Then the average expression of each gene in the cluster per genotype was calculated using the counts identity in Seurat. These average counts were compared between genotypes across lists of frontonasal marker genes, anterior arch marker genes, and all genes. GO enrichment analysis was performed using the GO enrichment analysis tool from the Gene Ontology Consortium^68–70. Differential expression analyses were performed using the Seurat tools, and differentially expressed genes were visualized using the violin plot command. A Wilcox test was used for the DEG analysis.

The raw, feature-barcode matrix can be accessed from the GEO database through GEO Series accession number GSE333598. Code available: https://github.com/AbiRMM/alx3-alx4-scRNA-seq-Analysis).

24 hpf zebrafish single-cell RNA sequencing analysis

Expression mapping on UMAPs of published zebrafish cranial neural crest cell scRNA-seq datasets at 24 h post-fertilization (GEO accession number GSE220081) was performed as previously described^12,71.

Zebrafish fluorescent in situ hybridization and imaging

Fluorescent in situ hybridization was performed as previous⁷². All probes have been previously described^22,72–75. Zebrafish embryos were then embedded in 0.2% agarose and imaged with an Andor Dragonfly 301 spinning disk confocal system.

Zebrafish in situ hybridization chain reaction (HCR), imaging, and Imaris quantification

For alx3 we used the same probe as published⁸. For alx1, alx4a, and dlx2a we ordered oligos from Molecular Instruments® using the NM001045074, XM001340930.6, and ENSDART00000111386.4 transcripts, respectively, and preferentially targeting the 3’ UTR sequence. However, the whole sequence was provided to the manufacturer, and ORFs may have also been targeted. The probe set size was 12 for alx1, 20 for alx4a, and 20 for dlx2a. The prrx1a probe was already available from Molecular Instruments®. We followed the manufacturer’s protocol for whole-mount zebrafish embryos and larvae. Briefly, after fixation overnight at 4 °C, embryos were dehydrated and permeabilized with a series of 100% methanol washes and stored overnight. Rehydration with a series of graded methanol in phosphate-buffered saline with 0.1% Tween 20 (PBST) was then performed. Samples were pre-hybridized for 30 min at 37 °C with the buffer provided by Molecular Instruments. Later, samples were hybridized in probe solution containing 2 pmol of each probe and incubated overnight at 37 °C. After two 5× SSCT washes, samples were amplified in Molecular Instruments-provided Amplification Buffer containing 30 pmol of snap-cooled hairpin pairs. Samples were left overnight in the dark at room temperature. The next day, a series of washes in 5× SSCT was performed to remove excess hairpins. Finally, the samples were continuously washed in 5x SCCT for a week on a nutator at 4 °C protected from light before imaging. Zebrafish embryos were then embedded in 0.2% agarose and imaged with an Andor Dragonfly 301 spinning disk confocal system. Acquisition parameters and fluorescence adjustments were applied linearly and equally to all samples. Digital mRNA absolute quantitation was performed using dot detection methods⁷¹. For each embryo, the total numbers of transcripts (Imaris spots tool) were captured for alx3 first, then for each of the other genes sequentially until all three signals overlapped. The total number of spots where two or more transcripts overlap was tabulated and used to determine the number of frontonasal cranial neural crest cells expressing alx genes combinations. For transgene quantification, Z-stacks of fluorescent transgene signal were analyzed using Imaris, with display adjustment settings were uniformly set across all samples. After generating a surface scene, a region of interest (ROI) was set to capture the frontonasal region, anterior to the choroid fissure and extending into the nare. Surface detail was set to 2.41 and the smooth surface feature was selected. Surface thresholds (absolute intensity) were held constant across samples and set manually to best represent the signal. The sum of the volume was used to assess transgene signal and significance was tested between genotypes using Welch’s two-sample t-test.

Skeletal preparation, bright field and Nomarski imaging, and fluorescent microscopy

Skeletal preparations were performed as previously described^76,77 and larval skeletons were scored for phenotype penetrance after genotyping. Whole mount imaging was performed in a Zeiss SteREO Discovery V8 coupled to Axiocam 105 color camera. Alcian Blue/Alizarin Red-stained skeletons were dissected and flat-mounted for imaging on a Leica DMi8 inverted microscope equipped with a Leica DMC2900 camera. Fluorescent images and time-lapse recordings were captured using a Leica DMi8 microscope equipped with an Andor Dragonfly 301 spinning disk confocal system. For time-lapse imaging, embryos were mounted in low-melt agarose with clove oil and continuously imaged at consistent intervals. Images were analyzed using Imaris and Fiji⁷⁸. Acquisition parameters and fluorescence adjustments were applied linearly and equally to all samples. A rolling ball background subtraction was applied to HCR images in Fiji to eliminate background fluorescence.

Lineage tracing and imaging

Kaede photoconversion was performed in a Zeiss 880 confocal microscope using one pulse of UV laser (405 nm) at 60% power, performed unilaterally so that the contralateral side serves as an internal control.

TUNEL assay

At 26 hpf, fish transgenic for fli1:EGFP were fixed with 4% paraformaldehyde in 1X phosphate-buffered saline (PBS) overnight at 4 °C. Fish were then rinsed (3 × 5 min in PBS), dehydrated (3 × 10 mins in methanol), and stored overnight at −20 °C. Next, samples were serially rehydrated in solutions containing methanol and PBS containing 0.1% Tween20 (PBST). Fish were incubated in 30 µg/ mL Proteinase K for 25 minutes, post fixed in 4% PFA for 20 minutes, and rinsed (3 × 10 min in PBST). A subset of fish were then treated with 0.04 U/µl of TURBO DNase in 1X reaction buffer for 30 minutes and rinsed (3 × 10 min in PBST) to serve as a positive control. Following these preparation steps, the Click-iT™ Plus TUNEL Assay Kit for In Situ Apoptosis Detection with Alexa Fluor 647 picolyl azide dye from ThermoFisher Scientific was used as described by the manufacturer. Samples were imaged via confocal microscopy. The number of cells double positive for fli1:EGFP and TUNEL anterior to the choroid fissure were manually quantified with image display settings uniform between wild-type and mutant samples. The image display settings for the TUNEL channel were reduced in samples treated with TURBO DNase to increase cellular resolution.

in situ hybridization chain reaction (HCR) for lamprey samples

After fixation in 4% PFA in 1X PBS overnight at 4 C, larvae were dehydrated and permeabilized with a series of 100% methanol washes (4×10 min followed by 1×50 min) and stored at 4 C overnight. Rehydration with a series of graded methanol (75, 50, 25%) in phosphate-buffered saline with 0.1% Tween 20 (PBST) was then performed and followed by 5 washes of 100% PBST. All washes were performed for 5 minutes each at room temperature. Larvae were then treated with 30 ug/ml proteinase K in PBST for 15 min at room temperature, washed two times in 1 ml PSBT without incubation, and post-fixed with 1 ml of 4% PFA for 10 min at room temperature. After fixation, larvae were washed 5 times for 5 minutes with 1 ml of PBST. Samples were pre-hybridized for 30 min at 37 °C with Molecular Instrument-provided buffer. Later, samples were hybridized in probe solution containing 2 pmol of each probe (probe set size DlxB:12 and Alx4:17) and left incubating overnight at 37 °C. After two 5× SSCT washes, samples were pre-amplified in 500 μl of Molecular Instruments-provided Amplification Buffer for 1 hour before moving into amplification buffer containing 30 pmol of snap-cooled hairpin pairs. Samples were left overnight in the dark at room temperature. The next day, a series of washes in 5× SSCT was performed to remove excess hairpins. Samples continued washing in 5x SCCT for a week on a nutator at 4 °C protected from light before imaging. Finally, samples were cleared and mounted in 2% agarose in CUBIC2⁷⁹.

Petromyzon marinus (sea lamprey) gene target sequence

>PmAlx4AACTCTGGGTGAAATTTAATTGAGGAGGGCTGATTTCTCTCGGGGCCGCGCCTGCTCGGATATATGCGGCGGTGCGTAGAAGGGCCGGCAGCGATTCGTGGCTGACGCGGAGCAGGAGATGACATCTCCGATCCCTCGCATCAGATAGGCTGGCCGGAGGTCGCCCCCTTTCAATAGCTCAGCAGCACCGCGACTGCATTGCCTCAGTCGCAGCGATTCGGCGGCGTGCGTACGCGTGCGTTATTAATACGAAGACAGCCACCCACCCGCCCGGTCCCTGTTCGGTACGGCCCGCTGCAACGCAGACTTTTTTTTCGACAACAGCCGAGTCTTGCCTCGCATAGCCACGCACGTAGAGAGAAAGGGACAGACAGACGACGAGGAGGAGGAATAGAGACAGAGACAGCGTTTTCACGACTTCAGAAGTACGGCTCAAACGCTCGCCGCGAATCGACTCGAGGGAAGTGAAGTCAGCGCAGATCGTTCTATAGCGCTCCGTACAATACGAAGTGGGGAAGGGTTGTGTGTGTGTGGGGGGGGGAGTTATTGCTCGTTGCAGAGACAGACGTTCAGGGAAAAGTTTGAGCTCGTGGCCGCTGGTGCCCGCCGTCGGCTTTTAAATGCAGTGTCGAGAAGATCTGTGCGTTTTGAACTCCGTGCAACTGACATCGTCCTCCTCTCGACCGTTACCCGACTCGTGCGCCACTGCAAATCTCTCGGCCATGAACCAGTCGCCCACGATGGACGCCTCCTCGTACCTGGCTCTCAAGCAGTCGTCGCTAAAGGATTCCCCGGACAAGTCGAGCGGCGAATCGGTGTCCGAGCGAGACGGAGGAGGCACGCCTGGCGTCAAGGCGGCCAAGAAGAGGCGCAATCGGACGACCTTCACGAGCTACCAGCTCGAGGAGCTGGAGAAGGTCTTCCAGAAGACCCACTACCCGGACGTGTACGCTCGAGAGCAGCTCGCACTGCGCACAGACCTCACGGAAGCGCGAGTGCAGGTCCGCTCATGACACGTGACATCATCCGCTCATGA

>PmDlxBATGACGGGCGGCTTCGACCAGCAGCAGTTGCGGAACTTCCACGCCGCCGCCTCTTCCTCTTCGTTCCATGGCATGCACCCGCTGCATCAGCAGGAGCAGCAGCAGCAGCACCTGCAGCAGCAGCGGCAGCAGCAGCAGCAGCTACAGCCCGAGAGCGCAGCGACTACGGGCGTTGGATTTTACGCATCAGCGCCGTACGCCGCTTACCAGCCATCCCCAGCCGCCTTCTATGGCCCGCCACCTCCGCAGGGCCAGCCGGCTATCTACAGCGAAGGGAGCTACTTAAACCATAGCCACCACCACCACCAACAACATCACCAGCAGCAGCAGCAGCAGCACGCGCACCATCATCAGCAGCAGCAGCAGCACCAACAACAGCAGCACCAACAACAGCAGCAACAGCAGTTGAGTGCGTATTACGCGGCTGCCGAAGCACTGGATGCGCCCTATGACGCCAAGGAGCGCGGCGACTACCCCGGGTATAACGCGGCTGCCGGAGATGCGCAAGACGGCAAAGCATTCGACAGAGATAAGGACGACTCTGAGCCGGAGGCGCGCCTCGTGAACGGCAAGCCCAAGAAAGTCCGCAAGCCCCGAACGATCTACTCGAGTTTCCAGCTCGCAGCCCTGCAGAGGCGCTTCGAGAGGACTCAGTACCTCGCCCTGCCCGAGCGAGCCGAGCTCGCGGCCTCTCTGGGACTCACGCAGACACAGGTCAAGATCTGGTTCCAAAACCGGCGCTCCAAGTTCAAGCGCCTCGTCAAGAGCGGCGAACTCGGCTCGGAGCACAGCCCCGGGGCCAGCGATTCCATGACCTGCGACTCCCCTCCCTCGCCGGCCCCCTGGTCCGACGCGATCCCCCACGGAGCCCCCGCGCCGCAGGGCCTCCCCACGGGGGGTCTCCAGACCCCGTCGACGGCGCTCGGCCTGCGCACCCCGTCGCCGACGCGGGCGCACCACGACGCGCCGCTGATGCAAGTCCACGTGGCGCCGTCGTCGACTCCCACGGCATCTTCCACGTCGTCGTCGTTGTCGTCGTCGTCGTCGTCATCGTCGTCGTCGTCGTCGTCCTCTTCGTCGACGTACCAGCCGTGGTACCGGCCGCCCCACAACTCGACCTTTGCTCCTCCGTCCTCGACTGCCGTCGCCGCGGCCATGCAGCCCCTGTGCGGTGGAGCTGCTGTCAGCGCCGCGGTCATTTATTGA

Statistics and reproducibility

No statistical tests were used to predetermine sample size. No data were excluded from the analyses. The investigators were not blinded to genotype during analyses. All statistical tests are indicated in the figure legend in which they are used. In brief, a Fisher’s exact test was used to compare differences in phenotype penetrance. A Wilcoxon rank sum test was used to compare differences in gene expression i the single cell RNA sequencing. Differences in gene expression volume were calculated used a two tailed T-test. All numbers, replicates, and p-values are indicated in each figure. All microscopy images are representative of multiple replicates as indicated in figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: J.T.N.; Methodology: J.M.M., J.T.N., R.B.Z., A.M.M., C.A.H., N.W., K.R., L.N, M.K.K.; Software: N/A; Validation: J.T.N.; Formal analysis: J.M.M., A.M.M., K.R., D.M. J.T.N.; Investigation: J.M.M., A.M.M., R.B.Z, N.W, N.D.M, J.T.N.; Data curation: J.M.M., A.M.M., K.R., J.T.N.; Writing—original draft: J.M.M., J.T.N.; Writing—review & editing: J.M.M., R.B.Z., A.M.M., G.G.; D.M., J.T.N. Visualization: J.M.M., A.M.M., R.B.Z., J.T.N.; Supervision: J.T.N.; Project administration: J.T.N.; Funding acquisition: J.T.N.

Peer review

Peer review information

Nature Communications thanks Thomas Schilling and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

The National Institutes of Health, National Institute for Dental and Craniofacial Research (NIDCR) provided support (F32DE029995 to J.M.M., R01DE030448 and R01DE029193 to J.T.N.), National Institute of General Medical Sciences (NIGMS) T32GM141742 (to M.K.K.) An RBI pilot grant (to J.T.N.) funded RNA sequencing. The National Science Foundation provided Graduate Research Fellowship Program support to A.M.M (1938058) and to G.G. (2439026).

Data availability

All data are available without restriction. Most data are contained in the figures and supplementary information. Source data is included providing raw data for image quantification. Single-cell RNA sequencing data has been deposited in NIH GEO using accession number GSE333598. Source data are provided with this paper.

Code availability

The code and filtered data used for scRNA-seq analyses can be found at: https://github.com/AbiRMM/alx3-alx4-scRNA-seq-Analysis.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74434-w.