Single-Cell RNA-seq Identifies Cell Diversity in Embryonic Salivary Glands

Abstract

Branching organs, including the salivary and mammary glands, lung, and kidney, arise as epithelial buds that are morphologically very similar. However, the mesenchyme is known to guide epithelial morphogenesis and to help govern cell fate and eventual organ specificity. We performed single-cell transcriptome analyses of 14,441 cells from embryonic day 12 submandibular and parotid salivary glands to characterize their molecular identities during bud initiation. The mesenchymal cells were considerably more heterogeneous by clustering analysis than the epithelial cells. Nonetheless, distinct clusters were evident among even the epithelial cells, where unique molecular markers separated presumptive bud and duct cells. Mesenchymal cells formed separate, well-defined clusters specific to each gland. Neuronal and muscle cells of the 2 glands in particular showed different markers and localization patterns. Several gland-specific genes were characteristic of different rhombomeres. A muscle cluster was prominent in the parotid, which was not myoepithelial or vascular smooth muscle. Instead, the muscle cluster expressed genes that mediate skeletal muscle differentiation and function. Striated muscle was indeed found later in development surrounding the parotid gland. Distinct spatial localization patterns of neuronal and muscle cells in embryonic stages appear to foreshadow later differences in adult organ function. These findings demonstrate that the establishment of transcriptional identities emerges early in development, primarily in the mesenchyme of developing salivary glands. We present the first comprehensive description of molecular signatures that define specific cellular landmarks for the bud initiation stage, when the neural crest–derived ectomesenchyme predominates in the salivary mesenchyme that immediately surrounds the budding epithelium. We also provide the first transcriptome data for the largely understudied embryonic parotid gland as compared with the submandibular gland, focusing on the mesenchymal cell populations.

Introduction

Embryonic development of branching organs starts with a localized thickening of the epithelium that progresses to buds and ducts. The morphologies of initial bud outgrowths are remarkably similar across many branching organs, suggesting that common as well as organ-specific mechanisms likely underlie this first step of organogenesis. Mammalian salivary glands are unique study models in that 3 major types of salivary gland—submandibular, sublingual, and parotid—differ in their acinar cell types, physiology, and disease susceptibility (Grundmann et al. 2009). Submandibular gland has been the predominant developmental research model; consequently, parotid gland development represents an understudied but important area for regenerative medicine research because of its high susceptibility to dry mouth disorders. The secretion of submandibular and sublingual glands is viscous due to mucins and accounts for resting saliva. In contrast, parotid saliva is watery and rich in digestive enzymes, and it is secreted primarily upon stimulation, such as eating. While myoepithelial cells, which have traditionally been presumed to expel saliva, cover the basal surface of adult murine submandibular and sublingual acini, they rarely localize in the parotid acini (Kawabe et al. 2016). This suggests fundamental differences in functions and secretory processes of serous glands (parotid) and mucous or mixed glands (sublingual and submandibular). Cell type–specific transcription can emerge early in embryonic development, as evidenced in pancreas, where early molecular signatures define what later become functionally distinct endocrine and exocrine cells (Byrnes et al. 2018). However, early organ- or cell type–specific transcription patterns that define salivary lineages have largely been uncharacterized.

The epithelium of developing organs has been the predominant focus, so another knowledge gap involves the transcriptional characterization of mesenchymal cells. Branching morphogenesis is known to be controlled by precisely choreographed intercellular interactions between epithelial and mesenchymal cells (Kadoya and Yamashina 2005; Kashimata and Hayashi 2018). Such communication occurs in specialized microenvironments that provide spatiotemporally controlled cues for stem/progenitor cells. The cranial neural crest provides diverse mesenchymal derivatives that are multipotent and collectively termed “ectomesenchyme” (Le Douarin et al. 2004). The particularly high repair potential of oral connective tissue (Page and Ammons 1974) suggests the contribution of its neural crest–derived cells that dictate development and homeostasis. In fact, the mesenchyme plays key roles in epithelial morphogenesis and cell fate decisions (Grobstein 1953; Lu et al. 2016). Accordingly, what guides cell fate and eventually organ specificity is likely to involve the mesenchyme. The ectomesenchyme is expected to be particularly important during the early single-bud stage of salivary glands, when the neural crest–derived ectomesenchymal cells represent most of early salivary mesenchyme (Jaskoll et al. 2002). While single-cell RNA sequencing (scRNA-seq) represents a promising method to unravel transcriptional heterogeneity of branching organs, most of these studies focus on the epithelial cells from later developmental or adult stages. Neural crest fate decisions have recently been revealed by scRNA profiling of embryonic day 8.5 (E8.5) to E10.5 murine embryos (Soldatov et al. 2019) at stages prior to the beginning of salivary gland development. Although this analysis provided valuable insight into broad neural crest lineages, cranial crest subpopulations that contribute to salivary gland development remain to be identified.

The objective of the present study was to characterize the molecular identities of cells that compose the 2 largest salivary glands—submandibular and parotid—during bud initiation. Even at a very early stage of morphogenesis, our findings reveal substantial transcriptional and spatial localization differences between the mesenchymal cells of the 2 glands, particularly neuronal and muscle precursor cells. Some gland-specific neuronal genes were characteristic of different rhombomeres (embryonic neural tube segments that are precursors of the rhombencephalon and migrating neural crest). These findings demonstrate the establishment of transcriptional identities that emerge quite early in development of these branching organs.

Materials and Methods

Detailed descriptions of methods for salivary gland isolation and immunofluorescence, as well as scRNA-seq and bulk library preparation, RNA sequencing (RNA-seq), and data analyses, are provided in the Appendix.

Animals and Tissues

Tissue samples were obtained from time-mated ICR mice (Envigo). All experimental animal work was conducted in accordance with National Institutes of Health animal study protocol 17-845 that was approved by the National Institute of Dental and Craniofacial Research Animal Care and Use Committee. Epithelial and mesenchymal tissue samples were prepared separately for bulk RNA-seq and scRNA-seq. Single-cell preparations from E12 submandibular and parotid glands, approximately 10 each, were prepared as previously described (Sekiguchi and Hauser 2019). E11.5 submandibular and E12 parotid glands were used for bulk RNA-seq.

Results

Molecular Heterogeneity of Salivary Mesenchymal Tissues

Bulk RNA-seq was conducted to molecularly characterize the epithelial and mesenchymal tissues at the bud initiation stage of embryonic submandibular and parotid glands (Fig. 1A). Principal component analysis revealed the marked transcriptional difference between submandibular and parotid mesenchymal tissues. Principal component 1 separated the epithelium and mesenchyme, whereas principal component 2 separated submandibular and parotid glands, with the mesenchymal clusters being more well separated than their epithelial counterparts (Fig. 1B). This early molecular heterogeneity was attributed to tissue-specific (epithelium or mesenchyme) and gland-specific (submandibular or parotid) marker genes (Appendix Fig. 1, Appendix Tables 1–4).

Figure 1.

tSNE plots, principal components, and molecular markers of submandibular and parotid glands. (A) Representative images of embryonic day 11.5 (E11.5) submandibular and E12 parotid epithelial and mesenchymal tissues. The epithelia of the 2 glands at this early developmental stage are morphologically similar, with round evaginating epithelial buds. In contrast, morphologic differences can be observed between the mesenchymal tissues, with the submandibular mesenchyme being denser and more well defined. (B) Principal component analysis based on the top 500 most variable genes detected with bulk RNA-seq. Principal component analysis is a dimensionality reduction technique that represents the global expression profiles of different samples in a 2-dimensional space. Principal component 1 separated samples according to differential gene expression between epithelial and mesenchymal tissues, while principal component 2 separated the gene expression patterns of submandibular and parotid glands, with much greater separation between the mesenchymal samples as compared with the epithelial counterparts. This finding indicates greater transcriptional heterogeneity between mesenchymal tissues as compared with epithelial tissues of the 2 glands. (C) tSNE plot of all single-cell RNA-seq samples, including submandibular and parotid gland epithelial and mesenchymal cells, colored by cell types. Clustering of the single-cell data sets, as implemented by Seurat analysis software, was represented with the tSNE algorithm, which is well suited for the representation of highly dimensional data by nonlinear dimensional reduction. The 8 clusters comprised 3 epithelia—bud (red), duct (orange), Krt19+ duct (magenta)—and 5 mesenchymal cell types—erythroid (cyan), neuronal (purple), muscle (forest green), as well as submandibular (light blue) and parotid mesenchyme (dark blue). A total of 14,441 E12 submandibular and parotid epithelial and mesenchymal cells, which contained on average 2,484 genes per cell, were included in the analyses. (D) tSNE plot of all single-cell RNA-seq samples, according to the type of samples. Cells are colored according to tissue type: submandibular mesenchyme (purple) and epithelium (blue), as well as parotid mesenchyme (green) and epithelium (light red). The formation of gland-specific mesenchymal cell clusters (light and dark blues) contrasted with the partial intermixing of epithelial cells from the 2 glands, suggesting that mesenchymal cells were considerably more transcriptionally heterogeneous than the epithelial cells. The neuronal cluster as indicated in panel C consisted primarily of cells originating from the submandibular gland, whereas the muscle-related cluster primarily contained parotid cells. (E) Dot plots for representative genes identified as differentially expressed (FDR < 0.05) among the clusters as indicated in panel A. In addition to bud and duct markers, gland-specific mesenchymal markers were identified (submandibular, light blue; parotid, dark blue). The size of the dot encodes the percentage of gene-expressing cells for each cell type. Known molecular markers, such as Epcam (epithelium), Col1a2 (mesenchyme), Sox10 (bud/neuronal), and Etv4/5 (bud), are included for comparison. FDR, false discovery rate; PG, parotid gland; RNA-seq, RNA sequencing; SMG, submandibular gland; tSNE, t-distributed stochastic neighbor embedding.

Differentially Expressed Genes in Embryonic and Adult Salivary Glands

To compare gene expression in early embryonic and adult salivary glands, our bulk RNA-seq data were compared with adult murine salivary gland RNA-seq data (Gao et al. 2018). Embryonic salivary glands expressed higher percentages of differentially expressed genes as compared with adult glands (25.7% vs. 10.9%) and transcription factors (1.8% vs. 0.6%; Appendix Table 5). This comparison method is by no means optimal given the differences in experimental and data analysis processes utilized in the 2 studies. Nonetheless, it suggests a higher complexity of transcriptional programs during development.

Genes differentially expressed between the glands at both early developmental and adult stages (Gao et al. 2018) were determined, since they may reinforce distinct submandibular or parotid identity. Tlx1 expression was enriched in the submandibular gland at both stages (Appendix Table 6). Among other functions, Tlx1 cooperates with a pan-autonomic determinant, Phox2b, to control neuronal cell fate (Borghini et al. 2006). One of the genes enriched in embryonic and adult submandibular glands was Mylk (myosin light chain kinase)—a myoepithelial marker (Nguyen et al. 2018; Appendix Table 7). In contrast, genes enriched in the parotid gland at embryonic and adult stages, such as troponins, are associated with striated muscle contraction.

Cellular Diversity in Early Submandibular and Parotid Salivary Glands

To determine cell types and to identify which cell types express gland-specific molecular markers, scRNA-seq was performed with 4 samples: epithelium and mesenchyme from E12 submandibular and parotid glands. Data validity was confirmed with high correlations observed for all sample pairs of scRNA-seq and bulk RNA-seq (Appendix Fig. 2). Appendix Table 8 provides scRNA-seq quality control statistics.

Differential gene expression analysis identified 3 epithelial and 5 mesenchymal cell types (Fig. 1C). Consistent with the findings from the bulk RNA-seq principal component analysis, the mesenchymal cells were considerably more transcriptionally heterogeneous than the epithelial cells (Fig. 1D). Nonetheless, distinct clusters were evident among even the epithelial cells, where unique molecular markers separated presumptive bud and duct cells. Known markers that defined these clusters—Sox10 for bud and Krt5/19 for duct (Lombaert and Hoffman 2010)—confirmed their identity (Fig. 1E). The epithelial bud clusters also expressed markers not previously identified, such as Col9a1 and Hpca in the bud as well as Cldn4 (claudin 4) and Anxa1 (annexin A1) in the Krt19+ duct clusters.

Mesenchymal cells formed 5 well-defined clusters, with 2 major clusters being separated according to the type of gland (Fig. 1C). Three smaller clusters comprised erythroid, neuronal, and muscle-related cells. Markers of blood vessel (Pecam1) or myoepithelial progenitors (Krt5/14) were absent in the muscle cluster (Fig. 1E). The absence of these markers suggests its nonmyoepithelial/vascular smooth muscle identity. Distinct markers defined submandibular and parotid mesenchymal clusters, indicating transcriptional heterogeneity in mesenchymal cells at even this early developmental stage (Appendix Fig. 3, Appendix Table 9).

Further clustering analyses were performed separately for each gland to determine gland-specific cell types and their markers. The parotid muscle cluster was more prominent than its submandibular counterpart (Fig. 2A, B, E, F), and molecular markers of the 2 glands were largely different (Appendix Table 10). Unlike the submandibular gland, the parotid gland lacked a mesenchymal subpopulation specialized in neuronal processes. Instead, the large mesenchymal cluster of the parotid showed generalized expression of a neuronal marker, Tubb3 (tubulin beta 3; Fig. 2D, F).

Figure 2.

tSNE plots and cluster expression of neuronal and muscle-related molecular markers in submandibular or parotid salivary gland. (A) tSNE plot of embryonic day 12 (E12) submandibular cells. The submandibular clusters contained a neuronal cell cluster (purple) that was molecularly distinct in its gene expression from the rest of the submandibular mesenchymal cells. (B) tSNE plot of parotid cells. The overall clustering pattern for the 2 glands was similar, except that the neuronal cell cluster was absent from parotid cells. (C) Submandibular-enriched neuronal-related gene expression from scRNA-seq. The submandibular neuronal cell cluster is enriched with noradrenergic neuron differentiation determinants, including Hand2 and Phox2b. (D) Parotid-enriched neuronal-related gene expression from scRNA-seq. Unlike the submandibular gland, the parotid gland lacked a mesenchymal subpopulation specialized in neuronal processes. Rather, the general parotid mesenchymal cell cluster expressed some neuronal genes, such as Tubb3 and Pou3f3. Ngfr was coexpressed with other neuronal genes in the submandibular neuronal cluster (purple). This contrasted with the parotid mesenchyme, in which Ngfr belonged to the muscle cluster (green). Cluster expression of Acta2+ muscle precursors and Tubb3 neuronal cells in (E) submandibular or (F) parotid gland. Each purple-colored dot represents a cell expressing Acta2 or Tubb3. While both glands had an Acta2+ muscle cell cluster, the parotid muscle cluster was larger and expressed some gland-specific molecular markers. Of note, the submandibular gland contained a specialized cluster of Tubb3+ neuronal cells, which was absent from the parotid gland. Instead, the large mesenchymal cluster of the parotid showed widespread expression of Tubb3. tSNE, t-distributed stochastic neighbor embedding.

Neural Crest–Lineage Gene Expression in Salivary Gland Development

To identify the potential neural crest–derived populations that contribute to each type of salivary gland, gene set enrichment analyses were performed, focusing on neuronal genes differentially expressed between the glands. Submandibular mesenchyme-enriched genes included noradrenergic neuron differentiation determinants, including Ascl1, Hand2, and Phox2b (Appendix Table 12). Notably, differentially expressed transcription factors were associated with phenotypes that affect development of different rhombomeres. Mutations of the submandibular-enriched transcription factors Hoxb1 and Hoxb2 are linked to patterning and neurogenesis of rhombomere 4 (Davenne et al. 1999; Rossel and Capecchi 1999; Appendix Tables 11 and 12, Appendix Figs. 1 and 4). In contrast, the parotid-upregulated genes—En1, En2, and Lmx1b—are associated with mutations that cause rhombomere 1 abnormalities (Suda et al. 1999; Liu and Joyner 2001; Guo et al. 2007). Parotid-specific expression of genes that control neural crest development, Msx1, Msx2, and Fgf8 (Fig. 2D, Appendix Fig. 1; Ishii et al. 2005), is also indicative of potential differences in neural crest populations that migrate to submandibular or parotid glands.

These neuronal genes were expressed in neuronal cells of the submandibular mesenchyme, where other neural crest or neuronal genes, such as Sox10, Ngfr (nerve growth factor receptor), Tubb3, Phox2b, Ascl1, and Hand2 were coexpressed (Fig. 2C). Interestingly, Ngfr was not coexpressed with Tubb3 in the parotid (Fig. 2D). Ngfr was instead expressed in the muscle cluster of the parotid mesenchyme. Its Ntf3/5 (neurotrophin 3/5) ligands showed similar cluster expression patterns in both glands. In contrast, the expression of another ligand, Bdnf (brain-derived neurotrophic factor), was specific to the muscle cells and was more enriched in the parotid gland, suggesting Bdnf autocrine signaling in muscle precursor cells. The cell type–specific enrichment of yet another ligand, Ngf, in the parotid differentiating duct but not in submandibular clusters suggests differences in cell types involved in receptor-ligand interactions for the 2 glands.

Localization of Neuronal and Muscle Markers in Salivary Glands

The variations in cluster expression were paralleled by differences in the localization of neuronal and muscle-related markers. The expression of Tubb3 emerged in the periductal area of the early embryonic submandibular gland, which extended into interlobular areas (Fig. 3A). No obvious spatial relationship was observed among epithelial buds and Acta2+ and Tubb3+ tissues. This contrasted with the E13 parotid gland, in which the expression of Tubb3 was restricted to the mesenchyme immediately surrounding the epithelial bud (Fig. 3B). Tubb3-enriched parotid mesenchyme was in turn surrounded by Acta2 (Fig. 4A). This neuronal expression pattern reversed later in embryonic development. During the active branching and differentiation stage at E16, the submandibular gland exhibited Tubb3+ neuronal extensions from the ducts to the buds (Fig. 3C). The basal layer of the developing submandibular acini was populated by emerging myoepithelial cells expressing Acta2 and another myoepithelial marker, Cnn1 (calponin). Conversely, myoepithelial cells positive for these genes were sparsely localized in the parotid acini at E16 (Fig. 3D). Instead, the parotid gland exhibited prominent muscle tissues in the mesenchyme. Striations in some of the fibers revealed that they were of skeletal origin (Fig. 4B). Unlike in submandibular gland, Tubb3 was expressed in thin filamentous processes around parotid epithelium and in a thick bundle in Acta2+ muscle tissues. Figure 5 presents a schematic summary of the relationship of salivary epithelia to their surrounding neuronal and muscle precursors.

Figure 3.

Immunofluorescence images of neuronal and muscle marker expression in embryonic salivary glands. In a single confocal z-plane of embryonic day 13 (E13) submandibular gland (SMG) (A), Tubb3 (cyan) was localized to Cdh1+ (magenta) periductal and interlobular areas (arrowheads). (B) In the E13 parotid gland, the expression of Tubb3 and Acta2 was markedly spatially organized. E13 parotid Cdh1+ epithelium was surrounded by Tubb3 staining. Tubb3 was in turn surrounded by Acta2 (yellow), with a clear junction between the tissues (closed arrowheads). Coexpression of Acta2 and Cnn1 (green) was observed only in presumptive blood vessels (open arrowheads). (C) Many E16 submandibular gland basal epithelial cells showed co-localization of Cnn1 and Acta2 (arrowheads) and prominent Tubb3+ nerve projections growing in close proximity to the epithelium. This finding corroborates the reported findings of nerve fibers arising from the parasympathetic ganglion in the submandibular gland. (D) In contrast, in the E16 parotid basal epithelial layer, Cnn1 and Acta2 expression was sparse, which is consistent with reported findings that demonstrated sparsity of myoepithelial cells lining mature parotid acini. Unlike in submandibular gland, the parotid gland lacked thick Tubb3+ extensions in the immediate vicinity of the epithelium. However, Tubb3 was expressed in thin filamentous processes around the E16 parotid epithelium (open arrowheads), while a thick bundle of Tubb3+ tissue in Acta2+ skeletal muscle in the periphery of the parotid gland (arrowheads). Scale bars: 100 μm.

Figure 4.

Confocal cross section and magnified image of developing skeletal muscle in embryonic parotid glands. (A) Cross section (right) of the embryonic day 13 (E13) parotid gland along the dotted line (magenta) shows the parotid epithelial bud wrapped in Tubb3+ neuronal tissues (cyan), which in turn are surrounded by Acta2+ muscle precursors (yellow). (B) Magnified confocal image of Acta2+ developing muscle surrounding E16 parotid gland shows striation. This striation likely represents the alignment of myofibrils containing dark A bands and light I bands, characteristic of developing skeletal muscle.

Figure 5.

Schematic summary of the relationship of salivary epithelia to their surrounding neuronal and muscle precursors in embryonic submandibular versus parotid glands. At the single-bud stage (left), differences in the localization of neuronal and muscle precursors are already evident in the 2 glands. The prominent submandibular ganglion emerges at the base of the submandibular duct. The parotid gland lacks such a ganglion, and its epithelial bud is instead wrapped by dense neuronal tissues, as indicated in the schematic diagram exposing the bud. Later in embryonic development (right), thick neuronal tissues extend to submandibular acini, which are surrounded by abundant myoepithelial precursors. Myoepithelial precursors that surround parotid acini are sparse, whereas skeletal muscle fibers develop in close proximity to the gland. The figures of embryonic day 16 (E16) salivary glands depict representative acinar units and their surrounding tissues and do not show the entire gland. Figures are not drawn to scale.

Discussion

Our transcriptome data have identified cell populations that were surprisingly heterogeneous for this early single-bud stage of embryonic salivary development. This stage precedes the active differentiation stage, which is initiated at approximately E15 (Harunaga et al. 2014). More conspicuous transcriptional differences were found among mesenchymal than epithelial cells of the 2 glands. Gland-enriched neuronal genes contained neural crest markers known to be characteristic of different rhombomeres. Many parotid-enriched genes are known to control skeletal muscle differentiation and contraction. Distinct spatial localization patterns of neuronal and muscle cells in embryonic stages seem to foreshadow later differences in adult organ function.

Our findings contrast with some previous single-cell analyses of early branching organs, which showed largely homogeneous cells without clear clustering patterns (Farmer et al. 2017; Pal et al. 2017). This discrepancy can be attributed to greater inherent transcriptional homogeneity of early epithelial versus mesenchymal tissues, as well as to differences in the number of captured cells and sequencing depth. Attaining a sufficient number of epithelial cells initially posed a challenge due to the extremely small size of early embryonic salivary glands (100 to 150 µm in diameter) and the disproportionately small epithelium. Nonetheless, we managed to improve the data resolution by capturing ~2,000 to 3,800 epithelial cells per gland.

Potential Role of Muscles in Parotid Gland Secretion

Since the muscle precursor cells were localized to the periphery of the parotid gland, they are likely not an integral part of the gland per se. Nevertheless, adult parotid glands were also enriched with genes associated with skeletal muscle contraction (Gao et al. 2018), suggesting that these cells are not mere contaminants. Furthermore, the validity of our transcriptome data is supported by the expression of embryonic salivary gland mesenchymal markers that are known to have important roles in development, including Fgf10 (Jaskoll et al. 2005), Kdr, and Cdh5 (Kwon et al. 2017; Appendix Fig. 1).

The patterns of neuronal and muscle-precursor marker expression suggest distinct secretory mechanisms for the 2 glands. Given its extremely soft texture and sparsity of myoepithelial cells lining acini, the parotid may not possess strong intrinsic contractile ability. Biological mechanisms are often a product of mechanical processes combined with molecular signaling (Shyer et al. 2017). Saliva secretion is indeed partly a mechanical process (Lazaridou et al. 2012), and muscle contraction may aid parotid saliva secretion through myokine-mediated signals, mechanoreceptors, and mechanical force. This is consistent with the parotid providing most of the volume of stimulated saliva (Grundmann et al. 2009). The submandibular gland produces viscous secretion, which likely requires locally exerted strong contractile forces directly on acini (i.e., by myoepithelial cells innervated by the adjacent submandibular ganglion). In contrast, the parotid secretory process may be more global in nature, assisted by contraction of surrounding skeletal muscles (i.e., sternocleidomastoid and masseter).

While most facial muscles originate from cranial mesoderm, neural crest cells and muscle progenitor cells are extensively mixed during first branchial arch development (Grenier et al. 2009). Neural crest cells in fact govern mesodermal patterning, including organization of the sternocleidomastoid muscle (Vieux-Rochas et al. 2013), in which the parotid gland is embedded. Notably, Ngfr belonged to different clusters in the 2 glands—in the submandibular neuronal cluster and in the parotid muscle cluster. Ngfr marks skeletal muscle progenitor cells, and Ngfr-expressing cells are responsive to paracrine signals from neural crest or neuronal cells (Hicks et al. 2018). Given the highly organized neuronal and muscular structures in the early parotid gland, skeletal muscle development may be supported by a local microenvironment that includes neuronal cells.

Differential Enrichment of Neuronal Genes in Submandibular and Parotid Glands

Our transcriptome analyses demonstrated differences in the neuronal progenitor markers associated with the 2 types of salivary gland. Consistent with the recent finding of noradrenergic markers expressed in some Tubb3+ submandibular tissues (Teshima et al. 2019), we observed the coexpression of Tubb3 and noradrenergic neuron differentiation marker genes such as Hand2 in the embryonic submandibular gland. However, such noradrenergic marker expression was absent from the early parotid gland. Protein immunolocalization analyses of the noradrenergic neuron marker tyrosine hydroxylase confirmed its presence within the submandibular ganglion and complete absence from the Tubb3+ tissue surrounding the parotid gland at E13 (Appendix Fig. 5). One explanation that may account for the differential expression of neuronal genes in the 2 embryonic salivary glands is the difference in the cranial nerves and autonomic ganglia that supply innervation (Liebgott 2001). The submandibular epithelium and adjacent neuronal tissues grow synchronously. The submandibular ganglion and its projections enhance branching morphogenesis of the submandibular gland (Knox et al. 2010). The lack of a neuronal ganglion in the immediate vicinity of the parotid epithelium may be compensated for through the dense infiltration of Tubb3+ neuronal filaments that wrap around the evaginating parotid epithelial bud. This highly specific localization pattern of neuronal tissues surrounding the parotid epithelium is lost by E16. Thus, the role of such dense neuronal filaments in parotid development is likely time dependent—specifically around bud evagination—when the multipotent ectomesenchyme has been found to predominate in the salivary mesenchyme (Jaskoll et al. 2002). Because nothing is known about the role of neuronal cells during early parotid development, the source and function of these Tubb3+ tissues, as well as the role of the autonomic nervous system in embryonic parotid gland development, warrant further investigation.

Another potential explanation for the differential expression of neuronal genes lies in the different sources of neural crest–derived cells associated with each gland. The cranial neural crest is the largest contributor to the developing face, giving rise to tissues that include cartilage, bone, dermis, smooth muscle, and sympathetic and parasympathetic neurons (Kaucka et al. 2016). Each segment of rhombomeres is specified molecularly (Minoux and Rijli 2010). The spatial segregation of neural crest cell migration streams from rhombomeres to individual pharyngeal arches plays an important role in craniofacial morphogenesis (Minoux and Rijli 2010). For instance, cranial nerve patterning largely depends on specific rhombomeres (Kuratani and Eichele 1993; Kurosaka et al. 2015). Although our findings are preliminary, distinct innervation patterns for the submandibular and parotid glands also suggest differences in neural crest populations contributing to the development of each gland. Future studies could characterize further the cell populations that orchestrate development of each type of salivary gland.

Implications for Regenerative Therapy

Development of effective regenerative therapy requires identification of key developmental signaling processes and organ-specific progenitors. Sox9 and Foxc1 were recently used to generate salivary glands from mouse embryonic stem cells (Tanaka et al. 2018). The expression of these genes was found to be enriched in submandibular and parotid epithelia in our data (see Appendix Fig. 1), suggesting that they may contribute to general salivary gland epithelial differentiation rather than gland- or acinar cell type–specific differentiation. The gland-specific molecules in mesenchyme and epithelium identified in this study likely represent differentiation processes more specific to each type of gland. Accordingly, our data provide a resource and foundation for future studies utilizing imaging, lineage tracing, and genetic perturbation analyses that characterize the roles of key molecules that mold the salivary glands into functionally specialized organs.

Author Contributions

R. Sekiguchi, contributed to conception, design, data acquisition, analysis, and interpretation, drafted and critically revised the manuscript; D. Martin, contributed to data acquisition and analysis, drafted and critically revised the manuscript; Genomics and Computational Biology Core, contributed to data acquisition, drafted the manuscript; K.M. Yamada, contributed to conception and design, drafted and critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.