Nkx2.7 is a conserved regulator of craniofacial development

Abstract

Craniofacial malformations arise from developmental defects in the head, face, and neck with phenotypes such as 22q11.2 deletion syndrome illustrating a developmental link between cardiovascular and craniofacial morphogenesis. NKX2-5 is a key cardiac transcription factor associated with congenital heart disease and mouse models of Nkx2-5 deficiency highlight roles in cardiac development. In zebrafish, nkx2.5 and nkx2.7 are paralogues in the NK4 family expressed in cardiomyocytes and pharyngeal arches. Despite shared cellular origins of cardiac and craniofacial tissues, the function of NK4 factors in head and neck patterning has not been elucidated. Molecular evolutionary analysis of NK4 genes shows that nkx2.5 and nkx2.7 are ohnologs resulting from whole genome duplication events. Nkx2.7 serves as a previously unappreciated regulator of branchiomeric muscle and cartilage formation for which nkx2.5 cannot fully compensate. Mechanistically, our results highlight that Nkx2.7 patterns the cranial neural crest and functions upstream of Endothelin1 to inhibit Notch signals. Together, our studies shed light on an evolutionarily conserved Nkx transcription factor with unique functions in vertebrate craniofacial development, advancing our understanding of congenital head and neck deformities.

Craniofacial malformations have been linked to congenital heart defects, as in 22q11.2 deletion syndrome, but the mechanisms linking these lineages remain unknown. Here they show that zebrafish nxk2.7 is expressed in cardiopharyngeal progenitors and has roles in craniofacial development that cannot be compensated for by nkx2.5.

Introduction

Craniofacial development is guided by complex interactions between mesodermal, endodermal, and ectodermal populations in the pharyngeal arches (PAs) and migrating cranial neural crest cells (CNCCs) from the neural tube¹. The mesodermal core of the anterior PAs is derived from the cardiopharyngeal field (CPF), a developmental pool that houses multipotent progenitors of cardiomyocytes, branchiomeric muscles (BMs), and connective tissues^2–7. Given this intertwined developmental trajectory, it is not surprising that cardiac and facial defects are often identified together in patients⁸. Yet, the regulatory hierarchy directing the CPF towards cardiac versus BM fates has not been fully decoded. Here, we exploit our understanding of cardiac transcriptional pathways in the CPF to uncover Nkx2.7 as a regulator of craniofacial morphogenesis.

A conserved mesodermal network involving Nkx2.5 and other key transcription factors, Tbx1, Pitx2, and Isl1, governs cardiac and skeletal muscle progenitor specification in mice, chick, and zebrafish^6,9–12. Moreover, vital insights from the tunicate Ciona robusta have deepened our understanding of the complex cross-antagonism between NK4 and Tbx1/10 that occurs in these lineages mediated through inhibition of GATAa activation¹³. Murine studies demonstrate a similar regulatory mechanism in which Tbx1 binds and activates a Fgf10 enhancer¹⁴. Yet, in different heart field progenitor populations, Nkx2-5 binds and represses the same enhancer activity and simultaneously competes with Isl1 for overlapping homeodomain binding sites. In humans, Tbx1 is haploinsufficient in 22q11.2 deletion syndrome (or DiGeorge Syndrome); these patients develop cardiovascular and craniofacial anomalies implicating an essential role in the CPF^2,15.

Despite extensive investigation of these CPF regulatory networks, open questions remain regarding the distinct, yet similar functions of murine NK4 transcription family members, Nkx2-5 and Nkx2-6, in cardiopharyngeal morphogenesis. Homozygous targeted interruption of Nkx2-5 produces defects in heart tube looping and ventricular myogenesis¹⁶. Surprisingly, mice homozygous for Nkx2-6 mutations are viable and develop normal cardiac and posterior pharyngeal structures¹⁷. However, Nkx2.5^−/− Nkx2.6^−/− embryos at E10.5 exhibit severe dilatation of the pharynx with reduced pharyngeal endodermal cell number along with exacerbation of the cardiac chamber defects observed in the absence of Nkx2.5¹⁸. Yet, these embryos were not examined for deformations of the anterior pharyngeal or jaw structures. Taken together, these results underscore the evolutionarily conserved transcriptional hierarchy directed by NK4 factors orchestrating cardiac and pharyngeal muscle development in chordates. However, the mechanisms by which NK4 family members underwent subfunctionalization and acquired discrete roles remain unknown.

Phylogenetic reconstruction reveals that derivatives of the NK4 family in zebrafish include nkx2.5, nkx2.7, and nkx2.3. nkx2.5 and nkx2.7 are expressed in the cardiomyocytes, anterior lateral plate mesoderm (LPM), and PA mesoderm, yet nkx2.7 is also expressed in the PA endoderm¹⁹. However, nkx2.3 is more commonly used as a pharyngeal pouch marker due to its robust expression primarily in PA endodermal epithelium^19–21. Our prior studies elucidate mechanisms by which Nkx family genes mediate first heart field (FHF) cardiomyocyte identity and pattern second heart field (SHF) progenitors^22–24. Moreover, these data underscore the importance of nkx2.7 in SHF identity maintenance^22,23 and point to subtle defects in jaw morphogenesis in the nkx2.7^−/− embryos²⁴. Despite these findings, the function of nkx2.7 in CPF progenitor differentiation essential for craniofacial development has yet to be investigated.

Given the continuous and reciprocal tissue-tissue interactions required during PA morphogenesis, endodermal nkx2.7 expression implicates additional roles in the development of non-mesodermal head structures²⁵. As CNCCs migrate into the pharyngeal region, pharyngeal endoderm buds laterally in an anterior-posterior wave to establish the endodermal pouches that separate the crest into distinct arches. Studies in zebrafish and chick have reported indispensable functions for endoderm in craniofacial chondrogenesis^26–30. Yet, further exploration is required to understand the transcriptional and signaling pathways originating in the endodermal pouches to direct cartilage patterning. The discovery of new chondrogenic regulators such as Nkx2.7, expressed in the pharyngeal endoderm, will enhance insights into the complex cellular interactions shaping craniofacial skeletal development.

In this study, we demonstrate the finding that nkx2.7, the zebrafish ortholog of murine Nkx2-6, regulates BM and cartilage development, functions for which nkx2.5 is unable to compensate. Our microsynteny and molecular phylogenetic studies propose that a split between nkx2.7 and nkx2.5 occurred early in vertebrate evolution and highlight the conserved and unique function of nkx2.7 in the development of anterior PA derivatives. Analysis of BM morphogenesis unveils a requirement for nkx2.7 in progenitor proliferation at a critical timepoint in development when these cells migrate from the LPM to the PA mesodermal cores. In the context of normal pharyngeal endoderm morphogenesis, nkx2.7^−/− embryos exhibit loss of anterior PA CNCC domains with a particular deficit in the ventral region. This neural crest tissue deficiency foreshadows the dysmorphic jaw cartilage elements in time series experiments comparing wild-type and nkx2.7^−/− embryos. Moreover, single-cell RNA-sequencing data from dissected PAs reveals differentially expressed gene profiles that corroborate the dorsal-ventral cartilage fate imbalance. Our data demonstrate spatially-restricted downregulation of endothelin 1 (edn1), barx1, and hand2 in PA1 of nkx2.7^−/− embryos which validates our bioinformatic analysis supporting diminished ventral neural crest fate. Upregulation of key components of the Notch signaling pathway implicate the necessity for nkx2.7 to repress dorsal CNCC fate. Finally, inhibition of Notch rescues cartilage defects in the nkx2.7^−/− embryos, pointing to a non-cell autonomous mechanism by which nkx2.7 patterns the jaw cartilage. Altogether, our work illustrates the phylogenetically conserved significance of Nkx2.7 as a critical regulator of BM proliferation and ventral CNCC patterning, thereby underscoring the requirement of nkx2.7 in constructing the vertebrate jaw.

Results

Evolution of the NK4 gene family

To understand how the distinct yet overlapping functions of Nkx2.7 and Nkx2.5 evolved, we analyzed the microsynteny around the NK4- and tightly-linked NK3-family genes in vertebrate and invertebrate genomes (Fig. 1A) and reconstructed the phylogeny of the NK4 family (Fig. 1B). The tight linkage of nk4 (tinman) and nk3 (bagpipe) genes in invertebrates is well known from studies of insect and lancelet genomes^31,32 and represents part of the NK cluster, tandemly duplicated in the protostome/deuterostome ancestor³¹. Furthermore, the independent loss of NK3 was previously reported in Ciona (Fig. 1A, C)³³.

Fig. 1. Evolution of NK4-family genes in bilaterian animals.

A Conserved microsynteny of NK4-family genes (nkx2.3, nkx2.6/7, nkx2.5, nkx2.5l) with neighboring genes from NK3 family (nkx3.1, nkx3.2, nkx3.3) and other families (cpeb2/3/4, got1/1l1, bod1/1l1, stc1/2, nkx3.1/2/3, slc25a28/37, entpd4/7, bnip1, crebrf/rfl, atp6v0e1/2). Single slashes indicate the presence of a single other gene, while a double slash indicates the presence of two or more other genes. One ancestral chromosomal locus with inferred genes was present in the vertebrate ancestor, prior to the first round (1 R) of vertebrate whole genome duplication (WGD) and two ancestral chromosomal loci with inferred genes are shown prior to the second round (2 R) of WGD. Four chromosomal loci were identified for extant vertebrate NK4-family genes: nkx2.3, nkx2.6/7 (not found in thorny skate and zebra shark), nkx2.5, and nkx2.5l (not found in investigated bony fishes). Zebrafish retained only one copy of each NK4- and NK3-family gene following 3 R, and zebrafish nkx2.7 shares the conserved synteny with vertebrate Nkx2.6. B The phylogenetic reconstruction of NK4-family genes places the tunicate Nk4 gene at the base of two nodes, one node leading to nkx2.3 and nkx2.6 subfamilies and another to nkx2.5 and nkx2.5l subfamilies. Zebrafish nkx2.7 and gar nkx2.6 cluster with the amniote Nkx2-6 genes. C Summary of whole genome duplications and NK4- and NK3-family gene losses in bilaterian animals.

Our analysis confirmed the orthology of zebrafish nkx2.7 with vertebrate Nkx2.6 as previously reported³² and identified a novel member of the NK4 family in the cartilaginous fish (chondrichthyans; zebra shark and thorny skate) named nkx2.5-like (nkx2.5l) (Fig. 1A–C). Altogether, we describe four distinct loci with NK4/NK3 genes in vertebrates. However, nkx2.6/7 was not found in investigated chondrichthyan genomes, nkx3.3 appears to be lost in amniote genomes, and nkx2.5l was not found in osteichthyan genomes (Fig. 1A, C). In examined vertebrate genomes, the four loci contain not only nk4 and/or nk3 gene paralogs, but also paralogous genes of other gene families (Fig. 1A). Moreover, in invertebrate genomes of the lancelet and vase tunicate, we identified homologous genes to several of these paralogs on the same chromosomes as the nk4/nk3 genes (Fig. 1A).

Taken together, our combined microsynteny and phylogenetic reconstruction indicates that nkx2.3, nkx2.6/7, nkx2.5, and nkx2.5l genes are ohnologs that arose from the ancestral NK4 gene through two rounds of vertebrate whole genome duplication (WGD) events. The first round (1 R) most probably resulted in two ancestral genes: nkx2.3/nkx2.6/7 and nkx2.5/nkx2.5l. A second round (2 R) yielded sister ohnologs: nkx2.3, nkx2.6/7, and nkx2.5, nkx2.5l (Fig. 1A). No additional copies of NK4- or NK3-family genes were maintained in the zebrafish genome following a third round (3 R) of teleost-specific WGD (Fig. S1).

nkx2.7 is necessary for jaw development

Given our identification of a previously unappreciated and conserved early split between nkx2.7 and nkx2.5 ohnologs, we decided to probe the separate and unique function of nkx2.7 in pharyngeal muscle development. We scrutinized its expression pattern from early somitogenesis through pharyngeal arch (PA) formation (Fig. 2A–F). In wild-type embryos, nkx2.7 is expressed in the lateral plate mesoderm (LPM) as early as 6 somites (so) (Fig. 2A). The onset of expression in the cardiac cone and PA primordia is evident by 21 so (Fig. 2B). Between 26 hours post fertilization (hpf) and 42 hpf, robust nkx2.7 expression is detected in the cardiac tube and more posteriorly in the cores of PA1 through PA6 (Fig. 2C–F), as previously described¹⁹. Given that branchiomeric muscles (BMs) are born from skeletal muscle progenitor cells located in these PA segments³⁴, nkx2.7 expression domains implicate a potential role in early jaw muscle formation.

Fig. 2. Pharyngeal arch expression of nkx2.7 is necessary for jaw formation and larval survival.

In situ hybridization (ISH) for nkx2.7 in wild-type embryos at 6 so (n = 5) (A), 21 so (n = 7) (B), 26 hpf (n = 5) (C), 32 hpf (n = 5) (D), 36 hpf (n = 20) (E), and 42 hpf (n = 20) (F) demarcates progressive expression in the lateral plate mesoderm, cardiac cone, pharyngeal arch (PA) primordia, cardiac tube, and mesendoderm of the PAs throughout early development. Numbers represent PA primordia 1-6. Black arrows and white dotted lines encircling PA1 and PA2 denote mesodermal cores (C). Dorsal views, anterior to the top (A–E) and lateral view, anterior to the left (F). Scale bar, 50 μm. Hybridization chain reaction (HCR) for nkx2.7 in Tg(tcf21:nucGFP) (n = 6) (G–I) and Tg(sox17:GFP) (n = 7) (J–L) wild-type embryos. nkx2.7 and dlx2a expression is reflected through HCR in wild-type embryos (n = 12) (M–O). Insets show single plane, higher magnification images of PA2. White arrows indicate double positive cells. Lateral views, anterior to the left. Scale bar, 50 μm. P Schematic depicts the zebrafish nkx2.7^vu413 allele and the predicted wild-type and mutant Nkx2.7 protein products. The vu413 point mutation generated through TILLING is noted in red with the nucleotide transversion indicated above the gene and the corresponding codon change indicated next to the protein. The C- > A transversion at position 321 of the open reading frame leads to a nonsense mutation that is predicted to cause truncation of the protein prior to the homeodomain. Q, R Representative images of 4 dpf wild-type (n = 11) (Q) and nkx2.7^−/− (n = 8) (R) embryos. Red arrows highlight normal jaw development in the wild-type embryo compared to a collapsed jaw in the nkx2.7^−/− embryo. Lateral view, anterior to the left. Scale bar, 200 μm. S Kaplan-Meier survival curve depicts the pattern of larval death based on genotypic assessment in wild-type (n = 225) and nkx2.7^−/− (n = 74) larvae.

Yet, the complex gene regulatory networks signal between multiple tissue layers to orchestrate PA segmentation and patterning. Therefore, we sought to delineate the specific populations with which nkx2.7 expression overlaps. Employing HCR, we examined nkx2.7 transcripts in wild-type embryos carrying Tg(tcf21:nucGFP) and Tg(sox17:GFP) to demarcate the mesoderm and endoderm, respectively. Visible co-expression with Tg(tcf21:nucGFP) and Tg(sox17:GFP) validates mesodermal (Fig. 2G–I) and endodermal (Fig. 2J–L) nkx2.7 domains. Moreover, dlx2 marks all cranial neural crest cells (CNCC)-derived ectomesenchyme precursors of the craniofacial skeleton³⁵. Following inspection of nkx2.7 and dlx2a probes with HCR, we detected no evidence of intersecting domains (Fig. 2M–O). Taken together, our data substantiate nkx2.7 expression territories in both pharyngeal mesoderm and endoderm tissues.

While nkx2.7 is known to play redundant roles with nkx2.5 in zebrafish cardiac morphogenesis^{22–24,36,37}, its function in PA patterning has not been revealed. To address this gap in knowledge, we examined nkx2.7^vu413 embryos which carry a C- > A transversion at position 321 of the open reading frame; this allele produces a nonsense mutation predicted to cause truncation of the protein prior to the homeodomain (Fig. 2P)²⁴. Homozygous null embryos for the vu413 allele display overt jaw defects (Fig. 2Q, R). Protuberance of the lower jaw is evident in wild-type embryos, whereas nkx2.7^−/− embryos demonstrate a retracted lower jaw. Survival analysis exhibits larval lethality of nkx2.7^−/− animals by 14 days post fertilization (dpf) (Fig. 2S). Moreover, heterozygotes are morphologically indistinguishable from wild-type, suggesting that this point mutation has loss-of-function, rather than dominant-negative, effects. Altogether, the nkx2.7 spatiotemporal expression pattern and loss-of-function model suggest previously unappreciated requirements for nkx2.7 in directing craniofacial development.

Branchiomeric muscle morphogenesis requires Nkx2.7

Given the distinct expression of nkx2.7 in the pharyngeal apparatus (Fig. 2A–F) and larval lethality soon after initiation of oral feeding (Fig. 2S)³⁸, we investigated the patterning of head and jaw musculature in wild-type and nkx2.7^−/− embryos. The mesodermal cores of the larger anterior segments, the first pharyngeal arch (PA1; mandibular) and the second pharyngeal arch (PA2; hyoid) give rise to the BMs (Fig. 2C)³⁹. Specifically, the ventral and middle muscles derived from PA1, the intermandibularis anterior (ima) and the intermandibularis posterior (imp), form a triangular structure between the eyes that generates the jaw infrastructure, a hallmark of gnathostomes (Fig. 3A)^40,41. The ventral muscles derived from PA2 include the interhyals (ih) and the hyohyals (hh) which both insert at the midline and provide additional support for jaw function (Fig. 3A). PAs 3-6 (posterior PAs) house mesodermal progenitor cells that give rise to the endothelial linings of the pharyngeal arch arteries (PAAs)⁴². Finally, mesoderm from all posterior PAs including PA7, along with neural crest and epithelia of endodermal and ectodermal origin, are responsible for generating the gills in teleost fish^43,44.

Fig. 3. Nkx2.7 is necessary for branchiomeric muscle morphogenesis.

A Schematic of the BMs derived from PA1 (green) and PA2 (orange) including intermandibularis anterior (ima), intermandibularis posterior (imp), hyohyals (hh), and interhyals (ih). The heart, delineated in grey, has two chambers: ventricle (V) and atrium (A). Confocal images of MF20 (magenta) and Elnb (green) immunofluorescence designates BMs and smooth muscle of the outflow tract, respectively, of wild-type (n = 16) (B), nkx2.5^−/− (n = 11) (C), nkx2.7^−/− (n = 16) (D), and nkx2.5^−/−;nkx2.7^−/− (n = 5) (E) embryos. The extraocular muscles include the inferior oblique (io), inferior rectus (ir), and the adductor mandibulae (am) along with an additional PA-derived cephalic muscle, the sternohyoideus (sh). nkx2.5^−/−;nkx2.7^−/− embryos exhibit exacerbation of the PA2-derived BMs defects in comparison to nkx2.7^−/− embryos, implicating partially redundant functions of nkx2.7 and nkx2.5. Ventral views, anterior to the top. Scale, 100 μm. Two-color fluorescent ISH of wild-type embryos illuminates overlapping expression of nkx2.7 (green) and nkx2.5 (magenta) in PA1-5 at 26 hpf (n = 12) (F). Higher magnification panels represent PA1 and PA2 domains (white dotted lines) in a merged (G) and single channel maximum intensity (H, I) projections. Lateral views, anterior to the left. Scale, 100 μm. J Quantification of the number of BMs derived from PAs in wild-type (n = 16), nkx2.5^−/− (n = 11), nkx2.7^−/− (n = 16), and nkx2.5^−/−;nkx2.7^−/− (n = 5) embryos. Mean and standard error of each data set are shown. Unpaired, two-tailed t-test yields statistically significant differences between wild-type and nkx2.7^−/− (p < 0.0001), wild-type and nkx2.5^−/−;nkx2.7^−/− (p < 0.0001), and nkx2.7^−/− and nkx2.5^−/−;nkx2.7^−/− (p = 0.0024) embryos.

Employing MF20 antibody to examine the BMs along with Elastin b (Elnb) to highlight the smooth muscle of the cardiac outflow tract (OFT) at the anterior pole of the heart, we observe a normal hourglass shape demarcated by the ima and imp (PA1-derived BMs) and the ih and hh (PA2-derived BMs) in wild-type embryos (Fig. 3B). nkx2.7^−/− embryos, in contrast, exhibit severe deficiencies in BM number and patterning (Fig. 3D); this finding is reinforced by a statistically significant decrement observed when quantifying PA-derived BMs between wild-type and nkx2.7^−/− embryos (Fig. 3J). To probe the requirement of nkx2.5 in BM morphogenesis, we also examined nkx2.5^−/− embryos with the same protocol and found no significant differences when compared with wild-type embryos (Fig. 3C). We turned our attention to the potentially redundant roles of nkx2.7 and nkx2.5 given prior data highlighting this interplay in cardiac morphogenesis^24,36,37. While nkx2.7^−/− and nkx2.7^−/−;nkx2.5^−/− animals lack the ima and imp indicating PA1-derived BM abnormalities (Fig. 3D, E), nkx2.7^−/−;nkx2.5^−/− embryos reveal worsened defects in the ih and hh, PA2-dervied BMs (Fig. 3E). Our data demonstrate further exacerbation of the nkx2.7^−/− phenotype following the loss of the nkx2.5, leading to a statistically significant deficit in PA-derived BMs in nkx2.7^−/−;nkx2.5^−/− compared with nkx2.7^−/− embryos (Fig. 3J). Furthermore, scrutiny of nkx2.7 and nkx2.5 pharyngeal expression domains with two-color fluorescent ISH reveals overlapping regions in PA1-5 at 26 hpf (Fig. 3F–I). Together, these results highlight the unique requirement of nkx2.7 in PA1 mesodermal derivatives and partial functional redundancy of nkx2.7 with nkx2.5 in formation of the BMs derived from PA2.

Nkx2.7 regulates branchiomeric muscle progenitor proliferation

We probed the origins of the BM phenotype observed in nkx2.7^−/− embryos to outline the precise mechanistic roles of Nkx2.7 in craniofacial mesoderm development (Figs. 2,3). Previous studies have elucidated the vital role of Tbx1 in regulating branchiomeric myogenesis⁴⁵. Thus, we evaluated nkx2.7 expression in wild-type and tbx1^−/− embryos at 26 hpf which uncovered a dramatic loss of nkx2.7 transcripts in the bilateral PAs (Fig. 4A, B). Tbx1 is required for transcriptional activation of the basic helix-loop-helix (bHLH) myogenic regulatory factors such as MyoD1, which is initially expressed in the PAs and serves as a key node in skeletal muscle lineage formation^46–48. Our data depict diminished myod1 expression in nkx2.7^−/− embryos at 36 hpf in the intermandibularis (IM) which differentiates into the ima and imp and represents the ventral head muscle primordia derivatives of PA1 (Fig. 4C, D)³⁹. Similarly, decreased myod1 expression in the nkx2.7^−/− compared to wild-type embryos is evident in the constrictor hyoideus ventralis (CHV) which ultimately produces the ih and hh, ventral derivatives of PA2 (Fig. 4C, D)³⁹. Taken together, these results place nkx2.7 downstream of tbx1 in establishing the muscles of mastication while also underscoring the critical function of nkx2.7 in activating myogenesis of the PA1 and PA2 derivatives.

Fig. 4. Branchiomeric muscle progenitor proliferation is disrupted in the absence of Nkx2.7.

ISH for nkx2.7 in wild-type (n = 4) (A) and tbx1^−/− (n = 4) (B) embryos illuminates significant loss of nkx2.7 expression in the PAs at 26 hpf. Scale bar, 50 μm. myod1 transcripts in wild-type (n = 20) (C) and nkx2.7^−/− (n = 10) (D) embryos are downregulated in the Intermandibularis (IM) and the Constrictor Hyoideus Ventralis (CHV) precursors at 36 hpf. Expression of myod1 is also detected in the superior rectus (sr), lateral rectus (lr), medical rectus (mr), inferior rectus (ir), Constrictor Hyoideus Dorsalis (CHD), and the Masticatory Plate (MP). Lateral views, anterior to the left. Scale bar, 50 μm. E, F tcf21 ISH shows no overt discrepancy in pharyngeal arch expression between wild-type (n = 8) and nkx2.7^−/− (n = 9) embryos. Dorsal views, anterior to the top. Scale, 50 μm. G–L Maximum projection images of confocal z-stacks of wild-type and nkx2.7^−/− embryos carrying Tg(tcf21:nucGFP) demarcate the BM progenitors in the PA1, PA2, and PA3-6 (posterior arches) at 18 so (wild-type, n = 10; nkx2.7^−/−, n = 9) (G,H), at 21 so (wild-type, n = 19; nkx2.7^−/−, n = 17) (I, J), at 26 hpf (wild-type, n = 15; nkx2.7^−/−, n = 15) (K, L). Insets show higher magnification images of PAs with PCNA immunostaining in both wild-type and nkx2.7^−/− embryos, respectively, at 18 so (G’,H’), 21 so (I’,J’), and 26 hpf (K’,L’). White arrows highlight PCNA⁺tcf21:nucGFP⁺ cells. Lateral view, anterior to the left. Scale, 100 μm (G–L) and scale, 50 μm (G’-L’). M–R Quantification of BM progenitors and percentage of PCNA⁺ cells in specific PAs reveal no statistically significant differences in specification or proliferation in the absence of nkx2.7 at 18 so (M, N). However, a statistically significant decrement in BM progenitors is evident at 21 so, with an associated proliferative defect (p < 0.0001 for O and p = 0.0037 for P) (O, P). While the statistically significant decrease in PA1 and PA2 BM progenitors persists at 26 hpf, there is no evidence of abnormal proliferation at this timepoint (PA1, p < 0.0001; PA2, p = 0.0102) (Q, R). The number of samples for each genotype and timepoint are noted above. Mean and standard error of each data set are shown. Unpaired, two-tailed Student’s t-test is used to demonstrate statistically significance.

Prior genetic studies stress the requirement of Tcf21, another myogenic bHLH transcription factor, activity for specification of pharyngeal head muscle progenitors and underscore its function upstream of myod1^49,50 and as a modifier of Tbx1 pathway⁵¹. While equivalent domains of tcf21 expression are noted in wild-type and nkx2.7^−/− embryos in anterior and posterior PAs at 26 hpf (Fig. 4E, F), it is difficult to delineate more subtle, yet quantifiable differences in numbers of the BM precursors through ISH. Thus, we took advantage of Tg(tcf21:nucGFP)^pd41 expressed in the mesodermal cores of the PAs to count progenitor cells in wild-type and nkx2.7^−/− embryos (Fig. 4G-L)⁵². tcf21:nucGFP⁺ populations in the anterior domain of the LPM stream towards and contribute to PA1 and PA2 which, in turn, give rise to the ventral BMs⁵³. At the posterior border of the cardiac LPM, tcf21:nucGFP⁺ progenitors are the source of posterior PA mesodermal cores and form the PA3-6-dervied ventral head muscles and the PAAs⁵³. We hypothesize that nkx2.7 is required either to support specification of tcf21:nucGFP⁺ BM progenitors or to promote their proliferation to generate the appropriate number of mandibular and hyoid muscles.

We initiated our studies in the LPM at 18 so and uncovered equal values for both the total number of tcf21:nucGFP⁺ cells and also the percentage of PCNA⁺tcf21:nucGFP⁺ double-positive cells in the anterior PAs of both genotypes (Fig. 4G, H, M, N). At 21 so, as Tg(tcf21:nucGFP) refines to the cores of the PAs, we identified a statistically significant difference in the number of tcf21:nucGFP⁺ progenitors (78 versus 61 cells) and PCNA indices (62% versus 52%) in the two anterior PAs when comparing wild-type and nkx2.7^−/− embryos (Fig. 4I, J, O, P). Finally, our data illuminate a statistically significant decrease in tcf21:nucGFP⁺ cells specifically in PA1 (58 versus 40 cells) and PA2 (36 versus 29 cells), with unchanged total cell counts for PA3-6, between wild-type and nkx2.7^−/− embryos at 26 hpf (Fig. 4K, L, Q). To test whether a deficit in proliferation of tcf21:nucGFP⁺ cells correlates with the decreased cell number, we again performed immunostaining directed against PCNA. Surprisingly, we found no significant difference in proliferative indices in PA1 and PA2 progenitor cells between wild-type and nkx2.7^−/− embryos at this stage (Fig. 4K, L, R). From this time course, our data indicate that BM progenitors are specified normally in the absence of nkx2.7, but a proliferative defect occurs as the tcf21:nucGFP⁺ cells reach the PA mesodermal cores (21 so) which results in diminished PA1- and PA2-derived ventral musculature of the jaw.

Nkx2.7 is indispensable for craniofacial cartilage development

To dissect the requirement of Nkx factors in facial skeletal formation, we next assessed the lengths of PA1-derived cartilages, Meckel’s cartilage (M) and palatoquadrate (Pq), and PA2-derived cartilage, ceratohyal (Ch) (Fig. 5A)^40,54,55. Alcian Blue (cartilage) staining reveals normal PA-derived chondrocyte elements in wild-type and nkx2.5^−/− embryos at 96 hpf, implying that nkx2.5 is not required for craniofacial cartilage morphogenesis (Fig. 5B, C). Interestingly, nkx2.7^−/− embryos exhibit dysmorphic and deficient cartilage structures (Fig. 5D). Quantification shows no significant differences in M, Ch, and Pq dimensions when comparing wild-type and nkx2.5^−/− (Fig. 5E). Yet, nkx2.7^−/− embryos illustrate a statistically significant decrease in these three parameters compared to wild-type embryos (Fig. 5E). We also measured the M-Pq angle given reports underscoring the function of this variable in detecting head and jaw deformations associated with craniofacial abnormalities⁵⁶. A statistically significant increase in the M-Pq angle between wild-type and nkx2.7^−/− embryos reflects the shorter and wider lower jaw structure that develops in the absence of Nkx2.7 (Fig. 5F). Interestingly, ectopic cartilage nodules (nubbins) are detected in nkx2.7^−/− embryos at a low (21%) penetrance rate (data now shown); these findings are consistent with models representing both partial and complete reduction in Edn1 signaling pathway^57–61. Finally, the normal appearance of the ceratobranchial (Cb) segments suggests appropriate ventral foregut endoderm patterning in all three genotypes (Fig. 5B-D)²⁷. Altogether, our data highlight the previously unrecognized function of nkx2.7 in the developing cartilages derived from the mandibular and hyoid arches.

Fig. 5. Craniofacial cartilage development requires nkx2.7 expression.

A Schematic illustrates the cartilage elements derived from PA1 (green) and PA2 (orange) along with associated morphometric parameters (gray dotted lines) measured in panels E and F. Cartilages, such as ceratobranchial elements, derived from PA3-7 are noted in gray. Dissected Alcian Blue staining in wild-type (n = 21) (B), nkx2.5^−/− (n = 11) (C), and nkx2.7^−/− (n = 15) (D) demonstrate shorter and wider orientation of the PA1- and PA2-derived cartilages in nkx2.7^−/− compared to wild-type and nkx2.5^−/− embryos. Viscerocranium cartilage elements are labeled as follows: Meckel’s (M), palatoquadrate (Pq), ceratohyal (Ch), hyosymplectic (Hs), basihyal (Bh), basibranchial (Bb), and ceratobranchial (Cb). Ventral views, anterior to the top. Scale, 100 μm. E Morphometric analysis of the cartilage segments reveals statistically significant decrease in lengths between wild-type (n = 21), nkx2.5^−/− (n = 11), and nkx2.7^−/− (n = 15) embryos in M (p < 0.0001), Ch (p < 0.0001), and Pq (p < 0.0001), employing an unpaired, two-tailed t-test. Mean and standard error of each data set are shown. F Measurement of the M-Pq angle delineates a statistically significant difference between wild-type (n = 14) and nkx2.7^−/− (n = 20) embryos with no difference detected when comparing wild-type and nkx2.5^−/− (n = 13) embryos. Unpaired, two-tailed t-test is used to demonstrate mean and standard error of each data set (p < 0.0001 between wild-type and nkx2.7^−/− embryos). G Schematic illustrates a lateral view (anterior to the left) of cartilage elements and joint articulations contributing to the lower jaw at 96 hpf. H–J Visualization of viscerocranium following Alcian Blue and Alizarin Red staining at 7 dpf reveals aggravation of the craniofacial abnormalities in nkx2.7^−/− embryos. Craniofacial skeletal elements identified as follows: maxilla bone (mx), entopterygoid (en), branchiostegal ray (br), and opercle (op). Ventral views (anterior to the top) illuminate widening and shortening of the lower jaw in nkx2.7^−/− (n = 7) (J) compared with wild-type (n = 35) (H) and nkx2.5^−/− (n = 16) (I) embryos. Scale, 100 μm. K Quantification of the statistically significant decline in br and op bones when evaluating wild-type and nkx2.5^−/− compared with nkx2.7^−/− embryos using an unpaired, two-tailed t-test (br, p < 0.0001; op, p = 0.0019). The number of samples for each genotype is noted above. Mean and standard error of each data set are shown. L–N Lateral views (anterior to the left) of dissected wild-type (n = 9) (L), nkx2.5^−/− (n = 7), and nkx2.7^−/− (n = 8) (N) embryos following Alcian Blue staining. PA2 jaw joint fusion with thickening of the connection between the ceratohyal and the hyosymplectic cartilages is visible in the loss-of-function model. Scale, 100 μm. O Quantification of the statistically significant enlargement in the Ch width at the juncture of the PA2 joint in nkx2.7^−/− compared to wild-type embryos employing an unpaired, two-tailed t-test (p < 0.0001). The number of samples for each genotype is noted above. Mean and standard error of each data set are shown.

Given that nkx2.7^−/−;nkx2.5^−/− embryos are difficult to scrutinize at this developmental timepoint due to the severity of cardiac edema, subsequent friability, and early lethality^22,24,36, we broadened our investigation of potentially redundant roles of nkx2.7 and nkx2.5 by examining cartilage chondrogenesis at 96 hpf (Fig. S2). In the nkx2.7^−/−;nkx2.5^−/− embryos, Alcian Blue staining uncovers aggravation of the defective cartilage infrastructure observed in the nkx2.7^−/− embryos with collapsed M and Ch elements (Fig. S2A–D). To probe possible earlier defects in CNCC formation in the absence of Nkx2.7 and Nkx2.5 that may be responsible for this perturbed jaw architecture, we inspected expression of hand2, a key regulatory gene in the ventral domain⁶², at 26 hpf. Our results show focally diminished transcripts in the anterior PAs with similar reductions in both nkx2.7^−/− and nkx2.7^−/−;nkx2.5^−/− compared with wild-type and nkx2.5^−/− embryos (Fig. S2E–I). Thus, our data suggest that nkx2.7 is the primary NK4 gene family member responsible for anterior PA-derived chondrogenesis since local swelling from cardiac dysfunction in nkx2.7^−/−;nkx2.5^−/− embryos is known to exaggerate craniofacial dysmorphology.

We next explored the skeletal features arising from the CNCCs of the mandibular and hyoid arches at 7 dpf. Specifically, we focused on the maxilla bone (mx), derived from PA1⁶³, which associates with the Pq. The finger-like branchiostegal ray (br) bones stem from the ventral domain and the fan-shaped opercle (op) bones, supporting the gill covering, originate from the dorsal domain of PA2⁶⁴. Moreover, the entopterygoid (en), articulating with the dorsal margin of the Pq, also arises from PA2⁶³. In wild-type, nkx2.5^−/−, and nkx2.7^−/− embryos, Alizarin Red (bone) staining accentuates normal mineralization of the mx and en CNCC-derived bones (Fig. 5H–J) Yet, our data illustrate that there is a noteworthy diminution in the number of nkx2.7^−/− embryos exhibiting a full complement of br and op bones (Fig. 5K). These results hint at the possibility that the loss of Nkx2.7 mimics irregularities in the levels of Edn1 and Jagged-Notch signals which produce myriad ventral br and dorsal op deficient phenotypes^58,59,64.

Nkx2.7 is required for mandibular and hyoid jaw joint morphogenesis

The origin of jaw joints is a pivotal, evolutionary milestone that permitted the jawed gnathostomes to shift from suspension feeding to an active predatory lifestyle^65,66. Two bilateral facial joints exist in the larval zebrafish, the mandibular (PA1-derived, articulating the M and the Pq) and the hyoid (PA2-derived, articulating the Hs and the Ch via the interhyal cartilage) (Fig. 5G)^67,68. The prospective jaw joint domain (interzone) originates as a single mesenchymal condensation through the convergence of pivotal developmental pathways⁶⁸. The intermediate domain of the PA CNCC gives rise to this interzone where chondrocyte maturation is repressed and joint cell fate is promoted^64,68–73. Specifically, dlx3b, dlx4b, and dlx5a redundantly pattern the intermediate region of the first two PAs and, along with hand2, impart discrete identities to construct the dorsal-intermediate-ventral axis⁶².

We investigated the impact of Nkx2.7 on the refinement of nascent cartilage condensations to generate joint mobility. Lateral dissections of Alcian Blue staining at 7 dpf elucidate defects in both PA1- and PA2-derived joints in nkx2.7^−/− compared to wild-type and nkx2.5^−/− embryos. We detect enriched Alcian cartilage matrix staining across the presumptive mandibular and hyoid joint articulation regions (Fig. 5L–N). Precisely, the hinge joint in PA1 is fused in nkx2.7^−/− larvae (Fig. 5N). In the PA2 hyoid joint, increased density and darker Alcian staining in the nkx2.7^−/− embryos is visible in opposition to the refinement and lightening of the joint chondrocytes in the interzones between the Ch and the interhyal and between the interhyal and the Hs (Fig. 5L–N). Quantification validates this observation by highlighting a statistically significant increase in the width of the PA2 joint in nkx2.7^−/− compared to wild-type embryos (Fig. 5P). Taken together, our data show a distinct requirement of nkx2.7 in establishing the ventral and intermediate CNCC populations in the mandibular and hyoid arches necessary to ensure normal chondrogenesis and bone maturation of the lower jaw and specification of the jaw joints.

Cranial neural crest formation is impaired in nkx2.7^−/− embryos

While nkx2.7 and nkx2.5 are both expressed in the LPM, the pharyngeal mesoderm, and the ventricular, OFT, and BM derivatives of the PAs^{19,42,74–76}, nkx2.7 alone is expressed in the pharyngeal endoderm (Fig. 2C–F, J–L)¹⁹. This ancillary expression domain within the pharyngeal apparatus suggests supplementary roles in craniofacial development. Moreover, the anatomical proximity of multiple lineages in the PAs sets the stage for an intricate network of cellular interactions⁷⁷. Thus, we expanded our investigation of the requirement of nkx2.7 through evaluation of PA endoderm and neural crest morphogenesis. PA mesodermal cores are encased by sox10⁺ neural crest-derived cells that are encircled medially by sox17⁺ endodermal pouches⁷⁸. Employing Tg(sox17:GFP)^{s870 79}, we examined PA endoderm development in wild-type embryos, noting the three to four endodermal pouches that materialize by 26 hpf (Fig. S3A)⁸⁰. Similarly, the architecture in nkx2.7^−/− embryos reflects this developmental rhythm and demonstrates no significant abnormalities in endoderm formation (Fig. S3B, C).

Given that pharyngeal endoderm signals promote survival and differentiation of CNCCs^30,81, we queried the sox10⁺ population in wild-type and nkx2.7^−/− embryos. Tg(sox10:DsRed)^{el10 82} labels CNCCs that migrate to the arches and then is re-activated during chondrocyte differentiation⁶⁹. To visualize CNCC morphogenesis in the context of the mesodermal lineage, we employed live imaging of double transgenic Tg(sox10:DsRed);Tg(tcf21:nucGFP) embryos to delineate the mandibular, hyoid, and posterior arches at 36 hpf (Fig. 6A–F). In nkx2.7^−/− embryos, there is a loss of the ventral domain of the Tg(sox10:DsRed)⁺ CNCCs in the first two PAs compared to wild-type embryos (Fig. 6D–F). Quantification of the neural crest fields encompassed in PA1 and PA2 uncovered a statistically significant diminution between wild-type and nkx2.7^−/− embryos, validating the observed deficiency in the ventral aspects of first and second arches (Fig. 6B, E, M). Repeated imaging of the same animals demonstrates that these CNCC defects predate the dysmorphic jaw cartilage development observed at 72 hpf in nkx2.7^−/− embryos (Fig. 6G–L). Consistent with the morphometric analysis of the Alcian Blue images acquired at 96 hpf (Fig. 5B–D), Tg(sox10:DsRed) delineates broadening of the lower jaw structures through Ch shortening and M-Pq angle widening in nkx2.7^−/− embryos (Fig. 6N, O).

Fig. 6. Ventral neural crest domains of the anterior pharyngeal arches fail to form in nkx2.7^−/− embryos.

A–F Maximum projection images of confocal z-stacks of wild-type (n = 8) and nkx2.7^−/− (n = 10) embryos carrying Tg(sox10:DsRed) and Tg(tcf21:nucGFP) highlight loss of ventral neural crest domains of PA1 and PA2 in the absence of nkx2.7 at 36 hpf. White dotted lines circumscribe PA1 and PA2 CNCC domains. Numbers represent CNCC domains derived from PA1-7. White arrows indicate deficient CNCCs in the ventral domains. Yellow arrows underscore a parallel shortage of the ventral mesoderm populations labeled with Tg(tcf21:nucGFP). Lateral views, anterior to the left. Scale, 50 μm. G–L Images of the same animals at 72 hpf underscore that the absence of ventral neural crest domains result in dysmorphic PA1- and PA2-derived cartilages in nkx2.7^−/− (n = 15) and wild-type (n = 14) embryos. Correspondingly, dysmorphic BM are visualized with Tg(tcf21:nucGFP) in nkx2.7^−/− embryos. Ventral views, anterior to the top. Scale, 100 μm. M Quantification of the area occupied by PA1 and PA2 CNCC lineage is measured indicating a statistically significant difference between wild-type and nkx2.7^−/− embryos at 36 hpf using an unpaired, two-tailed t-test (p = 0.0002). Mean and standard error of each data set are shown. N, O Measurements of Ch lengths and M-Pq angles demonstrate a statistically significant difference between wild-type and nkx2.7^−/− embryos at 72 hpf using an unpaired, two-tailed t-test (Ch Length, p = 0.0431; M-Pq Angle, p = 0.0003). Mean and standard error of each data set are shown.

We next explored the correlation between BM and cartilage defects given evidence of disruption of BM attachments in the context of disordered cartilage architecture in nkx2.7^−/− embryos (Fig. 6J–L). Given that Tg(tcf21:nucGFP) is less clearly defined at 96 hpf, we performed immunostaining with MF20 in wild-type and nkx2.7^−/− embryos carrying Tg(sox10:DsRed) to track both lineages (Fig. S4A–F). We executed a series of pairwise comparisons to uncover linear combinations of BM numbers and PA1- and PA2-derived cartilage lengths (Fig. S4G–I). PA1-derived BM number and M length are statistically significantly correlated (Fig. S4G), mirroring recent studies underlining that oriented growth of cartilage is necessary for BM patterning and directionality⁸³. However, PA1- and PA2-derived BM numbers are not correlated with Pq and Ch lengths, respectively (Fig. S4H, I). Taken together, our data support the conclusion that both biomechanical forces and genetic requirements for nkx2.7 in BM and cartilage morphogenesis contribute to the interplay of mesodermal and neural crest derived phenotypes.

Ventral neural crest fate is suppressed in the nkx2.7^−/− embryos

To understand the differentially regulated targets downstream of nkx2.7 in the distinct PA populations, we performed single-cell RNA-sequencing (scRNA-seq) on micro-dissected PA tissue from wild-type and nkx2.7^−/− embryos carrying Tg(tcf21:NTR-mCherry) at 26 hpf (Fig. 7A)⁸⁴. Although the tcf21:NTR-mCherry⁺ cells facilitated visualization of the PA structures for dissection, all germ layers were submitted for analysis. After excluding low-quality cells, we analyzed 7230 cells from wild-type and 4917 cells from nkx2.7^−/− embryos and identified a total of 16 distinct cell clusters using unbiased clustering of the single-cell transcriptomic data (Fig. 7B). We then explored and classified the diversity of cell types in our model, relying on the expression of specific marker genes obtained from the literature (Fig. 7C)^85,86. Next, we extracted a subset of the data to focus on clusters where nkx2.7 and nkx2.5 expression are identified in at least one cell. In these four clusters, differential expressed gene (DEG) analysis comparing wild-type and nkx2.7^−/− samples identified 2018 unique genes that are significantly differentially expressed in one or more of the selected cell types (Supplemental Data 5). We were particularly intrigued by the changes in expression within the cranial neural crest cluster (Fig. 7D) and uncovered a distinct set of perturbations related to dorsal-ventral (D-V) neural crest patterning^87,88. Among the significant DEGs, we identified upregulation of the dorsal genes expressed the CNCCs such as nr2f5, six1b, jag1b, hey1, and eya1 (Fig. 7D, E). We also observed downregulation of the intermediate and ventral genes, including hand2, dlx3b, dlx4b, dlx5a, dlx6a, barx1, and msx1b (Fig. 7D, E). These results reveal an important function for nkx2.7 in maintenance of the ventral PA1 and PA2 patterning program and provide a potential etiology for the jaw morphogenesis abnormalities observed in the nkx2.7^−/− embryos.

Fig. 7. Single-cell RNA-sequencing of pharyngeal arches uncovers misregulation of dorsal and ventral CNCC transcripts in nkx2.7^−/− embryos.

A Schematic portrays the approach for generation of the scRNA-seq data. Micro-dissected tissue from PAs of wild-type and nkx2.7^−/− embryos carrying Tg(tcf21:NTR-mCherry) were submitted for scRNA sequencing at 26 hpf. B UMAP of low-resolution clusters for wild-type and nkx2.7^−/− samples. C Expression matrix demonstrates genes that are used to classify the 16 different clusters. D Volcano plot compares the transcript expression in the cranial neural crest cluster in wild-type and nkx2.7^−/− embryos. Green dots represent significantly upregulated genes and orange dots represent downregulated genes (FDR-adjusted p-value < 0.05). Genes labeled in yellow are dorsal CNCC genes and those in purple are ventral CNCC genes. E Box plots indicate statistically significant upregulation of dorsal CNCC genes (nr2f5, six1b, jag1b, hey1, and eya1) and statistically significant downregulation of ventral CNCC genes (hand2, dlx3b, dlx5a, dlx6a, and barx1) when comparing wild-type and nkx2.7^−/− embryos in the cranial neural crest cluster. Wilcoxon Rank Sum Test was used to evaluate the expression levels of the pooled cells from two embryos in each genotype. Calculations are based on 1078 cells from the cranial neural crest clusters (642 wild-type and 436 nkx2.7^−/− cells). The boxes delimit the lower quartile (Q1) and upper quartile (Q3) of the distribution of the expression levels; the tick line within the boxes represents the median (Q2) value. Whiskers are extended to 1.5 Å~ interquartile range (Q3 - Q1) and outliers are represented as single points. The significance levels (nr2f5, p < 0.0001; six1b, p = 0.0203; jag1b, p < 0.0001; hey1, p = 0.0005; eya1, p = 0.0021; hand2, p < 0.0001; dlx3b, p < 0.0001; dlx5a, p < 0.0001; dlx6a, p < 0.0001; barx1, p < 0.0001) indicate the FDR-adjusted p-values from the statistical test.

Given this alignment of ventral cartilage defects (Fig. 5B–D), jaw joint fusions (Fig. 5L–N), and downregulation of ventral CNCC genes in nkx2.7^−/− embryos (Fig. 7D, E), we assessed in situ expression patterns of key D-V patterning genes to validate our in silico data. We first examined endothelin1 (edn1) which behaves as a permissive morphogen in the ventral arch ectoderm to partition the CNCCs along the D-V axis and activates expression of genes such as hand2 and the Dlx family^60,87–89. In addition to PA ectoderm, edn1 is expressed in the PA pouch endoderm and the core paraxial mesoderm^60,90–92. Moreover, Edn1 activates the Endothelin-A receptor, Ednra, in a limited window of CNCC post-migratory development from approximately 20-28 hpf⁶⁰ and regulates the dlx5/dlx6/hand2 hierarchical molecular circuitry⁸⁷. To test potential roles for edn1 downstream of Nkx2.7, we executed ISH and detected decreased expression of edn1 in nkx2.7^−/− compared to wild-type and nkx2.5^−/− embryos, with preferential loss in the PA1 (Fig. 8A–C). Given diminished edn1 expression at 26 hpf in nkx2.7^−/− embryos, we explored potential secondary defects in D-V patterning of the mandibular arch. An additional downstream effector of Edn1 signaling is barx1⁷⁰, a pre-cartilage condensation marker that is also known to be inhibited by Notch signals⁶⁹. Analogous to our findings with edn1, barx1 expression is lost in PA1 of nkx2.7^−/− embryos but is present in wild-type and nkx2.5^−/− embryos (Fig. 8D–F). Taken together, these data underscore the requirement of nkx2.7 in D-V patterning with decreased edn1 and barx1 expression in nkx2.7^−/− embryonic anterior PAs.

Fig. 8. Nkx2.7 is required for dorsal-ventral patterning of PA domains.

A–F ISH for edn1 and barx1 at 26 hpf demonstrates decreased expression of each respective ventral gene in PA1 of nkx2.7^−/− (n = 17 and 13, respectively) compared to wild-type (n = 26 and 22, respectively) and nkx2.5^−/− (n = 11 and 11, respectively) embryos. Black arrows point to significantly diminished expression of each gene in PA1. Dorsal views, anterior to the top. Scale bar, 50 μm. G–I nkx3.2 is expressed in the first arch mesenchyme and pharyngeal endoderm in wild-type (n = 34) and nkx2.5^−/− (n = 24) embryos at 52 hpf. However, its expression is eradicated in the pharyngeal arches of nkx2.7^−/− embryos (n = 20). Black arrow points to significantly diminished expression in PA1. Lateral views, anterior to the left. Scale bar, 50 μm.

In addition to the impact of Edn1 signals on D-V patterning in the mandibular arch, prior studies have illuminated the importance of Nkx3.2 (previously identified as Bapx1) as a downstream target of Edn1 in jaw joint formation^93,94. Therefore, we examined nkx3.2 expression at 52 hpf given that the upper and lower jaw cartilages have initiated chondrogenesis with nkx3.2-expressing cells within and around the jaw joint at this time point^93,95. We observe a conspicuous absence of nkx3.2 expression specifically in nkx2.7^−/− embryos (Fig. 8G–I). These data explain the aberrant PA1 jaw joint morphogenesis in the absence of nkx2.7 function, as Nkx3.2 is required for specification of the PA1-derived joint and the retroarticular process of Meckel’s cartilage^93,96. Altogether, our data support the finding that Nkx2.7 diverged early during evolution from Nkx2.5 acquiring a distinct role in establishing the D-V CNCC lineage configuration required for ventral jaw morphogenesis and jaw joint formation.

Nkx2.7 represses Notch signaling to pattern craniofacial cartilage

To dissect the mechanism underlying upregulation of dorsal CNCC markers and downregulation of ventral markers (Figs. 7,8), we directed our attention to Jagged-Notch signaling pathway given its essential function in establishing CNCC D-V identity in zebrafish⁶⁴. In dorsal arch cells, Jagged-Notch inhibits expression of Ednra-dependent targets such as hand2 and barx1^64,69,70,93. Moreover, Ednra signals repress jag1b expression^64,71, producing an antagonism between Edn1 and Jagged-Notch pathways that fine-tunes chondrogenic differentiation in CNCCs of the PAs⁶⁹. Given the specific endodermal expression domains of nkx2.7 (Fig. 2C–F, J–L), we hypothesize that reduced Edn1 signals in the nkx2.7^−/− embryonic endoderm lead to augmented expression of Notch ligands in the CNCC. Consistent with this model, jag1b is upregulated in PA1 in nkx2.7^−/− compared with wild-type embryos (Fig. 9A–D). Moreover, we interrogated the expression of nkx2.7 and edn1 in wild-type embryos employing HCR to document potential territories of overlap. Indeed, we discerned intersecting signals from both genes in the anterior PA regions indicating shared domains (Fig. S5A–C). In order to determine whether edn1 transcripts are depleted in the endoderm of nkx2.7^−/− embryos, we performed edn1 HCR in Tg(sox17:GFP) and nkx2.7^−/−;Tg(sox17:GFP) embryos. Aligned with our working hypothesis, our data reveal sox17:GFP⁺ endodermal cells expressing edn1 in wild-type embryos with reduced edn1 signal in the sox17:GFP⁺ population of nkx2.7^−/− embryos (Fig. S5D, E). From these observations, we conclude that nkx2.7 operates uniquely in the mandibular and hyoid endodermal pouches to modulate edn1 expression which, in turn, restricts Notch cues in the dorsal CNCC field.

Fig. 9. Notch signaling is repressed by Nkx2.7 in patterning craniofacial cartilage and joints.

ISH for jag1b at 26 hpf indicates enhanced expression (black arrow) in PA1 of nkx2.7^−/− (n = 12) (C) compared to wild-type (n = 17) (A) and nkx2.5^−/− (n = 15) (B) embryos. Dorsal views, anterior to the top. Scale bar, 50 μm. D Quantification of jag1b signal, detected by ISH, reveals statistically significant augmentation in PA1 in nkx2.7^−/− compared to wild-type and nkx2.5^−/− embryos (p < 0.0001). Mean and standard error of each data set are shown with analysis performed using an unpaired, two-tailed Student’s t-test. Alcian Blue staining in DMSO-exposed wild-type (n = 21) (E) and nkx2.7^−/− (n = 26) (F) embryos and in DBZ-exposed nkx2.7^−/− (n = 30) (G) embryos highlights lengthening and narrowing of anterior PA-derived cartilage elements in DBZ-exposed nkx2.7^−/− embryos compared to DMSO-exposed wild-type and nkx2.7^−/− embryos. Ventral views, anterior to the top. Scale, 100 μm. H Measurement of the M-Pq angle in embryos stained with Alcian Blue delineates a statistically significant decrease in the DBZ-exposed (n = 18) compared to the DMSO-exposed (n = 26) nkx2.7^−/− embryos indicating a partial rescue towards the angle displayed in the DMSO-exposed wild-type embryos (n = 26). Mean and standard error of each data set are shown. Unpaired, two-tailed Student’s t-test yields p < 0.0001 between DMSO-exposed wild-type and nkx2.7^−/− embryos, p = 0.0013 between DMSO-exposed wild-type and DBZ-exposed nkx2.7^−/− embryos, and p < 0.0001 between DMSO-exposed and DBZ-exposed nkx2.7^−/− embryos. I–L MF20 immunofluorescence depicts BMs in DMSO-exposed wild-type (n = 22) (I) and nkx2.7^−/− (n = 18) (J) embryos and in DBZ-exposed nkx2.7^−/− (n = 19) (K) embryos. No statistically significant difference in the number of PA-derived BMs is observed between DMSO- and DBZ-exposed nkx2.7^−/− embryos employing an unpaired, two-tailed Student’s t-test. Ventral views, anterior to the top. Scale, 100 μm. M–O ISH for nkx3.2 at 52 hpf demonstrates reappearance of expression in PA1 of DBZ-exposed nkx2.7^−/− (n = 9/9) (O) compared to DMSO-exposed nkx2.7^−/− (n = 7/9) (N) embryos. Moreover, comparable expression domains are evident in DMSO-exposed wild-type embryos (n = 10/10) (M) and DBZ-exposed nkx2.7^−/− embryos (O). Dorsal views, anterior to the top. Scale bar, 50 μm.

To test this model, we performed a rescue experiment to verify whether upregulation of Notch signaling is responsible for loss of the ventral CNCC genetic program. Mutant and sibling control embryos were incubated with dibenzazepine (DBZ), a γ-secretase inhibitor that prevents processing of the Notch receptor into an active form, from 19.5 hpf to 32 hpf, as previously described^69,97. We observed a statistically significant rescue of the M-Pq angle when comparing DBZ-exposed nkx2.7^−/− versus DMSO-exposed wild-type embryos (Fig. 9E–H). The upshot of this elongated PA1- and PA2-derived cartilage is improvement in the short and wide lower jaw morphology of the nkx2.7^−/− embryos. We then queried whether BM morphogenesis is analogously influenced by elevated Notch signals. Surprisingly, DBZ-exposed nkx2.7^−/− embryos exhibit an equivalent shortage of PA-derived BMs when compared with DMSO-exposed nkx2.7^−/− embryos (Fig. 9I–L). Our outcomes reinforce our earlier assessment that, in addition to synchronized biomechanical influences at play in the Nkx2.7 loss-of-function model, nkx2.7 functions through independent mechanisms in BM and CNCC morphogenesis. Specifically, elevated Notch signals contribute to the aberrant ventral craniofacial morphology and not the anomalous BM development in the absence of Nkx2.7.

Finally, we tested whether the nkx2.7^−/− jaw joint phenotype is equivalently mediated through augmented Notch pathway dynamics. Recent data suggest that Notch may function as an upstream regulator of nkx3.2 expression via Hey1 which negatively regulates nkx3.2 in the dorsal CNCC domain^64,95. Thus, we exposed wild-type and nkx2.7^−/− embryos to DBZ again at 12 hpf through 32 hpf and performed nkx3.2 ISH at 52 hpf. Our results indicate a definitive and consistent rescue of the suppressed nkx3.2 expression in the PA1 CNCCs (Fig. 9M-O). Altogether, we conclude that Nkx2.7 is required to establish Edn1 signaling which drives ventral barx1 and hand2 expression and confines Notch activity to the dorsal domain of the developing anterior PA-derived cartilage and jaw joint structures.

Discussion

Our data provide insights into the role of the NK4 family member, Nkx2.7, in fashioning the vertebrate jaw through discrete cellular and molecular mechanisms in the mesoderm- and CNCC-derived structures of the pharyngeal arches. In the mesoderm, Nkx2.7 promotes proliferation of the tcf21:nucGFP⁺ BM progenitors as they accumulate in the cores of the PAs. Nkx2.7 also regulates D-V patterning of anterior PA CNCCs. Upregulated dorsal and downregulated ventral domain CNCC genetic signatures in scRNA-seq data and in situ results point to a disrupted equilibrium between Jagged-Notch and Endothelin1 pathways. Inhibition of Notch signaling in nkx2.7^−/− embryos yields partial rescue of the cartilage and complete rescue of the jaw joint defects, reinforcing our conclusion that elevated activity of this pathway contributes to the dysmorphic cartilage architecture in the loss-of-function model. Synthesizing our findings, we conclude that Nkx2.7 serves as a phylogenetically conserved transcription factor essential for vertebrate jaw and jaw joint morphogenesis with instrumental roles in both BM and cartilage development.

Preceding studies have highlighted Nkx2.7 as a mediator of cardiac development that provides a redundant function to the master cardiac transcription factor, Nkx2.5^{18,23,24,36,37}. Emphasizing this cardiac-specific role, recent work has uncovered NKX2-6 mutations in patients with congenital heart diseases such as truncus arteriosus^98–101. Our current data reveal a previously unappreciated function for Nkx2.7 as a regulator of craniofacial development. Beyond the mechanistic insight delineated from our studies, the identification of an additional gene responsible for jaw formation in zebrafish opens doors to uncover genetic etiologies of mammalian middle ear defects¹⁰². Developmental and evolutionary examination has confirmed that middle ear bones and joints have their origins in the gnathostome jaw¹⁰³. Thus, exploiting the powerful genetic strategies embedded in zebrafish developmental biology research will augment our understanding of skeletal abnormalities causing hearing loss in patients, a field with myriad unresolved, clinically-relevant questions^104–106.

Our discovery of Nkx2.7 as a key transcription factor mediating BM and cartilage development is rooted in our molecular evolution analysis of the tightly linked NK4 and NK3 genes. As described in prior synteny analyses, nkx2.7 is the zebrafish ortholog of murine Nkx2.6 (Wotton et al 2009; Bartlett et al 2010); we echo the suggestions to rename zebrafish nkx2.7 to nkx2.6 to be consistent with the nomenclature conventions in other vertebrates. Moreover, our data highlight the ohnologous relationship of nkx2.7 and nkx2.5, resulting from two rounds (2 R) of WGD in vertebrates with divergence of nkx2.6/7 and nkx2.5 beginning after 1 R. In all examined vertebrate genomes, nkx2.5 is missing its linkage to a NK3-family gene and this loss most likely occurred before the chondrichthyan-osteichthyan divergence (Fig. 1C). In the osteichthyan lineage, nkx3.2 apparently lost its linkage to a NK4-family gene and it is fascinating to conjecture that this evolutionary trajectory may explain the prominent functions of Nkx2.5 in heart morphogenesis and Nkx3.2 in jaw patterning. On the other hand, the proposed ancestral nkx2.3/nkx2.6/7 gene preceding 2 R suggests the possibility of an evolutionarily-conserved shared role in the pharyngeal region for nkx2.3 and nkx2.6/7, where some redundant function in the heart is retained. Furthermore, nkx2.5l requires future investigation and, intriguingly, the absence of nkx2.6 in chondrichthyans might indicate a potential functional overlap between these two genes.

We also explored mechanisms by which Nkx2.7 co-regulates the BMs and cartilage elements of the vertebrate lower jaw. Our data support crucial roles of mesendodermal expression of nkx2.7 in mesodermal BM progenitor proliferation and CNCC D-V patterning. These findings are consistent with the concept that multiple cell types influence craniofacial development through modulation of pivotal signaling pathways such as Endothelin1 and Jagged-Notch^88,107. Furthermore, important PA transcription factors such as Tbx1 facilitate mesoderm to CNCC cell-cell communication through non-cell autonomous regulation of BMP and MAPK signals¹⁰⁸. While our experiments highlight a role for nkx2.7 downstream of tbx1 (Fig. 4A, B), it is unclear whether inhibiting BM progenitor proliferation participates in the evolution of ventral cartilage defects in the nkx2.7^−/− embryos. Yet, the identification of Tbx1 as an upstream regulator of nkx2.7 validates prior findings in mouse underscoring Nkx2-6 sensitivity to Tbx1 dosage¹⁰⁹ and pinpoints nkx2.7 as a newly defined player in the suite of transcription factors essential for branchiomeric muscle progenitor cell activation^5,110,111.

Our results also reveal that both genetic and environmental forces contribute to the orchestration of nkx2.7 functions in the mesoderm and CNCC. A recent investigation has revealed the importance of emerging cartilage interactions with individual BMs via mechanically coupled processes⁸³. In craniofacial enthesis formation, transcriptional regulation of key tenocyte genes, scxa and sox9a, is variable and responsive to biomechanical and mechanotransduction forces¹¹². Moreover, cyclic mechanical forces are the primary mechanoregulator of cell orientation in the developing joint; these local growth patterns are necessary to chisel the Meckel’s and palatoquadrate cartilage interface^113,114. These innovative studies elucidate the interplay between force transmission from muscle contraction and cartilage morphogenesis in the embryonic zebrafish jaw. Further inquiry is required to explore the contribution of BM mechanical disruption in nkx2.7^−/− embryos to the misregulated CNCC D-V patterning.

Through our investigation of molecular, cellular, and evolutionary mechanisms, we identify Nkx2.7 as a vital transcription factor in vertebrate craniofacial development. Our loss-of-function model exhibits a dramatic deficit in BM number and patterning along with defects in mandibular and hyoid arch-derived cartilage components. Both impaired proliferation in the mesodermal progenitors and Notch pathway upregulation in CNCCs of the anterior arches underlie morphological disruption of the BMs, cartilage, and joints of the ventral jaw. The divergence of Nkx2.7 from Nkx2.5 early during evolution buttresses our data separating its distinct function in lower jaw formation and underlines the potential for discoveries in human disease. Specifically, given conservation of essential gene networks, patients with congenital malformations of craniofacial architecture or middle ear bones will benefit from therapeutic insights gained through dissection of the genetic networks choreographing craniofacial morphogenesis¹¹⁵. Moreover, understanding the multifaceted genetic and developmental regulatory mechanisms underlying facial shape formation will advance opportunities for therapeutic interventions^116,117. Recent advances in single-cell technology in the arena of craniofacial biology has enriched our appreciation of the vast repertoire of regulatory elements and functionally significant genes that direct vertebrate head morphogenesis¹¹⁸. Integrating these datasets with GWAS studies have enriched the identification of causative mutations for patients with craniofacial abnormalities^119,120. Our work provides the initial steps to achieve this mission by uncovering how Nkx2.7 patterns pharyngeal arch progenitors to produce the BMs, cartilages, and joints that comprise the movable vertebrate jaw. Moreover, extension of this effort will reveal mechanisms to support stimulation of resident progenitors following injury in light of the powerful capacity for regeneration in the zebrafish jaw joint¹²¹. Given research in zebrafish highlighting the requirement for reactivation of the developmental program to ensure cardiac regenerative potential¹²², it would be energizing to consider whether the Nkx2.7 regulatory network can be leveraged for repair of craniofacial tissues in mammals.

Methods

Zebrafish mutant and transgenic lines

We used zebrafish carrying the previously described mutations and transgenes: nkx2.7^vu413 (RRID: ZFIN_ZDB-ALT-111213-12)²⁴, nkx2.5^vu179 (RRID: ZFIN_ZDB-ALT-131212-1)²⁴, tbx1^tu285 (RRID: ZFIN_ZDB-ALT-980203-1780)⁷⁸, Tg(tcf21:nucGFP)^pd41 (RRID: ZFIN_ZDB-ALT-110914-2)⁵², Tg(tcf21:NTR-mCherry)^pd108 (RRID: ZFIN_ZDB-ALT-150904-1)⁸⁴, Tg(sox17:GFP)^s870 (RRID: ZFIN_ZDB-ALT-061228-2)⁷⁹, and Tg(sox10:DsRed)^el10 (RRID: ZFIN_ZDB-ALT-120523-6)⁸². All zebrafish experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Columbia University.

In situ hybridization, hybridization chain reaction, and skeletal staining

We performed whole-mount in situ hybridization (ISH) as previously described¹²³. Formerly reported probes employed in this study include nkx2.5 (ZDB-GENE-980526-321), tcf21 (ZDB-GENE-051113-88), myod1 (ZDB-GENE-980526-561), hand2 (ZDB-GENE-000511-1), barx1 (ZDB-GENE-050522-28), and jag1b (ZDB-GENE-011128-4). New probes generated for these experiments include nkx2.7 (ZDB-GENE-990415-179), edn1 (ZDB-GENE-000920-1), and nkx3.2 (ZDB-GENE-030127-1). Primer sequences are noted in Supplemental Data 1.

Hybridization Chain Reaction (HCR 3.0) staining of embryos was executed following published protocols¹²⁴. HCR probes for nkx2.7, dlx2a, edn1 were synthesized by Molecular Instruments, Inc.

Two-color fluorescent in situ hybridization experiments were performed as previously described⁶⁴. nkx2.7 and nkx2.5 probes were labeled with digoxigenin and fluorescein, respectively.

Alcian Blue/Alizarin Red staining of embryonic facial skeletons was performed as previously described¹²⁵.

Immunofluorescence

Whole-mount immunofluorescence was conducted using a previously described protocol¹²⁶, with some modifications. To visualize BMs, double immunofluorescence was performed with primary antibodies: anti-sarcomeric myosin heavy chain (1:20; MF20, Developmental Studies Hybridoma Bank) and anti-Eln2 (1:500)¹²⁷, followed by secondary antibodies: goat anti-mouse IgG2b Alexa Fluor 568 (1:200; Invitrogen) and goat anti-rabbit IgG Alexa Fluor 488 (1:200; Invitrogen), respectively⁴². When incorporating immunostaining for sox10:DsRed⁺ cells, the following antibodies were employed: MF20 (1:20) and anti-DsRed (1:500; Clontech) along with goat anti-mouse IgG2b Alexa Fluor Cy5 (1:200; Invitrogen) and goat anti-rabbit IgG Alexa Fluor 568 (1:200; Invitrogen), respectively. When detecting tcf21:nucGFP⁺ cells, modifications were incorporated from a different protocol using the following antibodies: anti-GFP (1:250; Life Technologies) and anti-PCNA (1:250: Sigma-Aldrich), along with goat anti-mouse IgG1 Alexa Fluor 488 (1:250; Invitrogen) and goat anti-mouse IgG2a Alexa Fluor 568 (1:250; Invitrogen), respectively¹²⁸.

Notch inhibitor treatments

We applied a previously published protocol describing the use of the γ-secretase inhibitor, dibenzazepine (DBZ)⁶⁹, to block processing of the Notch receptor into its active intracellular form⁹⁷. Specifically, DBZ (Tocris #4489) was dissolved in dimethyl sulfoxide (DMSO) creating a 10 mM stock which was diluted to a final concentration of 10 μM in embryo medium. No more than 50 embryos were incubated in a single dish starting at 12 hpf (for ISH experiments) and 19.5 hpf (for Alcian Blue/Alizarin Red experiments). DBZ was thoroughly rinsed off at 32 hpf with embryo medium. Finally, embryos were allowed to grow until 52 hpf (for ISH experiments) or 96 hpf (for Alcian Blue/Alizarin Red experiments). Sibling controls were exposed to the same concentration of DMSO for the equivalent developmental windows for each respective experiment.

Imaging

Images for live embryos, Alcian Blue/Alizarin Red, and ISH were captured with a Zeiss M2Bio microscope and a Zeiss AxioCam digital camera, prior to processing with Zeiss AxioVision, ImageJ (version 1.53k, NIH), and Adobe Creative Suite software. Confocal images were acquired on a Zeiss 710 laser scanning confocal microscope with a Zeiss AXIO Observer Z1 inverted microscope stand with z-stacks analyzed using ImageJ (version 1.53k, NIH).

Genotyping

PCR genotyping was performed on genomic DNA extracted from individual embryos following live imaging, Alcian Blue/Alizarin Red staining, ISH, HCR, or immunofluorescence. Detection of nkx2.7^vu413 was performed using primers 5’-CTTTTTCAGGCATGTGTCCA-3’ and 5’-AAAGCGTCTTTCCAGCTCAA-3’ to generate a 146 bp fragment. Digestion of the mutant PCR product with MseI creates 111 bp and 35 bp fragments. Detection of nkx2.5^vu179 was performed using primers 5’-CAAACTCACCTCCACACAGG-3’ and 5’-TTACCATCCCGAACCAAAAC-3’ to generate a 144 bp fragment. Digestion of the mutant PCR product with Hinf1 generates 30 bp and 114 bp fragments. Detection of tbx1^tu285 was performed using primers 5’-TCCAACTCAGCACAAGCCCC-3’ and 5’-CCAATCAAGTGCATTGACGATG-3’ to generate a 438 bp fragment. Digestion of the PCR mutant product with PacI creates 276 bp and 162 bp fragments.

Quantification and statistical analysis

All quantifications were performed blinded. Results represent at least two independent experiments (technical replicates) in which multiple embryos from multiple independent matings are analyzed (biological replicates).

For quantification of the ISH signals, previously reported protocols were executed¹²⁹. 100% of embryos demonstrate the depicted phenotype unless otherwise specified. When counting BMs, each muscle body was identified as a unit and misplaced muscles were included with their respective PA1- or PA2-derivatives based on location. The number of tcf21:nucGFP⁺ cells within the first, second, and posterior PAs were counted manually in ImageJ. To measure proliferation, tcf21:nucGFP⁺ cells also positive for PCNA signal were counted and divided by the total tcf21:nucGFP⁺ cell number to obtain a proliferative index for each embryo. For cartilage morphometrics, the M-Pq angle was determined by assessing the angle established by drawing lines from the anterior end of the M to the bilateral PA2 joints⁵⁶.

Statistical analysis was performed using GraphPad Prism Version 9.5.0 to conduct two-tailed Student’s t-tests for data involving a continuous variable. Statistical values were displayed as mean ± standard error of the mean (SEM). Canonical Correlation Analysis was computed with a 95% confidence interval and graphs were depicted with a simple linear regression. The following nomenclature was employed to present results: ns, not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Genome database searches and phylogenetic reconstruction

Genome databases of human Homo sapiens (GRCh38.p14), mouse Mus musculus (GRCm39), chicken Gallus gallus (bGalGal1.mat.broiler.GRCg7b), frog Xenopus tropicalis (UCB_Xtro_10.0), zebrafish Danio rerio (GRCz11), spotted gar Lepisosteus oculatus (LepOcu1), zebra shark Stegostoma fasciatum (sSteFas1.1), thorny skate Amblyraja radiata (sAmbRad1.1.pri), vase tunicate Ciona intestinalis (KH), European lancelet Branchiostoma lanceolatum (BraLan3), acorn worm Saccoglossus kowalevskii (Skow1.1), and fruit fly Drosophila melanogaster (Release 6 plus ISO1 MT) were searched for NK4- and NK3-family genes using tblastn with zebrafish and gar nkx2.5 and nkx3.2 sequences as queries. The protein sequences for all collected NK4-family genes were aligned using Seaview with Clustal Omega¹³⁰ and reviewed to exclude the divergent sequences. The phylogenetic reconstruction was carried out on the entire protein sequences with IQ-TREE¹³¹ by applying the JTT + F + I + G4 model selected with ModelFinder¹³². Branch supports were calculated using UltraFast Bootstrap (UFBoot) method¹³³ with 1000 replicates. Consensus Maximum Likelihood tree was visualized with iOTL¹³⁴. Accession numbers used for phylogenetic tree reconstruction are listed in Supplemental Data 2.

Conserved synteny analysis

Several upstream and downstream genes from the NK4 (nkx2.3, nkx2.5, nkx2.6, nkx2.7) and adjacent NK3 (nkx3.1, nkx3.2, nkx3.3) family were analyzed for conserved synteny in the NCBI genome assembly viewer of the above-mentioned species. BLASTP and TBLASTN searches were used to search for conserved gene orthologs in poorly annotated genomes. Accession numbers identified in conserved synteny analyses are listed in Supplemental Data 2.

Single-cell RNA-sequencing

From the offspring of an intercross of nkx2.7^+/-;Tg(tcf21:NTR-mCherry)^pd108, two wild-type and two nkx2.7^−/− embryos were identified by genotyping. The PAs and surrounding tissue were manually dissected at 26 hpf, dissociated, and submitted for flow cytometry. Cellular dissociation was performed with liberase solution in PBS at 28 °C with pipetting every five minutes until adequately homogenized in solution. Fetal Bovine Serum (FBS) (5% of volume) was then added to each sample to arrest the dissociation, and each sample was centrifuged at 4 °C. The excess supernatant was removed, and the pellet was resuspended in PBS/1% FBS and filtered through a 40 μm cell strainer. Finally, DAPI and DRAQ5 were added and cells were sorted using a SORP FACSAria^TM Cell Sorter (BD Biosciences) under gentle conditions with the 130 μm nozzle at 12 PSI. NERL Diluent 2 (Thermo Fisher, DIL5522) solution was used for sheath fluid. The FACSAria was calibrated per the standard protocol in the CSCI Flow Cytometry Core Facility using Cytometer Setup and Tracking Beads (BD Biosciences, 655051), SPHERO Rainbow Calibration Particles (8 Peaks) 3.0 μm, 5 mL (Spherotech, RCP-30-5A), and then followed by optimization of the drop charge delay using BD FACS Accudrop Beads (BD Biosciences, 345249).

Cells were selected to be sorted based on the following scheme: first, forward and side light scatter (488 nm) were used to identify cells and establish an initial gate to remove debris. Following this selection, single cells were identified by plotting forward light scatter pulse area against forward scatter pulse height. Finally, signals from DRAQ5 (measured with 637 nm excitation at 140 mW and a 670/30 bandpass filter), a cell permeant nuclear dye, and DAPI (measured with 405 nm excitation at 100 mW and a 450/50 bandpass filter), a cell impermeant nuclear dye, were used to select live, nucleated cells based on high DRAQ5 signal and low DAPI signal. Finally, mCherry signal was collected alongside these other parameters using 561 nm excitation at 100 mW and a 610/20 bandpass filter.

Single cells were captured to prepare the library for sequencing using the Chromium Single Cell 3’ Reagent Kit (10x Genomics, version 3.1) and the 10X Genomics platform according to manufacturer’s instructions. Following successful quality control by Bioanalyzer, the cDNA libraries were sequenced on a NovaSeq 6000 at the JP Sulzberger Columbia Genome Center at Columbia University Medical Center.

Mapping and clustering of single-cell mRNA data

FASTQ files of wild-type and nkx2.7^−/− samples were aligned to the GRCz11 genome assembly (version 105) using the 10x Genomics CellRanger software (version 6.1.2)¹³⁵ with default parameters. The output of CellRanger was analyzed in the R programming environment using the Seurat package (version 4.1.0)^136–139. The feature-barcode matrices of the two samples were read into a R environment using the Read10X() function. A Seurat object was created for each sample using the CreateSeuratObject() function. The percentage of mitochondrial gene counts were computed using the PercentageFeatureSet() function on both Seurat objects. Cells having a minimum of 500 reads, expressing 125–8000 genes, along with an overall mitochondrial gene count lower than 20% of the total were used for downstream analysis. This filtering strategy yielded a total of 12,147 cells across both objects (Supplemental Data 3). The raw gene counts of the cells were normalised using NormalizeData() function, with LogNormalize as the selection method. FindVariableFeatures() function with variance stabilizing transformation (vst) method was used to identify the top 2000 variable genes. To integrate the two objects, features and anchors for integration were obtained using SelectIntegrationFeatures() and FindIntegrationAnchors() functions, respectively. This step was followed by integrating the two objects using IntegrateData() function. The integrated data was scaled and reduced to 30 principal components (PCs) using the ScaleData() and RunPCA() function. These PCs were passed as input to the RunUMAP() function. The Clustree package was used to determine the optimal resolution for clustering¹⁴⁰. Clustering was performed using FindNeighbors() and FindClusters() functions using the 30 PCs and a resolution parameter of 0.12, which resulted in 16 clusters (Supplemental Data 4). These data were visualized in a two-dimensional UMAP representation produced using the DimPlot() function.

Cell type identification in single-cell mRNA data

To identify the cell types present in the dataset, FindConservedMarkers() function of the Seurat package was used to detect putative marker genes whose expression is specific to only one cluster. A set of 3-5 markers per cluster were manually curated (employing lists generated from the literature), visualized, and used to annotate the clusters. The following cell types were identified in the dataset along with their respective markers: cluster 0 – neural 1 (sox3, sox19a, notch3, lrrn1), cluster 1 – neural 2 (onecut1, elavl3, elavl4), cluster 2 – pharyngeal mesenchyme (fmoda, col9a3, cxcl12a, col5a1), cluster 3 – neural 3 (sox3, mdka, lrrn1), cluster 4 – pharyngeal neural crest 1 (grem2b, twist1a, dlx2a, dlx5a), cluster 5 – unassigned (mix of cells, mitochondrial genes), cluster 6 – neural 4 (mdka, notch3, elavl3, sox3, sox19a), cluster 7 – pharyngeal endoderm (krt91, epcam, col1a1b, krt4), cluster 8 – neural 5 (mdka, notch3, sox3, lrrn1), cluster 9 – pharyngeal neural crest 2 (sox10, foxd3, crestin), cluster 10 – eye (six1b, neurod1), cluster 11 – periderm (cyt1l, cyt1, krt5), cluster 12 – cardiovascular mesoderm (aqp1a.1, cldn5b, sox7), cluster 13 – hematopoietic progenitors 1 (hbbe3, hbbe1.3, hbae3, hbae1.3, hbbe1.2, hbbe1.1, hbae1.1, hbae1.3.1, hbbe2), cluster 14 – hematopoietic progenitors 2 (lyz, spi1b), cluster 15 – muscle (mylpfa, actc1b, myl1, myhz1.1). The dot plot used to visualize the expression of marker genes was generated using the DotPlot() function of Seurat, in combination with viridis library¹⁴¹ for the color palette and ggplot2 library¹⁴².

Differential gene expression analysis

FindMarkers() function from Seurat with Wilcoxon rank sum test was used to perform differential expression analysis between wild-type and nkx2.7^−/− cells for each cluster. Genes expressed in at least 10% of the population were considered to be expressed in our experimental setting and selected for testing of differential expression. The p-value was adjusted using the “FDR” method and a threshold of adjusted p-value < = 0.05 was used to identify differentially expressed genes (DEGs). Volcano plots and box plots used to visualize DEGs were generated with ggplot2 library.

Reporting summary

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

Author contributions

C.F., C.d.S.-T., and K.L.T. conceived of the project and designed the experiments. C.F., C.d.S.-T., T.T.R.W., A.A., and M.I.N. performed the experiments. C.F., C.d.S.-T., T.T.R.W., U.R., J.L., U.C., J.W., T.H., R.S., and K.L.T. analyzed the data. C.F., T.T.R.W., U.R., J.L., T.H., R.S., and K.L.T. generated figures for data visualization. C.F., U.R., C.Z.G., H.T.N, M.S., M.R., and R.S. performed bioinformatic analysis. C.F., J.S., T.H., R.S., and K.L.T. wrote, reviewed, and edited the manuscript. C.F. and K.L.T. provided funding for the project.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNA sequencing data reported in this paper are available at the NIH Gene Expression Omnibus under accession number GSE240780. The gene level-normalized expression data for pharyngeal arch cell types from wild-type and nkx2.7^−/− embryos can be navigated at https://singlecell.broadinstitute.org/single_cell/study/SCP2431/nkx2-7-is-a-conserved-regulator-of-craniofacial-development. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-58821-3.